Intel IPU6 and IPU6 input system drivers
Subsystems (2)
v2 cover Intel IPU6 and IPU6 input system drivers
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
---
v1 -> v2:
- Add support with latest v4l2 async API
- Add support for latest bridge and Intel ivsc driver
- Fix lots of coding style issues not covered by checkpatch.pl
- Fix several bugs
- Fix several code issues found by Sparse and cppcheck
RFC -> v1:
- Add multiplexed streams support
- Use auxiliary bus to register IPU6 devices
- Add IPU6 hardware and driver overview documentation
- Updata IPU6 admin-guide documentation
- Update number of source pads and video nodes to support multiplexed streams
Bingbu Cao (15):
media: intel/ipu6: add Intel IPU6 PCI device driver
media: intel/ipu6: add IPU auxiliary devices
media: intel/ipu6: add IPU6 buttress interface driver
media: intel/ipu6: CPD parsing for get firmware components
media: intel/ipu6: add IPU6 DMA mapping API and MMU table
media: intel/ipu6: add syscom interfaces between firmware and driver
media: intel/ipu6: input system ABI between firmware and driver
media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-device
media: intel/ipu6: add the CSI2 DPHY implementation
media: intel/ipu6: add input system driver
media: intel/ipu6: input system video capture nodes
media: add Kconfig and Makefile for IPU6
MAINTAINERS: add maintainers for Intel IPU6 input system driver
Documentation: add Intel IPU6 ISYS driver admin-guide doc
Documentation: add documentation of Intel IPU6 driver and hardware
overview
Documentation/admin-guide/media/ipu6-isys.rst | 159 ++ .../admin-guide/media/ipu6_isys_graph.svg | 338 +++++ .../admin-guide/media/ipu6_isys_multi.svg | 1124 ++++++++++++++ .../admin-guide/media/v4l-drivers.rst | 1 + .../driver-api/media/drivers/index.rst | 1 + .../driver-api/media/drivers/ipu6.rst | 205 +++ MAINTAINERS | 10 + drivers/media/pci/intel/Kconfig | 1 + drivers/media/pci/intel/Makefile | 1 + drivers/media/pci/intel/ipu6/Kconfig | 17 + drivers/media/pci/intel/ipu6/Makefile | 23 + drivers/media/pci/intel/ipu6/ipu6-bus.c | 157 ++ drivers/media/pci/intel/ipu6/ipu6-bus.h | 50 + drivers/media/pci/intel/ipu6/ipu6-buttress.c | 900 +++++++++++ drivers/media/pci/intel/ipu6/ipu6-buttress.h | 107 ++ drivers/media/pci/intel/ipu6/ipu6-cpd.c | 355 +++++ drivers/media/pci/intel/ipu6/ipu6-cpd.h | 104 ++ drivers/media/pci/intel/ipu6/ipu6-dma.c | 491 ++++++ drivers/media/pci/intel/ipu6/ipu6-dma.h | 20 + drivers/media/pci/intel/ipu6/ipu6-fw-com.c | 411 +++++ drivers/media/pci/intel/ipu6/ipu6-fw-com.h | 47 + drivers/media/pci/intel/ipu6/ipu6-fw-isys.c | 484 ++++++ drivers/media/pci/intel/ipu6/ipu6-fw-isys.h | 570 +++++++ drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c | 654 ++++++++ drivers/media/pci/intel/ipu6/ipu6-isys-csi2.h | 81 + .../media/pci/intel/ipu6/ipu6-isys-dwc-phy.c | 533 +++++++ .../media/pci/intel/ipu6/ipu6-isys-jsl-phy.c | 237 +++ .../media/pci/intel/ipu6/ipu6-isys-mcd-phy.c | 712 +++++++++ .../media/pci/intel/ipu6/ipu6-isys-queue.c | 846 +++++++++++ .../media/pci/intel/ipu6/ipu6-isys-queue.h | 94 ++ .../media/pci/intel/ipu6/ipu6-isys-subdev.c | 379 +++++ .../media/pci/intel/ipu6/ipu6-isys-subdev.h | 58 + .../media/pci/intel/ipu6/ipu6-isys-video.c | 1239 +++++++++++++++ .../media/pci/intel/ipu6/ipu6-isys-video.h | 129 ++ drivers/media/pci/intel/ipu6/ipu6-isys.c | 1345 +++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-isys.h | 201 +++ drivers/media/pci/intel/ipu6/ipu6-mmu.c | 828 ++++++++++ drivers/media/pci/intel/ipu6/ipu6-mmu.h | 67 + .../intel/ipu6/ipu6-platform-buttress-regs.h | 230 +++ .../intel/ipu6/ipu6-platform-isys-csi2-reg.h | 187 +++ .../media/pci/intel/ipu6/ipu6-platform-regs.h | 179 +++ drivers/media/pci/intel/ipu6/ipu6.c | 955 ++++++++++++ drivers/media/pci/intel/ipu6/ipu6.h | 352 +++++ 43 files changed, 14882 insertions(+) create mode 100644 Documentation/admin-guide/media/ipu6-isys.rst create mode 100644 Documentation/admin-guide/media/ipu6_isys_graph.svg create mode 100644 Documentation/admin-guide/media/ipu6_isys_multi.svg create mode 100644 Documentation/driver-api/media/drivers/ipu6.rst create mode 100644 drivers/media/pci/intel/ipu6/Kconfig create mode 100644 drivers/media/pci/intel/ipu6/Makefile create mode 100644 drivers/media/pci/intel/ipu6/ipu6-bus.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-bus.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-buttress.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-buttress.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-cpd.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-cpd.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-dma.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-dma.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-isys.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-isys.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-csi2.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-jsl-phy.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-mcd-phy.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-queue.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-queue.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-subdev.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-subdev.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-video.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-video.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-mmu.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-mmu.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-platform-buttress-regs.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-platform-isys-csi2-reg.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-platform-regs.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6.h
Signature · 1 line
2.42.0
PATCH v2 1/15 media: intel/ipu6: add Intel IPU6 PCI device driver
From: Bingbu Cao <[email protected]>
Intel Image Processing Unit 6th Gen includes input and processing systems but the hardware presents itself as a single PCI device in system.
IPU6 PCI device driver basically does PCI configurations and load the firmware binary, initialises IPU virtual bus, and sets up platform specific variants to support multiple IPU6 devices in single device driver.
--- .../media/pci/intel/ipu6/ipu6-platform-regs.h | 179 ++++ drivers/media/pci/intel/ipu6/ipu6.c | 952 ++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6.h | 352 +++++++ 3 files changed, 1483 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-platform-regs.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6.h
drivers/media/pci/intel/ipu6/ipu6-platform-regs.h new file +179 -0
@@ -0,0 +1,179 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* Copyright (C) 2018 - 2023 Intel Corporation */ + +#ifndef IPU6_PLATFORM_REGS_H +#define IPU6_PLATFORM_REGS_H + +#include <linux/bits.h> + +/* +* IPU6 uses uniform address within IPU6, therefore all subsystem registers +* locates in one single space starts from 0 but in different sctions with +* different addresses, the subsystem offsets are defined to 0 as the +* register definition will have the address offset to 0. +*/ +#define IPU6_UNIFIED_OFFSET 0 + +#define IPU6_ISYS_IOMMU0_OFFSET 0x2e0000 +#define IPU6_ISYS_IOMMU1_OFFSET 0x2e0500 +#define IPU6_ISYS_IOMMUI_OFFSET 0x2e0a00 + +#define IPU6_PSYS_IOMMU0_OFFSET 0x1b0000 +#define IPU6_PSYS_IOMMU1_OFFSET 0x1b0700 +#define IPU6_PSYS_IOMMU1R_OFFSET 0x1b0e00 +#define IPU6_PSYS_IOMMUI_OFFSET 0x1b1500 + +/* the offset from IOMMU base register */ +#define IPU6_MMU_L1_STREAM_ID_REG_OFFSET 0x0c +#define IPU6_MMU_L2_STREAM_ID_REG_OFFSET 0x4c +#define IPU6_PSYS_MMU1W_L2_STREAM_ID_REG_OFFSET 0x8c + +#define IPU6_MMU_INFO_OFFSET 0x8 + +#define IPU6_ISYS_SPC_OFFSET 0x210000 + +#define IPU6SE_PSYS_SPC_OFFSET 0x110000 +#define IPU6_PSYS_SPC_OFFSET 0x118000 + +#define IPU6_ISYS_DMEM_OFFSET 0x200000 +#define IPU6_PSYS_DMEM_OFFSET 0x100000 + +#define IPU6_REG_ISYS_UNISPART_IRQ_EDGE 0x27c000 +#define IPU6_REG_ISYS_UNISPART_IRQ_MASK 0x27c004 +#define IPU6_REG_ISYS_UNISPART_IRQ_STATUS 0x27c008 +#define IPU6_REG_ISYS_UNISPART_IRQ_CLEAR 0x27c00c +#define IPU6_REG_ISYS_UNISPART_IRQ_ENABLE 0x27c010 +#define IPU6_REG_ISYS_UNISPART_IRQ_LEVEL_NOT_PULSE 0x27c014 +#define IPU6_REG_ISYS_UNISPART_SW_IRQ_REG 0x27c414 +#define IPU6_REG_ISYS_UNISPART_SW_IRQ_MUX_REG 0x27c418 +#define IPU6_ISYS_UNISPART_IRQ_CSI0 BIT(2) +#define IPU6_ISYS_UNISPART_IRQ_CSI1 BIT(3) +#define IPU6_ISYS_UNISPART_IRQ_SW BIT(22) + +#define IPU6_REG_ISYS_ISL_TOP_IRQ_EDGE 0x2b0200 +#define IPU6_REG_ISYS_ISL_TOP_IRQ_MASK 0x2b0204 +#define IPU6_REG_ISYS_ISL_TOP_IRQ_STATUS 0x2b0208 +#define IPU6_REG_ISYS_ISL_TOP_IRQ_CLEAR 0x2b020c +#define IPU6_REG_ISYS_ISL_TOP_IRQ_ENABLE 0x2b0210 +#define IPU6_REG_ISYS_ISL_TOP_IRQ_LEVEL_NOT_PULSE 0x2b0214 + +#define IPU6_REG_ISYS_CMPR_TOP_IRQ_EDGE 0x2d2100 +#define IPU6_REG_ISYS_CMPR_TOP_IRQ_MASK 0x2d2104 +#define IPU6_REG_ISYS_CMPR_TOP_IRQ_STATUS 0x2d2108 +#define IPU6_REG_ISYS_CMPR_TOP_IRQ_CLEAR 0x2d210c +#define IPU6_REG_ISYS_CMPR_TOP_IRQ_ENABLE 0x2d2110 +#define IPU6_REG_ISYS_CMPR_TOP_IRQ_LEVEL_NOT_PULSE 0x2d2114 + +/* CDC Burst collector thresholds for isys - 3 FIFOs i = 0..2 */ +#define IPU6_REG_ISYS_CDC_THRESHOLD (i) (0x27c400 + ((i) * 4)) + +#define IPU6_CSI_IRQ_NUM_PER_PIPE 4 +#define IPU6SE_ISYS_CSI_PORT_NUM 4 +#define IPU6_ISYS_CSI_PORT_NUM 8 + +#define IPU6_ISYS_CSI_PORT_IRQ (irq_num) BIT(irq_num) + +/* PKG DIR OFFSET in IMR in secure mode */ +#define IPU6_PKG_DIR_IMR_OFFSET 0x40 + +#define IPU6_ISYS_REG_SPC_STATUS_CTRL 0x0 + +#define IPU6_ISYS_SPC_STATUS_START BIT(1) +#define IPU6_ISYS_SPC_STATUS_RUN BIT(3) +#define IPU6_ISYS_SPC_STATUS_READY BIT(5) +#define IPU6_ISYS_SPC_STATUS_CTRL_ICACHE_INVALIDATE BIT(12) +#define IPU6_ISYS_SPC_STATUS_ICACHE_PREFETCH BIT(13) + +#define IPU6_PSYS_REG_SPC_STATUS_CTRL 0x0 +#define IPU6_PSYS_REG_SPC_START_PC 0x4 +#define IPU6_PSYS_REG_SPC_ICACHE_BASE 0x10 +#define IPU6_REG_PSYS_INFO_SEG_0_CONFIG_ICACHE_MASTER 0x14 + +#define IPU6_PSYS_SPC_STATUS_START BIT(1) +#define IPU6_PSYS_SPC_STATUS_RUN BIT(3) +#define IPU6_PSYS_SPC_STATUS_READY BIT(5) +#define IPU6_PSYS_SPC_STATUS_CTRL_ICACHE_INVALIDATE BIT(12) +#define IPU6_PSYS_SPC_STATUS_ICACHE_PREFETCH BIT(13) + +#define IPU6_PSYS_REG_SPP0_STATUS_CTRL 0x20000 + +#define IPU6_INFO_ENABLE_SNOOP BIT(0) +#define IPU6_INFO_DEC_FORCE_FLUSH BIT(1) +#define IPU6_INFO_DEC_PASS_THROUGH BIT(2) +#define IPU6_INFO_ZLW BIT(3) +#define IPU6_INFO_REQUEST_DESTINATION_IOSF BIT(9) +#define IPU6_INFO_IMR_BASE BIT(10) +#define IPU6_INFO_IMR_DESTINED BIT(11) + +#define IPU6_INFO_REQUEST_DESTINATION_PRIMARY IPU6_INFO_REQUEST_DESTINATION_IOSF + +/* +* s2m_pixel_soc_pixel_remapping is dedicated for the enabling of the +* pixel s2m remp ability.Remap here means that s2m rearange the order +* of the pixels in each 4 pixels group. +* For examle, mirroring remping means that if input's 4 first pixels +* are 1 2 3 4 then in output we should see 4 3 2 1 in this 4 first pixels. +* 0xE4 is from s2m MAS document. It means no remapping. +*/ +#define S2M_PIXEL_SOC_PIXEL_REMAPPING_FLAG_NO_REMAPPING 0xe4 +/* +* csi_be_soc_pixel_remapping is for the enabling of the pixel remapping. +* This remapping is exactly like the stream2mmio remapping. +*/ +#define CSI_BE_SOC_PIXEL_REMAPPING_FLAG_NO_REMAPPING 0xe4 + +#define IPU6_REG_DMA_TOP_AB_GROUP1_BASE_ADDR 0x1ae000 +#define IPU6_REG_DMA_TOP_AB_GROUP2_BASE_ADDR 0x1af000 +#define IPU6_REG_DMA_TOP_AB_RING_MIN_OFFSET (n) (0x4 + (n) * 0xc) +#define IPU6_REG_DMA_TOP_AB_RING_MAX_OFFSET (n) (0x8 + (n) * 0xc) +#define IPU6_REG_DMA_TOP_AB_RING_ACCESS_OFFSET (n) (0xc + (n) * 0xc) + +enum ipu6_device_ab_group1_target_id { +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R0_SPC_DMEM , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R1_SPC_DMEM , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R2_SPC_DMEM , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R3_SPC_STATUS_REG , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R4_SPC_MASTER_BASE_ADDR , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R5_SPC_PC_STALL , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R6_SPC_EQ , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R7_SPC_RESERVED , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R8_SPC_RESERVED , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R9_SPP0 , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R10_SPP1 , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R11_CENTRAL_R1 , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R12_IRQ , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R13_CENTRAL_R2 , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R14_DMA , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R15_DMA , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R16_GP , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R17_ZLW_INSERTER , +IPU6_DEVICE_AB_GROUP1_TARGET_ID_R18_AB , +}; + +enum nci_ab_access_mode { +NCI_AB_ACCESS_MODE_RW ,/* read & write */ +NCI_AB_ACCESS_MODE_RO ,/* read only */ +NCI_AB_ACCESS_MODE_WO ,/* write only */ +NCI_AB_ACCESS_MODE_NA ,/* No access at all */ +}; + +/* IRQ-related registers in PSYS */ +#define IPU6_REG_PSYS_GPDEV_IRQ_EDGE 0x1aa200 +#define IPU6_REG_PSYS_GPDEV_IRQ_MASK 0x1aa204 +#define IPU6_REG_PSYS_GPDEV_IRQ_STATUS 0x1aa208 +#define IPU6_REG_PSYS_GPDEV_IRQ_CLEAR 0x1aa20c +#define IPU6_REG_PSYS_GPDEV_IRQ_ENABLE 0x1aa210 +#define IPU6_REG_PSYS_GPDEV_IRQ_LEVEL_NOT_PULSE 0x1aa214 +/* There are 8 FW interrupts, n = 0..7 */ +#define IPU6_PSYS_GPDEV_FWIRQ0 5 +#define IPU6_PSYS_GPDEV_FWIRQ1 6 +#define IPU6_PSYS_GPDEV_FWIRQ2 7 +#define IPU6_PSYS_GPDEV_FWIRQ3 8 +#define IPU6_PSYS_GPDEV_FWIRQ4 9 +#define IPU6_PSYS_GPDEV_FWIRQ5 10 +#define IPU6_PSYS_GPDEV_FWIRQ6 11 +#define IPU6_PSYS_GPDEV_FWIRQ7 12 +#define IPU6_PSYS_GPDEV_IRQ_FWIRQ (n) BIT(n) +#define IPU6_REG_PSYS_GPDEV_FWIRQ (n) (4 * (n) + 0x1aa100) + +#endif /* IPU6_PLATFORM_REGS_H */
drivers/media/pci/intel/ipu6/ipu6.c new file +952 -0
952 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6.h new file +352 -0
352 lines not on this page · raw message
Signature · 1 line
2.42.0
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
Hi Bingbu,
Thanks for the updated series. I've rebased my work on this series, and ran into the following issue.
Quoted 1 line · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
+static int
ipu6_pci_config_setup(struct pci_dev *dev, u8 hw_ver)
...
drivers/media/pci/intel/ipu6/ipu6.c line 539
Quoted 6 lines
+ ret =
pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_MSI); + if (ret)
+ dev_err_probe(&dev->dev, ret, "Request msi failed");drivers/media/pci/intel/ipu6/ipu6.c new file +3
-0+ + return ret; +}
pci_alloc_irq_vectors returns number of irqs, so I think it should be something like this instead:
ret = pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_MSI);
if (ret<0) {
dev_err_probe(&dev->dev, ret, "Request msi failed");
return ret;
}
return 0;/Andreas
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
Quoted 1 line · Andreas Helbech Kleist
On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
...
Quoted 9 lines
drivers/media/pci/intel/ipu6/ipu6.c line 539
Quoted 6 lines
+ ret =
pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_MSI); + if (ret)
+ dev_err_probe(&dev->dev, ret, "Request msi failed");drivers/media/pci/intel/ipu6/ipu6.c new file +3
-0+ + return ret; +}
pci_alloc_irq_vectorsreturns number of irqs, so I think it should be something like this instead:
Indeed, good catch!
Quoted 4 lines
ret = pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_MSI);
if (ret<0) {
dev_err_probe(&dev->dev, ret, "Request msi failed");
return ret;return dev_err_probe();
saves a line.
Quoted 1 line
}
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
In reply to [email protected]Hi,
Quoted 5 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Intel Image Processing Unit 6th Gen includes input and processing systems but the hardware presents itself as a single PCI device in system.
Quoted 8 lines
IPU6 PCI device driver basically does PCI configurations and load the firmware binary, initialises IPU virtual bus, and sets up platform specific variants to support multiple IPU6 devices in single device driver.
Signed-off-by Bingbu Cao <[email protected]>Reported-by Hans de Goede <[email protected]>Suggested-by Andreas Helbech Kleist <[email protected]>
Thank you. Just one small remark, not a full review:
<snip>
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +0
-0
<snip>
Quoted 14 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +14
-0+static struct ipu6_bus_device * +ipu6_isys_init (struct pci_dev *pdev,struct device *parent, +struct ipu6_buttress_ctrl *ctrl,void __iomem *base, +const struct ipu6_isys_internal_pdata *ipdata) +{ +struct fwnode_handle *fwnode =dev_fwnode (&pdev->dev); +struct ipu6_bus_device *isys_adev; +struct ipu6_isys_pdata *pdata; +int ret; + +/* check fwnode at first, fallback into bridge if no fwnode graph */ + ret =ipu6_isys_check_fwnode_graph (fwnode); +if (ret) { +if (fwnode && !IS_ERR_OR_NULL (fwnode->secondary)) {
drivers/media/pci/intel/ipu6/ipu6.c line 430
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +10
-0+dev_err (&pdev->dev, +"fwnode graph has no endpoints connection\n" ); +return ERR_PTR (-EINVAL ); + } + + ret =ipu_bridge_init (&pdev->dev, ipu_bridge_parse_ssdb); +if (ret) { +if (ret != -EPROBE_DEFER ) +dev_err_probe (&pdev->dev, ret, +"IPU6 bridge init failed\n" );
dev_err_probe() already silences logging of -EPROBE_DEFER errors itself, so this "if (ret != -EPROBE_DEFER)" guard here is not necessary.
Regards,
Hans
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
Hans,
Quoted 49 lines · Hans de Goede
Hi,
Quoted 5 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Intel Image Processing Unit 6th Gen includes input and processing systems but the hardware presents itself as a single PCI device in system.
Quoted 8 lines
IPU6 PCI device driver basically does PCI configurations and load the firmware binary, initialises IPU virtual bus, and sets up platform specific variants to support multiple IPU6 devices in single device driver.
Signed-off-by Bingbu Cao <[email protected]>Reported-by Hans de Goede <[email protected]>Suggested-by Andreas Helbech Kleist <[email protected]>Thank you. Just one small remark, not a full review:
<snip>
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +0
-0<snip>
Quoted 19 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +19
-0+static struct ipu6_bus_device * +ipu6_isys_init (struct pci_dev *pdev,struct device *parent, +struct ipu6_buttress_ctrl *ctrl,void __iomem *base, +const struct ipu6_isys_internal_pdata *ipdata) +{ +struct fwnode_handle *fwnode =dev_fwnode (&pdev->dev); +struct ipu6_bus_device *isys_adev; +struct ipu6_isys_pdata *pdata; +int ret; + +/* check fwnode at first, fallback into bridge if no fwnode graph */ + ret =ipu6_isys_check_fwnode_graph (fwnode); +if (ret) { +if (fwnode && !IS_ERR_OR_NULL (fwnode->secondary)) { +dev_err (&pdev->dev, +"fwnode graph has no endpoints connection\n" ); +return ERR_PTR (-EINVAL ); + } +
Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6.c line 430
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +5
-0+ ret =ipu_bridge_init (&pdev->dev, ipu_bridge_parse_ssdb); +if (ret) { +if (ret != -EPROBE_DEFER ) +dev_err_probe (&pdev->dev, ret, +"IPU6 bridge init failed\n" );
dev_err_probe()already silences logging of -EPROBE_DEFER errors itself, so this "if (ret != -EPROBE_DEFER)" guard here is not necessary.
Good catch, thanks.
Quoted 8 lines
Regards,
Hans
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
In reply to [email protected]Hi,
Quoted 21 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Intel Image Processing Unit 6th Gen includes input and processing systems but the hardware presents itself as a single PCI device in system.
IPU6 PCI device driver basically does PCI configurations and load the firmware binary, initialises IPU virtual bus, and sets up platform specific variants to support multiple IPU6 devices in single device driver.
Signed-off-by Bingbu Cao <[email protected]>Reported-by Hans de Goede <[email protected]>Suggested-by Andreas Helbech Kleist <[email protected]>--- .../media/pci/intel/ipu6/ipu6-platform-regs.h | 179 ++++ drivers/media/pci/intel/ipu6/ipu6.c | 952 ++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6.h | 352 +++++++ 3 files changed, 1483 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-platform-regs.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6.h
<snip>
Quoted 6 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +0
-0@@ -0,0 +1,952 @@
<snip>
Quoted 7 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +7
-0+static int ipu6_pci_config_setup (struct pci_dev *dev,u8 hw_ver) +{ +int ret; + +/* disable IPU6 PCI ATS on mtl ES2 */ +if (is_ipu6ep_mtl (hw_ver) && boot_cpu_data.x86_stepping ==0x2 && +pci_ats_supported (dev))
drivers/media/pci/intel/ipu6/ipu6.c line 534
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +10
-0+pci_disable_ats (dev); + +/* No PCI msi capability for IPU6EP */ +if (is_ipu6ep (hw_ver) ||is_ipu6ep_mtl (hw_ver)) { +/* likely do nothing as msi not enabled by default */ +pci_disable_msi (dev); +return 0 ; + } + + ret =pci_alloc_irq_vectors (dev,1 ,1 ,PCI_IRQ_MSI );
This does not work on TGL systems (and is not reached on ADL and RPL).
The out of tree driver instead uses:
ret = pci_enable_msi(dev);
and that does work correctly on TGL.
Regards,
Hans
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
Quoted 1 line · Hans de Goede
On 10/24/23 13:29, [email protected] wrote:
...
Quoted 9 lines
drivers/media/pci/intel/ipu6/ipu6.c line 534
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6.c new file +1
-0+ ret =pci_alloc_irq_vectors (dev,1 ,1 ,PCI_IRQ_MSI );This does not work on TGL systems (and is not reached on ADL and RPL).
The out of tree driver instead uses:
ret = pci_enable_msi(dev);and that does work correctly on TGL.
Does it mean TGL uses legacy IRQ?
In that case we need to simply change the flag to allow that as the used API is what documentation says needs to be used and pci_enable_msi() is deprecated.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
In reply to Hans de GoedeHi Hans,
Quoted 64 lines · Hans de Goede
Hi,
Quoted 25 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Intel Image Processing Unit 6th Gen includes input and processing systems
but the hardware presents itself as a single PCI device in system.IPU6 PCI device driver basically does PCI configurations and load the firmware binary, initialises IPU virtual bus, and sets up platform
specific variants to support multiple IPU6 devices in single device driver.Signed-off-by Bingbu Cao <[email protected]>Reported-by Hans de Goede <[email protected]>Suggested-by Andreas Helbech Kleist <[email protected]>---
.../media/pci/intel/ipu6/ipu6-platform-regs.h | 179 ++++ drivers/media/pci/intel/ipu6/ipu6.c | 952
++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6.h | 352 +++++++3 files changed, 1483 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-platform-regs.h
create mode 100644drivers/media/pci/intel/ipu6/ipu6.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6.h<snip>
Quoted 7 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +0
-0@@ -0,0 +1,952 @@<snip>
Quoted 11 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +6
-0+static int ipu6_pci_config_setup(struct pci_dev *dev, u8 hw_ver) +{ + int ret; + + /* disable IPU6 PCI ATS on mtl ES2 */ + if (is_ipu6ep_mtl(hw_ver) && boot_cpu_data.x86_stepping ==
0x2&&
+pci_ats_supported(dev))
+ pci_disable_ats(dev);
+
+ /* No PCI msi capability for IPU6EP */Quoted 8 lines
+ if (
is_ipu6ep(hw_ver)||is_ipu6ep_mtl(hw_ver)) { + /* likely do nothing as msi not enabled by default */
+ pci_disable_msi(dev);
+ return 0;
+ }
+
+ ret =pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_MSI);This does not work on TGL systems (and is not reached on ADL and RPL).
Quoted 5 lines
The out of tree driver instead uses:
ret =
pci_enable_msi(dev);and that does work correctly on TGL.
Did you see my previous (25//10) comment on the same lines?
pci_alloc_irq_vectors returns number of irqs, so the "if (ret)" check following the quoted line is wrong.
/Andreas
RE: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
Andreas,
------------------------------------------------------------------------
BRs,
Bingbu Cao
Quoted 80 lines
-----Original Message-----
From: Andreas Helbech Kleist <[email protected]>
Sent: Wednesday, November 8, 2023 10:11 PM
To: Hans de Goede <[email protected]>; Cao, Bingbu <[email protected]>; [email protected]; [email protected]; [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Qiu, Tian Shu <[email protected]>; Wang, Hongju <[email protected]>
Subject: Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driverHi Hans,
Quoted 63 lines · Hans de Goede
Hi,
Quoted 24 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Intel Image Processing Unit 6th Gen includes input and processing systems but the hardware presents itself as a single PCI device in system.
IPU6 PCI device driver basically does PCI configurations and load the firmware binary, initialises IPU virtual bus, and sets up platform specific variants to support multiple IPU6 devices in single device driver.
Signed-off-by Bingbu Cao <[email protected]>Reported-by Hans de Goede <[email protected]>Suggested-by Andreas Helbech Kleist <[email protected]>---
.../media/pci/intel/ipu6/ipu6-platform-regs.h | 179 ++++ drivers/media/pci/intel/ipu6/ipu6.c | 952
++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6.h | 352 +++++++3 files changed, 1483 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-platform-regs.h
create mode 100644drivers/media/pci/intel/ipu6/ipu6.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6.h<snip>
Quoted 7 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +0
-0@@ -0,0 +1,952 @@<snip>
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +5
-0+static int ipu6_pci_config_setup(struct pci_dev *dev, u8 hw_ver) { + int ret; + + /* disable IPU6 PCI ATS on mtl ES2 */ + if (is_ipu6ep_mtl(hw_ver) && boot_cpu_data.x86_stepping ==
0x2&&
+pci_ats_supported(dev))
+ pci_disable_ats(dev);
+
+ /* No PCI msi capability for IPU6EP */Quoted 8 lines
+ if (
is_ipu6ep(hw_ver)||is_ipu6ep_mtl(hw_ver)) { + /* likely do nothing as msi not enabled by default */
+ pci_disable_msi(dev);
+ return 0;
+ }
+
+ ret =pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_MSI);This does not work on TGL systems (and is not reached on ADL and RPL).
The out of tree driver instead uses:
Quoted 8 lines
Quoted 3 lines
ret =
pci_enable_msi(dev);and that does work correctly on TGL.
Did you see my previous (25//10) comment on the same lines?
pci_alloc_irq_vectorsreturns number of irqs, so the "if (ret)" check following the quoted line is wrong.
Andreas,
You are right, the actual problem is the return value checking is wrong.
Hans, thanks for your report. I don't have TGL device now. ☹ I will fix it in next version.
Quoted 2 lines
/Andreas
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
In reply to Andreas Helbech KleistHi,
Quoted 69 lines · Andreas Helbech Kleist
Hi Hans,
Quoted 66 lines · Hans de Goede
Hi,
Quoted 25 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Intel Image Processing Unit 6th Gen includes input and processing systems
but the hardware presents itself as a single PCI device in system.IPU6 PCI device driver basically does PCI configurations and load the firmware binary, initialises IPU virtual bus, and sets up platform
specific variants to support multiple IPU6 devices in single device driver.Signed-off-by Bingbu Cao <[email protected]>Reported-by Hans de Goede <[email protected]>Suggested-by Andreas Helbech Kleist <[email protected]>---
.../media/pci/intel/ipu6/ipu6-platform-regs.h | 179 ++++ drivers/media/pci/intel/ipu6/ipu6.c | 952
++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6.h | 352 +++++++3 files changed, 1483 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-platform-regs.h
create mode 100644drivers/media/pci/intel/ipu6/ipu6.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6.h<snip>
Quoted 7 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +0
-0@@ -0,0 +1,952 @@<snip>
Quoted 11 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +6
-0+static int ipu6_pci_config_setup(struct pci_dev *dev, u8 hw_ver) +{ + int ret; + + /* disable IPU6 PCI ATS on mtl ES2 */ + if (is_ipu6ep_mtl(hw_ver) && boot_cpu_data.x86_stepping ==
0x2&&
+pci_ats_supported(dev))
+ pci_disable_ats(dev);
+
+ /* No PCI msi capability for IPU6EP */Quoted 8 lines
+ if (
is_ipu6ep(hw_ver)||is_ipu6ep_mtl(hw_ver)) { + /* likely do nothing as msi not enabled by default */
+ pci_disable_msi(dev);
+ return 0;
+ }
+
+ ret =pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_MSI);This does not work on TGL systems (and is not reached on ADL and RPL).
The out of tree driver instead uses:
Quoted 8 lines
Quoted 3 lines
ret =
pci_enable_msi(dev);and that does work correctly on TGL.
Did you see my previous (25//10) comment on the same lines?
pci_alloc_irq_vectorsreturns number of irqs, so the "if (ret)" check following the quoted line is wrong.
Ah right, so a proper fix for this would look something like this:
From 34de6611d3482d6695dd73eddf7bf3dc1f96883c Mon Sep 17 00:00:00 2001
From: Hans de Goede <[email protected]>
Date: Tue, 7 Nov 2023 11:40:29 +0100
Subject: [PATCH] media: ipu6: Fix interrupts not working on TGL
pci_alloc_irq_vectors() returns the number of successfully allocated interrupts. Fix the error checking to handle this correctly.
--- drivers/media/pci/intel/ipu6/ipu6.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/media/pci/intel/ipu6/ipu6.c +4 -2
@@ -532,10 +532,12 @@ static int ipu6_pci_config_setup(struct pci_dev *dev, u8 hw_ver) } ret =pci_alloc_irq_vectors (dev,1 ,1 ,PCI_IRQ_MSI ); -if (ret) +if (ret <0 ) {dev_err_probe (&dev->dev, ret,"Request msi failed" ); +return ret; + } -return ret; +return 0 ; }static void ipu6_configure_vc_mechanism (struct ipu6_device *isp)
Signature · 9 lines
2.41.0 Bingbu, please squash something like the above into the next version of this series. Regards, Hans
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
Quoted 3 lines · Hans de Goede
Quoted 2 lines · Andreas Helbech Kleist
Quoted 1 line · Hans de Goede
On 10/24/23 13:29, [email protected] wrote:
...
drivers/media/pci/intel/ipu6/ipu6.c static int ipu6_pci_config_setup(struct pci_dev *dev, u8 hw_ver) · line 537
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6.c +2
-1ret =pci_alloc_irq_vectors (dev,1 ,1 ,PCI_IRQ_MSI ); -if (ret) +if (ret <0 ) {dev_err_probe (&dev->dev, ret,"Request msi failed" ); +return ret;
Even further:
return dev_err_probe(&dev->dev, ret, "Request msi failed");
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6.c +1
-0+ }
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
Andy and Hans,
Thanks.
Quoted 15 lines · Andy Shevchenko
Quoted 3 lines · Hans de Goede
Quoted 2 lines · Andreas Helbech Kleist
Quoted 1 line · Hans de Goede
On 10/24/23 13:29, [email protected] wrote:
...
drivers/media/pci/intel/ipu6/ipu6.c static int ipu6_pci_config_setup(struct pci_dev *dev, u8 hw_ver) · line 537
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6.c +2
-1ret =pci_alloc_irq_vectors (dev,1 ,1 ,PCI_IRQ_MSI ); -if (ret) +if (ret <0 ) {dev_err_probe (&dev->dev, ret,"Request msi failed" ); +return ret;Even further:
Quoted 1 line
return dev_err_probe(&dev->dev, ret, "Request msi failed");
Yes, I have updated the code locally and will send in v3.
Quoted 3 lines
Quoted 2 lines
drivers/media/pci/intel/ipu6/ipu6.c +1
-0+ }
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 01/15] media: intel/ipu6: add Intel IPU6 PCI device driver
In reply to [email protected]Hi,
Quoted 24 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Intel Image Processing Unit 6th Gen includes input and processing systems
but the hardware presents itself as a single PCI device in system.IPU6 PCI device driver basically does PCI configurations and load the firmware binary, initialises IPU virtual bus, and sets up platform
specific variants to support multiple IPU6 devices in single device driver.Signed-off-by Bingbu Cao <[email protected]>Reported-by Hans de Goede <[email protected]>Suggested-by Andreas Helbech Kleist <[email protected]>---
.../media/pci/intel/ipu6/ipu6-platform-regs.h | 179 ++++ drivers/media/pci/intel/ipu6/ipu6.c | 952
++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6.h | 352 +++++++3 files changed, 1483 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-platform-regs.hcreate mode 100644drivers/media/pci/intel/ipu6/ipu6.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6.h
...
Quoted 9 lines
drivers/media/pci/intel/ipu6/ipu6.c new file +9
-0+static void ipu6_internal_pdata_init (struct ipu6_device *isp) +{ + u8 hw_ver = isp->hw_ver; + + isys_ipdata.num_parallel_streams = IPU6_ISYS_NUM_STREAMS; + isys_ipdata.sram_gran_shift = IPU6_SRAM_GRANULARITY_SHIFT; + isys_ipdata.sram_gran_size = IPU6_SRAM_GRANULARITY_SIZE; + isys_ipdata.max_sram_size = IPU6_MAX_SRAM_SIZE; + isys_ipdata.sensor_type_start =
Quoted 3 lines
IPU6_FW_ISYS_SENSOR_TYPE_START;
+ isys_ipdata.sensor_type_end =IPU6_FW_ISYS_SENSOR_TYPE_END; + isys_ipdata.max_streams =IPU6_ISYS_NUM_STREAMS;
num_parallel_streams and max_streams are set to the same value here and to another (but also equal) value for ipu6se.
num_parallel_streams is then passed through ipu6_fw_isys_init to ipu6_isys_fwcom_cfg_init to then be clamp'ed to 1..max_streams. It seems like they are the same, and thus a lot of code could be removed.
/Andreas
PATCH v2 2/15 media: intel/ipu6: add IPU auxiliary devices
In reply to [email protected]From: Bingbu Cao <[email protected]>
Even the IPU input system and processing system are in a single PCI device, each system has its own power sequence, the processing system power up depends on the input system power up.
Besides, input system and processing system have their own MMU hardware for IPU DMA address mapping.
Register the IS/PS devices on auxiliary bus and attach power domain to implement the power sequence dependency.
--- drivers/media/pci/intel/ipu6/ipu6-bus.c | 157 ++++++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-bus.h | 50 ++++++++ drivers/media/pci/intel/ipu6/ipu6.c | 5 +- 3 files changed, 211 insertions(+), 1 deletion(-) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-bus.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-bus.h
drivers/media/pci/intel/ipu6/ipu6-bus.c new file +157 -0
@@ -0,0 +1,157 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <linux/pci.h> +#include <linux/pm_domain.h> +#include <linux/pm_runtime.h> + +#include "ipu6.h" +#include "ipu6-bus.h" +#include "ipu6-dma.h" + +static int bus_pm_runtime_suspend (struct device *dev) +{ +struct ipu6_bus_device *adev =to_ipu6_bus_device (dev); +int ret; + + ret =pm_generic_runtime_suspend (dev); +if (ret) +return ret; + + ret =ipu6_buttress_power (dev, adev->ctrl, false); +if (!ret) +return 0 ; + +dev_err (dev,"power down failed!\n" ); + +/* Powering down failed, attempt to resume device now */ + ret =pm_generic_runtime_resume (dev); +if (!ret) +return -EBUSY ; + +return -EIO ; +} + +static int bus_pm_runtime_resume (struct device *dev) +{ +struct ipu6_bus_device *adev =to_ipu6_bus_device (dev); +int ret; + + ret =ipu6_buttress_power (dev, adev->ctrl, true); +if (ret) +return ret; + + ret =pm_generic_runtime_resume (dev); +if (ret) + goto out_err; + +return 0 ; + +out_err: +ipu6_buttress_power (dev, adev->ctrl, false); + +return -EBUSY ; +} + +static struct dev_pm_domain ipu6_bus_pm_domain = { + .ops = { + .runtime_suspend = bus_pm_runtime_suspend, + .runtime_resume = bus_pm_runtime_resume, + }, +}; + +static DEFINE_MUTEX (ipu6_bus_mutex); + +static void ipu6_bus_release (struct device *dev) +{ +struct ipu6_bus_device *adev =to_ipu6_bus_device (dev); + +kfree (adev->pdata); +kfree (adev); +} + +struct ipu6_bus_device * +ipu6_bus_initialize_device (struct pci_dev *pdev,struct device *parent, +void *pdata,struct ipu6_buttress_ctrl *ctrl, +char *name) +{ +struct auxiliary_device *auxdev; +struct ipu6_bus_device *adev; +struct ipu6_device *isp =pci_get_drvdata (pdev); +int ret; + + adev =kzalloc (sizeof (*adev),GFP_KERNEL ); +if (!adev) +return ERR_PTR (-ENOMEM ); + + adev->dma_mask =DMA_BIT_MASK (isp->secure_mode ?IPU6_MMU_ADDR_BITS : +IPU6_MMU_ADDR_BITS_NON_SECURE ); + adev->isp = isp; + adev->ctrl = ctrl; + adev->pdata = pdata; + auxdev = &adev->auxdev; + auxdev->name = name; + auxdev->id = (pci_domain_nr (pdev->bus) <<16 ) | +PCI_DEVID (pdev->bus->number, pdev->devfn); + + auxdev->dev.parent = parent; + auxdev->dev.release = ipu6_bus_release; + auxdev->dev.dma_ops = &ipu6_dma_ops; + auxdev->dev.dma_mask = &adev->dma_mask; + auxdev->dev.dma_parms = pdev->dev.dma_parms; + auxdev->dev.coherent_dma_mask = adev->dma_mask; + + ret =auxiliary_device_init (auxdev); +if (ret <0 ) { +dev_err (&isp->pdev->dev,"auxiliary device init failed (%d)\n" , + ret); +kfree (adev); +return ERR_PTR (ret); + } + +dev_pm_domain_set (&auxdev->dev, &ipu6_bus_pm_domain); + +pm_runtime_forbid (&adev->auxdev.dev); +pm_runtime_enable (&adev->auxdev.dev); + +return adev; +} + +int ipu6_bus_add_device (struct ipu6_bus_device *adev) +{ +struct auxiliary_device *auxdev = &adev->auxdev; +int ret; + + ret =auxiliary_device_add (auxdev); +if (ret) { +auxiliary_device_uninit (auxdev); +return ret; + } + +mutex_lock (&ipu6_bus_mutex); +list_add (&adev->list, &adev->isp->devices); +mutex_unlock (&ipu6_bus_mutex); + +pm_runtime_allow (&auxdev->dev); + +return 0 ; +} + +void ipu6_bus_del_devices (struct pci_dev *pdev) +{ +struct ipu6_device *isp =pci_get_drvdata (pdev); +struct ipu6_bus_device *adev, *save; + +mutex_lock (&ipu6_bus_mutex); + +list_for_each_entry_safe (adev, save, &isp->devices, list) { +pm_runtime_disable (&adev->auxdev.dev); +list_del (&adev->list); +auxiliary_device_delete (&adev->auxdev); +auxiliary_device_uninit (&adev->auxdev); + } + +mutex_unlock (&ipu6_bus_mutex); +}
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +50 -0
50 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6.c +4 -1
@@ -666,7 +666,10 @@ static int ipu6_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)return dev_err_probe (&pdev->dev, ret,"Failed to set DMA mask\n" ); -dma_set_max_seg_size (&pdev->dev,UINT_MAX ); + ret =dma_set_max_seg_size (&pdev->dev,UINT_MAX ); +if (ret) +return dev_err_probe (&pdev->dev, ret, +"Failed to set max_seg_size\n" ); ret =ipu6_pci_config_setup (pdev, isp->hw_ver);if (ret)
Signature · 1 line
2.42.0
Re: [PATCH v2 02/15] media: intel/ipu6: add IPU auxiliary devices
Quoted 11 lines · On Tue, Oct 24, 2023 at 07:29:11PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Even the IPU input system and processing system are in a single PCI device, each system has its own power sequence, the processing system power up depends on the input system power up.
Besides, input system and processing system have their own MMU hardware for IPU DMA address mapping.
Register the IS/PS devices on auxiliary bus and attach power domain to implement the power sequence dependency.
...
Seems again poor / random list of header inclusions...
Please, go through your files and use IWYU (Include What You Use) principle in them.
Quoted 1 line
+#include <
linux/pci.h>
auxiliary_bus.h, err.h, list.h, mutex.h, dev_printk.h, slab.h, dma-mapping.h are missing. You are lucky it does compile.
...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 7
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +4
-0+#ifndef IPU6_BUS_H +#define IPU6_BUS_H + +#include <linux/auxiliary_bus.h>
...Especially for headers which will affect the compilation time.
Quoted 1 line
+#include <
linux/pci.h>
This is not used.
...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 18
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+struct list_head list;
+ types.h
drivers/media/pci/intel/ipu6/ipu6-bus.h line 25
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+struct sg_table fw_sgt;
+ scatterlist.h
drivers/media/pci/intel/ipu6/ipu6-bus.h line 27
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+dma_addr_t pkg_dir_dma_addr;
types.h
...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 33
Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +3
-0+struct ipu6_auxdrv_data { +irqreturn_t (*isr)(struct ipu6_bus_device *adev); +irqreturn_t (*isr_threaded)(struct ipu6_bus_device *adev);
irqreturn.h
Quoted 2 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +2
-0+bool wake_isr_thread; +};
...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 40
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +4
-0+#define to_ipu6_bus_device (_dev) container_of(to_auxiliary_dev(_dev), \ + struct ipu6_bus_device, auxdev) +#define auxdev_to_adev (_auxdev) container_of(_auxdev, \ + struct ipu6_bus_device, auxdev)
container_of.h
Also, can you reformat to be more readable like
#define auxdev_to_adev(_auxdev) \
container_of(_auxdev, struct ipu6_bus_device, auxdev)
Quoted 1 line
+#define
ipu6_bus_get_drvdata(adev)dev_get_drvdata(&(adev)->auxdev.dev)
device.h
...
drivers/media/pci/intel/ipu6/ipu6.c static int ipu6_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id) · line 672
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6.c +4
-0+ ret =dma_set_max_seg_size (&pdev->dev,UINT_MAX ); +if (ret) +return dev_err_probe (&pdev->dev, ret, +"Failed to set max_seg_size\n" );
With the help of
struct device *dev = &pdev->dev;
this and other lines become neater
ret = dma_set_max_seg_size(dev, _DMA_SEGMENT_SIZE); // as I commented earlier about if (ret) return dev_err_probe(dev, ret, "Failed to set max_seg_size\n");
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 02/15] media: intel/ipu6: add IPU auxiliary devices
Andy,
Thanks for your review.
Quoted 15 lines · Andy Shevchenko
Quoted 11 lines · On Tue, Oct 24, 2023 at 07:29:11PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Even the IPU input system and processing system are in a single PCI device, each system has its own power sequence, the processing system power up depends on the input system power up.
Besides, input system and processing system have their own MMU hardware for IPU DMA address mapping.
Register the IS/PS devices on auxiliary bus and attach power domain to implement the power sequence dependency.
...
Quoted 8 lines
Seems again poor / random list of header inclusions...
Please, go through your files and use IWYU (Include What You Use) principle in them.Quoted 1 line
+#include <
linux/pci.h>
auxiliary_bus.h,err.h,list.h,mutex.h,dev_printk.h,slab.h,dma-mapping.hare missing. You are lucky it does compile.
auxiliary_bus.h is included in ipu6-bus.h, list.h, mutex.h dev_printk.h are included in device.h, dma-mapping.h and scatterlist.h are included in pci.h.
I am a little confused about the rule, do you mean we need include the generic headers we need even it is included in others header?
Quoted 10 lines
...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 7
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +4
-0+#ifndef IPU6_BUS_H +#define IPU6_BUS_H + +#include <linux/auxiliary_bus.h> ...Especially for headers which will affect the compilation time.
Quoted 3 lines
Quoted 1 line
+#include <
linux/pci.h>This is not used.
Do you mean it just need a 'struct pci_dev;' ?
Quoted 35 lines
...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 18
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+struct list_head list;+
types.hdrivers/media/pci/intel/ipu6/ipu6-bus.h line 25
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+struct sg_table fw_sgt;+
scatterlist.hdrivers/media/pci/intel/ipu6/ipu6-bus.h line 27
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+dma_addr_t pkg_dir_dma_addr;
types.h...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 33
Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +3
-0+struct ipu6_auxdrv_data { +irqreturn_t (*isr)(struct ipu6_bus_device *adev); +irqreturn_t (*isr_threaded)(struct ipu6_bus_device *adev);
irqreturn.hQuoted 2 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +2
-0+bool wake_isr_thread; +};...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 40
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +4
-0+#define to_ipu6_bus_device (_dev) container_of(to_auxiliary_dev(_dev), \ + struct ipu6_bus_device, auxdev) +#define auxdev_to_adev (_auxdev) container_of(_auxdev, \ + struct ipu6_bus_device, auxdev)
container_of.h
Quoted 4 lines
Also, can you reformat to be more readable like
#define
auxdev_to_adev(_auxdev)\
container_of(_auxdev, struct ipu6_bus_device, auxdev)
Ack.
Quoted 14 lines
Quoted 2 lines
+#define
ipu6_bus_get_drvdata(adev)dev_get_drvdata(&(adev)->auxdev.dev)
device.h...
drivers/media/pci/intel/ipu6/ipu6.c static int ipu6_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id) · line 672
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6.c +4
-0+ ret =dma_set_max_seg_size (&pdev->dev,UINT_MAX ); +if (ret) +return dev_err_probe (&pdev->dev, ret, +"Failed to set max_seg_size\n" );With the help of
Quoted 8 lines
struct device *dev = &pdev->dev;this and other lines become neater
ret = dma_set_max_seg_size(dev, _DMA_SEGMENT_SIZE); // as I commented earlier about if (ret) return dev_err_probe(dev, ret, "Failed to set max_seg_size\n");
Ack.
Quoted 1 line
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 02/15] media: intel/ipu6: add IPU auxiliary devices
Hi,
Quoted 26 lines · Bingbu Cao
Andy,
Thanks for your review.
Quoted 21 lines · Andy Shevchenko
Quoted 11 lines · On Tue, Oct 24, 2023 at 07:29:11PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Even the IPU input system and processing system are in a single PCI device, each system has its own power sequence, the processing system power up depends on the input system power up.
Besides, input system and processing system have their own MMU hardware for IPU DMA address mapping.
Register the IS/PS devices on auxiliary bus and attach power domain to implement the power sequence dependency.
...
Seems again poor / random list of header inclusions...
Please, go through your files and use IWYU (Include What You Use) principle in them.Quoted 1 line
+#include <
linux/pci.h>
Quoted 8 lines
Quoted 2 lines
auxiliary_bus.h,err.h,list.h,mutex.h,dev_printk.h,slab.h,dma-mapping.hare missing. You are lucky it does compile.
auxiliary_bus.his included inipu6-bus.h,list.h,mutex.hdev_printk.hare included indevice.h,dma-mapping.handscatterlist.hare included inpci.h.I am a little confused about the rule, do you mean we need include the generic headers we need even it is included in others header?
Yes you must *not* rely on other headers including headers from which you directly use symbols.
People are working on reducing the number of includes inside headers to speed up building the kernel because including unnecessary headers inside another header has a big impact on build time.
Regards,
Hans
Quoted 85 lines
Quoted 11 lines
...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 7
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +4
-0+#ifndef IPU6_BUS_H +#define IPU6_BUS_H + +#include <linux/auxiliary_bus.h> ...Especially for headers which will affect the compilation time.
Quoted 3 lines
Quoted 1 line
+#include <
linux/pci.h>This is not used.
Do you mean it just need a 'struct
pci_dev;' ?Quoted 35 lines
...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 18
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+struct list_head list;+
types.hdrivers/media/pci/intel/ipu6/ipu6-bus.h line 25
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+struct sg_table fw_sgt;+
scatterlist.hdrivers/media/pci/intel/ipu6/ipu6-bus.h line 27
Quoted 1 line
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +1
-0+dma_addr_t pkg_dir_dma_addr;
types.h...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 33
Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +3
-0+struct ipu6_auxdrv_data { +irqreturn_t (*isr)(struct ipu6_bus_device *adev); +irqreturn_t (*isr_threaded)(struct ipu6_bus_device *adev);
irqreturn.hQuoted 2 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +2
-0+bool wake_isr_thread; +};...
drivers/media/pci/intel/ipu6/ipu6-bus.h line 40
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +4
-0+#define to_ipu6_bus_device (_dev) container_of(to_auxiliary_dev(_dev), \ + struct ipu6_bus_device, auxdev) +#define auxdev_to_adev (_auxdev) container_of(_auxdev, \ + struct ipu6_bus_device, auxdev)
container_of.hQuoted 4 lines
Also, can you reformat to be more readable like
#define
auxdev_to_adev(_auxdev)\
container_of(_auxdev, struct ipu6_bus_device, auxdev)Ack.
Quoted 14 lines
Quoted 2 lines
+#define
ipu6_bus_get_drvdata(adev)dev_get_drvdata(&(adev)->auxdev.dev)
device.h...
drivers/media/pci/intel/ipu6/ipu6.c static int ipu6_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id) · line 672
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6.c +4
-0+ ret =dma_set_max_seg_size (&pdev->dev,UINT_MAX ); +if (ret) +return dev_err_probe (&pdev->dev, ret, +"Failed to set max_seg_size\n" );With the help of
Quoted 8 lines
struct device *dev = &pdev->dev;this and other lines become neater
ret = dma_set_max_seg_size(dev, _DMA_SEGMENT_SIZE); // as I commented earlier about if (ret) return dev_err_probe(dev, ret, "Failed to set max_seg_size\n");Ack.
Quoted 1 line
Re: [PATCH v2 02/15] media: intel/ipu6: add IPU auxiliary devices
In reply to Bingbu CaoQuoted 4 lines · Bingbu Cao
Quoted 1 line · Andy Shevchenko
On Tue, Oct 24, 2023 at 07:29:11PM +0800, [email protected] wrote:
auxiliary_bus.his included inipu6-bus.h,
So, you have to include it explicitly as IWYU, strictly speaking ipu6-bus.h does NOT guarantee that inclusion, even if you fully control it.
Quoted 2 lines
list.h,mutex.hdev_printk.hare
included indevice.h,
I ack for dev_printk.h, but for the rest see above.
Quoted 1 line
dma-mapping.handscatterlist.hare included inpci.h.
See above.
Quoted 2 lines
I am a little confused about the rule, do you mean we need include the generic headers we need even it is included in others header?
Yes. There are only few guarantees in the kernel, strictly speaking. And having spaghetti headers in your code is a bad idea from long run maintenance perspective.
...
Quoted 7 lines
Quoted 7 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h line 7
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +4
-0+#ifndef IPU6_BUS_H +#define IPU6_BUS_H + +#include <linux/auxiliary_bus.h> ...Especially for headers which will affect the compilation time.
Quoted 5 lines
Quoted 3 lines
Quoted 1 line
+#include <
linux/pci.h>This is not used.
Do you mean it just need a 'struct
pci_dev;' ?
Yes.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 02/15] media: intel/ipu6: add IPU auxiliary devices
Andy and Hans,
Thanks.
Quoted 35 lines · Andy Shevchenko
Quoted 4 lines · Bingbu Cao
Quoted 1 line · Andy Shevchenko
On Tue, Oct 24, 2023 at 07:29:11PM +0800, [email protected] wrote:
auxiliary_bus.his included inipu6-bus.h,So, you have to include it explicitly as IWYU, strictly speaking
ipu6-bus.hdoes NOT guarantee that inclusion, even if you fully control it.Quoted 2 lines
list.h,mutex.hdev_printk.hare
included indevice.h,I ack for
dev_printk.h, but for the rest see above.Quoted 1 line
dma-mapping.handscatterlist.hare included inpci.h.See above.
Quoted 2 lines
I am a little confused about the rule, do you mean we need include the generic headers we need even it is included in others header?
Yes. There are only few guarantees in the kernel, strictly speaking. And having spaghetti headers in your code is a bad idea from long run maintenance perspective.
...
Quoted 7 lines
Quoted 7 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h line 7
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-bus.h new file +4
-0+#ifndef IPU6_BUS_H +#define IPU6_BUS_H + +#include <linux/auxiliary_bus.h> ...Especially for headers which will affect the compilation time.
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Quoted 1 line
+#include <
linux/pci.h>This is not used.
Do you mean it just need a 'struct
pci_dev;' ?Yes.
Ack. Will address them in v3.
Signature · 2 lines
Best regards, Bingbu Cao
PATCH v2 3/15 media: intel/ipu6: add IPU6 buttress interface driver
In reply to [email protected]From: Bingbu Cao <[email protected]>
The IPU6 buttress is the interface between IPU device (input system and processing system) with rest of the SoC. It contains overall IPU hardware control registers, these control registers are used as the interfaces with the Intel Converged Security Engine and Punit to do firmware authentication and power management.
--- drivers/media/pci/intel/ipu6/ipu6-buttress.c | 900 ++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-buttress.h | 107 +++ .../intel/ipu6/ipu6-platform-buttress-regs.h | 230 +++++ 3 files changed, 1237 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-buttress.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-buttress.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-platform-buttress-regs.h
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +900 -0
@@ -0,0 +1,900 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <linux/bitfield.h> +#include <linux/completion.h> +#include <linux/firmware.h> +#include <linux/iopoll.h> +#include <linux/pci.h> +#include <linux/pm_runtime.h> + +#include "ipu6.h" +#include "ipu6-bus.h" +#include "ipu6-buttress.h" +#include "ipu6-platform-buttress-regs.h" + +#define BOOTLOADER_STATUS_OFFSET 0x15c + +#define BOOTLOADER_MAGIC_KEY 0xb00710ad + +#define ENTRY BUTTRESS_IU2CSECSR_IPC_PEER_COMP_ACTIONS_RST_PHASE1 +#define EXIT BUTTRESS_IU2CSECSR_IPC_PEER_COMP_ACTIONS_RST_PHASE2 +#define QUERY BUTTRESS_IU2CSECSR_IPC_PEER_QUERIED_IP_COMP_ACTIONS_RST_PHASE + +#define BUTTRESS_TSC_SYNC_RESET_TRIAL_MAX 10 + +#define BUTTRESS_POWER_TIMEOUT_US (200 * USEC_PER_MSEC) + +#define BUTTRESS_CSE_BOOTLOAD_TIMEOUT_US (5 * USEC_PER_SEC) +#define BUTTRESS_CSE_AUTHENTICATE_TIMEOUT_US (10 * USEC_PER_SEC) +#define BUTTRESS_CSE_FWRESET_TIMEOUT_US (100 * USEC_PER_MSEC) + +#define BUTTRESS_IPC_TX_TIMEOUT_MS MSEC_PER_SEC +#define BUTTRESS_IPC_RX_TIMEOUT_MS MSEC_PER_SEC +#define BUTTRESS_IPC_VALIDITY_TIMEOUT_US (1 * USEC_PER_SEC) +#define BUTTRESS_TSC_SYNC_TIMEOUT_US (5 * USEC_PER_MSEC) + +#define BUTTRESS_IPC_RESET_RETRY 2000 +#define BUTTRESS_CSE_IPC_RESET_RETRY 4 +#define BUTTRESS_IPC_CMD_SEND_RETRY 1 + +#define BUTTRESS_MAX_CONSECUTIVE_IRQS 100 + +static const u32 ipu6_adev_irq_mask[2 ] = { +BUTTRESS_ISR_IS_IRQ , +BUTTRESS_ISR_PS_IRQ +}; + +int ipu6_buttress_ipc_reset (struct ipu6_device *isp, +struct ipu6_buttress_ipc *ipc) +{ +unsigned int retries =BUTTRESS_IPC_RESET_RETRY ; +struct ipu6_buttress *b = &isp->buttress; +u32 val =0 , csr_in_clr; + +if (!isp->secure_mode) { +dev_dbg (&isp->pdev->dev,"Skip IPC reset for non-secure mode" ); +return 0 ; + } + +mutex_lock (&b->ipc_mutex); + +/* Clear-by-1 CSR (all bits), corresponding internal states. */ + val =readl (isp->base + ipc->csr_in); +writel (val, isp->base + ipc->csr_in); + +/* Set peer CSR bit IPC_PEER_COMP_ACTIONS_RST_PHASE1 */ +writel (ENTRY , isp->base + ipc->csr_out); +/* +* Clear-by-1 all CSR bits EXCEPT following +* bits: +* A. IPC_PEER_COMP_ACTIONS_RST_PHASE1. +* B. IPC_PEER_COMP_ACTIONS_RST_PHASE2. +* C. Possibly custom bits, depending on +* their role. +*/ + csr_in_clr =BUTTRESS_IU2CSECSR_IPC_PEER_DEASSERTED_REG_VALID_REQ | +BUTTRESS_IU2CSECSR_IPC_PEER_ACKED_REG_VALID | +BUTTRESS_IU2CSECSR_IPC_PEER_ASSERTED_REG_VALID_REQ |QUERY ; + +do { +usleep_range (400 ,500 ); + val =readl (isp->base + ipc->csr_in); +switch (val) { +case ENTRY |EXIT : +case ENTRY |EXIT |QUERY : +/* +* 1) Clear-by-1 CSR bits +* (IPC_PEER_COMP_ACTIONS_RST_PHASE1, +* IPC_PEER_COMP_ACTIONS_RST_PHASE2). +* 2) Set peer CSR bit +* IPC_PEER_QUERIED_IP_COMP_ACTIONS_RST_PHASE. +*/ +writel (ENTRY |EXIT , isp->base + ipc->csr_in); +writel (QUERY , isp->base + ipc->csr_out); +break ; +case ENTRY : +case ENTRY |QUERY : +/* +* 1) Clear-by-1 CSR bits +* (IPC_PEER_COMP_ACTIONS_RST_PHASE1, +* IPC_PEER_QUERIED_IP_COMP_ACTIONS_RST_PHASE). +* 2) Set peer CSR bit +* IPC_PEER_COMP_ACTIONS_RST_PHASE1. +*/ +writel (ENTRY |QUERY , isp->base + ipc->csr_in); +writel (ENTRY , isp->base + ipc->csr_out); +break ; +case EXIT : +case EXIT |QUERY : +/* +* Clear-by-1 CSR bit +* IPC_PEER_COMP_ACTIONS_RST_PHASE2. +* 1) Clear incoming doorbell. +* 2) Clear-by-1 all CSR bits EXCEPT following +* bits: +* A. IPC_PEER_COMP_ACTIONS_RST_PHASE1. +* B. IPC_PEER_COMP_ACTIONS_RST_PHASE2. +* C. Possibly custom bits, depending on +* their role. +* 3) Set peer CSR bit +* IPC_PEER_COMP_ACTIONS_RST_PHASE2. +*/ +writel (EXIT , isp->base + ipc->csr_in); +writel (0 , isp->base + ipc->db0_in); +writel (csr_in_clr, isp->base + ipc->csr_in); +writel (EXIT , isp->base + ipc->csr_out); + +/* +* Read csr_in again to make sure if RST_PHASE2 is done. +* If csr_in is QUERY, it should be handled again. +*/ +usleep_range (200 ,300 ); + val =readl (isp->base + ipc->csr_in); +if (val &QUERY ) { +dev_dbg (&isp->pdev->dev, +"RST_PHASE2 retry csr_in = %x\n" , val); +break ; + } +mutex_unlock (&b->ipc_mutex); +return 0 ; +case QUERY : +/* +* 1) Clear-by-1 CSR bit +* IPC_PEER_QUERIED_IP_COMP_ACTIONS_RST_PHASE. +* 2) Set peer CSR bit +* IPC_PEER_COMP_ACTIONS_RST_PHASE1 +*/ +writel (QUERY , isp->base + ipc->csr_in); +writel (ENTRY , isp->base + ipc->csr_out); +break ; +default : +dev_warn_ratelimited (&isp->pdev->dev, +"Unexpected CSR 0x%x\n" , val); +break ; + } + }while (retries--); + +mutex_unlock (&b->ipc_mutex); +dev_err (&isp->pdev->dev,"Timed out while waiting for CSE\n" ); + +return -ETIMEDOUT ; +} + +static void ipu6_buttress_ipc_validity_close (struct ipu6_device *isp, +struct ipu6_buttress_ipc *ipc) +{ +writel (BUTTRESS_IU2CSECSR_IPC_PEER_DEASSERTED_REG_VALID_REQ , + isp->base + ipc->csr_out); +} + +static int +ipu6_buttress_ipc_validity_open (struct ipu6_device *isp, +struct ipu6_buttress_ipc *ipc) +{ +unsigned int mask =BUTTRESS_IU2CSECSR_IPC_PEER_ACKED_REG_VALID ; +void __iomem *addr; +int ret; +u32 val; + +writel (BUTTRESS_IU2CSECSR_IPC_PEER_ASSERTED_REG_VALID_REQ , + isp->base + ipc->csr_out); + + addr = isp->base + ipc->csr_in; + ret =readl_poll_timeout (addr, val, val & mask,200 , +BUTTRESS_IPC_VALIDITY_TIMEOUT_US ); +if (ret) { +dev_err (&isp->pdev->dev,"CSE validity timeout 0x%x\n" , val); +ipu6_buttress_ipc_validity_close (isp, ipc); + } + +return ret; +} + +static void ipu6_buttress_ipc_recv (struct ipu6_device *isp, +struct ipu6_buttress_ipc *ipc,u32 *ipc_msg) +{ +if (ipc_msg) + *ipc_msg =readl (isp->base + ipc->data0_in); +writel (0 , isp->base + ipc->db0_in); +} + +static int ipu6_buttress_ipc_send_bulk (struct ipu6_device *isp, +enum ipu6_buttress_ipc_domain ipc_domain, +struct ipu6_ipc_buttress_bulk_msg *msgs, +u32 size) +{ +unsigned long tx_timeout_jiffies, rx_timeout_jiffies; +unsigned int i, retry =BUTTRESS_IPC_CMD_SEND_RETRY ; +struct ipu6_buttress *b = &isp->buttress; +struct ipu6_buttress_ipc *ipc; +u32 val; +int ret; +int tout; + + ipc = ipc_domain ==IPU6_BUTTRESS_IPC_CSE ? &b->cse : &b->ish; + +mutex_lock (&b->ipc_mutex); + + ret =ipu6_buttress_ipc_validity_open (isp, ipc); +if (ret) { +dev_err (&isp->pdev->dev,"IPC validity open failed\n" ); + goto out; + } + + tx_timeout_jiffies =msecs_to_jiffies (BUTTRESS_IPC_TX_TIMEOUT_MS ); + rx_timeout_jiffies =msecs_to_jiffies (BUTTRESS_IPC_RX_TIMEOUT_MS ); + +for (i =0 ; i < size; i++) { +reinit_completion (&ipc->send_complete); +if (msgs[i].require_resp) +reinit_completion (&ipc->recv_complete); + +dev_dbg (&isp->pdev->dev,"bulk IPC command: 0x%x\n" , + msgs[i].cmd); +writel (msgs[i].cmd, isp->base + ipc->data0_out); + val =BUTTRESS_IU2CSEDB0_BUSY | msgs[i].cmd_size; +writel (val, isp->base + ipc->db0_out); + + tout =wait_for_completion_timeout (&ipc->send_complete, + tx_timeout_jiffies); +if (!tout) { +dev_err (&isp->pdev->dev,"send IPC response timeout\n" ); +if (!retry--) { + ret = -ETIMEDOUT ; + goto out; + } + +/* Try again if CSE is not responding on first try */ +writel (0 , isp->base + ipc->db0_out); + i--; +continue ; + } + + retry =BUTTRESS_IPC_CMD_SEND_RETRY ; + +if (!msgs[i].require_resp) +continue ; + + tout =wait_for_completion_timeout (&ipc->recv_complete, + rx_timeout_jiffies); +if (!tout) { +dev_err (&isp->pdev->dev,"recv IPC response timeout\n" ); + ret = -ETIMEDOUT ; + goto out; + } + +if (ipc->nack_mask && + (ipc->recv_data & ipc->nack_mask) == ipc->nack) { +dev_err (&isp->pdev->dev, +"IPC NACK for cmd 0x%x\n" , msgs[i].cmd); + ret = -EIO ; + goto out; + } + +if (ipc->recv_data != msgs[i].expected_resp) { +dev_err (&isp->pdev->dev, +"expected resp: 0x%x, IPC response: 0x%x " , + msgs[i].expected_resp, ipc->recv_data); + ret = -EIO ; + goto out; + } + } + +dev_dbg (&isp->pdev->dev,"bulk IPC commands done\n" ); + +out: +ipu6_buttress_ipc_validity_close (isp, ipc); +mutex_unlock (&b->ipc_mutex); +return ret; +} + +static int +ipu6_buttress_ipc_send (struct ipu6_device *isp, +enum ipu6_buttress_ipc_domain ipc_domain, +u32 ipc_msg,u32 size,bool require_resp, +u32 expected_resp) +{ +struct ipu6_ipc_buttress_bulk_msg msg = { + .cmd = ipc_msg, + .cmd_size = size, + .require_resp = require_resp, + .expected_resp = expected_resp, + }; + +return ipu6_buttress_ipc_send_bulk (isp, ipc_domain, &msg,1 ); +} + +static irqreturn_t ipu6_buttress_call_isr (struct ipu6_bus_device *adev) +{ +irqreturn_t ret =IRQ_WAKE_THREAD ; + +if (!adev || !adev->auxdrv || !adev->auxdrv_data) +return IRQ_NONE ; + +if (adev->auxdrv_data->isr) + ret = adev->auxdrv_data->isr (adev); + +if (ret ==IRQ_WAKE_THREAD && !adev->auxdrv_data->isr_threaded) + ret =IRQ_NONE ; + +return ret; +} + +irqreturn_t ipu6_buttress_isr (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +struct ipu6_buttress *b = &isp->buttress; +u32 reg_irq_sts =BUTTRESS_REG_ISR_STATUS ; +irqreturn_t ret =IRQ_NONE ; +u32 disable_irqs =0 ; +u32 irq_status; +u32 i, count =0 ; + +pm_runtime_get_noresume (&isp->pdev->dev); + + irq_status =readl (isp->base + reg_irq_sts); +if (!irq_status) { +pm_runtime_put_noidle (&isp->pdev->dev); +return IRQ_NONE ; + } + +do { +writel (irq_status, isp->base +BUTTRESS_REG_ISR_CLEAR ); + +for (i =0 ; i <ARRAY_SIZE (ipu6_adev_irq_mask); i++) { +irqreturn_t r =ipu6_buttress_call_isr (adev[i]); + +if (!(irq_status & ipu6_adev_irq_mask[i])) +continue ; + +if (r ==IRQ_WAKE_THREAD ) { + ret =IRQ_WAKE_THREAD ; + disable_irqs |= ipu6_adev_irq_mask[i]; + }else if (ret ==IRQ_NONE && r ==IRQ_HANDLED ) { + ret =IRQ_HANDLED ; + } + } + +if ((irq_status &BUTTRESS_EVENT ) && ret ==IRQ_NONE ) + ret =IRQ_HANDLED ; + +if (irq_status &BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->cse, &b->cse.recv_data); +complete (&b->cse.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->ish, &b->ish.recv_data); +complete (&b->ish.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->cse.send_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_ISH ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->ish.send_complete); + } + +if (irq_status &BUTTRESS_ISR_SAI_VIOLATION && +ipu6_buttress_get_secure_mode (isp)) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_SAI_VIOLATION\n" ); + +if (irq_status & (BUTTRESS_ISR_IS_FATAL_MEM_ERR | +BUTTRESS_ISR_PS_FATAL_MEM_ERR )) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_FATAL_MEM_ERR\n" ); + +if (irq_status &BUTTRESS_ISR_UFI_ERROR ) +dev_err (&isp->pdev->dev,"BUTTRESS_ISR_UFI_ERROR\n" ); + +if (++count ==BUTTRESS_MAX_CONSECUTIVE_IRQS ) { +dev_err (&isp->pdev->dev,"too many consecutive IRQs\n" ); + ret =IRQ_NONE ; +break ; + } + + irq_status =readl (isp->base + reg_irq_sts); + }while (irq_status); + +if (disable_irqs) +writel (BUTTRESS_IRQS & ~disable_irqs, + isp->base +BUTTRESS_REG_ISR_ENABLE ); + +pm_runtime_put (&isp->pdev->dev); + +return ret; +} + +irqreturn_t ipu6_buttress_isr_threaded (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +const struct ipu6_auxdrv_data *drv_data =NULL ; +irqreturn_t ret =IRQ_NONE ; +unsigned int i; + +for (i =0 ; i <ARRAY_SIZE (ipu6_adev_irq_mask) && adev[i]; i++) { + drv_data = adev[i]->auxdrv_data; +if (!drv_data) +continue ; + +if (drv_data->wake_isr_thread && + drv_data->isr_threaded (adev[i]) ==IRQ_HANDLED ) + ret =IRQ_HANDLED ; + } + +writel (BUTTRESS_IRQS , isp->base +BUTTRESS_REG_ISR_ENABLE ); + +return ret; +} + +int ipu6_buttress_power (struct device *dev,struct ipu6_buttress_ctrl *ctrl, +bool on) +{ +struct ipu6_device *isp =to_ipu6_bus_device (dev)->isp; +u32 pwr_sts, val; +int ret; + +if (!ctrl) +return 0 ; + +mutex_lock (&isp->buttress.power_mutex); + +if (!on) { + val =0 ; + pwr_sts = ctrl->pwr_sts_off << ctrl->pwr_sts_shift; + }else { + val =BUTTRESS_FREQ_CTL_START | +FIELD_PREP (BUTTRESS_FREQ_CTL_RATIO_MASK , + ctrl->ratio) | +FIELD_PREP (BUTTRESS_FREQ_CTL_QOS_FLOOR_MASK , + ctrl->qos_floor) | +BUTTRESS_FREQ_CTL_ICCMAX_LEVEL ; + + pwr_sts = ctrl->pwr_sts_on << ctrl->pwr_sts_shift; + } + +writel (val, isp->base + ctrl->freq_ctl); + + ret =readl_poll_timeout (isp->base +BUTTRESS_REG_PWR_STATE , + val, (val & ctrl->pwr_sts_mask) == pwr_sts, +100 ,BUTTRESS_POWER_TIMEOUT_US ); +if (ret) +dev_err (&isp->pdev->dev, +"Change power status timeout with 0x%x\n" , val); + + ctrl->started = !ret && on; + +mutex_unlock (&isp->buttress.power_mutex); + +return ret; +} + +bool ipu6_buttress_get_secure_mode (struct ipu6_device *isp) +{ +u32 val; + + val =readl (isp->base +BUTTRESS_REG_SECURITY_CTL ); + +return val &BUTTRESS_SECURITY_CTL_FW_SECURE_MODE ; +} + +bool ipu6_buttress_auth_done (struct ipu6_device *isp) +{ +u32 val; + +if (!isp->secure_mode) +return true; + + val =readl (isp->base +BUTTRESS_REG_SECURITY_CTL ); + val =FIELD_GET (BUTTRESS_SECURITY_CTL_FW_SETUP_MASK , val); + +return val ==BUTTRESS_SECURITY_CTL_AUTH_DONE ; +} +EXPORT_SYMBOL_NS_GPL (ipu6_buttress_auth_done,INTEL_IPU6 ); + +int ipu6_buttress_reset_authentication (struct ipu6_device *isp) +{ +int ret; +u32 val; + +if (!isp->secure_mode) { +dev_dbg (&isp->pdev->dev,"Skip auth for non-secure mode\n" ); +return 0 ; + } + +writel (BUTTRESS_FW_RESET_CTL_START , isp->base + +BUTTRESS_REG_FW_RESET_CTL ); + + ret =readl_poll_timeout (isp->base +BUTTRESS_REG_FW_RESET_CTL , val, + val &BUTTRESS_FW_RESET_CTL_DONE ,500 , +BUTTRESS_CSE_FWRESET_TIMEOUT_US ); +if (ret) { +dev_err (&isp->pdev->dev, +"Time out while resetting authentication state\n" ); +return ret; + } + +dev_dbg (&isp->pdev->dev,"FW reset for authentication done\n" ); +writel (0 , isp->base +BUTTRESS_REG_FW_RESET_CTL ); +/* leave some time for HW restore */ +usleep_range (800 ,1000 ); + +return 0 ; +} + +int ipu6_buttress_map_fw_image (struct ipu6_bus_device *sys, +const struct firmware *fw,struct sg_table *sgt) +{ +struct page **pages; +const void *addr; +unsigned long n_pages; +unsigned int i; +int ret; + + n_pages =PHYS_PFN (PAGE_ALIGN (fw->size)); + + pages =kmalloc_array (n_pages,sizeof (*pages),GFP_KERNEL ); +if (!pages) +return -ENOMEM ; + + addr = fw->data; +for (i =0 ; i < n_pages; i++) { +struct page *p =vmalloc_to_page (addr); + +if (!p) { + ret = -ENOMEM ; + goto out; + } + pages[i] = p; + addr +=PAGE_SIZE ; + } + + ret =sg_alloc_table_from_pages (sgt, pages, n_pages,0 , fw->size, +GFP_KERNEL ); +if (ret) { + ret = -ENOMEM ; + goto out; + } + + ret =dma_map_sgtable (&sys->auxdev.dev, sgt,DMA_TO_DEVICE ,0 ); +if (ret <0 ) { + ret = -ENOMEM ; +sg_free_table (sgt); + goto out; + } + +dma_sync_sgtable_for_device (&sys->auxdev.dev, sgt,DMA_TO_DEVICE ); + +out: +kfree (pages); + +return ret; +} +EXPORT_SYMBOL_NS_GPL (ipu6_buttress_map_fw_image,INTEL_IPU6 ); + +void ipu6_buttress_unmap_fw_image (struct ipu6_bus_device *sys, +struct sg_table *sgt) +{ +dma_unmap_sgtable (&sys->auxdev.dev, sgt,DMA_TO_DEVICE ,0 ); +sg_free_table (sgt); +} +EXPORT_SYMBOL_NS_GPL (ipu6_buttress_unmap_fw_image,INTEL_IPU6 ); + +int ipu6_buttress_authenticate (struct ipu6_device *isp) +{ +struct ipu6_buttress *b = &isp->buttress; +struct ipu6_psys_pdata *psys_pdata; +u32 data, mask, done, fail; +int ret; + +if (!isp->secure_mode) { +dev_dbg (&isp->pdev->dev,"Skip auth for non-secure mode\n" ); +return 0 ; + } + + psys_pdata = isp->psys->pdata; + +mutex_lock (&b->auth_mutex); + +if (ipu6_buttress_auth_done (isp)) { + ret =0 ; + goto out_unlock; + } + +/* +* Write address of FIT table to FW_SOURCE register +* Let's use fw address. I.e. not using FIT table yet +*/ + data =lower_32_bits (isp->psys->pkg_dir_dma_addr); +writel (data, isp->base +BUTTRESS_REG_FW_SOURCE_BASE_LO ); + + data =upper_32_bits (isp->psys->pkg_dir_dma_addr); +writel (data, isp->base +BUTTRESS_REG_FW_SOURCE_BASE_HI ); + +/* +* Write boot_load into IU2CSEDATA0 +* Write sizeof(boot_load) | 0x2 << CLIENT_ID to +* IU2CSEDB.IU2CSECMD and set IU2CSEDB.IU2CSEBUSY as +*/ +dev_info (&isp->pdev->dev,"Sending BOOT_LOAD to CSE\n" ); + + ret =ipu6_buttress_ipc_send (isp,IPU6_BUTTRESS_IPC_CSE , +BUTTRESS_IU2CSEDATA0_IPC_BOOT_LOAD , +1 , true, +BUTTRESS_CSE2IUDATA0_IPC_BOOT_LOAD_DONE ); +if (ret) { +dev_err (&isp->pdev->dev,"CSE boot_load failed\n" ); + goto out_unlock; + } + + mask =BUTTRESS_SECURITY_CTL_FW_SETUP_MASK ; + done =BUTTRESS_SECURITY_CTL_FW_SETUP_DONE ; + fail =BUTTRESS_SECURITY_CTL_AUTH_FAILED ; + ret =readl_poll_timeout (isp->base +BUTTRESS_REG_SECURITY_CTL , data, + ((data & mask) == done || + (data & mask) == fail),500 , +BUTTRESS_CSE_BOOTLOAD_TIMEOUT_US ); +if (ret) { +dev_err (&isp->pdev->dev,"CSE boot_load timeout\n" ); + goto out_unlock; + } + +if ((data & mask) == fail) { +dev_err (&isp->pdev->dev,"CSE auth failed\n" ); + ret = -EINVAL ; + goto out_unlock; + } + + ret =readl_poll_timeout (psys_pdata->base +BOOTLOADER_STATUS_OFFSET , + data, data ==BOOTLOADER_MAGIC_KEY ,500 , +BUTTRESS_CSE_BOOTLOAD_TIMEOUT_US ); +if (ret) { +dev_err (&isp->pdev->dev,"Unexpected magic number 0x%x\n" , + data); + goto out_unlock; + } + +/* +* Write authenticate_run into IU2CSEDATA0 +* Write sizeof(boot_load) | 0x2 << CLIENT_ID to +* IU2CSEDB.IU2CSECMD and set IU2CSEDB.IU2CSEBUSY as +*/ +dev_info (&isp->pdev->dev,"Sending AUTHENTICATE_RUN to CSE\n" ); + ret =ipu6_buttress_ipc_send (isp,IPU6_BUTTRESS_IPC_CSE , +BUTTRESS_IU2CSEDATA0_IPC_AUTH_RUN , +1 , true, +BUTTRESS_CSE2IUDATA0_IPC_AUTH_RUN_DONE ); +if (ret) { +dev_err (&isp->pdev->dev,"CSE authenticate_run failed\n" ); + goto out_unlock; + } + + done =BUTTRESS_SECURITY_CTL_AUTH_DONE ; + ret =readl_poll_timeout (isp->base +BUTTRESS_REG_SECURITY_CTL , data, + ((data & mask) == done || + (data & mask) == fail),500 , +BUTTRESS_CSE_AUTHENTICATE_TIMEOUT_US ); +if (ret) { +dev_err (&isp->pdev->dev,"CSE authenticate timeout\n" ); + goto out_unlock; + } + +if ((data & mask) == fail) { +dev_err (&isp->pdev->dev,"CSE boot_load failed\n" ); + ret = -EINVAL ; + goto out_unlock; + } + +dev_info (&isp->pdev->dev,"CSE authenticate_run done\n" ); + +out_unlock: +mutex_unlock (&b->auth_mutex); + +return ret; +} + +static int ipu6_buttress_send_tsc_request (struct ipu6_device *isp) +{ +u32 val, mask, done; +int ret; + + mask =BUTTRESS_PWR_STATE_HH_STATUS_MASK ; + +writel (BUTTRESS_FABRIC_CMD_START_TSC_SYNC , + isp->base +BUTTRESS_REG_FABRIC_CMD ); + + val =readl (isp->base +BUTTRESS_REG_PWR_STATE ); + val =FIELD_GET (mask, val); +if (val ==BUTTRESS_PWR_STATE_HH_STATE_ERR ) { +dev_err (&isp->pdev->dev,"Start tsc sync failed\n" ); +return -EINVAL ; + } + + done =BUTTRESS_PWR_STATE_HH_STATE_DONE ; + ret =readl_poll_timeout (isp->base +BUTTRESS_REG_PWR_STATE , val, +FIELD_GET (mask, val) == done,500 , +BUTTRESS_TSC_SYNC_TIMEOUT_US ); +if (ret) +dev_err (&isp->pdev->dev,"Start tsc sync timeout\n" ); + +return ret; +} + +int ipu6_buttress_start_tsc_sync (struct ipu6_device *isp) +{ +unsigned int i; + +for (i =0 ; i <BUTTRESS_TSC_SYNC_RESET_TRIAL_MAX ; i++) { +u32 val; +int ret; + + ret =ipu6_buttress_send_tsc_request (isp); +if (ret != -ETIMEDOUT ) +return ret; + + val =readl (isp->base +BUTTRESS_REG_TSW_CTL ); + val = val |BUTTRESS_TSW_CTL_SOFT_RESET ; +writel (val, isp->base +BUTTRESS_REG_TSW_CTL ); + val = val & ~BUTTRESS_TSW_CTL_SOFT_RESET ; +writel (val, isp->base +BUTTRESS_REG_TSW_CTL ); + } + +dev_err (&isp->pdev->dev,"TSC sync failed (timeout)\n" ); + +return -ETIMEDOUT ; +} +EXPORT_SYMBOL_NS_GPL (ipu6_buttress_start_tsc_sync,INTEL_IPU6 ); + +void ipu6_buttress_tsc_read (struct ipu6_device *isp,u64 *val) +{ +u32 tsc_hi_1, tsc_hi_2, tsc_lo; +unsigned long flags; + +local_irq_save (flags); + tsc_hi_1 =readl (isp->base +BUTTRESS_REG_TSC_HI ); + tsc_lo =readl (isp->base +BUTTRESS_REG_TSC_LO ); + tsc_hi_2 =readl (isp->base +BUTTRESS_REG_TSC_HI ); +if (tsc_hi_1 == tsc_hi_2) { + *val = (u64 )tsc_hi_1 <<32 | tsc_lo; + }else { +/* Check if TSC has rolled over */ +if (tsc_lo &BIT (31 )) + *val = (u64 )tsc_hi_1 <<32 | tsc_lo; +else + *val = (u64 )tsc_hi_2 <<32 | tsc_lo; + } +local_irq_restore (flags); +} +EXPORT_SYMBOL_NS_GPL (ipu6_buttress_tsc_read,INTEL_IPU6 ); + +u64 ipu6_buttress_tsc_ticks_to_ns (u64 ticks,const struct ipu6_device *isp) +{ +u64 ns = ticks *10000 ; + +/* +* converting TSC tick count to ns is calculated by: +* Example (TSC clock frequency is 19.2MHz): +* ns = ticks * 1000 000 000 / 19.2Mhz +* = ticks * 1000 000 000 / 19200000Hz +* = ticks * 10000 / 192 ns +*/ +return div_u64 (ns, isp->buttress.ref_clk); +} +EXPORT_SYMBOL_NS_GPL (ipu6_buttress_tsc_ticks_to_ns,INTEL_IPU6 ); + +void ipu6_buttress_restore (struct ipu6_device *isp) +{ +struct ipu6_buttress *b = &isp->buttress; + +writel (BUTTRESS_IRQS , isp->base +BUTTRESS_REG_ISR_CLEAR ); +writel (BUTTRESS_IRQS , isp->base +BUTTRESS_REG_ISR_ENABLE ); +writel (b->wdt_cached_value, isp->base +BUTTRESS_REG_WDT ); +} + +int ipu6_buttress_init (struct ipu6_device *isp) +{ +int ret, ipc_reset_retry =BUTTRESS_CSE_IPC_RESET_RETRY ; +struct ipu6_buttress *b = &isp->buttress; +u32 val; + +mutex_init (&b->power_mutex); +mutex_init (&b->auth_mutex); +mutex_init (&b->cons_mutex); +mutex_init (&b->ipc_mutex); +init_completion (&b->ish.send_complete); +init_completion (&b->cse.send_complete); +init_completion (&b->ish.recv_complete); +init_completion (&b->cse.recv_complete); + + b->cse.nack =BUTTRESS_CSE2IUDATA0_IPC_NACK ; + b->cse.nack_mask =BUTTRESS_CSE2IUDATA0_IPC_NACK_MASK ; + b->cse.csr_in =BUTTRESS_REG_CSE2IUCSR ; + b->cse.csr_out =BUTTRESS_REG_IU2CSECSR ; + b->cse.db0_in =BUTTRESS_REG_CSE2IUDB0 ; + b->cse.db0_out =BUTTRESS_REG_IU2CSEDB0 ; + b->cse.data0_in =BUTTRESS_REG_CSE2IUDATA0 ; + b->cse.data0_out =BUTTRESS_REG_IU2CSEDATA0 ; + +/* no ISH on IPU6 */ +memset (&b->ish,0 ,sizeof (b->ish)); +INIT_LIST_HEAD (&b->constraints); + + isp->secure_mode =ipu6_buttress_get_secure_mode (isp); +dev_info (&isp->pdev->dev,"IPU6 in %s mode touch 0x%x mask 0x%x\n" , + isp->secure_mode ?"secure" :"non-secure" , +readl (isp->base +BUTTRESS_REG_SECURITY_TOUCH ), +readl (isp->base +BUTTRESS_REG_CAMERA_MASK )); + + b->wdt_cached_value =readl (isp->base +BUTTRESS_REG_WDT ); +writel (BUTTRESS_IRQS , isp->base +BUTTRESS_REG_ISR_CLEAR ); +writel (BUTTRESS_IRQS , isp->base +BUTTRESS_REG_ISR_ENABLE ); + +/* get ref_clk frequency by reading the indication in btrs control */ + val =readl (isp->base +BUTTRESS_REG_BTRS_CTRL ); + val =FIELD_GET (BUTTRESS_REG_BTRS_CTRL_REF_CLK_IND , val); + +switch (val) { +case 0x0 : + b->ref_clk =240 ; +break ; +case 0x1 : + b->ref_clk =192 ; +break ; +case 0x2 : + b->ref_clk =384 ; +break ; +default : +dev_warn (&isp->pdev->dev, +"Unsupported ref clock, use 19.2Mhz by default.\n" ); + b->ref_clk =192 ; +break ; + } + +/* Retry couple of times in case of CSE initialization is delayed */ +do { + ret =ipu6_buttress_ipc_reset (isp, &b->cse); +if (ret) { +dev_warn (&isp->pdev->dev, +"IPC reset protocol failed, retrying\n" ); + }else { +dev_dbg (&isp->pdev->dev,"IPC reset done\n" ); +return 0 ; + } + }while (ipc_reset_retry--); + +dev_err (&isp->pdev->dev,"IPC reset protocol failed\n" ); + +mutex_destroy (&b->power_mutex); +mutex_destroy (&b->auth_mutex); +mutex_destroy (&b->cons_mutex); +mutex_destroy (&b->ipc_mutex); + +return ret; +} + +void ipu6_buttress_exit (struct ipu6_device *isp) +{ +struct ipu6_buttress *b = &isp->buttress; + +writel (0 , isp->base +BUTTRESS_REG_ISR_ENABLE ); + +mutex_destroy (&b->power_mutex); +mutex_destroy (&b->auth_mutex); +mutex_destroy (&b->cons_mutex); +mutex_destroy (&b->ipc_mutex); +}
drivers/media/pci/intel/ipu6/ipu6-buttress.h new file +107 -0
107 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6-platform-buttress-regs.h new file +230 -0
230 lines not on this page · raw message
Signature · 1 line
2.42.0
Re: [PATCH v2 03/15] media: intel/ipu6: add IPU6 buttress interface driver
Hi Bingbu,
Quoted 10 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
The IPU6 buttress is the interface between IPU device (input system and processing system) with rest of the SoC. It contains overall IPU hardware control registers, these control registers are used as the interfaces with the Intel Converged Security Engine and Punit to do firmware authentication and power management.
Signed-off-by Bingbu Cao <[email protected]>---
...
Quoted 41 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +41
-0+static irqreturn_t ipu6_buttress_call_isr (struct ipu6_bus_device *adev) +{ +irqreturn_t ret =IRQ_WAKE_THREAD ; + +if (!adev || !adev->auxdrv || !adev->auxdrv_data) +return IRQ_NONE ; + +if (adev->auxdrv_data->isr) + ret = adev->auxdrv_data->isr (adev); + +if (ret ==IRQ_WAKE_THREAD && !adev->auxdrv_data->isr_threaded) + ret =IRQ_NONE ; + +return ret; +} + +irqreturn_t ipu6_buttress_isr (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +struct ipu6_buttress *b = &isp->buttress; +u32 reg_irq_sts =BUTTRESS_REG_ISR_STATUS ; +irqreturn_t ret =IRQ_NONE ; +u32 disable_irqs =0 ; +u32 irq_status; +u32 i, count =0 ; + +pm_runtime_get_noresume (&isp->pdev->dev); + + irq_status =readl (isp->base + reg_irq_sts); +if (!irq_status) { +pm_runtime_put_noidle (&isp->pdev->dev); +return IRQ_NONE ; + } + + do { + writel(irq_status, isp->base + BUTTRESS_REG_ISR_CLEAR); + + for (i = 0; i < ARRAY_SIZE(ipu6_adev_irq_mask); i++) { + irqreturn_t r = ipu6_buttress_call_isr(adev[i]); +
drivers/media/pci/intel/ipu6/ipu6-buttress.c line 360
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+if (!(irq_status & ipu6_adev_irq_mask[i])) +continue ; + +if (r ==IRQ_WAKE_THREAD ) { + ret =IRQ_WAKE_THREAD ; + disable_irqs |= ipu6_adev_irq_mask[i]; + }else if (ret ==IRQ_NONE && r ==IRQ_HANDLED ) { + ret =IRQ_HANDLED ; + } + }
It seems wrong to call the ISR for a adev[i] before checking the corresponding IRQ mask. If the mask is not set, the ISR is still called, but the result is thrown away.
I started investigating this because I'm seeing "general protection fault, probably for non-canonical address 0x6b6b6b6b6b6b6b6b" in this function when unbinding the IPU4 driver.
How do you ensure that the ISR is not called on a ipu6-bus device that has been deleted? Specifically in ipu6_pci_remove, ipu6_bus_del_devices is called before ipu6_buttress_exit (which disables buttress IRQs). Perhaps the above for loop should really be a "for each ipu6-bus device" loop?
Quoted 50 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +50
-0+ +if ((irq_status &BUTTRESS_EVENT ) && ret ==IRQ_NONE ) + ret =IRQ_HANDLED ; + +if (irq_status &BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->cse, &b->cse.recv_data); +complete (&b->cse.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->ish, &b->ish.recv_data); +complete (&b->ish.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->cse.send_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_ISH ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->ish.send_complete); + } + +if (irq_status &BUTTRESS_ISR_SAI_VIOLATION && +ipu6_buttress_get_secure_mode (isp)) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_SAI_VIOLATION\n" ); + +if (irq_status & (BUTTRESS_ISR_IS_FATAL_MEM_ERR | +BUTTRESS_ISR_PS_FATAL_MEM_ERR )) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_FATAL_MEM_ERR\n" ); + +if (irq_status &BUTTRESS_ISR_UFI_ERROR ) +dev_err (&isp->pdev->dev,"BUTTRESS_ISR_UFI_ERROR\n" ); + +if (++count ==BUTTRESS_MAX_CONSECUTIVE_IRQS ) { +dev_err (&isp->pdev->dev,"too many consecutive IRQs\n" ); + ret =IRQ_NONE ; +break ; + } + + irq_status =readl (isp->base + reg_irq_sts);
drivers/media/pci/intel/ipu6/ipu6-buttress.c line 420
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+ }while (irq_status); + +if (disable_irqs) +writel (BUTTRESS_IRQS & ~disable_irqs, + isp->base +BUTTRESS_REG_ISR_ENABLE ); + +pm_runtime_put (&isp->pdev->dev); + +return ret; +}
...
/Andreas
Re: [PATCH v2 03/15] media: intel/ipu6: add IPU6 buttress interface driver
Quoted 38 lines · Andreas Helbech Kleist
Hi Bingbu,
Quoted 10 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
The IPU6 buttress is the interface between IPU device (input system and processing system) with rest of the SoC. It contains overall IPU hardware control registers, these control registers are used as the interfaces with the Intel Converged Security Engine and Punit to do firmware authentication and power management.
Signed-off-by Bingbu Cao <[email protected]>---
...
Quoted 22 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +22
-0+static irqreturn_t ipu6_buttress_call_isr (struct ipu6_bus_device *adev) +{ +irqreturn_t ret =IRQ_WAKE_THREAD ; + +if (!adev || !adev->auxdrv || !adev->auxdrv_data) +return IRQ_NONE ; + +if (adev->auxdrv_data->isr) + ret = adev->auxdrv_data->isr (adev); + +if (ret ==IRQ_WAKE_THREAD && !adev->auxdrv_data->isr_threaded) + ret =IRQ_NONE ; + +return ret; +} + +irqreturn_t ipu6_buttress_isr (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +struct ipu6_buttress *b = &isp->buttress; +u32 reg_irq_sts =BUTTRESS_REG_ISR_STATUS ;
Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c line 339
Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +8
-0+irqreturn_t ret =IRQ_NONE ; +u32 disable_irqs =0 ; +u32 irq_status; +u32 i, count =0 ; + +pm_runtime_get_noresume (&isp->pdev->dev); + + irq_status =readl (isp->base + reg_irq_sts);
A drive-by comment: this seems dodgy. If someone calls pm_runtime_put*() just before the pm_runtime_get_noresume() above, the device won't be resumed when reading the register, which will likely not lead to the desired result.
Quoted 100 lines
Quoted 11 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +11
-0+if (!irq_status) { +pm_runtime_put_noidle (&isp->pdev->dev); +return IRQ_NONE ; + } + + do { + writel(irq_status, isp->base + BUTTRESS_REG_ISR_CLEAR); + + for (i = 0; i < ARRAY_SIZE(ipu6_adev_irq_mask); i++) { + irqreturn_t r = ipu6_buttress_call_isr(adev[i]); +drivers/media/pci/intel/ipu6/ipu6-buttress.c line 360
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+if (!(irq_status & ipu6_adev_irq_mask[i])) +continue ; + +if (r ==IRQ_WAKE_THREAD ) { + ret =IRQ_WAKE_THREAD ; + disable_irqs |= ipu6_adev_irq_mask[i]; + }else if (ret ==IRQ_NONE && r ==IRQ_HANDLED ) { + ret =IRQ_HANDLED ; + } + }It seems wrong to call the ISR for a adev[i] before checking the corresponding IRQ mask. If the mask is not set, the ISR is still called, but the result is thrown away.
I started investigating this because I'm seeing "general protection fault, probably for non-canonical address
0x6b6b6b6b6b6b6b6b" in this function when unbinding the IPU4 driver.How do you ensure that the ISR is not called on a ipu6-bus device that has been deleted? Specifically in
ipu6_pci_remove,ipu6_bus_del_devicesis called beforeipu6_buttress_exit(which disables buttress IRQs). Perhaps the above for loop should really be a "for each ipu6-bus device" loop?Quoted 50 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +50
-0+ +if ((irq_status &BUTTRESS_EVENT ) && ret ==IRQ_NONE ) + ret =IRQ_HANDLED ; + +if (irq_status &BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->cse, &b->cse.recv_data); +complete (&b->cse.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->ish, &b->ish.recv_data); +complete (&b->ish.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->cse.send_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_ISH ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->ish.send_complete); + } + +if (irq_status &BUTTRESS_ISR_SAI_VIOLATION && +ipu6_buttress_get_secure_mode (isp)) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_SAI_VIOLATION\n" ); + +if (irq_status & (BUTTRESS_ISR_IS_FATAL_MEM_ERR | +BUTTRESS_ISR_PS_FATAL_MEM_ERR )) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_FATAL_MEM_ERR\n" ); + +if (irq_status &BUTTRESS_ISR_UFI_ERROR ) +dev_err (&isp->pdev->dev,"BUTTRESS_ISR_UFI_ERROR\n" ); + +if (++count ==BUTTRESS_MAX_CONSECUTIVE_IRQS ) { +dev_err (&isp->pdev->dev,"too many consecutive IRQs\n" ); + ret =IRQ_NONE ; +break ; + } + + irq_status =readl (isp->base + reg_irq_sts);drivers/media/pci/intel/ipu6/ipu6-buttress.c line 420
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+ }while (irq_status); + +if (disable_irqs) +writel (BUTTRESS_IRQS & ~disable_irqs, + isp->base +BUTTRESS_REG_ISR_ENABLE ); + +pm_runtime_put (&isp->pdev->dev); + +return ret; +}...
/Andreas
Signature · 3 lines
Regards, Laurent Pinchart
Re: [PATCH v2 03/15] media: intel/ipu6: add IPU6 buttress interface driver
Laurent,
Quoted 44 lines · Laurent Pinchart
Quoted 43 lines · Andreas Helbech Kleist
Hi Bingbu,
Quoted 10 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
The IPU6 buttress is the interface between IPU device (input system and processing system) with rest of the SoC. It contains overall IPU hardware control registers, these control registers are used as the interfaces with the Intel Converged Security Engine and Punit to do firmware authentication and power management.
Signed-off-by Bingbu Cao <[email protected]>---
...
Quoted 27 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +27
-0+static irqreturn_t ipu6_buttress_call_isr (struct ipu6_bus_device *adev) +{ +irqreturn_t ret =IRQ_WAKE_THREAD ; + +if (!adev || !adev->auxdrv || !adev->auxdrv_data) +return IRQ_NONE ; + +if (adev->auxdrv_data->isr) + ret = adev->auxdrv_data->isr (adev); + +if (ret ==IRQ_WAKE_THREAD && !adev->auxdrv_data->isr_threaded) + ret =IRQ_NONE ; + +return ret; +} + +irqreturn_t ipu6_buttress_isr (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +struct ipu6_buttress *b = &isp->buttress; +u32 reg_irq_sts =BUTTRESS_REG_ISR_STATUS ; +irqreturn_t ret =IRQ_NONE ; +u32 disable_irqs =0 ; +u32 irq_status; +u32 i, count =0 ; +
Quoted 8 lines
Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c line 339
Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +3
-0+pm_runtime_get_noresume (&isp->pdev->dev); + + irq_status =readl (isp->base + reg_irq_sts);A drive-by comment: this seems dodgy. If someone calls pm_runtime_put*() just before the
pm_runtime_get_noresume()above, the device won't be resumed when reading the register, which will likely not lead to the desired result.
Thanks for your review.
What do you think using pm_runtime_get_if_in_use() here?
Quoted 102 lines
Quoted 101 lines
Quoted 12 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +11
-0+if (!irq_status) { +pm_runtime_put_noidle (&isp->pdev->dev); +return IRQ_NONE ; + } + + do { + writel(irq_status, isp->base + BUTTRESS_REG_ISR_CLEAR); + + for (i = 0; i < ARRAY_SIZE(ipu6_adev_irq_mask); i++) { + irqreturn_t r = ipu6_buttress_call_isr(adev[i]); +drivers/media/pci/intel/ipu6/ipu6-buttress.c line 360
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+if (!(irq_status & ipu6_adev_irq_mask[i])) +continue ; + +if (r ==IRQ_WAKE_THREAD ) { + ret =IRQ_WAKE_THREAD ; + disable_irqs |= ipu6_adev_irq_mask[i]; + }else if (ret ==IRQ_NONE && r ==IRQ_HANDLED ) { + ret =IRQ_HANDLED ; + } + }It seems wrong to call the ISR for a adev[i] before checking the corresponding IRQ mask. If the mask is not set, the ISR is still called, but the result is thrown away.
I started investigating this because I'm seeing "general protection fault, probably for non-canonical address
0x6b6b6b6b6b6b6b6b" in this function when unbinding the IPU4 driver.How do you ensure that the ISR is not called on a ipu6-bus device that has been deleted? Specifically in
ipu6_pci_remove,ipu6_bus_del_devicesis called beforeipu6_buttress_exit(which disables buttress IRQs). Perhaps the above for loop should really be a "for each ipu6-bus device" loop?Quoted 50 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +50
-0+ +if ((irq_status &BUTTRESS_EVENT ) && ret ==IRQ_NONE ) + ret =IRQ_HANDLED ; + +if (irq_status &BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->cse, &b->cse.recv_data); +complete (&b->cse.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->ish, &b->ish.recv_data); +complete (&b->ish.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->cse.send_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_ISH ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->ish.send_complete); + } + +if (irq_status &BUTTRESS_ISR_SAI_VIOLATION && +ipu6_buttress_get_secure_mode (isp)) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_SAI_VIOLATION\n" ); + +if (irq_status & (BUTTRESS_ISR_IS_FATAL_MEM_ERR | +BUTTRESS_ISR_PS_FATAL_MEM_ERR )) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_FATAL_MEM_ERR\n" ); + +if (irq_status &BUTTRESS_ISR_UFI_ERROR ) +dev_err (&isp->pdev->dev,"BUTTRESS_ISR_UFI_ERROR\n" ); + +if (++count ==BUTTRESS_MAX_CONSECUTIVE_IRQS ) { +dev_err (&isp->pdev->dev,"too many consecutive IRQs\n" ); + ret =IRQ_NONE ; +break ; + } + + irq_status =readl (isp->base + reg_irq_sts);drivers/media/pci/intel/ipu6/ipu6-buttress.c line 420
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+ }while (irq_status); + +if (disable_irqs) +writel (BUTTRESS_IRQS & ~disable_irqs, + isp->base +BUTTRESS_REG_ISR_ENABLE ); + +pm_runtime_put (&isp->pdev->dev); + +return ret; +}...
/Andreas
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 03/15] media: intel/ipu6: add IPU6 buttress interface driver
Hi Bingbu,
Quoted 52 lines · Bingbu Cao
Quoted 42 lines · Laurent Pinchart
Quoted 41 lines · Andreas Helbech Kleist
Quoted 10 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
The IPU6 buttress is the interface between IPU device (input system and processing system) with rest of the SoC. It contains overall IPU hardware control registers, these control registers are used as the interfaces with the Intel Converged Security Engine and Punit to do firmware authentication and power management.
Signed-off-by Bingbu Cao <[email protected]>---
...
Quoted 27 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +27
-0+static irqreturn_t ipu6_buttress_call_isr (struct ipu6_bus_device *adev) +{ +irqreturn_t ret =IRQ_WAKE_THREAD ; + +if (!adev || !adev->auxdrv || !adev->auxdrv_data) +return IRQ_NONE ; + +if (adev->auxdrv_data->isr) + ret = adev->auxdrv_data->isr (adev); + +if (ret ==IRQ_WAKE_THREAD && !adev->auxdrv_data->isr_threaded) + ret =IRQ_NONE ; + +return ret; +} + +irqreturn_t ipu6_buttress_isr (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +struct ipu6_buttress *b = &isp->buttress; +u32 reg_irq_sts =BUTTRESS_REG_ISR_STATUS ; +irqreturn_t ret =IRQ_NONE ; +u32 disable_irqs =0 ; +u32 irq_status; +u32 i, count =0 ; +Quoted 8 lines
Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c line 339
Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +3
-0+pm_runtime_get_noresume (&isp->pdev->dev); + + irq_status =readl (isp->base + reg_irq_sts);A drive-by comment: this seems dodgy. If someone calls pm_runtime_put*() just before the
pm_runtime_get_noresume()above, the device won't be resumed when reading the register, which will likely not lead to the desired result.
Quoted 2 lines
Thanks for your review.
What do you think usingpm_runtime_get_if_in_use()here?
This is the second discussion in a few weeks about this topic. See
https://lore.kernel.org/linux-media/[email protected]
Sakari, should we try to handle this issue with shared helpers ?
Quoted 98 lines
Quoted 98 lines
Quoted 98 lines
Quoted 11 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +11
-0+if (!irq_status) { +pm_runtime_put_noidle (&isp->pdev->dev); +return IRQ_NONE ; + } + + do { + writel(irq_status, isp->base + BUTTRESS_REG_ISR_CLEAR); + + for (i = 0; i < ARRAY_SIZE(ipu6_adev_irq_mask); i++) { + irqreturn_t r = ipu6_buttress_call_isr(adev[i]); +drivers/media/pci/intel/ipu6/ipu6-buttress.c line 360
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+if (!(irq_status & ipu6_adev_irq_mask[i])) +continue ; + +if (r ==IRQ_WAKE_THREAD ) { + ret =IRQ_WAKE_THREAD ; + disable_irqs |= ipu6_adev_irq_mask[i]; + }else if (ret ==IRQ_NONE && r ==IRQ_HANDLED ) { + ret =IRQ_HANDLED ; + } + }It seems wrong to call the ISR for a adev[i] before checking the corresponding IRQ mask. If the mask is not set, the ISR is still called, but the result is thrown away.
I started investigating this because I'm seeing "general protection fault, probably for non-canonical address
0x6b6b6b6b6b6b6b6b" in this function when unbinding the IPU4 driver.How do you ensure that the ISR is not called on a ipu6-bus device that has been deleted? Specifically in
ipu6_pci_remove,ipu6_bus_del_devicesis called beforeipu6_buttress_exit(which disables buttress IRQs). Perhaps the above for loop should really be a "for each ipu6-bus device" loop?Quoted 60 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +60
-0+ +if ((irq_status &BUTTRESS_EVENT ) && ret ==IRQ_NONE ) + ret =IRQ_HANDLED ; + +if (irq_status &BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->cse, &b->cse.recv_data); +complete (&b->cse.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->ish, &b->ish.recv_data); +complete (&b->ish.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->cse.send_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_ISH ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->ish.send_complete); + } + +if (irq_status &BUTTRESS_ISR_SAI_VIOLATION && +ipu6_buttress_get_secure_mode (isp)) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_SAI_VIOLATION\n" ); + +if (irq_status & (BUTTRESS_ISR_IS_FATAL_MEM_ERR | +BUTTRESS_ISR_PS_FATAL_MEM_ERR )) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_FATAL_MEM_ERR\n" ); + +if (irq_status &BUTTRESS_ISR_UFI_ERROR ) +dev_err (&isp->pdev->dev,"BUTTRESS_ISR_UFI_ERROR\n" ); + +if (++count ==BUTTRESS_MAX_CONSECUTIVE_IRQS ) { +dev_err (&isp->pdev->dev,"too many consecutive IRQs\n" ); + ret =IRQ_NONE ; +break ; + } + + irq_status =readl (isp->base + reg_irq_sts); + }while (irq_status); + +if (disable_irqs) +writel (BUTTRESS_IRQS & ~disable_irqs, + isp->base +BUTTRESS_REG_ISR_ENABLE ); + +pm_runtime_put (&isp->pdev->dev); + +return ret; +}...
Signature · 3 lines
Regards, Laurent Pinchart
Re: [PATCH v2 03/15] media: intel/ipu6: add IPU6 buttress interface driver
In reply to Andreas Helbech KleistAndreas,
Quoted 76 lines · Andreas Helbech Kleist
Hi Bingbu,
Quoted 10 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
The IPU6 buttress is the interface between IPU device (input system and processing system) with rest of the SoC. It contains overall IPU hardware control registers, these control registers are used as the interfaces with the Intel Converged Security Engine and Punit to do firmware authentication and power management.
Signed-off-by Bingbu Cao <[email protected]>---
...
Quoted 41 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +41
-0+static irqreturn_t ipu6_buttress_call_isr (struct ipu6_bus_device *adev) +{ +irqreturn_t ret =IRQ_WAKE_THREAD ; + +if (!adev || !adev->auxdrv || !adev->auxdrv_data) +return IRQ_NONE ; + +if (adev->auxdrv_data->isr) + ret = adev->auxdrv_data->isr (adev); + +if (ret ==IRQ_WAKE_THREAD && !adev->auxdrv_data->isr_threaded) + ret =IRQ_NONE ; + +return ret; +} + +irqreturn_t ipu6_buttress_isr (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +struct ipu6_buttress *b = &isp->buttress; +u32 reg_irq_sts =BUTTRESS_REG_ISR_STATUS ; +irqreturn_t ret =IRQ_NONE ; +u32 disable_irqs =0 ; +u32 irq_status; +u32 i, count =0 ; + +pm_runtime_get_noresume (&isp->pdev->dev); + + irq_status =readl (isp->base + reg_irq_sts); +if (!irq_status) { +pm_runtime_put_noidle (&isp->pdev->dev); +return IRQ_NONE ; + } + + do { + writel(irq_status, isp->base + BUTTRESS_REG_ISR_CLEAR); + + for (i = 0; i < ARRAY_SIZE(ipu6_adev_irq_mask); i++) { + irqreturn_t r = ipu6_buttress_call_isr(adev[i]); +drivers/media/pci/intel/ipu6/ipu6-buttress.c line 360
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+if (!(irq_status & ipu6_adev_irq_mask[i])) +continue ; + +if (r ==IRQ_WAKE_THREAD ) { + ret =IRQ_WAKE_THREAD ; + disable_irqs |= ipu6_adev_irq_mask[i]; + }else if (ret ==IRQ_NONE && r ==IRQ_HANDLED ) { + ret =IRQ_HANDLED ; + } + }It seems wrong to call the ISR for a adev[i] before checking the corresponding IRQ mask. If the mask is not set, the ISR is still called, but the result is thrown away.
I started investigating this because I'm seeing "general protection fault, probably for non-canonical address
0x6b6b6b6b6b6b6b6b" in this function when unbinding the IPU4 driver.
Quoted 5 lines
How do you ensure that the ISR is not called on a ipu6-bus device that has been deleted? Specifically in
ipu6_pci_remove,ipu6_bus_del_devicesis called beforeipu6_buttress_exit(which disables buttress IRQs). Perhaps the above for loop should really be a "for each ipu6-bus device" loop?
You are right, the buttress_exit() should be called before ipu6_bus_del_devices().
Even with this, driver cannot guarantee that hardware irq was actually disabled on IPU4. In some error cases, HW could report unmasked errors on IPU4 (no such case on IPU6), if I remember correctly.
Have you check whether devm_free_irq() in ipu6_pci_remove() can help you?
Quoted 66 lines
Quoted 51 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +50
-0+ +if ((irq_status &BUTTRESS_EVENT ) && ret ==IRQ_NONE ) + ret =IRQ_HANDLED ; + +if (irq_status &BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->cse, &b->cse.recv_data); +complete (&b->cse.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->ish, &b->ish.recv_data); +complete (&b->ish.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->cse.send_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_ISH ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->ish.send_complete); + } + +if (irq_status &BUTTRESS_ISR_SAI_VIOLATION && +ipu6_buttress_get_secure_mode (isp)) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_SAI_VIOLATION\n" ); + +if (irq_status & (BUTTRESS_ISR_IS_FATAL_MEM_ERR | +BUTTRESS_ISR_PS_FATAL_MEM_ERR )) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_FATAL_MEM_ERR\n" ); + +if (irq_status &BUTTRESS_ISR_UFI_ERROR ) +dev_err (&isp->pdev->dev,"BUTTRESS_ISR_UFI_ERROR\n" ); + +if (++count ==BUTTRESS_MAX_CONSECUTIVE_IRQS ) { +dev_err (&isp->pdev->dev,"too many consecutive IRQs\n" ); + ret =IRQ_NONE ; +break ; + } + + irq_status =readl (isp->base + reg_irq_sts);drivers/media/pci/intel/ipu6/ipu6-buttress.c line 420
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+ }while (irq_status); + +if (disable_irqs) +writel (BUTTRESS_IRQS & ~disable_irqs, + isp->base +BUTTRESS_REG_ISR_ENABLE ); + +pm_runtime_put (&isp->pdev->dev); + +return ret; +}...
/Andreas
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 03/15] media: intel/ipu6: add IPU6 buttress interface driver
Andreas,
Quoted 85 lines · Bingbu Cao
Andreas,
Quoted 76 lines · Andreas Helbech Kleist
Hi Bingbu,
Quoted 10 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
The IPU6 buttress is the interface between IPU device (input system and processing system) with rest of the SoC. It contains overall IPU hardware control registers, these control registers are used as the interfaces with the Intel Converged Security Engine and Punit to do firmware authentication and power management.
Signed-off-by Bingbu Cao <[email protected]>---
...
Quoted 41 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +41
-0+static irqreturn_t ipu6_buttress_call_isr (struct ipu6_bus_device *adev) +{ +irqreturn_t ret =IRQ_WAKE_THREAD ; + +if (!adev || !adev->auxdrv || !adev->auxdrv_data) +return IRQ_NONE ; + +if (adev->auxdrv_data->isr) + ret = adev->auxdrv_data->isr (adev); + +if (ret ==IRQ_WAKE_THREAD && !adev->auxdrv_data->isr_threaded) + ret =IRQ_NONE ; + +return ret; +} + +irqreturn_t ipu6_buttress_isr (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +struct ipu6_buttress *b = &isp->buttress; +u32 reg_irq_sts =BUTTRESS_REG_ISR_STATUS ; +irqreturn_t ret =IRQ_NONE ; +u32 disable_irqs =0 ; +u32 irq_status; +u32 i, count =0 ; + +pm_runtime_get_noresume (&isp->pdev->dev); + + irq_status =readl (isp->base + reg_irq_sts); +if (!irq_status) { +pm_runtime_put_noidle (&isp->pdev->dev); +return IRQ_NONE ; + } + + do { + writel(irq_status, isp->base + BUTTRESS_REG_ISR_CLEAR); + + for (i = 0; i < ARRAY_SIZE(ipu6_adev_irq_mask); i++) { + irqreturn_t r = ipu6_buttress_call_isr(adev[i]); +drivers/media/pci/intel/ipu6/ipu6-buttress.c line 360
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+if (!(irq_status & ipu6_adev_irq_mask[i])) +continue ; + +if (r ==IRQ_WAKE_THREAD ) { + ret =IRQ_WAKE_THREAD ; + disable_irqs |= ipu6_adev_irq_mask[i]; + }else if (ret ==IRQ_NONE && r ==IRQ_HANDLED ) { + ret =IRQ_HANDLED ; + } + }It seems wrong to call the ISR for a adev[i] before checking the corresponding IRQ mask. If the mask is not set, the ISR is still called, but the result is thrown away.
I started investigating this because I'm seeing "general protection fault, probably for non-canonical address
0x6b6b6b6b6b6b6b6b" in this function when unbinding the IPU4 driver.Quoted 5 lines
How do you ensure that the ISR is not called on a ipu6-bus device that has been deleted? Specifically in
ipu6_pci_remove,ipu6_bus_del_devicesis called beforeipu6_buttress_exit(which disables buttress IRQs). Perhaps the above for loop should really be a "for each ipu6-bus device" loop?
Quoted 8 lines
You are right, the
buttress_exit()should be called beforeipu6_bus_del_devices().Even with this, driver cannot guarantee that hardware irq was actually disabled on IPU4. In some error cases, HW could report unmasked errors on IPU4 (no such case on IPU6), if I remember correctly.
Have you check whether
devm_free_irq()inipu6_pci_remove()can help you?
BTW, could you share the irq_status in your case?
Quoted 68 lines
Quoted 67 lines
Quoted 52 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +50
-0+ +if ((irq_status &BUTTRESS_EVENT ) && ret ==IRQ_NONE ) + ret =IRQ_HANDLED ; + +if (irq_status &BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_CSE_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->cse, &b->cse.recv_data); +complete (&b->cse.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_FROM_ISH_IS_WAITING\n" ); +ipu6_buttress_ipc_recv (isp, &b->ish, &b->ish.recv_data); +complete (&b->ish.recv_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->cse.send_complete); + } + +if (irq_status &BUTTRESS_ISR_IPC_EXEC_DONE_BY_ISH ) { +dev_dbg (&isp->pdev->dev, +"BUTTRESS_ISR_IPC_EXEC_DONE_BY_CSE\n" ); +complete (&b->ish.send_complete); + } + +if (irq_status &BUTTRESS_ISR_SAI_VIOLATION && +ipu6_buttress_get_secure_mode (isp)) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_SAI_VIOLATION\n" ); + +if (irq_status & (BUTTRESS_ISR_IS_FATAL_MEM_ERR | +BUTTRESS_ISR_PS_FATAL_MEM_ERR )) +dev_err (&isp->pdev->dev, +"BUTTRESS_ISR_FATAL_MEM_ERR\n" ); + +if (irq_status &BUTTRESS_ISR_UFI_ERROR ) +dev_err (&isp->pdev->dev,"BUTTRESS_ISR_UFI_ERROR\n" ); + +if (++count ==BUTTRESS_MAX_CONSECUTIVE_IRQS ) { +dev_err (&isp->pdev->dev,"too many consecutive IRQs\n" ); + ret =IRQ_NONE ; +break ; + } + + irq_status =readl (isp->base + reg_irq_sts);drivers/media/pci/intel/ipu6/ipu6-buttress.c line 420
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+ }while (irq_status); + +if (disable_irqs) +writel (BUTTRESS_IRQS & ~disable_irqs, + isp->base +BUTTRESS_REG_ISR_ENABLE ); + +pm_runtime_put (&isp->pdev->dev); + +return ret; +}...
/Andreas
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 03/15] media: intel/ipu6: add IPU6 buttress interface driver
Bingbu,
Quoted 94 lines · Bingbu Cao
Andreas,
Quoted 91 lines · Bingbu Cao
Andreas,
Quoted 76 lines · Andreas Helbech Kleist
Hi Bingbu,
Quoted 10 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
The IPU6 buttress is the interface between IPU device (input system and processing system) with rest of the SoC. It contains overall IPU hardware control registers, these control registers are used as the interfaces with the Intel Converged Security Engine and Punit to do firmware authentication and power management.
Signed-off-by Bingbu Cao <[email protected]>---
...
Quoted 41 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +41
-0+static irqreturn_t ipu6_buttress_call_isr (struct ipu6_bus_device *adev) +{ +irqreturn_t ret =IRQ_WAKE_THREAD ; + +if (!adev || !adev->auxdrv || !adev->auxdrv_data) +return IRQ_NONE ; + +if (adev->auxdrv_data->isr) + ret = adev->auxdrv_data->isr (adev); + +if (ret ==IRQ_WAKE_THREAD && !adev->auxdrv_data->isr_threaded) + ret =IRQ_NONE ; + +return ret; +} + +irqreturn_t ipu6_buttress_isr (int irq,void *isp_ptr) +{ +struct ipu6_device *isp = isp_ptr; +struct ipu6_bus_device *adev[] = { isp->isys, isp->psys }; +struct ipu6_buttress *b = &isp->buttress; +u32 reg_irq_sts =BUTTRESS_REG_ISR_STATUS ; +irqreturn_t ret =IRQ_NONE ; +u32 disable_irqs =0 ; +u32 irq_status; +u32 i, count =0 ; + +pm_runtime_get_noresume (&isp->pdev->dev); + + irq_status =readl (isp->base + reg_irq_sts); +if (!irq_status) { +pm_runtime_put_noidle (&isp->pdev->dev); +return IRQ_NONE ; + } + + do { + writel(irq_status, isp->base + BUTTRESS_REG_ISR_CLEAR); + + for (i = 0; i < ARRAY_SIZE(ipu6_adev_irq_mask); i++) { + irqreturn_t r = ipu6_buttress_call_isr(adev[i]); +drivers/media/pci/intel/ipu6/ipu6-buttress.c line 360
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-buttress.c new file +10
-0+if (!(irq_status & ipu6_adev_irq_mask[i])) +continue ; + +if (r ==IRQ_WAKE_THREAD ) { + ret =IRQ_WAKE_THREAD ; + disable_irqs |= ipu6_adev_irq_mask[i]; + }else if (ret ==IRQ_NONE && r ==IRQ_HANDLED ) { + ret =IRQ_HANDLED ; + } + }It seems wrong to call the ISR for a adev[i] before checking the corresponding IRQ mask. If the mask is not set, the ISR is still called, but the result is thrown away.
I started investigating this because I'm seeing "general protection fault, probably for non-canonical address
0x6b6b6b6b6b6b6b6b" in this function when unbinding the IPU4 driver.Quoted 5 lines
How do you ensure that the ISR is not called on a ipu6-bus device that has been deleted? Specifically in
ipu6_pci_remove,ipu6_bus_del_devicesis called beforeipu6_buttress_exit(which disables buttress IRQs). Perhaps the above for loop should really be a "for each ipu6-bus device" loop?You are right, the
buttress_exit()should be called beforeipu6_bus_del_devices().Even with this, driver cannot guarantee that hardware irq was actually disabled on IPU4. In some error cases, HW could report unmasked errors on IPU4 (no such case on IPU6), if I remember correctly.
Thanks for the info, that's good to know.
Quoted 1 line
Quoted 1 line
Have you check whether
devm_free_irq()inipu6_pci_remove()can help you?
It looks like it might help.
Quoted 1 line
BTW, could you share the
irq_statusin your case?
Sorry, I don't have a log of it. And it seems to be a bit heisenbug- like - it doesn't always happen when I add logging :/
/Andreas
PATCH v2 4/15 media: intel/ipu6: CPD parsing for get firmware components
In reply to [email protected]From: Bingbu Cao <[email protected]>
For IPU6, firmware is generated and released as signed Code Partition Directory (CPD) format file, which is aligned with the SPI flash code partition definition. CPD format include CPD header, manifest, metadata and module data. Driver can parse them according to the CPD layout to acquire each component.
--- drivers/media/pci/intel/ipu6/ipu6-cpd.c | 355 ++++++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-cpd.h | 104 +++++++ 2 files changed, 459 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-cpd.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-cpd.h
drivers/media/pci/intel/ipu6/ipu6-cpd.c new file +355 -0
@@ -0,0 +1,355 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <linux/bitfield.h> + +#include "ipu6.h" +#include "ipu6-bus.h" +#include "ipu6-cpd.h" + +/* 15 entries + header*/ +#define MAX_PKG_DIR_ENT_CNT 16 +/* 2 qword per entry/header */ +#define PKG_DIR_ENT_LEN 2 +/* PKG_DIR size in bytes */ +#define PKG_DIR_SIZE ((MAX_PKG_DIR_ENT_CNT) * \ + (PKG_DIR_ENT_LEN) * sizeof(u64)) +/* _IUPKDR_ */ +#define PKG_DIR_HDR_MARK 0x5f4955504b44525fULL + +/* $CPD */ +#define CPD_HDR_MARK 0x44504324 + +#define MAX_MANIFEST_SIZE (SZ_2K * sizeof(u32)) +#define MAX_METADATA_SIZE SZ_64K + +#define MAX_COMPONENT_ID 127 +#define MAX_COMPONENT_VERSION 0xffff + +#define MANIFEST_IDX 0 +#define METADATA_IDX 1 +#define MODULEDATA_IDX 2 +/* +* PKG_DIR Entry (type == id) +* 63:56 55 54:48 47:32 31:24 23:0 +* Rsvd Rsvd Type Version Rsvd Size +*/ +#define PKG_DIR_SIZE_MASK GENMASK(23, 0) +#define PKG_DIR_VERSION_MASK GENMASK(47, 32) +#define PKG_DIR_TYPE_MASK GENMASK(54, 48) + +static inline const struct ipu6_cpd_ent *ipu6_cpd_get_entry (const void *cpd, +u8 idx) +{ +const struct ipu6_cpd_hdr *cpd_hdr = cpd; +const struct ipu6_cpd_ent *ent; + + ent = (const struct ipu6_cpd_ent *)((const u8 *)cpd + cpd_hdr->hdr_len); +return ent + idx; +} + +#define ipu6_cpd_get_manifest (cpd) ipu6_cpd_get_entry(cpd, MANIFEST_IDX) +#define ipu6_cpd_get_metadata (cpd) ipu6_cpd_get_entry(cpd, METADATA_IDX) +#define ipu6_cpd_get_moduledata (cpd) ipu6_cpd_get_entry(cpd, MODULEDATA_IDX) + +static const struct ipu6_cpd_metadata_cmpnt_hdr * +ipu6_cpd_metadata_get_cmpnt (struct ipu6_device *isp,const void *metadata, +unsigned int metadata_size,u8 idx) +{ +size_t extn_size =sizeof (struct ipu6_cpd_metadata_extn ); +size_t cmpnt_count = metadata_size - extn_size; + + cmpnt_count =div_u64 (cmpnt_count, isp->cpd_metadata_cmpnt_size); + +if (idx >MAX_COMPONENT_ID || idx >= cmpnt_count) { +dev_err (&isp->pdev->dev,"Component index out of range (%d)\n" , + idx); +return ERR_PTR (-EINVAL ); + } + +return metadata + extn_size + idx * isp->cpd_metadata_cmpnt_size; +} + +static u32 ipu6_cpd_metadata_cmpnt_version (struct ipu6_device *isp, +const void *metadata, +unsigned int metadata_size,u8 idx) +{ +const struct ipu6_cpd_metadata_cmpnt_hdr *cmpnt; + + cmpnt =ipu6_cpd_metadata_get_cmpnt (isp, metadata, metadata_size, idx); +if (IS_ERR (cmpnt)) +return PTR_ERR (cmpnt); + +return cmpnt->ver; +} + +static int ipu6_cpd_metadata_get_cmpnt_id (struct ipu6_device *isp, +const void *metadata, +unsigned int metadata_size,u8 idx) +{ +const struct ipu6_cpd_metadata_cmpnt_hdr *cmpnt; + + cmpnt =ipu6_cpd_metadata_get_cmpnt (isp, metadata, metadata_size, idx); +if (IS_ERR (cmpnt)) +return PTR_ERR (cmpnt); + +return cmpnt->id; +} + +static int ipu6_cpd_parse_module_data (struct ipu6_device *isp, +const void *module_data, +unsigned int module_data_size, +dma_addr_t dma_addr_module_data, +u64 *pkg_dir,const void *metadata, +unsigned int metadata_size) +{ +const struct ipu6_cpd_module_data_hdr *module_data_hdr; +const struct ipu6_cpd_hdr *dir_hdr; +const struct ipu6_cpd_ent *dir_ent; +unsigned int i; +u8 len; + +if (!module_data) +return -EINVAL ; + + module_data_hdr = module_data; + dir_hdr = module_data + module_data_hdr->hdr_len; + len = dir_hdr->hdr_len; + dir_ent = (const struct ipu6_cpd_ent *)(((u8 *)dir_hdr) + len); + + pkg_dir[0 ] =PKG_DIR_HDR_MARK ; +/* pkg_dir entry count = component count + pkg_dir header */ + pkg_dir[1 ] = dir_hdr->ent_cnt +1 ; + +for (i =0 ; i < dir_hdr->ent_cnt; i++, dir_ent++) { +u64 *p = &pkg_dir[PKG_DIR_ENT_LEN * (1 + i)]; +int ver, id; + + *p++ = dma_addr_module_data + dir_ent->offset; + id =ipu6_cpd_metadata_get_cmpnt_id (isp, metadata, + metadata_size, i); +if (id <0 || id >MAX_COMPONENT_ID ) { +dev_err (&isp->pdev->dev,"Invalid CPD component id\n" ); +return -EINVAL ; + } + + ver =ipu6_cpd_metadata_cmpnt_version (isp, metadata, + metadata_size, i); +if (ver <0 || ver >MAX_COMPONENT_VERSION ) { +dev_err (&isp->pdev->dev, +"Invalid CPD component version\n" ); +return -EINVAL ; + } + + *p =FIELD_PREP (PKG_DIR_SIZE_MASK , dir_ent->len) | +FIELD_PREP (PKG_DIR_TYPE_MASK , id) | +FIELD_PREP (PKG_DIR_VERSION_MASK , ver); + } + +return 0 ; +} + +int ipu6_cpd_create_pkg_dir (struct ipu6_bus_device *adev,const void *src) +{ +dma_addr_t dma_addr_src =sg_dma_address (adev->fw_sgt.sgl); +const struct ipu6_cpd_ent *ent, *man_ent, *met_ent; +struct device *dev = &adev->auxdev.dev; +struct ipu6_device *isp = adev->isp; +unsigned int man_sz, met_sz; +void *pkg_dir_pos; +int ret; + + man_ent =ipu6_cpd_get_manifest (src); + man_sz = man_ent->len; + + met_ent =ipu6_cpd_get_metadata (src); + met_sz = met_ent->len; + + adev->pkg_dir_size =PKG_DIR_SIZE + man_sz + met_sz; + adev->pkg_dir =dma_alloc_attrs (dev, adev->pkg_dir_size, + &adev->pkg_dir_dma_addr,GFP_KERNEL ,0 ); +if (!adev->pkg_dir) +return -ENOMEM ; + +/* +* pkg_dir entry/header: +* qword | 63:56 | 55 | 54:48 | 47:32 | 31:24 | 23:0 +* N Address/Offset/"_IUPKDR_" +* N + 1 | rsvd | rsvd | type | ver | rsvd | size +* +* We can ignore other fields that size in N + 1 qword as they +* are 0 anyway. Just setting size for now. +*/ + + ent =ipu6_cpd_get_moduledata (src); + + ret =ipu6_cpd_parse_module_data (isp, src + ent->offset, + ent->len, dma_addr_src + ent->offset, + adev->pkg_dir, src + met_ent->offset, + met_ent->len); +if (ret) { +dev_err (&isp->pdev->dev,"Failed to parse module data\n" ); +dma_free_attrs (dev, adev->pkg_dir_size, + adev->pkg_dir, adev->pkg_dir_dma_addr,0 ); +return ret; + } + +/* Copy manifest after pkg_dir */ + pkg_dir_pos = adev->pkg_dir +PKG_DIR_ENT_LEN *MAX_PKG_DIR_ENT_CNT ; +memcpy (pkg_dir_pos, src + man_ent->offset, man_sz); + +/* Copy metadata after manifest */ + pkg_dir_pos += man_sz; +memcpy (pkg_dir_pos, src + met_ent->offset, met_sz); + +dma_sync_single_range_for_device (dev, adev->pkg_dir_dma_addr, +0 , adev->pkg_dir_size,DMA_TO_DEVICE ); + +return 0 ; +} +EXPORT_SYMBOL_NS_GPL (ipu6_cpd_create_pkg_dir,INTEL_IPU6 ); + +void ipu6_cpd_free_pkg_dir (struct ipu6_bus_device *adev) +{ +dma_free_attrs (&adev->auxdev.dev, adev->pkg_dir_size, adev->pkg_dir, + adev->pkg_dir_dma_addr,0 ); +} +EXPORT_SYMBOL_NS_GPL (ipu6_cpd_free_pkg_dir,INTEL_IPU6 ); + +static int ipu6_cpd_validate_cpd (struct ipu6_device *isp,const void *cpd, +unsigned long cpd_size, +unsigned long data_size) +{ +const struct ipu6_cpd_hdr *cpd_hdr = cpd; +const struct ipu6_cpd_ent *ent; +unsigned int i; +u8 len; + + len = cpd_hdr->hdr_len; + +/* Ensure cpd hdr is within moduledata */ +if (cpd_size < len) { +dev_err (&isp->pdev->dev,"Invalid CPD moduledata size\n" ); +return -EINVAL ; + } + +/* Sanity check for CPD header */ +if ((cpd_size - len) /sizeof (*ent) < cpd_hdr->ent_cnt) { +dev_err (&isp->pdev->dev,"Invalid CPD header\n" ); +return -EINVAL ; + } + +/* Ensure that all entries are within moduledata */ + ent = (const struct ipu6_cpd_ent *)(((const u8 *)cpd_hdr) + len); +for (i =0 ; i < cpd_hdr->ent_cnt; i++, ent++) { +if (data_size < ent->offset || + data_size - ent->offset < ent->len) { +dev_err (&isp->pdev->dev,"Invalid CPD entry (%d)\n" , i); +return -EINVAL ; + } + } + +return 0 ; +} + +static int ipu6_cpd_validate_moduledata (struct ipu6_device *isp, +const void *moduledata, +u32 moduledata_size) +{ +const struct ipu6_cpd_module_data_hdr *mod_hdr = moduledata; +int ret; + +/* Ensure moduledata hdr is within moduledata */ +if (moduledata_size <sizeof (*mod_hdr) || + moduledata_size < mod_hdr->hdr_len) { +dev_err (&isp->pdev->dev,"Invalid CPD moduledata size\n" ); +return -EINVAL ; + } + +dev_info (&isp->pdev->dev,"FW version: %x\n" , mod_hdr->fw_pkg_date); + ret =ipu6_cpd_validate_cpd (isp, moduledata + mod_hdr->hdr_len, + moduledata_size - mod_hdr->hdr_len, + moduledata_size); +if (ret) { +dev_err (&isp->pdev->dev,"Invalid CPD in moduledata\n" ); +return ret; + } + +return 0 ; +} + +static int ipu6_cpd_validate_metadata (struct ipu6_device *isp, +const void *metadata,u32 meta_size) +{ +const struct ipu6_cpd_metadata_extn *extn = metadata; + +/* Sanity check for metadata size */ +if (meta_size <sizeof (*extn) || meta_size >MAX_METADATA_SIZE ) { +dev_err (&isp->pdev->dev,"Invalid CPD metadata\n" ); +return -EINVAL ; + } + +/* Validate extension and image types */ +if (extn->extn_type !=IPU6_CPD_METADATA_EXTN_TYPE_IUNIT || + extn->img_type !=IPU6_CPD_METADATA_IMAGE_TYPE_MAIN_FIRMWARE ) { +dev_err (&isp->pdev->dev, +"Invalid CPD metadata descriptor img_type (%d)\n" , + extn->img_type); +return -EINVAL ; + } + +/* Validate metadata size multiple of metadata components */ +if ((meta_size -sizeof (*extn)) % isp->cpd_metadata_cmpnt_size) { +dev_err (&isp->pdev->dev,"Invalid CPD metadata size\n" ); +return -EINVAL ; + } + +return 0 ; +} + +int ipu6_cpd_validate_cpd_file (struct ipu6_device *isp,const void *cpd_file, +unsigned long cpd_file_size) +{ +const struct ipu6_cpd_hdr *hdr = cpd_file; +const struct ipu6_cpd_ent *ent; +int ret; + + ret =ipu6_cpd_validate_cpd (isp, cpd_file, cpd_file_size, + cpd_file_size); +if (ret) { +dev_err (&isp->pdev->dev,"Invalid CPD in file\n" ); +return ret; + } + +/* Check for CPD file marker */ +if (hdr->hdr_mark !=CPD_HDR_MARK ) { +dev_err (&isp->pdev->dev,"Invalid CPD header\n" ); +return -EINVAL ; + } + +/* Sanity check for manifest size */ + ent =ipu6_cpd_get_manifest (cpd_file); +if (ent->len >MAX_MANIFEST_SIZE ) { +dev_err (&isp->pdev->dev,"Invalid CPD manifest size\n" ); +return -EINVAL ; + } + +/* Validate metadata */ + ent =ipu6_cpd_get_metadata (cpd_file); + ret =ipu6_cpd_validate_metadata (isp, cpd_file + ent->offset, ent->len); +if (ret) { +dev_err (&isp->pdev->dev,"Invalid CPD metadata\n" ); +return ret; + } + +/* Validate moduledata */ + ent =ipu6_cpd_get_moduledata (cpd_file); + ret =ipu6_cpd_validate_moduledata (isp, cpd_file + ent->offset, + ent->len); +if (ret) +dev_err (&isp->pdev->dev,"Invalid CPD moduledata\n" ); + +return ret; +}
drivers/media/pci/intel/ipu6/ipu6-cpd.h new file +104 -0
104 lines not on this page · raw message
Signature · 1 line
2.42.0
PATCH v2 5/15 media: intel/ipu6: add IPU6 DMA mapping API and MMU table
In reply to [email protected]From: Bingbu Cao <[email protected]>
he Intel IPU6 has an internal microcontroller (scalar processor, SP) which is used to execute the firmware. The SP can access IPU internal memory and map system DRAM to its an internal 32-bit virtual address space.
This patch adds a driver for the IPU MMU and a DMA mapping implementation using the internal MMU. The system IOMMU may be used besides the IPU MMU.
--- drivers/media/pci/intel/ipu6/ipu6-dma.c | 491 ++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-dma.h | 20 + drivers/media/pci/intel/ipu6/ipu6-mmu.c | 828 ++++++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-mmu.h | 67 ++ 4 files changed, 1406 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-dma.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-dma.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-mmu.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-mmu.h
drivers/media/pci/intel/ipu6/ipu6-dma.c new file +491 -0
@@ -0,0 +1,491 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <linux/cacheflush.h> +#include <linux/vmalloc.h> + +#include "ipu6.h" +#include "ipu6-bus.h" +#include "ipu6-dma.h" +#include "ipu6-mmu.h" + +struct vm_info { +struct list_head list; +struct page **pages; +dma_addr_t ipu6_iova; +void *vaddr; +unsigned long size; +}; + +static struct vm_info *get_vm_info (struct ipu6_mmu *mmu,dma_addr_t iova) +{ +struct vm_info *info, *save; + +list_for_each_entry_safe (info, save, &mmu->vma_list, list) { +if (iova >= info->ipu6_iova && + iova < (info->ipu6_iova + info->size)) +return info; + } + +return NULL ; +} + +static void __dma_clear_buffer (struct page *page,size_t size, +unsigned long attrs) +{ +void *ptr; + +if (!page) +return ; +/* +* Ensure that the allocated pages are zeroed, and that any data +* lurking in the kernel direct-mapped region is invalidated. +*/ + ptr =page_address (page); +memset (ptr,0 , size); +if ((attrs &DMA_ATTR_SKIP_CPU_SYNC ) ==0 ) +clflush_cache_range (ptr, size); +} + +static struct page **__dma_alloc_buffer (struct device *dev,size_t size, +gfp_t gfp,unsigned long attrs) +{ +int count =PHYS_PFN (size); +int array_size = count *sizeof (struct page *); +struct page **pages; +int i =0 ; + + pages =kvzalloc (array_size,GFP_KERNEL ); +if (!pages) +return NULL ; + + gfp |= __GFP_NOWARN; + +while (count) { +int j, order =__fls (count); + + pages[i] =alloc_pages (gfp, order); +while (!pages[i] && order) + pages[i] =alloc_pages (gfp, --order); +if (!pages[i]) + goto error; + +if (order) { +split_page (pages[i], order); + j =1 << order; +while (j--) + pages[i + j] = pages[i] + j; + } + +__dma_clear_buffer (pages[i],PAGE_SIZE << order, attrs); + i +=1 << order; + count -=1 << order; + } + +return pages; +error: +while (i--) +if (pages[i]) +__free_pages (pages[i],0 ); +kvfree (pages); +return NULL ; +} + +static void __dma_free_buffer (struct device *dev,struct page **pages, +size_t size,unsigned long attrs) +{ +int count =PHYS_PFN (size); +unsigned int i; + +for (i =0 ; i < count && pages[i]; i++) { +__dma_clear_buffer (pages[i],PAGE_SIZE , attrs); +__free_pages (pages[i],0 ); + } + +kvfree (pages); +} + +static void ipu6_dma_sync_single_for_cpu (struct device *dev, +dma_addr_t dma_handle, +size_t size, +enum dma_data_direction dir) +{ +void *vaddr; +u32 offset; +struct vm_info *info; +struct ipu6_mmu *mmu =to_ipu6_bus_device (dev)->mmu; + + info =get_vm_info (mmu, dma_handle); +if (WARN_ON (!info)) +return ; + + offset = dma_handle - info->ipu6_iova; +if (WARN_ON (size > (info->size - offset))) +return ; + + vaddr = info->vaddr + offset; +clflush_cache_range (vaddr, size); +} + +static void ipu6_dma_sync_sg_for_cpu (struct device *dev, +struct scatterlist *sglist, +int nents,enum dma_data_direction dir) +{ +struct scatterlist *sg; +int i; + +for_each_sg (sglist, sg, nents, i) +clflush_cache_range (page_to_virt (sg_page (sg)), sg->length); +} + +static void *ipu6_dma_alloc (struct device *dev,size_t size, +dma_addr_t *dma_handle,gfp_t gfp, +unsigned long attrs) +{ +struct ipu6_mmu *mmu =to_ipu6_bus_device (dev)->mmu; +struct pci_dev *pdev =to_ipu6_bus_device (dev)->isp->pdev; +dma_addr_t pci_dma_addr, ipu6_iova; +struct vm_info *info; +unsigned long count; +struct page **pages; +struct iova *iova; +unsigned int i; +int ret; + + info =kzalloc (sizeof (*info),GFP_KERNEL ); +if (!info) +return NULL ; + + size =PAGE_ALIGN (size); + count =PHYS_PFN (size); + + iova =alloc_iova (&mmu->dmap->iovad, count, +PHYS_PFN (dma_get_mask (dev)),0 ); +if (!iova) + goto out_kfree; + + pages =__dma_alloc_buffer (dev, size, gfp, attrs); +if (!pages) + goto out_free_iova; + +dev_dbg (dev,"dma_alloc: size %zu iova low pfn %lu, high pfn %lu\n" , + size, iova->pfn_lo, iova->pfn_hi); +for (i =0 ; iova->pfn_lo + i <= iova->pfn_hi; i++) { + pci_dma_addr =dma_map_page_attrs (&pdev->dev, pages[i],0 , +PAGE_SIZE ,DMA_BIDIRECTIONAL , + attrs); +dev_dbg (dev,"dma_alloc: mapped pci_dma_addr %pad\n" , + &pci_dma_addr); +if (dma_mapping_error (&pdev->dev, pci_dma_addr)) { +dev_err (dev,"pci_dma_mapping for page[%d] failed" , i); + goto out_unmap; + } + + ret =ipu6_mmu_map (mmu->dmap->mmu_info, +PFN_PHYS (iova->pfn_lo + i), pci_dma_addr, +PAGE_SIZE ); +if (ret) { +dev_err (dev,"ipu6_mmu_map for pci_dma[%d] %pad failed" , + i, &pci_dma_addr); +dma_unmap_page_attrs (&pdev->dev, pci_dma_addr, +PAGE_SIZE ,DMA_BIDIRECTIONAL , + attrs); + goto out_unmap; + } + } + + info->vaddr =vmap (pages, count,VM_USERMAP ,PAGE_KERNEL ); +if (!info->vaddr) + goto out_unmap; + + *dma_handle =PFN_PHYS (iova->pfn_lo); + + info->pages = pages; + info->ipu6_iova = *dma_handle; + info->size = size; +list_add (&info->list, &mmu->vma_list); + +return info->vaddr; + +out_unmap: +while (i--) { + ipu6_iova =PFN_PHYS (iova->pfn_lo + i); + pci_dma_addr =ipu6_mmu_iova_to_phys (mmu->dmap->mmu_info, + ipu6_iova); +dma_unmap_page_attrs (&pdev->dev, pci_dma_addr,PAGE_SIZE , +DMA_BIDIRECTIONAL , attrs); + +ipu6_mmu_unmap (mmu->dmap->mmu_info, ipu6_iova,PAGE_SIZE ); + } + +__dma_free_buffer (dev, pages, size, attrs); + +out_free_iova: +__free_iova (&mmu->dmap->iovad, iova); +out_kfree: +kfree (info); + +return NULL ; +} + +static void ipu6_dma_free (struct device *dev,size_t size,void *vaddr, +dma_addr_t dma_handle, +unsigned long attrs) +{ +struct ipu6_mmu *mmu =to_ipu6_bus_device (dev)->mmu; +struct pci_dev *pdev =to_ipu6_bus_device (dev)->isp->pdev; +struct iova *iova =find_iova (&mmu->dmap->iovad,PHYS_PFN (dma_handle)); +dma_addr_t pci_dma_addr, ipu6_iova; +struct vm_info *info; +struct page **pages; +unsigned int i; + +if (WARN_ON (!iova)) +return ; + + info =get_vm_info (mmu, dma_handle); +if (WARN_ON (!info)) +return ; + +if (WARN_ON (!info->vaddr)) +return ; + +if (WARN_ON (!info->pages)) +return ; + +list_del (&info->list); + + size =PAGE_ALIGN (size); + + pages = info->pages; + +vunmap (vaddr); + +for (i =0 ; i <PHYS_PFN (size); i++) { + ipu6_iova =PFN_PHYS (iova->pfn_lo + i); + pci_dma_addr =ipu6_mmu_iova_to_phys (mmu->dmap->mmu_info, + ipu6_iova); +dma_unmap_page_attrs (&pdev->dev, pci_dma_addr,PAGE_SIZE , +DMA_BIDIRECTIONAL , attrs); + } + +ipu6_mmu_unmap (mmu->dmap->mmu_info,PFN_PHYS (iova->pfn_lo), +PFN_PHYS (iova_size (iova))); + +__dma_free_buffer (dev, pages, size, attrs); + + mmu->tlb_invalidate (mmu); + +__free_iova (&mmu->dmap->iovad, iova); + +kfree (info); +} + +static int ipu6_dma_mmap (struct device *dev,struct vm_area_struct *vma, +void *addr,dma_addr_t iova,size_t size, +unsigned long attrs) +{ +struct ipu6_mmu *mmu =to_ipu6_bus_device (dev)->mmu; +size_t count =PHYS_PFN (PAGE_ALIGN (size)); +struct vm_info *info; +size_t i; + + info =get_vm_info (mmu, iova); +if (!info) +return -EFAULT ; + +if (!info->vaddr) +return -EFAULT ; + +if (vma->vm_start & ~PAGE_MASK ) +return -EINVAL ; + +if (size > info->size) +return -EFAULT ; + +for (i =0 ; i < count; i++) +vm_insert_page (vma, vma->vm_start +PFN_PHYS (i), + info->pages[i]); + +return 0 ; +} + +static void ipu6_dma_unmap_sg (struct device *dev, +struct scatterlist *sglist, +int nents,enum dma_data_direction dir, +unsigned long attrs) +{ +struct pci_dev *pdev =to_ipu6_bus_device (dev)->isp->pdev; +struct ipu6_mmu *mmu =to_ipu6_bus_device (dev)->mmu; +struct iova *iova =find_iova (&mmu->dmap->iovad, +PHYS_PFN (sg_dma_address (sglist))); +int i, npages, count; +struct scatterlist *sg; +dma_addr_t pci_dma_addr; + +if (!nents) +return ; + +if (WARN_ON (!iova)) +return ; + +if ((attrs &DMA_ATTR_SKIP_CPU_SYNC ) ==0 ) +ipu6_dma_sync_sg_for_cpu (dev, sglist, nents,DMA_BIDIRECTIONAL ); + +/* get the nents as orig_nents given by caller */ + count =0 ; + npages =iova_size (iova); +for_each_sg (sglist, sg, nents, i) { +if (sg_dma_len (sg) ==0 || +sg_dma_address (sg) ==DMA_MAPPING_ERROR ) +break ; + + npages -=PHYS_PFN (PAGE_ALIGN (sg_dma_len (sg))); + count++; +if (npages <=0 ) +break ; + } + +/* +* Before IPU6 mmu unmap, return the pci dma address back to sg +* assume the nents is less than orig_nents as the least granule +* is 1 SZ_4K page +*/ +dev_dbg (dev,"trying to unmap concatenated %u ents\n" , count); +for_each_sg (sglist, sg, count, i) { +dev_dbg (dev,"ipu unmap sg[%d] %pad\n" , i, &sg_dma_address (sg)); + pci_dma_addr =ipu6_mmu_iova_to_phys (mmu->dmap->mmu_info, +sg_dma_address (sg)); +dev_dbg (dev,"return pci_dma_addr %pad back to sg[%d]\n" , + &pci_dma_addr, i); +sg_dma_address (sg) = pci_dma_addr; + } + +dev_dbg (dev,"ipu6_mmu_unmap low pfn %lu high pfn %lu\n" , + iova->pfn_lo, iova->pfn_hi); +ipu6_mmu_unmap (mmu->dmap->mmu_info,PFN_PHYS (iova->pfn_lo), +PFN_PHYS (iova_size (iova))); + + mmu->tlb_invalidate (mmu); + +dma_unmap_sg_attrs (&pdev->dev, sglist, nents, dir, attrs); + +__free_iova (&mmu->dmap->iovad, iova); +} + +static int ipu6_dma_map_sg (struct device *dev,struct scatterlist *sglist, +int nents,enum dma_data_direction dir, +unsigned long attrs) +{ +struct ipu6_mmu *mmu =to_ipu6_bus_device (dev)->mmu; +struct pci_dev *pdev =to_ipu6_bus_device (dev)->isp->pdev; +struct scatterlist *sg; +struct iova *iova; +size_t npages =0 ; +unsigned long iova_addr; +int i, count; + +for_each_sg (sglist, sg, nents, i) { +if (sg->offset) { +dev_err (dev,"Unsupported non-zero sg[%d].offset %x\n" , + i, sg->offset); +return -EFAULT ; + } + } + +dev_dbg (dev,"pci_dma_map_sg trying to map %d ents\n" , nents); + count =dma_map_sg_attrs (&pdev->dev, sglist, nents, dir, attrs); +if (count <=0 ) { +dev_err (dev,"pci_dma_map_sg %d ents failed\n" , nents); +return 0 ; + } + +dev_dbg (dev,"pci_dma_map_sg %d ents mapped\n" , count); + +for_each_sg (sglist, sg, count, i) + npages +=PHYS_PFN (PAGE_ALIGN (sg_dma_len (sg))); + + iova =alloc_iova (&mmu->dmap->iovad, npages, +PHYS_PFN (dma_get_mask (dev)),0 ); +if (!iova) +return 0 ; + +dev_dbg (dev,"dmamap: iova low pfn %lu, high pfn %lu\n" , iova->pfn_lo, + iova->pfn_hi); + + iova_addr = iova->pfn_lo; +for_each_sg (sglist, sg, count, i) { +int ret; + +dev_dbg (dev,"mapping entry %d: iova 0x%llx phy %pad size %d\n" , + i,PFN_PHYS (iova_addr), &sg_dma_address (sg), +sg_dma_len (sg)); + + ret =ipu6_mmu_map (mmu->dmap->mmu_info,PFN_PHYS (iova_addr), +sg_dma_address (sg), +PAGE_ALIGN (sg_dma_len (sg))); +if (ret) + goto out_fail; + +sg_dma_address (sg) =PFN_PHYS (iova_addr); + + iova_addr +=PHYS_PFN (PAGE_ALIGN (sg_dma_len (sg))); + } + +if ((attrs &DMA_ATTR_SKIP_CPU_SYNC ) ==0 ) +ipu6_dma_sync_sg_for_cpu (dev, sglist, nents,DMA_BIDIRECTIONAL ); + +return count; + +out_fail: +ipu6_dma_unmap_sg (dev, sglist, i, dir, attrs); + +return 0 ; +} + +/* +* Create scatter-list for the already allocated DMA buffer +*/ +static int ipu6_dma_get_sgtable (struct device *dev,struct sg_table *sgt, +void *cpu_addr,dma_addr_t handle,size_t size, +unsigned long attrs) +{ +struct ipu6_mmu *mmu =to_ipu6_bus_device (dev)->mmu; +struct vm_info *info; +int n_pages; +int ret =0 ; + + info =get_vm_info (mmu, handle); +if (!info) +return -EFAULT ; + +if (!info->vaddr) +return -EFAULT ; + +if (WARN_ON (!info->pages)) +return -ENOMEM ; + + n_pages =PHYS_PFN (PAGE_ALIGN (size)); + + ret =sg_alloc_table_from_pages (sgt, info->pages, n_pages,0 , size, +GFP_KERNEL ); +if (ret) +dev_warn (dev,"IPU6 get sgt table failed\n" ); + +return ret; +} + +const struct dma_map_ops ipu6_dma_ops = { + .alloc = ipu6_dma_alloc, + .free = ipu6_dma_free, + .mmap = ipu6_dma_mmap, + .map_sg = ipu6_dma_map_sg, + .unmap_sg = ipu6_dma_unmap_sg, + .sync_single_for_cpu = ipu6_dma_sync_single_for_cpu, + .sync_single_for_device = ipu6_dma_sync_single_for_cpu, + .sync_sg_for_cpu = ipu6_dma_sync_sg_for_cpu, + .sync_sg_for_device = ipu6_dma_sync_sg_for_cpu, + .get_sgtable = ipu6_dma_get_sgtable, +};
drivers/media/pci/intel/ipu6/ipu6-dma.h new file +20 -0
@@ -0,0 +1,20 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* Copyright (C) 2013 - 2023 Intel Corporation */ + +#ifndef IPU6_DMA_H +#define IPU6_DMA_H + +#include <linux/dma-map-ops.h> +#include <linux/iova.h> + +struct ipu6_mmu_info ; + +struct ipu6_dma_mapping { +struct ipu6_mmu_info *mmu_info; +struct iova_domain iovad; +struct kref ref; +}; + +extern const struct dma_map_ops ipu6_dma_ops; + +#endif /* IPU6_DMA_H */
drivers/media/pci/intel/ipu6/ipu6-mmu.c new file +828 -0
828 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6-mmu.h new file +67 -0
67 lines not on this page · raw message
Signature · 1 line
2.42.0
Re: [PATCH v2 05/15] media: intel/ipu6: add IPU6 DMA mapping API and MMU table
Hi,
Quoted 25 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
he Intel IPU6 has an internal microcontroller (scalar processor, SP) which
is used to execute the firmware. The SP can access IPU internal memory and
map system DRAM to its an internal 32-bit virtual address space.This patch adds a driver for the IPU MMU and a DMA mapping implementation
using the internal MMU. The system IOMMU may be used besides the IPU MMU.Signed-off-by Bingbu Cao <[email protected]>---
drivers/media/pci/intel/ipu6/ipu6-dma.c | 491 ++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-dma.h | 20 + drivers/media/pci/intel/ipu6/ipu6-mmu.c | 828 ++++++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-mmu.h | 67 ++ 4 files changed, 1406 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-dma.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-dma.hcreate mode 100644drivers/media/pci/intel/ipu6/ipu6-mmu.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-mmu.h
...
Quoted 1 line
+++
b/drivers/media/pci/intel/ipu6/ipu6-mmu.c
...
Quoted 55 lines
drivers/media/pci/intel/ipu6/ipu6-mmu.c new file +13
-0+static struct ipu6_mmu_info *ipu6_mmu_alloc (struct ipu6_device *isp) +{ + struct ipu6_mmu_info *mmu_info; + int ret; + + mmu_info = kzalloc(sizeof(*mmu_info), GFP_KERNEL); + if (!mmu_info) + return NULL; + + mmu_info->aperture_start = 0; + mmu_info->aperture_end = DMA_BIT_MASK(isp->secure_mode ? + IPU6_MMU_ADDR_BITS : +
IPU6_MMU_ADDR_BITS_NON_SECURE);
+ mmu_info->pgsize_bitmap =SZ_4K;
+ mmu_info->dev =&isp->pdev->dev;drivers/media/pci/intel/ipu6/ipu6-mmu.c new file +10
-0+ + ret = get_dummy_page(mmu_info); + if (ret) + goto err_free_info; + + ret = alloc_dummy_l2_pt(mmu_info); + if (ret) + goto err_free_dummy_page; + + mmu_info->l2_pts = vzalloc(ISP_L2PT_PTES * sizeof(*mmu_info-Quoted 1 line
l2_pts));+ if (!mmu_info->l2_pts)
+ gotoerr_free_dummy_l2_pt;drivers/media/pci/intel/ipu6/ipu6-mmu.c new file +26
-0+ + /* + * We always map the L1 page table (a single page as well as + * the L2 page tables). + */ + mmu_info->l1_pt = alloc_l1_pt(mmu_info); + if (!mmu_info->l1_pt) + goto err_free_l2_pts; + + spin_lock_init(&mmu_info->lock); + + dev_dbg(mmu_info->dev, "domain initialised\n"); + + return mmu_info; + +err_free_l2_pts: + vfree(mmu_info->l2_pts); +err_free_dummy_l2_pt: + free_dummy_l2_pt(mmu_info); +err_free_dummy_page: + free_dummy_page(mmu_info); +err_free_info: + kfree(mmu_info); + + return NULL; +}
...
Quoted 43 lines
drivers/media/pci/intel/ipu6/ipu6-mmu.c new file +9
-0+static void ipu6_mmu_destroy (struct ipu6_mmu *mmu) +{ + struct ipu6_dma_mapping *dmap = mmu->dmap; + struct ipu6_mmu_info *mmu_info = dmap->mmu_info; + struct iova *iova; + u32 l1_idx; + + if (mmu->iova_trash_page) { + iova = find_iova(&dmap->iovad, PHYS_PFN(mmu-Quoted 1 line
iova_trash_page));+ if (iova) {
+ /* unmap and free the trash buffer iova */ + ipu6_mmu_unmap(mmu_info, PFN_PHYS(iova-Quoted 1 line
pfn_lo),+ PFN_PHYS(iova_size(iova))); + __free_iova(&dmap->iovad, iova); + } else {
+ dev_err(mmu->dev, "trash buffer iova not found.\n");
+ }drivers/media/pci/intel/ipu6/ipu6-mmu.c new file +10
-0+ + mmu->iova_trash_page = 0; + dma_unmap_page(mmu_info->dev, mmu->pci_trash_page, + PAGE_SIZE, DMA_BIDIRECTIONAL); + mmu->pci_trash_page = 0; + __free_page(mmu->trash_page); + } + + for (l1_idx = 0; l1_idx < ISP_L1PT_PTES; l1_idx++) { + if (mmu_info->l1_pt[l1_idx] != mmu_info-Quoted 1 line
dummy_l2_pteval) {+ dma_unmap_single(mmu_info->dev, + TBL_PHYS_ADDR(mmu_info-
Quoted 1 line
l1_pt[l1_idx]),
+
PAGE_SIZE,DMA_BIDIRECTIONAL);
+ free_page((unsigned long)mmu_info-Quoted 1 line
l2_pts[l1_idx]);
+ }
+ }
+
+ free_dummy_page(mmu_info);
+ dma_unmap_single(mmu_info->dev, TBL_PHYS_ADDR(mmu_info-
Quoted 6 lines
Quoted 1 line
l1_pt_dma),+
PAGE_SIZE,DMA_BIDIRECTIONAL); + free_page((unsigned long)mmu_info->dummy_l2_pt); + free_page((unsigned long)mmu_info->l1_pt); + kfree(mmu_info);
+}
mmu_info->l2_pts is leaked here. It is allocated in ipu6_mmu_alloc and freed in the error path of that function along with the other stuff you free here, but not anywhere else.
/Andreas
PATCH v2 6/15 media: intel/ipu6: add syscom interfaces between firmware and driver
In reply to [email protected]From: Bingbu Cao <[email protected]>
Syscom is an inter-process(or) communication mechanism between an IPU and host. Syscom uses message queues for message exchange between IPU and host. Each message queue has its consumer and producer, host queue messages to firmware as the producer and then firmware to dequeue the messages as consumer and vice versa. IPU and host use shared registers or memory to reside the read and write indices which are updated by consumer and producer.
--- drivers/media/pci/intel/ipu6/ipu6-fw-com.c | 411 +++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-fw-com.h | 47 +++ 2 files changed, 458 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.h
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +411 -0
@@ -0,0 +1,411 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include "ipu6.h" +#include "ipu6-bus.h" +#include "ipu6-fw-com.h" + +/* +* FWCOM layer is a shared resource between FW and driver. It consist +* of token queues to both send and receive directions. Queue is simply +* an array of structures with read and write indexes to the queue. +* There are 1...n queues to both directions. Queues locates in +* system RAM and are mapped to ISP MMU so that both CPU and ISP can +* see the same buffer. Indexes are located in ISP DMEM so that FW code +* can poll those with very low latency and cost. CPU access to indexes is +* more costly but that happens only at message sending time and +* interrupt triggered message handling. CPU doesn't need to poll indexes. +* wr_reg / rd_reg are offsets to those dmem location. They are not +* the indexes itself. +*/ + +/* Shared structure between driver and FW - do not modify */ +struct ipu6_fw_sys_queue { +u64 host_address; +u32 vied_address; +u32 size; +u32 token_size; +u32 wr_reg;/* reg number in subsystem's regmem */ +u32 rd_reg; +u32 _align; +} __packed; + +struct ipu6_fw_sys_queue_res { +u64 host_address; +u32 vied_address; +u32 reg; +} __packed; + +enum syscom_state { +/* Program load or explicit host setting should init to this */ +SYSCOM_STATE_UNINIT =0x57A7E000 , +/* SP Syscom sets this when it is ready for use */ +SYSCOM_STATE_READY =0x57A7E001 , +/* SP Syscom sets this when no more syscom accesses will happen */ +SYSCOM_STATE_INACTIVE =0x57A7E002 +}; + +enum syscom_cmd { +/* Program load or explicit host setting should init to this */ +SYSCOM_COMMAND_UNINIT =0x57A7F000 , +/* Host Syscom requests syscom to become inactive */ +SYSCOM_COMMAND_INACTIVE =0x57A7F001 +}; + +/* firmware config: data that sent from the host to SP via DDR */ +/* Cell copies data into a context */ + +struct ipu6_fw_syscom_config { +u32 firmware_address; + +u32 num_input_queues; +u32 num_output_queues; + +/* ISP pointers to an array of ipu6_fw_sys_queue structures */ +u32 input_queue; +u32 output_queue; + +/* ISYS / PSYS private data */ +u32 specific_addr; +u32 specific_size; +}; + +struct ipu6_fw_com_context { +struct ipu6_bus_device *adev; +void __iomem *dmem_addr; +int (*cell_ready)(struct ipu6_bus_device *adev); +void (*cell_start)(struct ipu6_bus_device *adev); + +void *dma_buffer; +dma_addr_t dma_addr; +unsigned int dma_size; +unsigned long attrs; + +struct ipu6_fw_sys_queue *input_queue;/* array of host to SP queues */ +struct ipu6_fw_sys_queue *output_queue;/* array of SP to host */ + +u32 config_vied_addr; + +unsigned int buttress_boot_offset; +void __iomem *base_addr; +}; + +#define FW_COM_WR_REG 0 +#define FW_COM_RD_REG 4 + +#define REGMEM_OFFSET 0 +#define TUNIT_MAGIC_PATTERN 0x5a5a5a5a + +enum regmem_id { +/* pass pkg_dir address to SPC in non-secure mode */ +PKG_DIR_ADDR_REG =0 , +/* Tunit CFG blob for secure - provided by host.*/ +TUNIT_CFG_DWR_REG =1 , +/* syscom commands - modified by the host */ +SYSCOM_COMMAND_REG =2 , +/* Store interrupt status - updated by SP */ +SYSCOM_IRQ_REG =3 , +/* first syscom queue pointer register */ +SYSCOM_QPR_BASE_REG =4 +}; + +enum message_direction { +DIR_RECV =0 , +DIR_SEND +}; + +#define BUTTRESS_FW_BOOT_PARAMS_0 0x4000 +#define BUTTRESS_FW_BOOT_PARAM_REG (base, offset, id) \ + ((base) + BUTTRESS_FW_BOOT_PARAMS_0 + ((offset) + (id)) * 4) + +enum buttress_syscom_id { +/* pass syscom configuration to SPC */ +SYSCOM_CONFIG_ID =0 , +/* syscom state - modified by SP */ +SYSCOM_STATE_ID =1 , +/* syscom vtl0 addr mask */ +SYSCOM_VTL0_ADDR_MASK_ID =2 , +SYSCOM_ID_MAX +}; + +static void ipu6_sys_queue_init (struct ipu6_fw_sys_queue *q,unsigned int size, +unsigned int token_size, +struct ipu6_fw_sys_queue_res *res) +{ +unsigned int buf_size = (size +1 ) * token_size; + + q->size = size +1 ; + q->token_size = token_size; + +/* acquire the shared buffer space */ + q->host_address = res->host_address; + res->host_address += buf_size; + q->vied_address = res->vied_address; + res->vied_address += buf_size; + +/* acquire the shared read and writer pointers */ + q->wr_reg = res->reg; + res->reg++; + q->rd_reg = res->reg; + res->reg++; +} + +void *ipu6_fw_com_prepare (struct ipu6_fw_com_cfg *cfg, +struct ipu6_bus_device *adev,void __iomem *base) +{ +size_t conf_size, inq_size, outq_size, specific_size; +struct ipu6_fw_syscom_config *config_host_addr; +unsigned int sizeinput =0 , sizeoutput =0 ; +struct ipu6_fw_sys_queue_res res; +struct ipu6_fw_com_context *ctx; +struct device *dev = &adev->auxdev.dev; +size_t sizeall, offset; +unsigned long attrs =0 ; +void *specific_host_addr; +unsigned int i; + +if (!cfg || !cfg->cell_start || !cfg->cell_ready) +return NULL ; + + ctx =kzalloc (sizeof (*ctx),GFP_KERNEL ); +if (!ctx) +return NULL ; + ctx->dmem_addr = base + cfg->dmem_addr +REGMEM_OFFSET ; + ctx->adev = adev; + ctx->cell_start = cfg->cell_start; + ctx->cell_ready = cfg->cell_ready; + ctx->buttress_boot_offset = cfg->buttress_boot_offset; + ctx->base_addr = base; + +/* +* Allocate DMA mapped memory. Allocate one big chunk. +*/ +/* Base cfg for FW */ + conf_size =roundup (sizeof (struct ipu6_fw_syscom_config ),8 ); +/* Descriptions of the queues */ + inq_size =size_mul (cfg->num_input_queues, +sizeof (struct ipu6_fw_sys_queue )); + outq_size =size_mul (cfg->num_output_queues, +sizeof (struct ipu6_fw_sys_queue )); +/* FW specific information structure */ + specific_size =roundup (cfg->specific_size,8 ); + + sizeall = conf_size + inq_size + outq_size + specific_size; + +for (i =0 ; i < cfg->num_input_queues; i++) + sizeinput +=size_mul (cfg->input[i].queue_size +1 , + cfg->input[i].token_size); + +for (i =0 ; i < cfg->num_output_queues; i++) + sizeoutput +=size_mul (cfg->output[i].queue_size +1 , + cfg->output[i].token_size); + + sizeall += sizeinput + sizeoutput; + + ctx->dma_buffer =dma_alloc_attrs (dev, sizeall, &ctx->dma_addr, +GFP_KERNEL , attrs); + ctx->attrs = attrs; +if (!ctx->dma_buffer) { +dev_err (dev,"failed to allocate dma memory\n" ); +kfree (ctx); +return NULL ; + } + + ctx->dma_size = sizeall; + + config_host_addr = ctx->dma_buffer; + ctx->config_vied_addr = ctx->dma_addr; + + offset = conf_size; + ctx->input_queue = ctx->dma_buffer + offset; + config_host_addr->input_queue = ctx->dma_addr + offset; + config_host_addr->num_input_queues = cfg->num_input_queues; + + offset += inq_size; + ctx->output_queue = ctx->dma_buffer + offset; + config_host_addr->output_queue = ctx->dma_addr + offset; + config_host_addr->num_output_queues = cfg->num_output_queues; + +/* copy firmware specific data */ + offset += outq_size; + specific_host_addr = ctx->dma_buffer + offset; + config_host_addr->specific_addr = ctx->dma_addr + offset; + config_host_addr->specific_size = cfg->specific_size; +if (cfg->specific_addr && cfg->specific_size) +memcpy (specific_host_addr, cfg->specific_addr, + cfg->specific_size); + +/* initialize input queues */ + offset += specific_size; + res.reg =SYSCOM_QPR_BASE_REG ; + res.host_address = (u64)(ctx->dma_buffer + offset); + res.vied_address = ctx->dma_addr + offset; +for (i =0 ; i < cfg->num_input_queues; i++) +ipu6_sys_queue_init (ctx->input_queue + i, + cfg->input[i].queue_size, + cfg->input[i].token_size, &res); + +/* initialize output queues */ + offset += sizeinput; + res.host_address = (u64)(ctx->dma_buffer + offset); + res.vied_address = ctx->dma_addr + offset; +for (i =0 ; i < cfg->num_output_queues; i++) { +ipu6_sys_queue_init (ctx->output_queue + i, + cfg->output[i].queue_size, + cfg->output[i].token_size, &res); + } + +return ctx; +} +EXPORT_SYMBOL_NS_GPL (ipu6_fw_com_prepare,INTEL_IPU6 ); + +int ipu6_fw_com_open (struct ipu6_fw_com_context *ctx) +{ +/* write magic pattern to disable the tunit trace */ +writel (TUNIT_MAGIC_PATTERN , ctx->dmem_addr +TUNIT_CFG_DWR_REG *4 ); +/* Check if SP is in valid state */ +if (!ctx->cell_ready (ctx->adev)) +return -EIO ; + +/* store syscom uninitialized command */ +writel (SYSCOM_COMMAND_UNINIT , ctx->dmem_addr +SYSCOM_COMMAND_REG *4 ); + +/* store syscom uninitialized state */ +writel (SYSCOM_STATE_UNINIT , +BUTTRESS_FW_BOOT_PARAM_REG (ctx->base_addr, + ctx->buttress_boot_offset, +SYSCOM_STATE_ID )); + +/* store firmware configuration address */ +writel (ctx->config_vied_addr, +BUTTRESS_FW_BOOT_PARAM_REG (ctx->base_addr, + ctx->buttress_boot_offset, +SYSCOM_CONFIG_ID )); + ctx->cell_start (ctx->adev); + +return 0 ; +} +EXPORT_SYMBOL_NS_GPL (ipu6_fw_com_open,INTEL_IPU6 ); + +int ipu6_fw_com_close (struct ipu6_fw_com_context *ctx) +{ +int state; + + state =readl (BUTTRESS_FW_BOOT_PARAM_REG (ctx->base_addr, + ctx->buttress_boot_offset, +SYSCOM_STATE_ID )); +if (state !=SYSCOM_STATE_READY ) +return -EBUSY ; + +/* set close request flag */ +writel (SYSCOM_COMMAND_INACTIVE , ctx->dmem_addr + +SYSCOM_COMMAND_REG *4 ); + +return 0 ; +} +EXPORT_SYMBOL_NS_GPL (ipu6_fw_com_close,INTEL_IPU6 ); + +int ipu6_fw_com_release (struct ipu6_fw_com_context *ctx,unsigned int force) +{ +/* check if release is forced, an verify cell state if it is not */ +if (!force && !ctx->cell_ready (ctx->adev)) +return -EBUSY ; + +dma_free_attrs (&ctx->adev->auxdev.dev, ctx->dma_size, + ctx->dma_buffer, ctx->dma_addr, ctx->attrs); +kfree (ctx); +return 0 ; +} +EXPORT_SYMBOL_NS_GPL (ipu6_fw_com_release,INTEL_IPU6 ); + +bool ipu6_fw_com_ready (struct ipu6_fw_com_context *ctx) +{ +int state; + + state =readl (BUTTRESS_FW_BOOT_PARAM_REG (ctx->base_addr, + ctx->buttress_boot_offset, +SYSCOM_STATE_ID )); + +return state ==SYSCOM_STATE_READY ; +} +EXPORT_SYMBOL_NS_GPL (ipu6_fw_com_ready,INTEL_IPU6 ); + +void *ipu6_send_get_token (struct ipu6_fw_com_context *ctx,int q_nbr) +{ +struct ipu6_fw_sys_queue *q = &ctx->input_queue[q_nbr]; +void __iomem *q_dmem = ctx->dmem_addr + q->wr_reg *4 ; +unsigned int wr, rd; +unsigned int packets; +unsigned int index; + + wr =readl (q_dmem +FW_COM_WR_REG ); + rd =readl (q_dmem +FW_COM_RD_REG ); + +if (WARN_ON_ONCE (wr >= q->size || rd >= q->size)) +return NULL ; + +if (wr < rd) + packets = rd - wr -1 ; +else + packets = q->size - (wr - rd +1 ); + +if (!packets) +return NULL ; + + index =readl (q_dmem +FW_COM_WR_REG ); + +return (void *)(q->host_address + index * q->token_size); +} +EXPORT_SYMBOL_NS_GPL (ipu6_send_get_token,INTEL_IPU6 ); + +void ipu6_send_put_token (struct ipu6_fw_com_context *ctx,int q_nbr) +{ +struct ipu6_fw_sys_queue *q = &ctx->input_queue[q_nbr]; +void __iomem *q_dmem = ctx->dmem_addr + q->wr_reg *4 ; +unsigned int wr =readl (q_dmem +FW_COM_WR_REG ) +1 ; + +if (wr >= q->size) + wr =0 ; + +writel (wr, q_dmem +FW_COM_WR_REG ); +} +EXPORT_SYMBOL_NS_GPL (ipu6_send_put_token,INTEL_IPU6 ); + +void *ipu6_recv_get_token (struct ipu6_fw_com_context *ctx,int q_nbr) +{ +struct ipu6_fw_sys_queue *q = &ctx->output_queue[q_nbr]; +void __iomem *q_dmem = ctx->dmem_addr + q->wr_reg *4 ; +unsigned int wr, rd; +unsigned int packets; + + wr =readl (q_dmem +FW_COM_WR_REG ); + rd =readl (q_dmem +FW_COM_RD_REG ); + +if (WARN_ON_ONCE (wr >= q->size || rd >= q->size)) +return NULL ; + +if (wr < rd) + wr += q->size; + + packets = wr - rd; +if (!packets) +return NULL ; + +return (void *)(q->host_address + rd * q->token_size); +} +EXPORT_SYMBOL_NS_GPL (ipu6_recv_get_token,INTEL_IPU6 ); + +void ipu6_recv_put_token (struct ipu6_fw_com_context *ctx,int q_nbr) +{ +struct ipu6_fw_sys_queue *q = &ctx->output_queue[q_nbr]; +void __iomem *q_dmem = ctx->dmem_addr + q->wr_reg *4 ; +unsigned int rd =readl (q_dmem +FW_COM_RD_REG ) +1 ; + +if (rd >= q->size) + rd =0 ; + +writel (rd, q_dmem +FW_COM_RD_REG ); +} +EXPORT_SYMBOL_NS_GPL (ipu6_recv_put_token,INTEL_IPU6 );
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +47 -0
47 lines not on this page · raw message
Signature · 1 line
2.42.0
Re: [PATCH v2 06/15] media: intel/ipu6: add syscom interfaces between firmware and driver
Hi,
Quoted 20 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Syscom is an inter-process(or) communication mechanism between an IPU and host. Syscom uses message queues for message exchange between IPU and host. Each message queue has its consumer and producer, host queue
messages to firmware as the producer and then firmware to dequeue the messages as consumer and vice versa. IPU and host use shared registers
or memory to reside the read and write indices which are updated by consumer and producer.Signed-off-by Bingbu Cao <[email protected]>---
drivers/media/pci/intel/ipu6/ipu6-fw-com.c | 411
+++++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6-fw-com.h | 47 +++2 files changed, 458 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-fw-com.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-fw-com.h
...
Quoted 5 lines
+struct
ipu6_fw_com_context{
+ structipu6_bus_device*adev;
+ void__iomem*dmem_addr;
+ int (*cell_ready)(structipu6_bus_device*adev); + void (*cell_start)(structipu6_bus_device*adev);
Why are cell_ready and cell_start function pointers? They seem to always be set to query_sp and start_sp.
Quoted 30 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +7
-0+ + void *dma_buffer; + dma_addr_t dma_addr; + unsigned int dma_size; + unsigned long attrs; + + struct ipu6_fw_sys_queue *input_queue; /* array of host toSP queues */
+ structipu6_fw_sys_queue*output_queue; /* array of SP to host */drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +20
-0+ + u32 config_vied_addr; + + unsigned int buttress_boot_offset; + void __iomem *base_addr; +}; + +#define FW_COM_WR_REG 0 +#define FW_COM_RD_REG 4 + +#define REGMEM_OFFSET 0 +#define TUNIT_MAGIC_PATTERN 0x5a5a5a5a + +enum regmem_id { + /* pass pkg_dir address to SPC in non-secure mode */ + PKG_DIR_ADDR_REG = 0, + /* Tunit CFG blob for secure - provided by host.*/ + TUNIT_CFG_DWR_REG = 1, + /* syscom commands - modified by the host */ + SYSCOM_COMMAND_REG = 2,
drivers/media/pci/intel/ipu6/ipu6-fw-com.c line 117
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +10
-0+ /* Store interrupt status - updated by SP */ + SYSCOM_IRQ_REG = 3, + /* first syscom queue pointer register */ + SYSCOM_QPR_BASE_REG = 4 +}; + +enum message_direction { + DIR_RECV = 0, + DIR_SEND +};
Not used?
...
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +5
-0+struct ipu6_fw_com_cfg { + unsigned int num_input_queues; + unsigned int num_output_queues; + struct ipu6_fw_syscom_queue_config *input; + struct ipu6_fw_syscom_queue_config *output;
drivers/media/pci/intel/ipu6/ipu6-fw-com.h line 31
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +10
-0+ + unsigned int dmem_addr; + + /* firmware-specific configuration data */ + void *specific_addr; + unsigned int specific_size; + int (*cell_ready)(struct ipu6_bus_device *adev); + void (*cell_start)(struct ipu6_bus_device *adev); + + unsigned int buttress_boot_offset;
This seems to always be 0 (set by ipu6-fw-isys.c), seems to be trivial to remove.
/Andreas
Re: [PATCH v2 06/15] media: intel/ipu6: add syscom interfaces between firmware and driver
Andreas,
Quoted 67 lines · Andreas Helbech Kleist
Hi,
Quoted 20 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Syscom is an inter-process(or) communication mechanism between an IPU and host. Syscom uses message queues for message exchange between IPU and host. Each message queue has its consumer and producer, host queue
messages to firmware as the producer and then firmware to dequeue the messages as consumer and vice versa. IPU and host use shared registers
or memory to reside the read and write indices which are updated by consumer and producer.Signed-off-by Bingbu Cao <[email protected]>--- drivers/media/pci/intel/ipu6/ipu6-fw-com.c | 411 +++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-fw-com.h | 47 +++ 2 files changed, 458 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.h...
drivers/media/pci/intel/ipu6/ipu6-fw-com.c line 79
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +5
-0+struct ipu6_fw_com_context { +struct ipu6_bus_device *adev; +void __iomem *dmem_addr; +int (*cell_ready)(struct ipu6_bus_device *adev); +void (*cell_start)(struct ipu6_bus_device *adev);Why are
cell_readyandcell_startfunction pointers? They seem to always be set toquery_spandstart_sp.Quoted 34 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +7
-0+ +void *dma_buffer; +dma_addr_t dma_addr; +unsigned int dma_size; +unsigned long attrs; + +struct ipu6_fw_sys_queue *input_queue;/* array of host to SP queues */
+ structipu6_fw_sys_queue*output_queue; /* array of SP to
host */drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +24
-0+ +u32 config_vied_addr; + +unsigned int buttress_boot_offset; +void __iomem *base_addr; +}; + +#define FW_COM_WR_REG 0 +#define FW_COM_RD_REG 4 + +#define REGMEM_OFFSET 0 +#define TUNIT_MAGIC_PATTERN 0x5a5a5a5a + +enum regmem_id { +/* pass pkg_dir address to SPC in non-secure mode */ +PKG_DIR_ADDR_REG =0 , +/* Tunit CFG blob for secure - provided by host.*/ +TUNIT_CFG_DWR_REG =1 , +/* syscom commands - modified by the host */ +SYSCOM_COMMAND_REG =2 , +/* Store interrupt status - updated by SP */ +SYSCOM_IRQ_REG =3 , +/* first syscom queue pointer register */ +SYSCOM_QPR_BASE_REG =4
Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c line 117
Quoted 6 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +6
-0+}; + +enum message_direction { +DIR_RECV =0 , +DIR_SEND +};Not used?
Ack, will remove.
Quoted 13 lines
...
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +10
-0+struct ipu6_fw_com_cfg { +unsigned int num_input_queues; +unsigned int num_output_queues; +struct ipu6_fw_syscom_queue_config *input; +struct ipu6_fw_syscom_queue_config *output; + +unsigned int dmem_addr; + +/* firmware-specific configuration data */ +void *specific_addr;
Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h line 31
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +5
-0+unsigned int specific_size; +int (*cell_ready)(struct ipu6_bus_device *adev); +void (*cell_start)(struct ipu6_bus_device *adev); + +unsigned int buttress_boot_offset;This seems to always be 0 (set by
ipu6-fw-isys.c), seems to be trivial to remove.
All these fields are all used to extend for psys driver.
Quoted 3 lines
/Andreas
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 06/15] media: intel/ipu6: add syscom interfaces between firmware and driver
Hi,
Quoted 103 lines · Bingbu Cao
Andreas,
Quoted 67 lines · Andreas Helbech Kleist
Hi,
Quoted 20 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Syscom is an inter-process(or) communication mechanism between an IPU and host. Syscom uses message queues for message exchange between IPU and host. Each message queue has its consumer and producer, host queue
messages to firmware as the producer and then firmware to dequeue the messages as consumer and vice versa. IPU and host use shared registers
or memory to reside the read and write indices which are updated by consumer and producer.Signed-off-by Bingbu Cao <[email protected]>--- drivers/media/pci/intel/ipu6/ipu6-fw-com.c | 411 +++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-fw-com.h | 47 +++ 2 files changed, 458 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.h...
drivers/media/pci/intel/ipu6/ipu6-fw-com.c line 79
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +5
-0+struct ipu6_fw_com_context { +struct ipu6_bus_device *adev; +void __iomem *dmem_addr; +int (*cell_ready)(struct ipu6_bus_device *adev); +void (*cell_start)(struct ipu6_bus_device *adev);Why are
cell_readyandcell_startfunction pointers? They seem to always be set toquery_spandstart_sp.Quoted 34 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +7
-0+ +void *dma_buffer; +dma_addr_t dma_addr; +unsigned int dma_size; +unsigned long attrs; + +struct ipu6_fw_sys_queue *input_queue;/* array of host to SP queues */
+ structipu6_fw_sys_queue*output_queue; /* array of SP to
host */drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +24
-0+ +u32 config_vied_addr; + +unsigned int buttress_boot_offset; +void __iomem *base_addr; +}; + +#define FW_COM_WR_REG 0 +#define FW_COM_RD_REG 4 + +#define REGMEM_OFFSET 0 +#define TUNIT_MAGIC_PATTERN 0x5a5a5a5a + +enum regmem_id { +/* pass pkg_dir address to SPC in non-secure mode */ +PKG_DIR_ADDR_REG =0 , +/* Tunit CFG blob for secure - provided by host.*/ +TUNIT_CFG_DWR_REG =1 , +/* syscom commands - modified by the host */ +SYSCOM_COMMAND_REG =2 , +/* Store interrupt status - updated by SP */ +SYSCOM_IRQ_REG =3 , +/* first syscom queue pointer register */ +SYSCOM_QPR_BASE_REG =4 Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c line 117
Quoted 6 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +6
-0+}; + +enum message_direction { +DIR_RECV =0 , +DIR_SEND +};Not used?
Ack, will remove.
Quoted 13 lines
...
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +10
-0+struct ipu6_fw_com_cfg { +unsigned int num_input_queues; +unsigned int num_output_queues; +struct ipu6_fw_syscom_queue_config *input; +struct ipu6_fw_syscom_queue_config *output; + +unsigned int dmem_addr; + +/* firmware-specific configuration data */ +void *specific_addr;Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h line 31
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +5
-0+unsigned int specific_size; +int (*cell_ready)(struct ipu6_bus_device *adev); +void (*cell_start)(struct ipu6_bus_device *adev); + +unsigned int buttress_boot_offset;This seems to always be 0 (set by
ipu6-fw-isys.c), seems to be trivial to remove.
Quoted 1 line
All these fields are all used to extend for psys driver.
Then I suppose they can be added if a psys driver is submitted upstream at some point.
Is is it the intention to submit a psys driver as well?
/Andreas
Re: [PATCH v2 06/15] media: intel/ipu6: add syscom interfaces between firmware and driver
Andreas,
Quoted 106 lines · Andreas Helbech Kleist
Hi,
Quoted 103 lines · Bingbu Cao
Andreas,
Quoted 67 lines · Andreas Helbech Kleist
Hi,
Quoted 20 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Syscom is an inter-process(or) communication mechanism between an IPU and host. Syscom uses message queues for message exchange between IPU and host. Each message queue has its consumer and producer, host queue
messages to firmware as the producer and then firmware to dequeue the messages as consumer and vice versa. IPU and host use shared registers
or memory to reside the read and write indices which are updated by consumer and producer.Signed-off-by Bingbu Cao <[email protected]>--- drivers/media/pci/intel/ipu6/ipu6-fw-com.c | 411 +++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-fw-com.h | 47 +++ 2 files changed, 458 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-com.h...
drivers/media/pci/intel/ipu6/ipu6-fw-com.c line 79
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +5
-0+struct ipu6_fw_com_context { +struct ipu6_bus_device *adev; +void __iomem *dmem_addr; +int (*cell_ready)(struct ipu6_bus_device *adev); +void (*cell_start)(struct ipu6_bus_device *adev);Why are
cell_readyandcell_startfunction pointers? They seem to always be set toquery_spandstart_sp.Quoted 34 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +7
-0+ +void *dma_buffer; +dma_addr_t dma_addr; +unsigned int dma_size; +unsigned long attrs; + +struct ipu6_fw_sys_queue *input_queue;/* array of host to SP queues */
+ structipu6_fw_sys_queue*output_queue; /* array of SP to
host */drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +24
-0+ +u32 config_vied_addr; + +unsigned int buttress_boot_offset; +void __iomem *base_addr; +}; + +#define FW_COM_WR_REG 0 +#define FW_COM_RD_REG 4 + +#define REGMEM_OFFSET 0 +#define TUNIT_MAGIC_PATTERN 0x5a5a5a5a + +enum regmem_id { +/* pass pkg_dir address to SPC in non-secure mode */ +PKG_DIR_ADDR_REG =0 , +/* Tunit CFG blob for secure - provided by host.*/ +TUNIT_CFG_DWR_REG =1 , +/* syscom commands - modified by the host */ +SYSCOM_COMMAND_REG =2 , +/* Store interrupt status - updated by SP */ +SYSCOM_IRQ_REG =3 , +/* first syscom queue pointer register */ +SYSCOM_QPR_BASE_REG =4 Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c line 117
Quoted 6 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.c new file +6
-0+}; + +enum message_direction { +DIR_RECV =0 , +DIR_SEND +};Not used?
Ack, will remove.
Quoted 21 lines
...
Quoted 5 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +5
-0+struct ipu6_fw_com_cfg { +unsigned int num_input_queues; +unsigned int num_output_queues; +struct ipu6_fw_syscom_queue_config *input; +struct ipu6_fw_syscom_queue_config *output;drivers/media/pci/intel/ipu6/ipu6-fw-com.h line 31
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-fw-com.h new file +10
-0+ +unsigned int dmem_addr; + +/* firmware-specific configuration data */ +void *specific_addr; +unsigned int specific_size; +int (*cell_ready)(struct ipu6_bus_device *adev); +void (*cell_start)(struct ipu6_bus_device *adev); + +unsigned int buttress_boot_offset;This seems to always be 0 (set by
ipu6-fw-isys.c), seems to be trivial to remove.
Quoted 4 lines
Quoted 1 line
All these fields are all used to extend for psys driver.
Then I suppose they can be added if a psys driver is submitted upstream at some point.
For IPU subsystem, the fwcom interface (IPU6 driver) is a common interface for both isys and psys. It makes sense that making the common interface scalable even it is only set by isys driver now.
Quoted 2 lines
Is is it the intention to submit a psys driver as well?
Yes, we are going to submit a IPU6 psys driver, but it will take more time to get ready.
Quoted 4 lines
/Andreas
Signature · 2 lines
Best regards, Bingbu Cao
PATCH v2 7/15 media: intel/ipu6: input system ABI between firmware and driver
In reply to [email protected]From: Bingbu Cao <[email protected]>
Implement the input system firmware ABIs between the firmware and driver - include stream configuration, control command, capture request and response, etc.
--- drivers/media/pci/intel/ipu6/ipu6-fw-isys.c | 484 +++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-fw-isys.h | 570 ++++++++++++++++++++ 2 files changed, 1054 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-isys.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-fw-isys.h
drivers/media/pci/intel/ipu6/ipu6-fw-isys.c new file +484 -0
@@ -0,0 +1,484 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <linux/cacheflush.h> +#include <linux/delay.h> + +#include <media/mipi-csi2.h> + +#include "ipu6-fw-com.h" +#include "ipu6-isys.h" +#include "ipu6-platform-isys-csi2-reg.h" +#include "ipu6-platform-regs.h" + +static const char send_msg_types[N_IPU6_FW_ISYS_SEND_TYPE ][32 ] = { +"STREAM_OPEN" , +"STREAM_START" , +"STREAM_START_AND_CAPTURE" , +"STREAM_CAPTURE" , +"STREAM_STOP" , +"STREAM_FLUSH" , +"STREAM_CLOSE" +}; + +static int handle_proxy_response (struct ipu6_isys *isys,unsigned int req_id) +{ +struct device *dev = &isys->adev->auxdev.dev; +struct ipu6_fw_isys_proxy_resp_info_abi *resp; +int ret; + + resp =ipu6_recv_get_token (isys->fwcom,IPU6_BASE_PROXY_RECV_QUEUES ); +if (!resp) +return 1 ; + +dev_dbg (dev,"Proxy response: id %u, error %u, details %u\n" , + resp->request_id, resp->error_info.error, + resp->error_info.error_details); + + ret = req_id == resp->request_id ?0 : -EIO ; + +ipu6_recv_put_token (isys->fwcom,IPU6_BASE_PROXY_RECV_QUEUES ); + +return ret; +} + +int ipu6_fw_isys_send_proxy_token (struct ipu6_isys *isys, +unsigned int req_id, +unsigned int index, +unsigned int offset,u32 value) +{ +struct ipu6_fw_com_context *ctx = isys->fwcom; +struct device *dev = &isys->adev->auxdev.dev; +struct ipu6_fw_proxy_send_queue_token *token; +unsigned int timeout =1000 ; +int ret; + +dev_dbg (dev, +"proxy send: req_id 0x%x, index %d, offset 0x%x, value 0x%x\n" , + req_id, index, offset, value); + + token =ipu6_send_get_token (ctx,IPU6_BASE_PROXY_SEND_QUEUES ); +if (!token) +return -EBUSY ; + + token->request_id = req_id; + token->region_index = index; + token->offset = offset; + token->value = value; +ipu6_send_put_token (ctx,IPU6_BASE_PROXY_SEND_QUEUES ); + +do { +usleep_range (100 ,110 ); + ret =handle_proxy_response (isys, req_id); +if (!ret) +break ; +if (ret == -EIO ) { +dev_err (dev,"Proxy respond with unexpected id\n" ); +break ; + } + timeout--; + }while (ret && timeout); + +if (!timeout) +dev_err (dev,"Proxy response timed out\n" ); + +return ret; +} + +int ipu6_fw_isys_complex_cmd (struct ipu6_isys *isys, +const unsigned int stream_handle, +void *cpu_mapped_buf, +dma_addr_t dma_mapped_buf, +size_t size,u16 send_type) +{ +struct ipu6_fw_com_context *ctx = isys->fwcom; +struct device *dev = &isys->adev->auxdev.dev; +struct ipu6_fw_send_queue_token *token; + +if (send_type >=N_IPU6_FW_ISYS_SEND_TYPE ) +return -EINVAL ; + +dev_dbg (dev,"send_token: %s\n" , send_msg_types[send_type]); + +/* +* Time to flush cache in case we have some payload. Not all messages +* have that +*/ +if (cpu_mapped_buf) +clflush_cache_range (cpu_mapped_buf, size); + + token =ipu6_send_get_token (ctx, + stream_handle +IPU6_BASE_MSG_SEND_QUEUES ); +if (!token) +return -EBUSY ; + + token->payload = dma_mapped_buf; + token->buf_handle = (unsigned long )cpu_mapped_buf; + token->send_type = send_type; + +ipu6_send_put_token (ctx, stream_handle +IPU6_BASE_MSG_SEND_QUEUES ); + +return 0 ; +} + +int ipu6_fw_isys_simple_cmd (struct ipu6_isys *isys, +const unsigned int stream_handle,u16 send_type) +{ +return ipu6_fw_isys_complex_cmd (isys, stream_handle,NULL ,0 ,0 , + send_type); +} + +int ipu6_fw_isys_close (struct ipu6_isys *isys) +{ +struct device *dev = &isys->adev->auxdev.dev; +int retry =IPU6_ISYS_CLOSE_RETRY ; +unsigned long flags; +void *fwcom; +int ret; + +/* +* Stop the isys fw. Actual close takes +* some time as the FW must stop its actions including code fetch +* to SP icache. +* spinlock to wait the interrupt handler to be finished +*/ +spin_lock_irqsave (&isys->power_lock, flags); + ret =ipu6_fw_com_close (isys->fwcom); + fwcom = isys->fwcom; + isys->fwcom =NULL ; +spin_unlock_irqrestore (&isys->power_lock, flags); +if (ret) +dev_err (dev,"Device close failure: %d\n" , ret); + +/* release probably fails if the close failed. Let's try still */ +do { +usleep_range (400 ,500 ); + ret =ipu6_fw_com_release (fwcom,0 ); + retry--; + }while (ret && retry); + +if (ret) { +dev_err (dev,"Device release time out %d\n" , ret); +spin_lock_irqsave (&isys->power_lock, flags); + isys->fwcom = fwcom; +spin_unlock_irqrestore (&isys->power_lock, flags); + } + +return ret; +} + +void ipu6_fw_isys_cleanup (struct ipu6_isys *isys) +{ +int ret; + + ret =ipu6_fw_com_release (isys->fwcom,1 ); +if (ret <0 ) +dev_warn (&isys->adev->auxdev.dev, +"Device busy, fw_com release failed." ); + isys->fwcom =NULL ; +} + +static void start_sp (struct ipu6_bus_device *adev) +{ +struct ipu6_isys *isys =ipu6_bus_get_drvdata (adev); +void __iomem *spc_regs_base = isys->pdata->base + + isys->pdata->ipdata->hw_variant.spc_offset; +u32 val =IPU6_ISYS_SPC_STATUS_START | +IPU6_ISYS_SPC_STATUS_RUN | +IPU6_ISYS_SPC_STATUS_CTRL_ICACHE_INVALIDATE ; + + val |= isys->icache_prefetch ?IPU6_ISYS_SPC_STATUS_ICACHE_PREFETCH :0 ; + +writel (val, spc_regs_base +IPU6_ISYS_REG_SPC_STATUS_CTRL ); +} + +static int query_sp (struct ipu6_bus_device *adev) +{ +struct ipu6_isys *isys =ipu6_bus_get_drvdata (adev); +void __iomem *spc_regs_base = isys->pdata->base + + isys->pdata->ipdata->hw_variant.spc_offset; +u32 val; + + val =readl (spc_regs_base +IPU6_ISYS_REG_SPC_STATUS_CTRL ); +/* return true when READY == 1, START == 0 */ + val &=IPU6_ISYS_SPC_STATUS_READY |IPU6_ISYS_SPC_STATUS_START ; + +return val ==IPU6_ISYS_SPC_STATUS_READY ; +} + +static int ipu6_isys_fwcom_cfg_init (struct ipu6_isys *isys, +struct ipu6_fw_com_cfg *fwcom, +unsigned int num_streams) +{ +unsigned int max_send_queues, max_sram_blocks, max_devq_size; +struct ipu6_fw_syscom_queue_config *input_queue_cfg; +struct ipu6_fw_syscom_queue_config *output_queue_cfg; +struct device *dev = &isys->adev->auxdev.dev; +int type_proxy =IPU6_FW_ISYS_QUEUE_TYPE_PROXY ; +int type_dev =IPU6_FW_ISYS_QUEUE_TYPE_DEV ; +int type_msg =IPU6_FW_ISYS_QUEUE_TYPE_MSG ; +int base_dev_send =IPU6_BASE_DEV_SEND_QUEUES ; +int base_msg_send =IPU6_BASE_MSG_SEND_QUEUES ; +int base_msg_recv =IPU6_BASE_MSG_RECV_QUEUES ; +struct ipu6_fw_isys_fw_config *isys_fw_cfg; +u32 num_in_message_queues; +unsigned int max_streams; +unsigned int size; +unsigned int i; + + max_streams = isys->pdata->ipdata->max_streams; + max_send_queues = isys->pdata->ipdata->max_send_queues; + max_sram_blocks = isys->pdata->ipdata->max_sram_blocks; + max_devq_size = isys->pdata->ipdata->max_devq_size; + num_in_message_queues =clamp (num_streams,1U , max_streams); + isys_fw_cfg =devm_kzalloc (dev,sizeof (*isys_fw_cfg),GFP_KERNEL ); +if (!isys_fw_cfg) +return -ENOMEM ; + + isys_fw_cfg->num_send_queues[type_proxy] =IPU6_N_MAX_PROXY_SEND_QUEUES ; + isys_fw_cfg->num_send_queues[type_dev] =IPU6_N_MAX_DEV_SEND_QUEUES ; + isys_fw_cfg->num_send_queues[type_msg] = num_in_message_queues; + isys_fw_cfg->num_recv_queues[type_proxy] =IPU6_N_MAX_PROXY_RECV_QUEUES ; +/* Common msg/dev return queue */ + isys_fw_cfg->num_recv_queues[type_dev] =0 ; + isys_fw_cfg->num_recv_queues[type_msg] =1 ; + + size =sizeof (*input_queue_cfg) * max_send_queues; + input_queue_cfg =devm_kzalloc (dev, size,GFP_KERNEL ); +if (!input_queue_cfg) +return -ENOMEM ; + + size =sizeof (*output_queue_cfg) *IPU6_N_MAX_RECV_QUEUES ; + output_queue_cfg =devm_kzalloc (dev, size,GFP_KERNEL ); +if (!output_queue_cfg) +return -ENOMEM ; + + fwcom->input = input_queue_cfg; + fwcom->output = output_queue_cfg; + + fwcom->num_input_queues = isys_fw_cfg->num_send_queues[type_proxy] + + isys_fw_cfg->num_send_queues[type_dev] + + isys_fw_cfg->num_send_queues[type_msg]; + + fwcom->num_output_queues = isys_fw_cfg->num_recv_queues[type_proxy] + + isys_fw_cfg->num_recv_queues[type_dev] + + isys_fw_cfg->num_recv_queues[type_msg]; + +/* SRAM partitioning. Equal partitioning is set. */ +for (i =0 ; i < max_sram_blocks; i++) { +if (i < num_in_message_queues) + isys_fw_cfg->buffer_partition.num_gda_pages[i] = + (IPU6_DEVICE_GDA_NR_PAGES * +IPU6_DEVICE_GDA_VIRT_FACTOR ) / + num_in_message_queues; +else + isys_fw_cfg->buffer_partition.num_gda_pages[i] =0 ; + } + +/* FW assumes proxy interface at fwcom queue 0 */ +for (i =0 ; i < isys_fw_cfg->num_send_queues[type_proxy]; i++) { + input_queue_cfg[i].token_size = +sizeof (struct ipu6_fw_proxy_send_queue_token ); + input_queue_cfg[i].queue_size =IPU6_ISYS_SIZE_PROXY_SEND_QUEUE ; + } + +for (i =0 ; i < isys_fw_cfg->num_send_queues[type_dev]; i++) { + input_queue_cfg[base_dev_send + i].token_size = +sizeof (struct ipu6_fw_send_queue_token ); + input_queue_cfg[base_dev_send + i].queue_size = max_devq_size; + } + +for (i =0 ; i < isys_fw_cfg->num_send_queues[type_msg]; i++) { + input_queue_cfg[base_msg_send + i].token_size = +sizeof (struct ipu6_fw_send_queue_token ); + input_queue_cfg[base_msg_send + i].queue_size = +IPU6_ISYS_SIZE_SEND_QUEUE ; + } + +for (i =0 ; i < isys_fw_cfg->num_recv_queues[type_proxy]; i++) { + output_queue_cfg[i].token_size = +sizeof (struct ipu6_fw_proxy_resp_queue_token ); + output_queue_cfg[i].queue_size = +IPU6_ISYS_SIZE_PROXY_RECV_QUEUE ; + } +/* There is no recv DEV queue */ +for (i =0 ; i < isys_fw_cfg->num_recv_queues[type_msg]; i++) { + output_queue_cfg[base_msg_recv + i].token_size = +sizeof (struct ipu6_fw_resp_queue_token ); + output_queue_cfg[base_msg_recv + i].queue_size = +IPU6_ISYS_SIZE_RECV_QUEUE ; + } + + fwcom->dmem_addr = isys->pdata->ipdata->hw_variant.dmem_offset; + fwcom->specific_addr = isys_fw_cfg; + fwcom->specific_size =sizeof (*isys_fw_cfg); + +return 0 ; +} + +int ipu6_fw_isys_init (struct ipu6_isys *isys,unsigned int num_streams) +{ +struct device *dev = &isys->adev->auxdev.dev; +int retry =IPU6_ISYS_OPEN_RETRY ; +struct ipu6_fw_com_cfg fwcom = { + .cell_start = start_sp, + .cell_ready = query_sp, + .buttress_boot_offset =SYSCOM_BUTTRESS_FW_PARAMS_ISYS_OFFSET , + }; +int ret; + +ipu6_isys_fwcom_cfg_init (isys, &fwcom, num_streams); + + isys->fwcom =ipu6_fw_com_prepare (&fwcom, isys->adev, + isys->pdata->base); +if (!isys->fwcom) { +dev_err (dev,"isys fw com prepare failed\n" ); +return -EIO ; + } + + ret =ipu6_fw_com_open (isys->fwcom); +if (ret) { +dev_err (dev,"isys fw com open failed %d\n" , ret); +return ret; + } + +do { +usleep_range (400 ,500 ); +if (ipu6_fw_com_ready (isys->fwcom)) +break ; + retry--; + }while (retry >0 ); + +if (!retry) { +dev_err (dev,"isys port open ready failed %d\n" , ret); +ipu6_fw_isys_close (isys); + ret = -EIO ; + } + +return ret; +} + +struct ipu6_fw_isys_resp_info_abi * +ipu6_fw_isys_get_resp (void *context,unsigned int queue) +{ +return ipu6_recv_get_token (context, queue); +} + +void ipu6_fw_isys_put_resp (void *context,unsigned int queue) +{ +ipu6_recv_put_token (context, queue); +} + +void ipu6_fw_isys_dump_stream_cfg (struct device *dev, +struct ipu6_fw_isys_stream_cfg_data_abi *cfg) +{ +unsigned int i; + +dev_dbg (dev,"-----------------------------------------------------\n" ); +dev_dbg (dev,"IPU6_FW_ISYS_STREAM_CFG_DATA\n" ); + +dev_dbg (dev,"compfmt = %d\n" , cfg->vc); +dev_dbg (dev,"src = %d\n" , cfg->src); +dev_dbg (dev,"vc = %d\n" , cfg->vc); +dev_dbg (dev,"isl_use = %d\n" , cfg->isl_use); +dev_dbg (dev,"sensor_type = %d\n" , cfg->sensor_type); + +dev_dbg (dev,"send_irq_sof_discarded = %d\n" , + cfg->send_irq_sof_discarded); +dev_dbg (dev,"send_irq_eof_discarded = %d\n" , + cfg->send_irq_eof_discarded); +dev_dbg (dev,"send_resp_sof_discarded = %d\n" , + cfg->send_resp_sof_discarded); +dev_dbg (dev,"send_resp_eof_discarded = %d\n" , + cfg->send_resp_eof_discarded); + +dev_dbg (dev,"crop:\n" ); +dev_dbg (dev,"\t.left_top = [%d, %d]\n" , cfg->crop.left_offset, + cfg->crop.top_offset); +dev_dbg (dev,"\t.right_bottom = [%d, %d]\n" , cfg->crop.right_offset, + cfg->crop.bottom_offset); + +dev_dbg (dev,"nof_input_pins = %d\n" , cfg->nof_input_pins); +for (i =0 ; i < cfg->nof_input_pins; i++) { +dev_dbg (dev,"input pin[%d]:\n" , i); +dev_dbg (dev,"\t.dt = 0x%0x\n" , cfg->input_pins[i].dt); +dev_dbg (dev,"\t.mipi_store_mode = %d\n" , + cfg->input_pins[i].mipi_store_mode); +dev_dbg (dev,"\t.bits_per_pix = %d\n" , + cfg->input_pins[i].bits_per_pix); +dev_dbg (dev,"\t.mapped_dt = 0x%0x\n" , + cfg->input_pins[i].mapped_dt); +dev_dbg (dev,"\t.input_res = %dx%d\n" , + cfg->input_pins[i].input_res.width, + cfg->input_pins[i].input_res.height); +dev_dbg (dev,"\t.mipi_decompression = %d\n" , + cfg->input_pins[i].mipi_decompression); +dev_dbg (dev,"\t.capture_mode = %d\n" , + cfg->input_pins[i].capture_mode); + } + +dev_dbg (dev,"nof_output_pins = %d\n" , cfg->nof_output_pins); +for (i =0 ; i < cfg->nof_output_pins; i++) { +dev_dbg (dev,"output_pin[%d]:\n" , i); +dev_dbg (dev,"\t.input_pin_id = %d\n" , + cfg->output_pins[i].input_pin_id); +dev_dbg (dev,"\t.output_res = %dx%d\n" , + cfg->output_pins[i].output_res.width, + cfg->output_pins[i].output_res.height); +dev_dbg (dev,"\t.stride = %d\n" , cfg->output_pins[i].stride); +dev_dbg (dev,"\t.pt = %d\n" , cfg->output_pins[i].pt); +dev_dbg (dev,"\t.payload_buf_size = %d\n" , + cfg->output_pins[i].payload_buf_size); +dev_dbg (dev,"\t.ft = %d\n" , cfg->output_pins[i].ft); +dev_dbg (dev,"\t.watermark_in_lines = %d\n" , + cfg->output_pins[i].watermark_in_lines); +dev_dbg (dev,"\t.send_irq = %d\n" , + cfg->output_pins[i].send_irq); +dev_dbg (dev,"\t.reserve_compression = %d\n" , + cfg->output_pins[i].reserve_compression); +dev_dbg (dev,"\t.snoopable = %d\n" , + cfg->output_pins[i].snoopable); +dev_dbg (dev,"\t.error_handling_enable = %d\n" , + cfg->output_pins[i].error_handling_enable); +dev_dbg (dev,"\t.sensor_type = %d\n" , + cfg->output_pins[i].sensor_type); + } +dev_dbg (dev,"-----------------------------------------------------\n" ); +} + +void +ipu6_fw_isys_dump_frame_buff_set (struct device *dev, +struct ipu6_fw_isys_frame_buff_set_abi *buf, +unsigned int outputs) +{ +unsigned int i; + +dev_dbg (dev,"-----------------------------------------------------\n" ); +dev_dbg (dev,"IPU6_FW_ISYS_FRAME_BUFF_SET\n" ); + +for (i =0 ; i < outputs; i++) { +dev_dbg (dev,"output_pin[%d]:\n" , i); +dev_dbg (dev,"\t.out_buf_id = %llu\n" , + buf->output_pins[i].out_buf_id); +dev_dbg (dev,"\t.addr = 0x%x\n" , buf->output_pins[i].addr); +dev_dbg (dev,"\t.compress = %d\n" , + buf->output_pins[i].compress); + } + +dev_dbg (dev,"send_irq_sof = 0x%x\n" , buf->send_irq_sof); +dev_dbg (dev,"send_irq_eof = 0x%x\n" , buf->send_irq_eof); +dev_dbg (dev,"send_resp_sof = 0x%x\n" , buf->send_resp_sof); +dev_dbg (dev,"send_resp_eof = 0x%x\n" , buf->send_resp_eof); +dev_dbg (dev,"send_irq_capture_ack = 0x%x\n" , + buf->send_irq_capture_ack); +dev_dbg (dev,"send_irq_capture_done = 0x%x\n" , + buf->send_irq_capture_done); +dev_dbg (dev,"send_resp_capture_ack = 0x%x\n" , + buf->send_resp_capture_ack); +dev_dbg (dev,"send_resp_capture_done = 0x%x\n" , + buf->send_resp_capture_done); + +dev_dbg (dev,"-----------------------------------------------------\n" ); +}
drivers/media/pci/intel/ipu6/ipu6-fw-isys.h new file +570 -0
570 lines not on this page · raw message
Signature · 1 line
2.42.0
PATCH v2 8/15 media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-device
In reply to [email protected]From: Bingbu Cao <[email protected]>
Input system CSI2 receiver is exposed as a v4l2 sub-device. Each CSI2 sub-device represent one single CSI2 hardware port which be linked with external sub-device such camera sensor by linked with ISYS CSI2's sink pad. CSI2 source pad is linked to the sink pad of video capture device.
--- drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c | 654 ++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-isys-csi2.h | 81 +++ .../media/pci/intel/ipu6/ipu6-isys-subdev.c | 379 ++++++++++ .../media/pci/intel/ipu6/ipu6-isys-subdev.h | 58 ++ .../intel/ipu6/ipu6-platform-isys-csi2-reg.h | 187 +++++ 5 files changed, 1359 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-csi2.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-subdev.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-subdev.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-platform-isys-csi2-reg.h
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c new file +654 -0
@@ -0,0 +1,654 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <linux/bitfield.h> +#include <linux/delay.h> + +#include <media/media-entity.h> +#include <media/v4l2-device.h> +#include <media/v4l2-event.h> + +#include "ipu6-isys.h" +#include "ipu6-isys-subdev.h" +#include "ipu6-platform-buttress-regs.h" +#include "ipu6-platform-isys-csi2-reg.h" +#include "ipu6-platform-regs.h" + +static const u32 csi2_supported_codes[] = { +MEDIA_BUS_FMT_RGB565_1X16 , +MEDIA_BUS_FMT_RGB888_1X24 , +MEDIA_BUS_FMT_UYVY8_1X16 , +MEDIA_BUS_FMT_YUYV8_1X16 , +MEDIA_BUS_FMT_SBGGR10_1X10 , +MEDIA_BUS_FMT_SGBRG10_1X10 , +MEDIA_BUS_FMT_SGRBG10_1X10 , +MEDIA_BUS_FMT_SRGGB10_1X10 , +MEDIA_BUS_FMT_SBGGR12_1X12 , +MEDIA_BUS_FMT_SGBRG12_1X12 , +MEDIA_BUS_FMT_SGRBG12_1X12 , +MEDIA_BUS_FMT_SRGGB12_1X12 , +MEDIA_BUS_FMT_SBGGR8_1X8 , +MEDIA_BUS_FMT_SGBRG8_1X8 , +MEDIA_BUS_FMT_SGRBG8_1X8 , +MEDIA_BUS_FMT_SRGGB8_1X8 , +0 +}; + +/* +* Strings corresponding to CSI-2 receiver errors are here. +* Corresponding macros are defined in the header file. +*/ +static const struct ipu6_csi2_error dphy_rx_errors[] = { + {"Single packet header error corrected" , true }, + {"Multiple packet header errors detected" , true }, + {"Payload checksum (CRC) error" , true }, + {"Transfer FIFO overflow" , false }, + {"Reserved short packet data type detected" , true }, + {"Reserved long packet data type detected" , true }, + {"Incomplete long packet detected" , false }, + {"Frame sync error" , false }, + {"Line sync error" , false }, + {"DPHY recoverable synchronization error" , true }, + {"DPHY fatal error" , false }, + {"DPHY elastic FIFO overflow" , false }, + {"Inter-frame short packet discarded" , true }, + {"Inter-frame long packet discarded" , true }, + {"MIPI pktgen overflow" , false }, + {"MIPI pktgen data loss" , false }, + {"FIFO overflow" , false }, + {"Lane deskew" , false }, + {"SOT sync error" , false }, + {"HSIDLE detected" , false } +}; + +s64 ipu6_isys_csi2_get_link_freq (struct ipu6_isys_csi2 *csi2) +{ +struct media_pad *src_pad; +struct v4l2_subdev *ext_sd; +struct device *dev; + +if (!csi2) +return -EINVAL ; + + dev = &csi2->isys->adev->auxdev.dev; + src_pad =media_entity_remote_source_pad_unique (&csi2->asd.sd.entity); +if (IS_ERR_OR_NULL (src_pad)) { +dev_err (dev,"can't get source pad of %s\n" , csi2->asd.sd.name); +return -ENOLINK ; + } + + ext_sd =media_entity_to_v4l2_subdev (src_pad->entity); +if (WARN (!ext_sd,"Failed to get subdev for %s\n" , csi2->asd.sd.name)) +return -ENODEV ; + +return v4l2_get_link_freq (ext_sd->ctrl_handler,0 ,0 ); +} + +static int csi2_subscribe_event (struct v4l2_subdev *sd,struct v4l2_fh *fh, +struct v4l2_event_subscription *sub) +{ +struct ipu6_isys_csi2 *csi2 =to_ipu6_isys_csi2 (sd); +struct device *dev = &csi2->isys->adev->auxdev.dev; + +dev_dbg (dev,"csi2 subscribe event(type %u id %u)\n" , + sub->type, sub->id); + +switch (sub->type) { +case V4L2_EVENT_FRAME_SYNC : +return v4l2_event_subscribe (fh, sub,10 ,NULL ); +case V4L2_EVENT_CTRL : +return v4l2_ctrl_subscribe_event (fh, sub); +default : +return -EINVAL ; + } +} + +static const struct v4l2_subdev_core_ops csi2_sd_core_ops = { + .subscribe_event = csi2_subscribe_event, + .unsubscribe_event = v4l2_event_subdev_unsubscribe, +}; + +/* +* The input system CSI2+ receiver has several +* parameters affecting the receiver timings. These depend +* on the MIPI bus frequency F in Hz (sensor transmitter rate) +* as follows: +* register value = (A/1e9 + B * UI) / COUNT_ACC +* where +* UI = 1 / (2 * F) in seconds +* COUNT_ACC = counter accuracy in seconds +* COUNT_ACC = 0.125 ns = 1 / 8 ns, ACCINV = 8. +* +* A and B are coefficients from the table below, +* depending whether the register minimum or maximum value is +* calculated. +* Minimum Maximum +* Clock lane A B A B +* reg_rx_csi_dly_cnt_termen_clane 0 0 38 0 +* reg_rx_csi_dly_cnt_settle_clane 95 -8 300 -16 +* Data lanes +* reg_rx_csi_dly_cnt_termen_dlane0 0 0 35 4 +* reg_rx_csi_dly_cnt_settle_dlane0 85 -2 145 -6 +* reg_rx_csi_dly_cnt_termen_dlane1 0 0 35 4 +* reg_rx_csi_dly_cnt_settle_dlane1 85 -2 145 -6 +* reg_rx_csi_dly_cnt_termen_dlane2 0 0 35 4 +* reg_rx_csi_dly_cnt_settle_dlane2 85 -2 145 -6 +* reg_rx_csi_dly_cnt_termen_dlane3 0 0 35 4 +* reg_rx_csi_dly_cnt_settle_dlane3 85 -2 145 -6 +* +* We use the minimum values of both A and B. +*/ + +#define DIV_SHIFT 8 +#define CSI2_ACCINV 8 + +static u32 calc_timing (s32 a,s32 b,s64 link_freq,u32 accinv) +{ +return accinv * a + (accinv * b * (500000000 >>DIV_SHIFT ) + / (int32_t )(link_freq >>DIV_SHIFT )); +} + +static int +ipu6_isys_csi2_calc_timing (struct ipu6_isys_csi2 *csi2, +struct ipu6_isys_csi2_timing *timing,u32 accinv) +{ +struct device *dev = &csi2->isys->adev->auxdev.dev; +s64 link_freq; + + link_freq =ipu6_isys_csi2_get_link_freq (csi2); +if (link_freq <0 ) +return link_freq; + + timing->ctermen =calc_timing (CSI2_CSI_RX_DLY_CNT_TERMEN_CLANE_A , +CSI2_CSI_RX_DLY_CNT_TERMEN_CLANE_B , + link_freq, accinv); + timing->csettle =calc_timing (CSI2_CSI_RX_DLY_CNT_SETTLE_CLANE_A , +CSI2_CSI_RX_DLY_CNT_SETTLE_CLANE_B , + link_freq, accinv); + timing->dtermen =calc_timing (CSI2_CSI_RX_DLY_CNT_TERMEN_DLANE_A , +CSI2_CSI_RX_DLY_CNT_TERMEN_DLANE_B , + link_freq, accinv); + timing->dsettle =calc_timing (CSI2_CSI_RX_DLY_CNT_SETTLE_DLANE_A , +CSI2_CSI_RX_DLY_CNT_SETTLE_DLANE_B , + link_freq, accinv); + +dev_dbg (dev,"ctermen %u csettle %u dtermen %u dsettle %u\n" , + timing->ctermen, timing->csettle, + timing->dtermen, timing->dsettle); + +return 0 ; +} + +void ipu6_isys_register_errors (struct ipu6_isys_csi2 *csi2) +{ +u32 irq =readl (csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_STATUS_OFFSET ); +struct ipu6_isys *isys = csi2->isys; +u32 mask; + + mask = isys->pdata->ipdata->csi2.irq_mask; +writel (irq & mask, csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_CLEAR_OFFSET ); + csi2->receiver_errors |= irq & mask; +} + +void ipu6_isys_csi2_error (struct ipu6_isys_csi2 *csi2) +{ +struct device *dev = &csi2->isys->adev->auxdev.dev; +const struct ipu6_csi2_error *errors; +u32 status; +u32 i; + +/* register errors once more in case of interrupts are disabled */ +ipu6_isys_register_errors (csi2); + status = csi2->receiver_errors; + csi2->receiver_errors =0 ; + errors = dphy_rx_errors; + +for (i =0 ; i <CSI_RX_NUM_ERRORS_IN_IRQ ; i++) { +if (status &BIT (i)) +dev_err_ratelimited (dev,"csi2-%i error: %s\n" , + csi2->port, errors[i].error_string); + } +} + +static int ipu6_isys_csi2_set_stream (struct v4l2_subdev *sd, +const struct ipu6_isys_csi2_timing *timing, +unsigned int nlanes,int enable) +{ +struct ipu6_isys_csi2 *csi2 =to_ipu6_isys_csi2 (sd); +struct ipu6_isys *isys = csi2->isys; +struct device *dev = &isys->adev->auxdev.dev; +struct ipu6_isys_csi2_config cfg; +unsigned int nports; +int ret =0 ; +u32 mask =0 ; +u32 i; + +dev_dbg (dev,"stream %s CSI2-%u with %u lanes\n" , enable ?"on" :"off" , + csi2->port, nlanes); + + cfg.port = csi2->port; + cfg.nlanes = nlanes; + + mask = isys->pdata->ipdata->csi2.irq_mask; + nports = isys->pdata->ipdata->csi2.nports; + +if (!enable) { +writel (0 , csi2->base +CSI_REG_CSI_FE_ENABLE ); +writel (0 , csi2->base +CSI_REG_PPI2CSI_ENABLE ); + +writel (0 , + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_ENABLE_OFFSET ); +writel (mask, + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_CLEAR_OFFSET ); +writel (0 , + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_ENABLE_OFFSET ); +writel (0xffffffff , + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_CLEAR_OFFSET ); + + isys->phy_set_power (isys, &cfg, timing, false); + +writel (0 , isys->pdata->base +CSI_REG_HUB_FW_ACCESS_PORT + (isys->pdata->ipdata->csi2.fw_access_port_ofs, + csi2->port)); +writel (0 , isys->pdata->base + +CSI_REG_HUB_DRV_ACCESS_PORT (csi2->port)); + +return ret; + } + +/* reset port reset */ +writel (0x1 , csi2->base +CSI_REG_PORT_GPREG_SRST ); +usleep_range (100 ,200 ); +writel (0x0 , csi2->base +CSI_REG_PORT_GPREG_SRST ); + +/* enable port clock */ +for (i =0 ; i < nports; i++) { +writel (1 , isys->pdata->base +CSI_REG_HUB_DRV_ACCESS_PORT (i)); +writel (1 , isys->pdata->base +CSI_REG_HUB_FW_ACCESS_PORT + (isys->pdata->ipdata->csi2.fw_access_port_ofs, i)); + } + +/* enable all error related irq */ +writel (mask, + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_STATUS_OFFSET ); +writel (mask, + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_MASK_OFFSET ); +writel (mask, + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_CLEAR_OFFSET ); +writel (mask, + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_LEVEL_NOT_PULSE_OFFSET ); +writel (mask, + csi2->base +CSI_PORT_REG_BASE_IRQ_CSI + +CSI_PORT_REG_BASE_IRQ_ENABLE_OFFSET ); + +/* +* Using event from firmware instead of irq to handle CSI2 sync event +* which can reduce system wakeups. If CSI2 sync irq enabled, we need +* disable the firmware CSI2 sync event to avoid duplicate handling. +*/ +writel (0xffffffff , csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_STATUS_OFFSET ); +writel (0 , csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_MASK_OFFSET ); +writel (0xffffffff , csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_CLEAR_OFFSET ); +writel (0 , csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_LEVEL_NOT_PULSE_OFFSET ); +writel (0xffffffff , csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_ENABLE_OFFSET ); + +/* configure to enable FE and PPI2CSI */ +writel (0 , csi2->base +CSI_REG_CSI_FE_MODE ); +writel (CSI_SENSOR_INPUT , csi2->base +CSI_REG_CSI_FE_MUX_CTRL ); +writel (CSI_CNTR_SENSOR_LINE_ID |CSI_CNTR_SENSOR_FRAME_ID , + csi2->base +CSI_REG_CSI_FE_SYNC_CNTR_SEL ); +writel (FIELD_PREP (PPI_INTF_CONFIG_NOF_ENABLED_DLANES_MASK , nlanes -1 ), + csi2->base +CSI_REG_PPI2CSI_CONFIG_PPI_INTF ); + +writel (1 , csi2->base +CSI_REG_PPI2CSI_ENABLE ); +writel (1 , csi2->base +CSI_REG_CSI_FE_ENABLE ); + + ret = isys->phy_set_power (isys, &cfg, timing, true); +if (ret) +dev_err (dev,"csi-%d phy power up failed %d\n" , csi2->port, + ret); + +return ret; +} + +static int set_stream (struct v4l2_subdev *sd,int enable) +{ +struct ipu6_isys_csi2 *csi2 =to_ipu6_isys_csi2 (sd); +struct device *dev = &csi2->isys->adev->auxdev.dev; +struct ipu6_isys_csi2_timing timing = { }; +unsigned int nlanes; +int ret; + +dev_dbg (dev,"csi2 stream %s callback\n" , enable ?"on" :"off" ); + +if (!enable) { + csi2->stream_count--; +if (csi2->stream_count) +return 0 ; + +ipu6_isys_csi2_set_stream (sd, &timing,0 , enable); +return 0 ; + } + +if (csi2->stream_count) { + csi2->stream_count++; +return 0 ; + } + + nlanes = csi2->nlanes; + + ret =ipu6_isys_csi2_calc_timing (csi2, &timing,CSI2_ACCINV ); +if (ret) +return ret; + + ret =ipu6_isys_csi2_set_stream (sd, &timing, nlanes, enable); +if (ret) +return ret; + + csi2->stream_count++; + +return 0 ; +} + +static int ipu6_isys_csi2_set_sel (struct v4l2_subdev *sd, +struct v4l2_subdev_state *state, +struct v4l2_subdev_selection *sel) +{ +struct ipu6_isys_subdev *asd =to_ipu6_isys_subdev (sd); +struct device *dev = &asd->isys->adev->auxdev.dev; +struct v4l2_mbus_framefmt *sink_ffmt; +struct v4l2_mbus_framefmt *src_ffmt; +struct v4l2_rect *crop; + +if (sel->pad ==CSI2_PAD_SINK || sel->target !=V4L2_SEL_TGT_CROP ) +return -EINVAL ; + + sink_ffmt =v4l2_subdev_state_get_opposite_stream_format (state, + sel->pad, + sel->stream); +if (!sink_ffmt) +return -EINVAL ; + + src_ffmt =v4l2_subdev_state_get_stream_format (state, sel->pad, + sel->stream); +if (!src_ffmt) +return -EINVAL ; + + crop =v4l2_subdev_state_get_stream_crop (state, sel->pad, sel->stream); +if (!crop) +return -EINVAL ; + +/* Only vertical cropping is supported */ + sel->r.left =0 ; + sel->r.width = sink_ffmt->width; +/* Non-bayer formats can't be single line cropped */ +if (!ipu6_isys_is_bayer_format (sink_ffmt->code)) + sel->r.top &= ~1 ; + sel->r.height =clamp (sel->r.height & ~1 ,IPU6_ISYS_MIN_HEIGHT , + sink_ffmt->height - sel->r.top); + *crop = sel->r; + +/* update source pad format */ + src_ffmt->width = sel->r.width; + src_ffmt->height = sel->r.height; +if (ipu6_isys_is_bayer_format (sink_ffmt->code)) + src_ffmt->code =ipu6_isys_convert_bayer_order (sink_ffmt->code, + sel->r.left, + sel->r.top); +dev_dbg (dev,"set crop for %s sel: %d,%d,%d,%d code: 0x%x\n" , + sd->name, sel->r.left, sel->r.top, sel->r.width, sel->r.height, + src_ffmt->code); + +return 0 ; +} + +static int ipu6_isys_csi2_get_sel (struct v4l2_subdev *sd, +struct v4l2_subdev_state *state, +struct v4l2_subdev_selection *sel) +{ +struct v4l2_mbus_framefmt *sink_ffmt; +struct v4l2_rect *crop; +int ret =0 ; + +if (sd->entity.pads[sel->pad].flags &MEDIA_PAD_FL_SINK ) +return -EINVAL ; + + sink_ffmt =v4l2_subdev_state_get_opposite_stream_format (state, + sel->pad, + sel->stream); +if (!sink_ffmt) +return -EINVAL ; + + crop =v4l2_subdev_state_get_stream_crop (state, sel->pad, sel->stream); +if (!crop) +return -EINVAL ; + +switch (sel->target) { +case V4L2_SEL_TGT_CROP_DEFAULT : +case V4L2_SEL_TGT_CROP_BOUNDS : + sel->r.left =0 ; + sel->r.top =0 ; + sel->r.width = sink_ffmt->width; + sel->r.height = sink_ffmt->height; +break ; +case V4L2_SEL_TGT_CROP : + sel->r = *crop; +break ; +default : + ret = -EINVAL ; + } + +return ret; +} + +static const struct v4l2_subdev_video_ops csi2_sd_video_ops = { + .s_stream = set_stream, +}; + +static const struct v4l2_subdev_pad_ops csi2_sd_pad_ops = { + .init_cfg = ipu6_isys_subdev_init_cfg, + .get_fmt = v4l2_subdev_get_fmt, + .set_fmt = ipu6_isys_subdev_set_fmt, + .get_selection = ipu6_isys_csi2_get_sel, + .set_selection = ipu6_isys_csi2_set_sel, + .enum_mbus_code = ipu6_isys_subdev_enum_mbus_code, + .set_routing = ipu6_isys_subdev_set_routing, +}; + +static const struct v4l2_subdev_ops csi2_sd_ops = { + .core = &csi2_sd_core_ops, + .video = &csi2_sd_video_ops, + .pad = &csi2_sd_pad_ops, +}; + +static const struct media_entity_operations csi2_entity_ops = { + .link_validate = v4l2_subdev_link_validate, + .has_pad_interdep = v4l2_subdev_has_pad_interdep, +}; + +void ipu6_isys_csi2_cleanup (struct ipu6_isys_csi2 *csi2) +{ +if (!csi2->isys) +return ; + +v4l2_device_unregister_subdev (&csi2->asd.sd); +v4l2_subdev_cleanup (&csi2->asd.sd); +ipu6_isys_subdev_cleanup (&csi2->asd); + csi2->isys =NULL ; +} + +int ipu6_isys_csi2_init (struct ipu6_isys_csi2 *csi2, +struct ipu6_isys *isys, +void __iomem *base,unsigned int index) +{ +struct device *dev = &isys->adev->auxdev.dev; +int ret; + + csi2->isys = isys; + csi2->base = base; + csi2->port = index; + + csi2->asd.sd.entity.ops = &csi2_entity_ops; + csi2->asd.isys = isys; + ret =ipu6_isys_subdev_init (&csi2->asd, &csi2_sd_ops,0 , +NR_OF_CSI2_SINK_PADS ,NR_OF_CSI2_SRC_PADS ); +if (ret) + goto fail; + + csi2->asd.source =IPU6_FW_ISYS_STREAM_SRC_CSI2_PORT0 + index; + csi2->asd.supported_codes = csi2_supported_codes; +snprintf (csi2->asd.sd.name,sizeof (csi2->asd.sd.name), +IPU6_ISYS_ENTITY_PREFIX " CSI2 %u" , index); +v4l2_set_subdevdata (&csi2->asd.sd, &csi2->asd); + ret =v4l2_subdev_init_finalize (&csi2->asd.sd); +if (ret) { +dev_err (dev,"failed to init v4l2 subdev\n" ); + goto fail; + } + + ret =v4l2_device_register_subdev (&isys->v4l2_dev, &csi2->asd.sd); +if (ret) { +dev_err (dev,"failed to register v4l2 subdev\n" ); + goto fail; + } + +return 0 ; + +fail: +ipu6_isys_csi2_cleanup (csi2); + +return ret; +} + +void ipu6_isys_csi2_sof_event_by_stream (struct ipu6_isys_stream *stream) +{ +struct video_device *vdev = stream->asd->sd.devnode; +struct device *dev = &stream->isys->adev->auxdev.dev; +struct ipu6_isys_csi2 *csi2 =ipu6_isys_subdev_to_csi2 (stream->asd); +struct v4l2_event ev = { + .type =V4L2_EVENT_FRAME_SYNC , + }; + + ev.u.frame_sync.frame_sequence =atomic_inc_return (&stream->sequence); +v4l2_event_queue (vdev, &ev); + +dev_dbg (dev,"sof_event::csi2-%i sequence: %i, vc: %d\n" , + csi2->port, ev.u.frame_sync.frame_sequence, stream->vc); +} + +void ipu6_isys_csi2_eof_event_by_stream (struct ipu6_isys_stream *stream) +{ +struct device *dev = &stream->isys->adev->auxdev.dev; +struct ipu6_isys_csi2 *csi2 =ipu6_isys_subdev_to_csi2 (stream->asd); +u32 frame_sequence =atomic_read (&stream->sequence); + +dev_dbg (dev,"eof_event::csi2-%i sequence: %i\n" , + csi2->port, frame_sequence); +} + +int ipu6_isys_csi2_get_remote_desc (u32 source_stream, +struct ipu6_isys_csi2 *csi2, +struct media_entity *source_entity, +struct v4l2_mbus_frame_desc_entry *entry, +int *nr_queues) +{ +struct v4l2_mbus_frame_desc_entry *desc_entry =NULL ; +struct device *dev = &csi2->isys->adev->auxdev.dev; +struct v4l2_mbus_frame_desc desc; +struct v4l2_subdev *source; +struct media_pad *pad; +unsigned int i; +int count =0 ; +int ret; + + source =media_entity_to_v4l2_subdev (source_entity); +if (!source) +return -EPIPE ; + + pad =media_pad_remote_pad_first (&csi2->asd.pad[CSI2_PAD_SINK ]); +if (!pad) +return -EPIPE ; + + ret =v4l2_subdev_call (source, pad, get_frame_desc, pad->index, &desc); +if (ret) +return ret; + +if (desc.type !=V4L2_MBUS_FRAME_DESC_TYPE_CSI2 ) { +dev_err (dev,"Unsupported frame descriptor type\n" ); +return -EINVAL ; + } + +for (i =0 ; i < desc.num_entries; i++) { +if (source_stream == desc.entry[i].stream) { + desc_entry = &desc.entry[i]; +break ; + } + } + +if (!desc_entry) { +dev_err (dev,"Failed to find stream %u from remote subdev\n" , + source_stream); +return -EINVAL ; + } + +if (desc_entry->bus.csi2.vc >=NR_OF_CSI2_VC ) { +dev_err (dev,"invalid vc %d\n" , desc_entry->bus.csi2.vc); +return -EINVAL ; + } + + *entry = *desc_entry; +for (i =0 ; i < desc.num_entries; i++) { +if (entry->bus.csi2.vc == desc.entry[i].bus.csi2.vc) + count++; + } + + *nr_queues = count; +return 0 ; +} + +void ipu6_isys_set_csi2_streams_status (struct ipu6_isys_video *av,bool status) +{ +struct ipu6_isys_stream *stream = av->stream; +struct v4l2_subdev *sd = &stream->asd->sd; +struct v4l2_subdev_state *state; +struct media_pad *r_pad; +unsigned int i; +u32 r_stream; + + r_pad =media_pad_remote_pad_first (&av->pad); + r_stream =ipu6_isys_get_src_stream_by_src_pad (sd, r_pad->index); + + state =v4l2_subdev_lock_and_get_active_state (sd); + +for (i =0 ; i < state->stream_configs.num_configs; i++) { +struct v4l2_subdev_stream_config *cfg = + &state->stream_configs.configs[i]; + +if (cfg->pad == r_pad->index && r_stream == cfg->stream) { +dev_dbg (&av->isys->adev->auxdev.dev, +"%s: pad:%u, stream:%u, status:%u\n" , + sd->entity.name, r_pad->index, r_stream, + status); + cfg->enabled = status; + } + } + +v4l2_subdev_unlock_state (state); +}
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.h new file +81 -0
81 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6-isys-subdev.c new file +379 -0
379 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6-isys-subdev.h new file +58 -0
58 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6-platform-isys-csi2-reg.h new file +187 -0
187 lines not on this page · raw message
Signature · 1 line
2.42.0
Re: [PATCH v2 08/15] media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-device
Hi Bingbu,
Quoted 30 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Input system CSI2 receiver is exposed as a v4l2 sub-device. Each CSI2 sub-device represent one single CSI2 hardware port which be linked with external sub-device such camera sensor by linked with ISYS CSI2's sink pad. CSI2 source pad is linked to the sink pad of video capture device.
Signed-off-by Bingbu Cao <[email protected]>---
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c | 654
++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.h | 81 +++ .../media/pci/intel/ipu6/ipu6-isys-subdev.c | 379 ++++++++++ .../media/pci/intel/ipu6/ipu6-isys-subdev.h | 58 ++ .../intel/ipu6/ipu6-platform-isys-csi2-reg.h | 187 +++++5 files changed, 1359 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-csi2.hcreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-subdev.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-subdev.hcreate mode 100644drivers/media/pci/intel/ipu6/ipu6-platform-isys-csi2-reg.hdrivers/media/pci/intel/ipu6/ipu6-isys-csi2.c new file +0
-0
...
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c line 151
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c new file +9
-0+ +#define DIV_SHIFT 8 +#define CSI2_ACCINV 8 + +static u32 calc_timing (s32 a,s32 b,s64 link_freq,u32 accinv) +{ + return accinv * a + (accinv * b * (500000000 >> DIV_SHIFT) + / (int32_t)(link_freq >> DIV_SHIFT)); +}
I think accinv should be s32 here. When accinv is unsigned, the expression (accinv * b) becomes unsigned, and thus negative values of b gives incorrect results.
This is different from e.g. https://github.com/intel/ipu6-drivers.
/Andreas
RE: [PATCH v2 08/15] media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-device
Andreas,
------------------------------------------------------------------------
BRs,
Bingbu Cao
Quoted 48 lines
-----Original Message-----
From: Andreas Helbech Kleist <[email protected]>
Sent: Wednesday, November 8, 2023 7:26 PM
To: Cao, Bingbu <[email protected]>; [email protected]; [email protected]; [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Qiu, Tian Shu <[email protected]>; Wang, Hongju <[email protected]>
Subject: Re: [PATCH v2 08/15] media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-deviceHi Bingbu,
Quoted 30 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Input system CSI2 receiver is exposed as a v4l2 sub-device. Each CSI2 sub-device represent one single CSI2 hardware port which be linked with external sub-device such camera sensor by linked with ISYS CSI2's sink pad. CSI2 source pad is linked to the sink pad of video capture device.
Signed-off-by Bingbu Cao <[email protected]>---
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c | 654
++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.h | 81 +++ .../media/pci/intel/ipu6/ipu6-isys-subdev.c | 379 ++++++++++ .../media/pci/intel/ipu6/ipu6-isys-subdev.h | 58 ++ .../intel/ipu6/ipu6-platform-isys-csi2-reg.h | 187 +++++5 files changed, 1359 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-csi2.hcreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-subdev.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-subdev.hcreate mode 100644drivers/media/pci/intel/ipu6/ipu6-platform-isys-csi2-reg.hdrivers/media/pci/intel/ipu6/ipu6-isys-csi2.c new file +0
-0...
Quoted 1 line
Quoted 11 lines
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c
Quoted 7 lines
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c new file +7
-0+ +#define DIV_SHIFT 8 +#define CSI2_ACCINV 8 + +static u32 calc_timing (s32 a,s32 b,s64 link_freq,u32 accinv) { + return accinv * a + (accinv * b * (500000000 >> DIV_SHIFT) + / (int32_t)(link_freq >> DIV_SHIFT)); }I think accinv should be s32 here. When accinv is unsigned, the expression (accinv * b) becomes unsigned, and thus negative values of b gives incorrect results.
accinv is always an unsigned value, so I think we don't need change the type of argument. Following usual arithmetic conversions, I think it needs a signed integer cast here. What do you think?
Quoted 4 lines
This is different from e.g. https://github.com/intel/ipu6-drivers.
/Andreas
Re: [PATCH v2 08/15] media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-device
Quoted 66 lines · Bingbu
Andreas,
------------------------------------------------------------------------
BRs,
Bingbu CaoQuoted 48 lines
-----Original Message-----
From: Andreas Helbech Kleist <[email protected]>
Sent: Wednesday, November 8, 2023 7:26 PM
To: Cao, Bingbu <[email protected]>; [email protected]; [email protected]; [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Qiu, Tian Shu <[email protected]>; Wang, Hongju <[email protected]>
Subject: Re: [PATCH v2 08/15] media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-deviceHi Bingbu,
Quoted 30 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Input system CSI2 receiver is exposed as a v4l2 sub-device. Each CSI2 sub-device represent one single CSI2 hardware port which be linked with external sub-device such camera sensor by linked with ISYS CSI2's sink pad. CSI2 source pad is linked to the sink pad of video capture device.
Signed-off-by Bingbu Cao <[email protected]>---
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c | 654
++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.h | 81 +++ .../media/pci/intel/ipu6/ipu6-isys-subdev.c | 379 ++++++++++ .../media/pci/intel/ipu6/ipu6-isys-subdev.h | 58 ++ .../intel/ipu6/ipu6-platform-isys-csi2-reg.h | 187 +++++5 files changed, 1359 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-csi2.hcreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-subdev.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys-subdev.hcreate mode 100644drivers/media/pci/intel/ipu6/ipu6-platform-isys-csi2-reg.hdrivers/media/pci/intel/ipu6/ipu6-isys-csi2.c new file +0
-0...
Quoted 1 line
Quoted 11 lines
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c
Quoted 7 lines
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c new file +7
-0+ +#define DIV_SHIFT 8 +#define CSI2_ACCINV 8 + +static u32 calc_timing (s32 a,s32 b,s64 link_freq,u32 accinv) { + return accinv * a + (accinv * b * (500000000 >> DIV_SHIFT) + / (int32_t)(link_freq >> DIV_SHIFT)); }I think accinv should be s32 here. When accinv is unsigned, the expression (accinv * b) becomes unsigned, and thus negative values of b gives incorrect results.
Quoted 3 lines
accinv is always an unsigned value, so I think we don't need change the type of argument. Following usual arithmetic conversions, I think it needs a signed integer cast here. What do you think?
However, I can also follow your comments to change to s32. Thanks!
Quoted 6 lines
Quoted 6 lines
This is different from e.g. https://github.com/intel/ipu6-drivers.
/Andreas
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 08/15] media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-device
In reply to Cao, BingbuBingbu,
Quoted 18 lines · Bingbu
Quoted 17 lines
-----Original Message-----
Quoted 9 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c new file +4
-0+#define DIV_SHIFT 8 +#define CSI2_ACCINV 8 + +static u32 calc_timing (s32 a,s32 b,s64 link_freq,u32 accinv){
+ return accinv * a + (accinv * b * (500000000 >>DIV_SHIFT)
+ / (int32_t)(link_freq >>DIV_SHIFT)); }I think accinv should be s32 here. When accinv is unsigned, the expression
(accinv * b) becomes unsigned, and thus negative values of b gives incorrect results.
Quoted 5 lines
accinv is always an unsigned value, so I think we don't need change the
type of argument. Following usual arithmetic conversions, I think it needs
a signed integer cast here. What do you think?
I think changing the type of argument is clearer. A cast just adds noise in my opinion. Of course the argument to ipu6_isys_csi2_calc_timing should also be changed to match, which would also remove any implicit casts as far as I can tell.
Now we're talking about casts. There is a cast to int32_t, but the rest of the code uses the s32/u32 typedefs. I don't know if one or the other is recommended for new code, but I think it should at least be consistent across the driver.
/Andreas
Re: [PATCH v2 08/15] media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-device
Quoted 1 line · Andreas Helbech Kleist
On Wed, 2023-11-08 at 14:50 +0000, Cao, Bingbu wrote:
...
Quoted 4 lines
Now we're talking about casts. There is a cast to
int32_t, but the rest of the code uses the s32/u32 typedefs. I don't know if one or the other is recommended for new code, but I think it should at least be consistent across the driver.
s32/u32 are preferable in the kernel code.
Also if you need to extend the sign bit, we have sign_extendXX() helpers.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 08/15] media: intel/ipu6: add IPU6 CSI2 receiver v4l2 sub-device
Andy,
Quoted 5 lines · On 11/8/23 11:32 PM, [email protected] wrote:
Quoted 1 line · Andreas Helbech Kleist
On Wed, 2023-11-08 at 14:50 +0000, Cao, Bingbu wrote:
...
Quoted 8 lines
Quoted 4 lines
Now we're talking about casts. There is a cast to
int32_t, but the rest of the code uses the s32/u32 typedefs. I don't know if one or the other is recommended for new code, but I think it should at least be consistent across the driver.s32/u32 are preferable in the kernel code.
Also if you need to extend the sign bit, we have
sign_extendXX()helpers.
Ack, thanks. I will walk through the code and fix the casts in the driver.
Quoted 1 line
Signature · 2 lines
Best regards, Bingbu Cao
PATCH v2 9/15 media: intel/ipu6: add the CSI2 DPHY implementation
In reply to [email protected]From: Bingbu Cao <[email protected]>
IPU6 CSI2 DPHY hardware varies on different platforms, current IPU6 has three DPHY hardware instance which maybe used on tigerlake, alderlake, metorlake and jasperlake. MCD DPHY is shipped on tigerlake and alderlake, DWC DPHY is shipped on metorlake.
Each PHY has its own register space, input system driver call the DPHY callback which was set at isys_probe().
--- .../media/pci/intel/ipu6/ipu6-isys-dwc-phy.c | 533 +++++++++++++ .../media/pci/intel/ipu6/ipu6-isys-jsl-phy.c | 237 ++++++ .../media/pci/intel/ipu6/ipu6-isys-mcd-phy.c | 712 ++++++++++++++++++ 3 files changed, 1482 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-jsl-phy.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-mcd-phy.c
drivers/media/pci/intel/ipu6/ipu6-isys-dwc-phy.c new file +533 -0
@@ -0,0 +1,533 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <linux/bitfield.h> +#include <linux/delay.h> +#include <linux/iopoll.h> + +#include "ipu6-isys.h" +#include "ipu6-platform-isys-csi2-reg.h" +#include "ipu6-platform-regs.h" + +#define IPU6_DWC_DPHY_BASE (i) (0x238038 + 0x34 * (i)) +#define IPU6_DWC_DPHY_RSTZ 0x00 +#define IPU6_DWC_DPHY_SHUTDOWNZ 0x04 +#define IPU6_DWC_DPHY_HSFREQRANGE 0x08 +#define IPU6_DWC_DPHY_CFGCLKFREQRANGE 0x0c +#define IPU6_DWC_DPHY_TEST_IFC_ACCESS_MODE 0x10 +#define IPU6_DWC_DPHY_TEST_IFC_REQ 0x14 +#define IPU6_DWC_DPHY_TEST_IFC_REQ_COMPLETION 0x18 +#define IPU6_DWC_DPHY_DFT_CTRL0 0x28 +#define IPU6_DWC_DPHY_DFT_CTRL1 0x2c +#define IPU6_DWC_DPHY_DFT_CTRL2 0x30 + +/* +* test IFC request definition: +* - req: 0 for read, 1 for write +* - 12 bits address +* - 8bits data (will ignore for read) +* --24----16------4-----0 +* --|-data-|-addr-|-req-| +*/ +#define IFC_REQ (req, addr, data) (FIELD_PREP(GENMASK(23, 16), data) | \ + FIELD_PREP(GENMASK(15, 4), addr) | \ + FIELD_PREP(GENMASK(1, 0), req)) + +#define TEST_IFC_REQ_READ 0 +#define TEST_IFC_REQ_WRITE 1 +#define TEST_IFC_REQ_RESET 2 + +#define TEST_IFC_ACCESS_MODE_FSM 0 +#define TEST_IFC_ACCESS_MODE_IFC_CTL 1 + +enum phy_fsm_state { +PHY_FSM_STATE_POWERON =0 , +PHY_FSM_STATE_BGPON =1 , +PHY_FSM_STATE_CAL_TYPE =2 , +PHY_FSM_STATE_BURNIN_CAL =3 , +PHY_FSM_STATE_TERMCAL =4 , +PHY_FSM_STATE_OFFSETCAL =5 , +PHY_FSM_STATE_OFFSET_LANE =6 , +PHY_FSM_STATE_IDLE =7 , +PHY_FSM_STATE_ULP =8 , +PHY_FSM_STATE_DDLTUNNING =9 , +PHY_FSM_STATE_SKEW_BACKWARD =10 , +PHY_FSM_STATE_INVALID , +}; + +static void dwc_dphy_write (struct ipu6_isys *isys,u32 phy_id,u32 addr, +u32 data) +{ +struct device *dev = &isys->adev->auxdev.dev; +void __iomem *isys_base = isys->pdata->base; +void __iomem *base = isys_base +IPU6_DWC_DPHY_BASE (phy_id); + +dev_dbg (dev,"write: reg 0x%lx = data 0x%x" , base + addr - isys_base, + data); +writel (data, base + addr); +} + +static u32 dwc_dphy_read (struct ipu6_isys *isys,u32 phy_id,u32 addr) +{ +struct device *dev = &isys->adev->auxdev.dev; +void __iomem *isys_base = isys->pdata->base; +void __iomem *base = isys_base +IPU6_DWC_DPHY_BASE (phy_id); +u32 data; + + data =readl (base + addr); +dev_dbg (dev,"read: reg 0x%lx = data 0x%x" , base + addr - isys_base, + data); + +return data; +} + +static void dwc_dphy_write_mask (struct ipu6_isys *isys,u32 phy_id,u32 addr, +u32 data,u8 shift,u8 width) +{ +u32 temp; +u32 mask; + + mask = (1 << width) -1 ; + temp =dwc_dphy_read (isys, phy_id, addr); + temp &= ~(mask << shift); + temp |= (data & mask) << shift; +dwc_dphy_write (isys, phy_id, addr, temp); +} + +static u32 __maybe_unused dwc_dphy_read_mask (struct ipu6_isys *isys,u32 phy_id, +u32 addr,u8 shift,u8 width) +{ +u32 val; + + val =dwc_dphy_read (isys, phy_id, addr) >> shift; +return val & ((1 << width) -1 ); +} + +#define DWC_DPHY_TIMEOUT (5 * USEC_PER_SEC) +static int dwc_dphy_ifc_read (struct ipu6_isys *isys,u32 phy_id,u32 addr, +u32 *val) +{ +struct device *dev = &isys->adev->auxdev.dev; +void __iomem *isys_base = isys->pdata->base; +void __iomem *base = isys_base +IPU6_DWC_DPHY_BASE (phy_id); +void __iomem *reg; +u32 completion; +int ret; + +dwc_dphy_write (isys, phy_id,IPU6_DWC_DPHY_TEST_IFC_REQ , +IFC_REQ (TEST_IFC_REQ_READ , addr,0 )); + reg = base +IPU6_DWC_DPHY_TEST_IFC_REQ_COMPLETION ; + ret =readl_poll_timeout (reg, completion, !(completion &BIT (0 )), +10 ,DWC_DPHY_TIMEOUT ); +if (ret) { +dev_err (dev,"DWC ifc request read timeout\n" ); +return ret; + } + + *val = completion >>8 &0xff ; + *val =FIELD_GET (GENMASK (15 ,8 ), completion); +dev_dbg (dev,"DWC ifc read 0x%x = 0x%x" , addr, *val); + +return 0 ; +} + +static int dwc_dphy_ifc_write (struct ipu6_isys *isys,u32 phy_id,u32 addr, +u32 data) +{ +struct device *dev = &isys->adev->auxdev.dev; +void __iomem *isys_base = isys->pdata->base; +void __iomem *base = isys_base +IPU6_DWC_DPHY_BASE (phy_id); +void __iomem *reg; +u32 completion; +int ret; + +dwc_dphy_write (isys, phy_id,IPU6_DWC_DPHY_TEST_IFC_REQ , +IFC_REQ (TEST_IFC_REQ_WRITE , addr, data)); + completion =readl (base +IPU6_DWC_DPHY_TEST_IFC_REQ_COMPLETION ); + reg = base +IPU6_DWC_DPHY_TEST_IFC_REQ_COMPLETION ; + ret =readl_poll_timeout (reg, completion, !(completion &BIT (0 )), +10 ,DWC_DPHY_TIMEOUT ); +if (ret) +dev_err (dev,"DWC ifc request write timeout\n" ); + +return ret; +} + +static void dwc_dphy_ifc_write_mask (struct ipu6_isys *isys,u32 phy_id, +u32 addr,u32 data,u8 shift,u8 width) +{ +u32 temp, mask; +int ret; + + ret =dwc_dphy_ifc_read (isys, phy_id, addr, &temp); +if (ret) +return ; + + mask = (1 << width) -1 ; + temp &= ~(mask << shift); + temp |= (data & mask) << shift; +dwc_dphy_ifc_write (isys, phy_id, addr, temp); +} + +static u32 dwc_dphy_ifc_read_mask (struct ipu6_isys *isys,u32 phy_id,u32 addr, +u8 shift,u8 width) +{ +int ret; +u32 val; + + ret =dwc_dphy_ifc_read (isys, phy_id, addr, &val); +if (ret) +return 0 ; + +return ((val >> shift) & ((1 << width) -1 )); +} + +static int dwc_dphy_pwr_up (struct ipu6_isys *isys,u32 phy_id) +{ +struct device *dev = &isys->adev->auxdev.dev; +u32 fsm_state; +int ret; + +dwc_dphy_write (isys, phy_id,IPU6_DWC_DPHY_RSTZ ,1 ); +usleep_range (10 ,20 ); +dwc_dphy_write (isys, phy_id,IPU6_DWC_DPHY_SHUTDOWNZ ,1 ); + + ret =read_poll_timeout (dwc_dphy_ifc_read_mask, fsm_state, + (fsm_state ==PHY_FSM_STATE_IDLE || + fsm_state ==PHY_FSM_STATE_ULP ), +100 ,DWC_DPHY_TIMEOUT , false, isys, + phy_id,0x1e ,0 ,4 ); + +if (ret) +dev_err (dev,"Dphy %d power up failed, state 0x%x" , phy_id, + fsm_state); + +return ret; +} + +struct dwc_dphy_freq_range { +u8 hsfreq; +u16 min; +u16 max; +u16 default_mbps; +u16 osc_freq_target; +}; + +#define DPHY_FREQ_RANGE_NUM (63) +#define DPHY_FREQ_RANGE_INVALID_INDEX (0xff) +static const struct dwc_dphy_freq_range freqranges[DPHY_FREQ_RANGE_NUM ] = { + {0x00 ,80 ,97 ,80 ,335 }, + {0x10 ,80 ,107 ,90 ,335 }, + {0x20 ,84 ,118 ,100 ,335 }, + {0x30 ,93 ,128 ,110 ,335 }, + {0x01 ,103 ,139 ,120 ,335 }, + {0x11 ,112 ,149 ,130 ,335 }, + {0x21 ,122 ,160 ,140 ,335 }, + {0x31 ,131 ,170 ,150 ,335 }, + {0x02 ,141 ,181 ,160 ,335 }, + {0x12 ,150 ,191 ,170 ,335 }, + {0x22 ,160 ,202 ,180 ,335 }, + {0x32 ,169 ,212 ,190 ,335 }, + {0x03 ,183 ,228 ,205 ,335 }, + {0x13 ,198 ,244 ,220 ,335 }, + {0x23 ,212 ,259 ,235 ,335 }, + {0x33 ,226 ,275 ,250 ,335 }, + {0x04 ,250 ,301 ,275 ,335 }, + {0x14 ,274 ,328 ,300 ,335 }, + {0x25 ,297 ,354 ,325 ,335 }, + {0x35 ,321 ,380 ,350 ,335 }, + {0x05 ,369 ,433 ,400 ,335 }, + {0x16 ,416 ,485 ,450 ,335 }, + {0x26 ,464 ,538 ,500 ,335 }, + {0x37 ,511 ,590 ,550 ,335 }, + {0x07 ,559 ,643 ,600 ,335 }, + {0x18 ,606 ,695 ,650 ,335 }, + {0x28 ,654 ,748 ,700 ,335 }, + {0x39 ,701 ,800 ,750 ,335 }, + {0x09 ,749 ,853 ,800 ,335 }, + {0x19 ,796 ,905 ,850 ,335 }, + {0x29 ,844 ,958 ,900 ,335 }, + {0x3a ,891 ,1010 ,950 ,335 }, + {0x0a ,939 ,1063 ,1000 ,335 }, + {0x1a ,986 ,1115 ,1050 ,335 }, + {0x2a ,1034 ,1168 ,1100 ,335 }, + {0x3b ,1081 ,1220 ,1150 ,335 }, + {0x0b ,1129 ,1273 ,1200 ,335 }, + {0x1b ,1176 ,1325 ,1250 ,335 }, + {0x2b ,1224 ,1378 ,1300 ,335 }, + {0x3c ,1271 ,1430 ,1350 ,335 }, + {0x0c ,1319 ,1483 ,1400 ,335 }, + {0x1c ,1366 ,1535 ,1450 ,335 }, + {0x2c ,1414 ,1588 ,1500 ,335 }, + {0x3d ,1461 ,1640 ,1550 ,208 }, + {0x0d ,1509 ,1693 ,1600 ,214 }, + {0x1d ,1556 ,1745 ,1650 ,221 }, + {0x2e ,1604 ,1798 ,1700 ,228 }, + {0x3e ,1651 ,1850 ,1750 ,234 }, + {0x0e ,1699 ,1903 ,1800 ,241 }, + {0x1e ,1746 ,1955 ,1850 ,248 }, + {0x2f ,1794 ,2008 ,1900 ,255 }, + {0x3f ,1841 ,2060 ,1950 ,261 }, + {0x0f ,1889 ,2113 ,2000 ,268 }, + {0x40 ,1936 ,2165 ,2050 ,275 }, + {0x41 ,1984 ,2218 ,2100 ,281 }, + {0x42 ,2031 ,2270 ,2150 ,288 }, + {0x43 ,2079 ,2323 ,2200 ,294 }, + {0x44 ,2126 ,2375 ,2250 ,302 }, + {0x45 ,2174 ,2428 ,2300 ,308 }, + {0x46 ,2221 ,2480 ,2350 ,315 }, + {0x47 ,2269 ,2500 ,2400 ,321 }, + {0x48 ,2316 ,2500 ,2450 ,328 }, + {0x49 ,2364 ,2500 ,2500 ,335 } +}; + +static u16 get_hsfreq_by_mbps (u32 mbps) +{ +unsigned int i =DPHY_FREQ_RANGE_NUM ; + +while (i--) { +if (freqranges[i].default_mbps == mbps || + (mbps >= freqranges[i].min && mbps <= freqranges[i].max)) +return i; + } + +return DPHY_FREQ_RANGE_INVALID_INDEX ; +} + +static int ipu6_isys_dwc_phy_config (struct ipu6_isys *isys, +u32 phy_id,u32 mbps) +{ +struct ipu6_bus_device *adev = isys->adev; +struct device *dev = &adev->auxdev.dev; +struct ipu6_device *isp = adev->isp; +u32 cfg_clk_freqrange; +u32 osc_freq_target; +u32 index; + +dev_dbg (dev,"config Dphy %u with %u mbps" , phy_id, mbps); + + index =get_hsfreq_by_mbps (mbps); +if (index ==DPHY_FREQ_RANGE_INVALID_INDEX ) { +dev_err (dev,"link freq not found for mbps %u" , mbps); +return -EINVAL ; + } + +dwc_dphy_write_mask (isys, phy_id,IPU6_DWC_DPHY_HSFREQRANGE , + freqranges[index].hsfreq,0 ,7 ); + +/* Force termination Calibration */ +if (isys->phy_termcal_val) { +dwc_dphy_ifc_write_mask (isys, phy_id,0x20a ,0x1 ,0 ,1 ); +dwc_dphy_ifc_write_mask (isys, phy_id,0x209 ,0x3 ,0 ,2 ); +dwc_dphy_ifc_write_mask (isys, phy_id,0x209 , + isys->phy_termcal_val,4 ,4 ); + } + +/* +* Enable override to configure the DDL target oscillation +* frequency on bit 0 of register 0xe4 +*/ +dwc_dphy_ifc_write_mask (isys, phy_id,0xe4 ,0x1 ,0 ,1 ); +/* +* configure registers 0xe2, 0xe3 with the +* appropriate DDL target oscillation frequency +* 0x1cc(460) +*/ + osc_freq_target = freqranges[index].osc_freq_target; +dwc_dphy_ifc_write_mask (isys, phy_id,0xe2 , + osc_freq_target &0xff ,0 ,8 ); +dwc_dphy_ifc_write_mask (isys, phy_id,0xe3 , + (osc_freq_target >>8 ) &0xf ,0 ,4 ); + +if (mbps <1500 ) { +/* deskew_polarity_rw, for < 1.5Gbps */ +dwc_dphy_ifc_write_mask (isys, phy_id,0x8 ,0x1 ,5 ,1 ); + } + +/* +* Set cfgclkfreqrange[5:0] = round[(Fcfg_clk(MHz)-17)*4] +* (38.4 - 17) * 4 = ~85 (0x55) +*/ + cfg_clk_freqrange = (isp->buttress.ref_clk -170 ) *4 /10 ; +dev_dbg (dev,"ref_clk = %u clk_freqrange = %u" , + isp->buttress.ref_clk, cfg_clk_freqrange); +dwc_dphy_write_mask (isys, phy_id,IPU6_DWC_DPHY_CFGCLKFREQRANGE , + cfg_clk_freqrange,0 ,8 ); + +dwc_dphy_write_mask (isys, phy_id,IPU6_DWC_DPHY_DFT_CTRL2 ,0x1 ,4 ,1 ); +dwc_dphy_write_mask (isys, phy_id,IPU6_DWC_DPHY_DFT_CTRL2 ,0x1 ,8 ,1 ); + +return 0 ; +} + +static void ipu6_isys_dwc_phy_aggr_setup (struct ipu6_isys *isys,u32 master, +u32 slave,u32 mbps) +{ +/* Config mastermacro */ +dwc_dphy_ifc_write_mask (isys, master,0x133 ,0x1 ,0 ,1 ); +dwc_dphy_ifc_write_mask (isys, slave,0x133 ,0x0 ,0 ,1 ); + +/* Config master PHY clk lane to drive long channel clk */ +dwc_dphy_ifc_write_mask (isys, master,0x307 ,0x1 ,2 ,1 ); +dwc_dphy_ifc_write_mask (isys, slave,0x307 ,0x0 ,2 ,1 ); + +/* Config both PHYs data lanes to get clk from long channel */ +dwc_dphy_ifc_write_mask (isys, master,0x508 ,0x1 ,5 ,1 ); +dwc_dphy_ifc_write_mask (isys, slave,0x508 ,0x1 ,5 ,1 ); +dwc_dphy_ifc_write_mask (isys, master,0x708 ,0x1 ,5 ,1 ); +dwc_dphy_ifc_write_mask (isys, slave,0x708 ,0x1 ,5 ,1 ); + +/* Config slave PHY clk lane to bypass long channel clk to DDR clk */ +dwc_dphy_ifc_write_mask (isys, master,0x308 ,0x0 ,3 ,1 ); +dwc_dphy_ifc_write_mask (isys, slave,0x308 ,0x1 ,3 ,1 ); + +/* Override slave PHY clk lane enable (DPHYRXCLK_CLL_demux module) */ +dwc_dphy_ifc_write_mask (isys, slave,0xe0 ,0x3 ,0 ,2 ); + +/* Override slave PHY DDR clk lane enable (DPHYHSRX_div124 module) */ +dwc_dphy_ifc_write_mask (isys, slave,0xe1 ,0x1 ,1 ,1 ); +dwc_dphy_ifc_write_mask (isys, slave,0x307 ,0x1 ,3 ,1 ); + +/* Turn off slave PHY LP-RX clk lane */ +dwc_dphy_ifc_write_mask (isys, slave,0x304 ,0x1 ,7 ,1 ); +dwc_dphy_ifc_write_mask (isys, slave,0x305 ,0xa ,0 ,5 ); +} + +#define PHY_E 4 +static int ipu6_isys_dwc_phy_powerup_ack (struct ipu6_isys *isys,u32 phy_id) +{ +struct device *dev = &isys->adev->auxdev.dev; +u32 rescal_done; +int ret; + + ret =dwc_dphy_pwr_up (isys, phy_id); +if (ret !=0 ) { +dev_err (dev,"Dphy %u power up failed(%d)" , phy_id, ret); +return ret; + } + +/* reset forcerxmode */ +dwc_dphy_write_mask (isys, phy_id,IPU6_DWC_DPHY_DFT_CTRL2 ,0 ,4 ,1 ); +dwc_dphy_write_mask (isys, phy_id,IPU6_DWC_DPHY_DFT_CTRL2 ,0 ,8 ,1 ); + +dev_dbg (dev,"Dphy %u is ready!" , phy_id); + +if (phy_id !=PHY_E || isys->phy_termcal_val) +return 0 ; + +usleep_range (100 ,200 ); + rescal_done =dwc_dphy_ifc_read_mask (isys, phy_id,0x221 ,7 ,1 ); +if (rescal_done) { + isys->phy_termcal_val =dwc_dphy_ifc_read_mask (isys, phy_id, +0x220 ,2 ,4 ); +dev_dbg (dev,"termcal done with value = %u" , + isys->phy_termcal_val); + } + +return 0 ; +} + +static void ipu6_isys_dwc_phy_reset (struct ipu6_isys *isys,u32 phy_id) +{ +dev_dbg (&isys->adev->auxdev.dev,"Reset phy %u" , phy_id); + +dwc_dphy_write (isys, phy_id,IPU6_DWC_DPHY_SHUTDOWNZ ,0 ); +dwc_dphy_write (isys, phy_id,IPU6_DWC_DPHY_RSTZ ,0 ); +dwc_dphy_write (isys, phy_id,IPU6_DWC_DPHY_TEST_IFC_ACCESS_MODE , +TEST_IFC_ACCESS_MODE_FSM ); +dwc_dphy_write (isys, phy_id,IPU6_DWC_DPHY_TEST_IFC_REQ , +TEST_IFC_REQ_RESET ); +} + +int ipu6_isys_dwc_phy_set_power (struct ipu6_isys *isys, +struct ipu6_isys_csi2_config *cfg, +const struct ipu6_isys_csi2_timing *timing, +bool on) +{ +struct device *dev = &isys->adev->auxdev.dev; +void __iomem *isys_base = isys->pdata->base; +u32 phy_id, primary, secondary; +u32 nlanes, port, mbps; +s64 link_freq; +int ret; + + port = cfg->port; + +if (!isys_base || port >= isys->pdata->ipdata->csi2.nports) { +dev_warn (dev,"invalid port ID %d\n" , port); +return -EINVAL ; + } + + nlanes = cfg->nlanes; +/* only port 0, 2 and 4 support 4 lanes */ +if (nlanes ==4 && port %2 ) { +dev_err (dev,"invalid csi-port %u with %u lanes\n" , port, + nlanes); +return -EINVAL ; + } + + link_freq =ipu6_isys_csi2_get_link_freq (&isys->csi2[port]); +if (link_freq <0 ) { +dev_err (dev,"get link freq failed(%lld).\n" , link_freq); +return link_freq; + } + + mbps =div_u64 (link_freq,500000 ); + + phy_id = port; + primary = port & ~1 ; + secondary = primary +1 ; +if (on) { +if (nlanes ==4 ) { +dev_dbg (dev,"config phy %u and %u in aggr mode\n" , + primary, secondary); + +ipu6_isys_dwc_phy_reset (isys, primary); +ipu6_isys_dwc_phy_reset (isys, secondary); +ipu6_isys_dwc_phy_aggr_setup (isys, primary, + secondary, mbps); + + ret =ipu6_isys_dwc_phy_config (isys, primary, mbps); +if (ret) +return ret; + ret =ipu6_isys_dwc_phy_config (isys, secondary, mbps); +if (ret) +return ret; + + ret =ipu6_isys_dwc_phy_powerup_ack (isys, primary); +if (ret) +return ret; + + ret =ipu6_isys_dwc_phy_powerup_ack (isys, secondary); +return ret; + } + +dev_dbg (dev,"config phy %u with %u lanes in non-aggr mode\n" , + phy_id, nlanes); + +ipu6_isys_dwc_phy_reset (isys, phy_id); + ret =ipu6_isys_dwc_phy_config (isys, phy_id, mbps); +if (ret) +return ret; + + ret =ipu6_isys_dwc_phy_powerup_ack (isys, phy_id); +return ret; + } + +if (nlanes ==4 ) { +dev_dbg (dev,"Power down phy %u and phy %u for port %u\n" , + primary, secondary, port); +ipu6_isys_dwc_phy_reset (isys, secondary); +ipu6_isys_dwc_phy_reset (isys, primary); + +return 0 ; + } + +dev_dbg (dev,"Powerdown phy %u with %u lanes\n" , phy_id, nlanes); + +ipu6_isys_dwc_phy_reset (isys, phy_id); + +return 0 ; +}
drivers/media/pci/intel/ipu6/ipu6-isys-jsl-phy.c new file +237 -0
237 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6-isys-mcd-phy.c new file +712 -0
712 lines not on this page · raw message
Signature · 1 line
2.42.0
PATCH v2 10/15 media: intel/ipu6: add input system driver
In reply to [email protected]From: Bingbu Cao <[email protected]>
Input system driver do basic isys hardware setup and irq handling and work with fwnode and v4l2 to register the ISYS v4l2 devices.
--- drivers/media/pci/intel/ipu6/ipu6-isys.c | 1345 ++++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-isys.h | 201 ++++ 2 files changed, 1546 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys.h
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +1345 -0
@@ -0,0 +1,1345 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <linux/bitfield.h> +#include <linux/delay.h> +#include <linux/dma-mapping.h> +#include <linux/firmware.h> +#include <linux/module.h> +#include <linux/pci.h> +#include <linux/pm_runtime.h> + +#include <media/v4l2-ctrls.h> +#include <media/v4l2-device.h> +#include <media/v4l2-event.h> +#include <media/v4l2-fwnode.h> +#include <media/v4l2-ioctl.h> + +#include "ipu6-cpd.h" +#include "ipu6-dma.h" +#include "ipu6-isys.h" +#include "ipu6-mmu.h" +#include "ipu6-platform-buttress-regs.h" +#include "ipu6-platform-isys-csi2-reg.h" +#include "ipu6-platform-regs.h" + +#define IPU6_BUTTRESS_FABIC_CONTROL 0x68 +#define GDA_ENABLE_IWAKE_INDEX 2 +#define GDA_IWAKE_THRESHOLD_INDEX 1 +#define GDA_IRQ_CRITICAL_THRESHOLD_INDEX 0 +#define GDA_MEMOPEN_THRESHOLD_INDEX 3 +#define DEFAULT_DID_RATIO 90 +#define DEFAULT_IWAKE_THRESHOLD 0x42 +#define DEFAULT_MEM_OPEN_TIME 10 +#define ONE_THOUSAND_MICROSECOND 1000 +/* One page is 2KB, 8 x 16 x 16 = 2048B = 2KB */ +#define ISF_DMA_TOP_GDA_PROFERTY_PAGE_SIZE 0x800 + +/* LTR & DID value are 10 bit at most */ +#define LTR_DID_VAL_MAX 1023 +#define LTR_DEFAULT_VALUE 0x70503C19 +#define FILL_TIME_DEFAULT_VALUE 0xFFF0783C +#define LTR_DID_PKGC_2R 20 +#define LTR_SCALE_DEFAULT 5 +#define LTR_SCALE_1024NS 2 +#define DID_SCALE_1US 2 +#define DID_SCALE_32US 3 +#define REG_PKGC_PMON_CFG 0xB00 + +#define VAL_PKGC_PMON_CFG_RESET 0x38 +#define VAL_PKGC_PMON_CFG_START 0x7 + +#define IS_PIXEL_BUFFER_PAGES 0x80 +/* +* when iwake mode is disabled, the critical threshold is statically set +* to 75% of the IS pixel buffer, criticalThreshold = (128 * 3) / 4 +*/ +#define CRITICAL_THRESHOLD_IWAKE_DISABLE (IS_PIXEL_BUFFER_PAGES * 3 / 4) + +union fabric_ctrl { +struct { +u16 ltr_val :10 ; +u16 ltr_scale :3 ; +u16 reserved :3 ; +u16 did_val :10 ; +u16 did_scale :3 ; +u16 reserved2 :1 ; +u16 keep_power_in_D0 :1 ; +u16 keep_power_override :1 ; + } bits; +u32 value; +}; + +enum ltr_did_type { +LTR_IWAKE_ON , +LTR_IWAKE_OFF , +LTR_ISYS_ON , +LTR_ISYS_OFF , +LTR_ENHANNCE_IWAKE , +LTR_TYPE_MAX +}; + +#define ISYS_PM_QOS_VALUE 300 + +static int isys_isr_one (struct ipu6_bus_device *adev); + +static int +isys_complete_ext_device_registration (struct ipu6_isys *isys, +struct v4l2_subdev *sd, +struct ipu6_isys_csi2_config *csi2) +{ +struct device *dev = &isys->adev->auxdev.dev; +unsigned int i; +int ret; + +for (i =0 ; i < sd->entity.num_pads; i++) { +if (sd->entity.pads[i].flags &MEDIA_PAD_FL_SOURCE ) +break ; + } + +if (i == sd->entity.num_pads) { +dev_warn (dev,"no src pad in external entity\n" ); + ret = -ENOENT ; + goto unregister_subdev; + } + + ret =media_create_pad_link (&sd->entity, i, + &isys->csi2[csi2->port].asd.sd.entity, +0 ,0 ); +if (ret) { +dev_warn (dev,"can't create link\n" ); + goto unregister_subdev; + } + + isys->csi2[csi2->port].nlanes = csi2->nlanes; + +return 0 ; + +unregister_subdev: +v4l2_device_unregister_subdev (sd); + +return ret; +} + +static void isys_stream_init (struct ipu6_isys *isys) +{ +u32 i; + +for (i =0 ; i <IPU6_ISYS_MAX_STREAMS ; i++) { +mutex_init (&isys->streams[i].mutex); +init_completion (&isys->streams[i].stream_open_completion); +init_completion (&isys->streams[i].stream_close_completion); +init_completion (&isys->streams[i].stream_start_completion); +init_completion (&isys->streams[i].stream_stop_completion); +INIT_LIST_HEAD (&isys->streams[i].queues); + isys->streams[i].isys = isys; + isys->streams[i].stream_handle = i; + isys->streams[i].vc =INVALID_VC_ID ; + } +} + +static void isys_csi2_unregister_subdevices (struct ipu6_isys *isys) +{ +const struct ipu6_isys_internal_csi2_pdata *csi2 = + &isys->pdata->ipdata->csi2; +unsigned int i; + +for (i =0 ; i < csi2->nports; i++) +ipu6_isys_csi2_cleanup (&isys->csi2[i]); +} + +static int isys_csi2_register_subdevices (struct ipu6_isys *isys) +{ +const struct ipu6_isys_internal_csi2_pdata *csi2_pdata = + &isys->pdata->ipdata->csi2; +struct device *dev = &isys->adev->auxdev.dev; +unsigned int i; +int ret; + + isys->csi2 =devm_kcalloc (dev, csi2_pdata->nports, +sizeof (*isys->csi2),GFP_KERNEL ); +if (!isys->csi2) +return -ENOMEM ; + +for (i =0 ; i < csi2_pdata->nports; i++) { + ret =ipu6_isys_csi2_init (&isys->csi2[i], isys, + isys->pdata->base + + csi2_pdata->offsets[i], i); +if (ret) + goto fail; + + isys->isr_csi2_bits |=IPU6_ISYS_UNISPART_IRQ_CSI2 (i); + } + +return 0 ; + +fail: +while (i--) +ipu6_isys_csi2_cleanup (&isys->csi2[i]); + +return ret; +} + +static int isys_csi2_create_media_links (struct ipu6_isys *isys) +{ +const struct ipu6_isys_internal_csi2_pdata *csi2_pdata = + &isys->pdata->ipdata->csi2; +struct device *dev = &isys->adev->auxdev.dev; +unsigned int i, j, k; +int ret; + +for (i =0 ; i < csi2_pdata->nports; i++) { +struct media_entity *sd = &isys->csi2[i].asd.sd.entity; + +for (j =0 ; j <NR_OF_VIDEO_DEVICE ; j++) { +struct media_entity *v = &isys->av[j].vdev.entity; +u32 flag =MEDIA_LNK_FL_DYNAMIC ; + +for (k =CSI2_PAD_SRC ; k <NR_OF_CSI2_PADS ; k++) { + ret =media_create_pad_link (sd, k, v,0 , flag); +if (ret) { +dev_err (dev,"CSI2 can't create link\n" ); +return ret; + } + } + } + } + +return 0 ; +} + +static void isys_unregister_video_devices (struct ipu6_isys *isys) +{ +unsigned int i; + +for (i =0 ; i <NR_OF_VIDEO_DEVICE ; i++) +ipu6_isys_video_cleanup (&isys->av[i]); +} + +static int isys_register_video_devices (struct ipu6_isys *isys) +{ +unsigned int i; +int ret; + +for (i =0 ; i <NR_OF_VIDEO_DEVICE ; i++) { +snprintf (isys->av[i].vdev.name,sizeof (isys->av[i].vdev.name), +IPU6_ISYS_ENTITY_PREFIX " ISYS Capture %u" , i); + isys->av[i].isys = isys; + isys->av[i].aq.buf_prepare = ipu6_isys_buf_prepare; + isys->av[i].aq.fill_frame_buf_set = + ipu6_isys_buf_to_fw_frame_buf_pin; + isys->av[i].aq.link_fmt_validate = ipu6_isys_link_fmt_validate; + isys->av[i].aq.vbq.buf_struct_size = +sizeof (struct ipu6_isys_video_buffer ); + isys->av[i].pfmt = &ipu6_isys_pfmts[0 ]; + + ret =ipu6_isys_video_init (&isys->av[i]); +if (ret) + goto fail; + } + +return 0 ; + +fail: +while (i--) +ipu6_isys_video_cleanup (&isys->av[i]); + +return ret; +} + +void isys_setup_hw (struct ipu6_isys *isys) +{ +void __iomem *base = isys->pdata->base; +const u8 *thd = isys->pdata->ipdata->hw_variant.cdc_fifo_threshold; +u32 irqs =0 ; +unsigned int i, nports; + + nports = isys->pdata->ipdata->csi2.nports; + +/* Enable irqs for all MIPI ports */ +for (i =0 ; i < nports; i++) + irqs |=IPU6_ISYS_UNISPART_IRQ_CSI2 (i); + +writel (irqs, base + isys->pdata->ipdata->csi2.ctrl0_irq_edge); +writel (irqs, base + isys->pdata->ipdata->csi2.ctrl0_irq_lnp); +writel (irqs, base + isys->pdata->ipdata->csi2.ctrl0_irq_mask); +writel (irqs, base + isys->pdata->ipdata->csi2.ctrl0_irq_enable); +writel (GENMASK (19 ,0 ), + base + isys->pdata->ipdata->csi2.ctrl0_irq_clear); + + irqs =ISYS_UNISPART_IRQS ; +writel (irqs, base +IPU6_REG_ISYS_UNISPART_IRQ_EDGE ); +writel (irqs, base +IPU6_REG_ISYS_UNISPART_IRQ_LEVEL_NOT_PULSE ); +writel (GENMASK (28 ,0 ), base +IPU6_REG_ISYS_UNISPART_IRQ_CLEAR ); +writel (irqs, base +IPU6_REG_ISYS_UNISPART_IRQ_MASK ); +writel (irqs, base +IPU6_REG_ISYS_UNISPART_IRQ_ENABLE ); + +writel (0 , base +IPU6_REG_ISYS_UNISPART_SW_IRQ_REG ); +writel (0 , base +IPU6_REG_ISYS_UNISPART_SW_IRQ_MUX_REG ); + +/* Write CDC FIFO threshold values for isys */ +for (i =0 ; i < isys->pdata->ipdata->hw_variant.cdc_fifos; i++) +writel (thd[i], base +IPU6_REG_ISYS_CDC_THRESHOLD (i)); +} + +static void ipu6_isys_csi2_isr (struct ipu6_isys_csi2 *csi2) +{ +struct ipu6_isys_stream *stream; +unsigned int i; +u32 status; +int source; + +ipu6_isys_register_errors (csi2); + + status =readl (csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_STATUS_OFFSET ); + +writel (status, csi2->base +CSI_PORT_REG_BASE_IRQ_CSI_SYNC + +CSI_PORT_REG_BASE_IRQ_CLEAR_OFFSET ); + + source = csi2->asd.source; +for (i =0 ; i <NR_OF_CSI2_VC ; i++) { +if (status &IPU_CSI_RX_IRQ_FS_VC (i)) { + stream =ipu6_isys_query_stream_by_source (csi2->isys, + source, i); +if (stream) { +ipu6_isys_csi2_sof_event_by_stream (stream); +ipu6_isys_put_stream (stream); + } + } + +if (status &IPU_CSI_RX_IRQ_FE_VC (i)) { + stream =ipu6_isys_query_stream_by_source (csi2->isys, + source, i); +if (stream) { +ipu6_isys_csi2_eof_event_by_stream (stream); +ipu6_isys_put_stream (stream); + } + } + } +} + +irqreturn_t isys_isr (struct ipu6_bus_device *adev) +{ +struct ipu6_isys *isys =ipu6_bus_get_drvdata (adev); +void __iomem *base = isys->pdata->base; +u32 status_sw, status_csi; +u32 ctrl0_status, ctrl0_clear; + +spin_lock (&isys->power_lock); +if (!isys->power) { +spin_unlock (&isys->power_lock); +return IRQ_NONE ; + } + + ctrl0_status = isys->pdata->ipdata->csi2.ctrl0_irq_status; + ctrl0_clear = isys->pdata->ipdata->csi2.ctrl0_irq_clear; + + status_csi =readl (isys->pdata->base + ctrl0_status); + status_sw =readl (isys->pdata->base + +IPU6_REG_ISYS_UNISPART_IRQ_STATUS ); + +writel (ISYS_UNISPART_IRQS & ~IPU6_ISYS_UNISPART_IRQ_SW , + base +IPU6_REG_ISYS_UNISPART_IRQ_MASK ); + +do { +writel (status_csi, isys->pdata->base + ctrl0_clear); + +writel (status_sw, isys->pdata->base + +IPU6_REG_ISYS_UNISPART_IRQ_CLEAR ); + +if (isys->isr_csi2_bits & status_csi) { +unsigned int i; + +for (i =0 ; i < isys->pdata->ipdata->csi2.nports; i++) { +/* irq from not enabled port */ +if (!isys->csi2[i].base) +continue ; +if (status_csi &IPU6_ISYS_UNISPART_IRQ_CSI2 (i)) +ipu6_isys_csi2_isr (&isys->csi2[i]); + } + } + +writel (0 , base +IPU6_REG_ISYS_UNISPART_SW_IRQ_REG ); + +if (!isys_isr_one (adev)) + status_sw =IPU6_ISYS_UNISPART_IRQ_SW ; +else + status_sw =0 ; + + status_csi =readl (isys->pdata->base + ctrl0_status); + status_sw |=readl (isys->pdata->base + +IPU6_REG_ISYS_UNISPART_IRQ_STATUS ); + }while ((status_csi & isys->isr_csi2_bits) || + (status_sw &IPU6_ISYS_UNISPART_IRQ_SW )); + +writel (ISYS_UNISPART_IRQS , base +IPU6_REG_ISYS_UNISPART_IRQ_MASK ); + +spin_unlock (&isys->power_lock); + +return IRQ_HANDLED ; +} + +static void get_lut_ltrdid (struct ipu6_isys *isys,struct ltr_did *pltr_did) +{ +struct isys_iwake_watermark *iwake_watermark = &isys->iwake_watermark; +struct ltr_did ltrdid_default; + + ltrdid_default.lut_ltr.value =LTR_DEFAULT_VALUE ; + ltrdid_default.lut_fill_time.value =FILL_TIME_DEFAULT_VALUE ; + +if (iwake_watermark->ltrdid.lut_ltr.value) + *pltr_did = iwake_watermark->ltrdid; +else + *pltr_did = ltrdid_default; +} + +static int set_iwake_register (struct ipu6_isys *isys,u32 index,u32 value) +{ +struct device *dev = &isys->adev->auxdev.dev; +u32 req_id = index; +u32 offset =0 ; +int ret; + + ret =ipu6_fw_isys_send_proxy_token (isys, req_id, index, offset, value); +if (ret) +dev_err (dev,"write %d failed %d" , index, ret); + +return ret; +} + +/* +* When input system is powered up and before enabling any new sensor capture, +* or after disabling any sensor capture the following values need to be set: +* LTR_value = LTR(usec) from calculation; +* LTR_scale = 2; +* DID_value = DID(usec) from calculation; +* DID_scale = 2; +* +* When input system is powered down, the LTR and DID values +* must be returned to the default values: +* LTR_value = 1023; +* LTR_scale = 5; +* DID_value = 1023; +* DID_scale = 2; +*/ +static void set_iwake_ltrdid (struct ipu6_isys *isys,u16 ltr,u16 did, +enum ltr_did_type use) +{ +struct device *dev = &isys->adev->auxdev.dev; +u16 ltr_val, ltr_scale =LTR_SCALE_1024NS ; +u16 did_val, did_scale =DID_SCALE_1US ; +struct ipu6_device *isp = isys->adev->isp; +union fabric_ctrl fc; + +switch (use) { +case LTR_IWAKE_ON : + ltr_val =min_t (u16, ltr, (u16 )LTR_DID_VAL_MAX ); + did_val =min_t (u16, did, (u16 )LTR_DID_VAL_MAX ); + ltr_scale = (ltr ==LTR_DID_VAL_MAX && + did ==LTR_DID_VAL_MAX ) ? +LTR_SCALE_DEFAULT :LTR_SCALE_1024NS ; +break ; +case LTR_ISYS_ON : +case LTR_IWAKE_OFF : + ltr_val =LTR_DID_PKGC_2R ; + did_val =LTR_DID_PKGC_2R ; +break ; +case LTR_ISYS_OFF : + ltr_val =LTR_DID_VAL_MAX ; + did_val =LTR_DID_VAL_MAX ; + ltr_scale =LTR_SCALE_DEFAULT ; +break ; +case LTR_ENHANNCE_IWAKE : +if (ltr ==LTR_DID_VAL_MAX && did ==LTR_DID_VAL_MAX ) { + ltr_val =LTR_DID_VAL_MAX ; + did_val =LTR_DID_VAL_MAX ; + ltr_scale =LTR_SCALE_DEFAULT ; + }else if (did <ONE_THOUSAND_MICROSECOND ) { + ltr_val = ltr; + did_val = did; + }else { + ltr_val = ltr; +/* div 90% value by 32 to account for scale change */ + did_val = did /32 ; + did_scale =DID_SCALE_32US ; + } +break ; +default : + ltr_val =LTR_DID_VAL_MAX ; + did_val =LTR_DID_VAL_MAX ; + ltr_scale =LTR_SCALE_DEFAULT ; +break ; + } + + fc.value =readl (isp->base +IPU6_BUTTRESS_FABIC_CONTROL ); + fc.bits.ltr_val = ltr_val; + fc.bits.ltr_scale = ltr_scale; + fc.bits.did_val = did_val; + fc.bits.did_scale = did_scale; + +dev_dbg (dev,"ltr: value %u scale %u, did: value %u scale %u\n" , + ltr_val, ltr_scale, did_val, did_scale); +writel (fc.value, isp->base +IPU6_BUTTRESS_FABIC_CONTROL ); +} + +/* +* Driver may clear register GDA_ENABLE_IWAKE before FW configures the +* stream for debug purpose. Otherwise driver should not access this register. +*/ +static void enable_iwake (struct ipu6_isys *isys,bool enable) +{ +struct isys_iwake_watermark *iwake_watermark = &isys->iwake_watermark; +int ret; + +mutex_lock (&iwake_watermark->mutex); + +if (iwake_watermark->iwake_enabled == enable) { +mutex_unlock (&iwake_watermark->mutex); +return ; + } + + ret =set_iwake_register (isys,GDA_ENABLE_IWAKE_INDEX , enable); +if (!ret) + iwake_watermark->iwake_enabled = enable; + +mutex_unlock (&iwake_watermark->mutex); +} + +void update_watermark_setting (struct ipu6_isys *isys) +{ +struct isys_iwake_watermark *iwake_watermark = &isys->iwake_watermark; +u32 iwake_threshold, iwake_critical_threshold, page_num; +struct device *dev = &isys->adev->auxdev.dev; +u32 calc_fill_time_us =0 , ltr =0 , did =0 ; +struct video_stream_watermark *p_watermark; +enum ltr_did_type ltr_did_type; +struct list_head *stream_node; +u64 isys_pb_datarate_mbs =0 ; +u32 mem_open_threshold =0 ; +struct ltr_did ltrdid; +u64 threshold_bytes; +u32 max_sram_size; +u32 shift; + + shift = isys->pdata->ipdata->sram_gran_shift; + max_sram_size = isys->pdata->ipdata->max_sram_size; + +mutex_lock (&iwake_watermark->mutex); +if (iwake_watermark->force_iwake_disable) { +set_iwake_ltrdid (isys,0 ,0 ,LTR_IWAKE_OFF ); +set_iwake_register (isys,GDA_IRQ_CRITICAL_THRESHOLD_INDEX , +CRITICAL_THRESHOLD_IWAKE_DISABLE ); + goto unlock_exit; + } + +if (list_empty (&iwake_watermark->video_list)) { + isys_pb_datarate_mbs =0 ; + }else { +list_for_each (stream_node, &iwake_watermark->video_list) { + p_watermark =list_entry (stream_node, + struct video_stream_watermark, + stream_node); + isys_pb_datarate_mbs += p_watermark->stream_data_rate; + } + } +mutex_unlock (&iwake_watermark->mutex); + +if (!isys_pb_datarate_mbs) { +enable_iwake (isys, false); +set_iwake_ltrdid (isys,0 ,0 ,LTR_IWAKE_OFF ); +mutex_lock (&iwake_watermark->mutex); +set_iwake_register (isys,GDA_IRQ_CRITICAL_THRESHOLD_INDEX , +CRITICAL_THRESHOLD_IWAKE_DISABLE ); + goto unlock_exit; + } + +enable_iwake (isys, true); + calc_fill_time_us = max_sram_size / isys_pb_datarate_mbs; + +if (isys->pdata->ipdata->enhanced_iwake) { + ltr = isys->pdata->ipdata->ltr; + did = calc_fill_time_us *DEFAULT_DID_RATIO /100 ; + ltr_did_type =LTR_ENHANNCE_IWAKE ; + }else { +get_lut_ltrdid (isys, <rdid); + +if (calc_fill_time_us <= ltrdid.lut_fill_time.bits.th0) + ltr =0 ; +else if (calc_fill_time_us <= ltrdid.lut_fill_time.bits.th1) + ltr = ltrdid.lut_ltr.bits.val0; +else if (calc_fill_time_us <= ltrdid.lut_fill_time.bits.th2) + ltr = ltrdid.lut_ltr.bits.val1; +else if (calc_fill_time_us <= ltrdid.lut_fill_time.bits.th3) + ltr = ltrdid.lut_ltr.bits.val2; +else + ltr = ltrdid.lut_ltr.bits.val3; + + did = calc_fill_time_us - ltr; + ltr_did_type =LTR_IWAKE_ON ; + } + +set_iwake_ltrdid (isys, ltr, did, ltr_did_type); + +/* calculate iwake threshold with 2KB granularity pages */ + threshold_bytes = did * isys_pb_datarate_mbs; + iwake_threshold =max_t (u32,1 , threshold_bytes >> shift); + iwake_threshold =min_t (u32, iwake_threshold, max_sram_size); + +mutex_lock (&iwake_watermark->mutex); +if (isys->pdata->ipdata->enhanced_iwake) { +set_iwake_register (isys,GDA_IWAKE_THRESHOLD_INDEX , +DEFAULT_IWAKE_THRESHOLD ); +/* calculate number of pages that will be filled in 10 usec */ + page_num = (DEFAULT_MEM_OPEN_TIME * isys_pb_datarate_mbs) / +ISF_DMA_TOP_GDA_PROFERTY_PAGE_SIZE ; + page_num += ((DEFAULT_MEM_OPEN_TIME * isys_pb_datarate_mbs) % +ISF_DMA_TOP_GDA_PROFERTY_PAGE_SIZE ) ?1 :0 ; + mem_open_threshold = isys->pdata->ipdata->memopen_threshold; + mem_open_threshold =max_t (u32, mem_open_threshold, page_num); +dev_dbg (dev,"mem_open_threshold: %u\n" , mem_open_threshold); +set_iwake_register (isys,GDA_MEMOPEN_THRESHOLD_INDEX , + mem_open_threshold); + }else { +set_iwake_register (isys,GDA_IWAKE_THRESHOLD_INDEX , + iwake_threshold); + } + + iwake_critical_threshold = iwake_threshold + + (IS_PIXEL_BUFFER_PAGES - iwake_threshold) /2 ; + +dev_dbg (dev,"threshold: %u critical: %u\n" , iwake_threshold, + iwake_critical_threshold); + +set_iwake_register (isys,GDA_IRQ_CRITICAL_THRESHOLD_INDEX , + iwake_critical_threshold); + +writel (VAL_PKGC_PMON_CFG_RESET , + isys->adev->isp->base +REG_PKGC_PMON_CFG ); +writel (VAL_PKGC_PMON_CFG_START , + isys->adev->isp->base +REG_PKGC_PMON_CFG ); +unlock_exit: +mutex_unlock (&iwake_watermark->mutex); +} + +static void isys_iwake_watermark_init (struct ipu6_isys *isys) +{ +struct isys_iwake_watermark *iwake_watermark = &isys->iwake_watermark; + +INIT_LIST_HEAD (&iwake_watermark->video_list); +mutex_init (&iwake_watermark->mutex); + + iwake_watermark->ltrdid.lut_ltr.value =0 ; + iwake_watermark->isys = isys; + iwake_watermark->iwake_enabled = false; + iwake_watermark->force_iwake_disable = false; +} + +static void isys_iwake_watermark_cleanup (struct ipu6_isys *isys) +{ +struct isys_iwake_watermark *iwake_watermark = &isys->iwake_watermark; + +mutex_lock (&iwake_watermark->mutex); +list_del (&iwake_watermark->video_list); +mutex_unlock (&iwake_watermark->mutex); + +mutex_destroy (&iwake_watermark->mutex); +} + +/* The .bound() notifier callback when a match is found */ +static int isys_notifier_bound (struct v4l2_async_notifier *notifier, +struct v4l2_subdev *sd, +struct v4l2_async_connection *asc) +{ +struct ipu6_isys *isys = + container_of(notifier, struct ipu6_isys, notifier); +struct sensor_async_sd *s_asd = + container_of(asc, struct sensor_async_sd, asc); +int ret; + +dev_dbg (&isys->adev->auxdev.dev,"bind %s nlanes is %d port is %d\n" , + sd->name, s_asd->csi2.nlanes, s_asd->csi2.port); + ret =isys_complete_ext_device_registration (isys, sd, &s_asd->csi2); +if (ret) +return ret; + +return v4l2_device_register_subdev_nodes (&isys->v4l2_dev); +} + +static int isys_notifier_complete (struct v4l2_async_notifier *notifier) +{ +struct ipu6_isys *isys = + container_of(notifier, struct ipu6_isys, notifier); + +return v4l2_device_register_subdev_nodes (&isys->v4l2_dev); +} + +static const struct v4l2_async_notifier_operations isys_async_ops = { + .bound = isys_notifier_bound, + .complete = isys_notifier_complete, +}; + +#define ISYS_MAX_PORTS 8 +static int isys_notifier_init (struct ipu6_isys *isys) +{ +struct ipu6_device *isp = isys->adev->isp; +struct device *dev = &isp->pdev->dev; +unsigned int i; +int ret; + +v4l2_async_nf_init (&isys->notifier, &isys->v4l2_dev); + +for (i =0 ; i <ISYS_MAX_PORTS ; i++) { +struct v4l2_fwnode_endpoint vep = { + .bus_type =V4L2_MBUS_CSI2_DPHY + }; +struct sensor_async_sd *s_asd; +struct fwnode_handle *ep; + + ep =fwnode_graph_get_endpoint_by_id (dev_fwnode (dev), i,0 , +FWNODE_GRAPH_ENDPOINT_NEXT ); +if (!ep) +continue ; + + ret =v4l2_fwnode_endpoint_parse (ep, &vep); +if (ret) { +dev_err (dev,"fwnode endpoint parse failed: %d\n" , ret); + goto err_parse; + } + + s_asd =v4l2_async_nf_add_fwnode_remote (&isys->notifier, ep, + struct sensor_async_sd); +if (IS_ERR (s_asd)) { + ret =PTR_ERR (s_asd); +dev_err (dev,"add remove fwnode failed: %d\n" , ret); + goto err_parse; + } + + s_asd->csi2.port = vep.base.port; + s_asd->csi2.nlanes = vep.bus.mipi_csi2.num_data_lanes; + +dev_dbg (dev,"remote endpoint port %d with %d lanes added\n" , + s_asd->csi2.port, s_asd->csi2.nlanes); + +fwnode_handle_put (ep); + +continue ; + +err_parse: +fwnode_handle_put (ep); +return ret; + } + + isys->notifier.ops = &isys_async_ops; + ret =v4l2_async_nf_register (&isys->notifier); +if (ret) { +dev_err (dev,"failed to register async notifier : %d\n" , ret); +v4l2_async_nf_cleanup (&isys->notifier); + } + +return ret; +} + +static void isys_notifier_cleanup (struct ipu6_isys *isys) +{ +v4l2_async_nf_unregister (&isys->notifier); +v4l2_async_nf_cleanup (&isys->notifier); +} + +static int isys_register_devices (struct ipu6_isys *isys) +{ +struct device *dev = &isys->adev->auxdev.dev; +struct pci_dev *pdev = isys->adev->isp->pdev; +int ret; + + isys->media_dev.dev = dev; +media_device_pci_init (&isys->media_dev, + pdev,IPU6_MEDIA_DEV_MODEL_NAME ); + +strscpy (isys->v4l2_dev.name, isys->media_dev.model, +sizeof (isys->v4l2_dev.name)); + + ret =media_device_register (&isys->media_dev); +if (ret <0 ) + goto out_media_device_unregister; + + isys->v4l2_dev.mdev = &isys->media_dev; + isys->v4l2_dev.ctrl_handler =NULL ; + + ret =v4l2_device_register (dev->parent, &isys->v4l2_dev); +if (ret <0 ) + goto out_media_device_unregister; + + ret =isys_register_video_devices (isys); +if (ret) + goto out_v4l2_device_unregister; + + ret =isys_csi2_register_subdevices (isys); +if (ret) + goto out_isys_unregister_video_device; + + ret =isys_csi2_create_media_links (isys); +if (ret) + goto out_isys_unregister_subdevices; + + ret =isys_notifier_init (isys); +if (ret) + goto out_isys_unregister_subdevices; + +return 0 ; + +out_isys_unregister_subdevices: +isys_csi2_unregister_subdevices (isys); + +out_isys_unregister_video_device: +isys_unregister_video_devices (isys); + +out_v4l2_device_unregister: +v4l2_device_unregister (&isys->v4l2_dev); + +out_media_device_unregister: +media_device_unregister (&isys->media_dev); +media_device_cleanup (&isys->media_dev); + +dev_err (dev,"failed to register isys devices\n" ); + +return ret; +} + +static void isys_unregister_devices (struct ipu6_isys *isys) +{ +isys_unregister_video_devices (isys); +isys_csi2_unregister_subdevices (isys); +v4l2_device_unregister (&isys->v4l2_dev); +media_device_unregister (&isys->media_dev); +media_device_cleanup (&isys->media_dev); +} + +static int isys_runtime_pm_resume (struct device *dev) +{ +struct ipu6_bus_device *adev =to_ipu6_bus_device (dev); +struct ipu6_isys *isys =ipu6_bus_get_drvdata (adev); +struct ipu6_device *isp = adev->isp; +unsigned long flags; +int ret; + +if (!isys) +return 0 ; + + ret =ipu6_mmu_hw_init (adev->mmu); +if (ret) +return ret; + +cpu_latency_qos_update_request (&isys->pm_qos,ISYS_PM_QOS_VALUE ); + + ret =ipu6_buttress_start_tsc_sync (isp); +if (ret) +return ret; + +spin_lock_irqsave (&isys->power_lock, flags); + isys->power =1 ; +spin_unlock_irqrestore (&isys->power_lock, flags); + +isys_setup_hw (isys); + +set_iwake_ltrdid (isys,0 ,0 ,LTR_ISYS_ON ); + +return 0 ; +} + +static int isys_runtime_pm_suspend (struct device *dev) +{ +struct ipu6_bus_device *adev =to_ipu6_bus_device (dev); +struct ipu6_isys *isys; +unsigned long flags; + + isys =dev_get_drvdata (dev); +if (!isys) +return 0 ; + +spin_lock_irqsave (&isys->power_lock, flags); + isys->power =0 ; +spin_unlock_irqrestore (&isys->power_lock, flags); + +mutex_lock (&isys->mutex); + isys->need_reset = false; +mutex_unlock (&isys->mutex); + + isys->phy_termcal_val =0 ; +cpu_latency_qos_update_request (&isys->pm_qos,PM_QOS_DEFAULT_VALUE ); + +set_iwake_ltrdid (isys,0 ,0 ,LTR_ISYS_OFF ); + +ipu6_mmu_hw_cleanup (adev->mmu); + +return 0 ; +} + +static int isys_suspend (struct device *dev) +{ +struct ipu6_isys *isys =dev_get_drvdata (dev); + +/* If stream is open, refuse to suspend */ +if (isys->stream_opened) +return -EBUSY ; + +return 0 ; +} + +static int isys_resume (struct device *dev) +{ +return 0 ; +} + +static const struct dev_pm_ops isys_pm_ops = { + .runtime_suspend = isys_runtime_pm_suspend, + .runtime_resume = isys_runtime_pm_resume, + .suspend = isys_suspend, + .resume = isys_resume, +}; + +static void isys_remove (struct auxiliary_device *auxdev) +{ +struct ipu6_bus_device *adev =auxdev_to_adev (auxdev); +struct ipu6_isys *isys =dev_get_drvdata (&auxdev->dev); +struct ipu6_device *isp = adev->isp; +struct isys_fw_msgs *fwmsg, *safe; +unsigned int i; + +for (i =0 ; i <IPU6_ISYS_MAX_STREAMS ; i++) +mutex_destroy (&isys->streams[i].mutex); + +list_for_each_entry_safe (fwmsg, safe, &isys->framebuflist, head) +dma_free_attrs (&auxdev->dev,sizeof (struct isys_fw_msgs ), + fwmsg, fwmsg->dma_addr,0 ); + +list_for_each_entry_safe (fwmsg, safe, &isys->framebuflist_fw, head) +dma_free_attrs (&auxdev->dev,sizeof (struct isys_fw_msgs ), + fwmsg, fwmsg->dma_addr,0 ); + +isys_iwake_watermark_cleanup (isys); +isys_notifier_cleanup (isys); +isys_unregister_devices (isys); + +cpu_latency_qos_remove_request (&isys->pm_qos); + +if (!isp->secure_mode) { +ipu6_cpd_free_pkg_dir (adev); +ipu6_buttress_unmap_fw_image (adev, &adev->fw_sgt); +release_firmware (adev->fw); + } + +mutex_destroy (&isys->stream_mutex); +mutex_destroy (&isys->mutex); +} + +static int alloc_fw_msg_bufs (struct ipu6_isys *isys,int amount) +{ +struct device *dev = &isys->adev->auxdev.dev; +struct isys_fw_msgs *addr; +dma_addr_t dma_addr; +unsigned long flags; +unsigned int i; + +for (i =0 ; i < amount; i++) { + addr =dma_alloc_attrs (dev,sizeof (struct isys_fw_msgs ), + &dma_addr,GFP_KERNEL ,0 ); +if (!addr) +break ; + addr->dma_addr = dma_addr; + +spin_lock_irqsave (&isys->listlock, flags); +list_add (&addr->head, &isys->framebuflist); +spin_unlock_irqrestore (&isys->listlock, flags); + } + +if (i == amount) +return 0 ; + +spin_lock_irqsave (&isys->listlock, flags); +while (!list_empty (&isys->framebuflist)) { + addr =list_first_entry (&isys->framebuflist, + struct isys_fw_msgs, head); +list_del (&addr->head); +spin_unlock_irqrestore (&isys->listlock, flags); +dma_free_attrs (dev,sizeof (struct isys_fw_msgs ), addr, + addr->dma_addr,0 ); +spin_lock_irqsave (&isys->listlock, flags); + } +spin_unlock_irqrestore (&isys->listlock, flags); + +return -ENOMEM ; +} + +struct isys_fw_msgs *ipu6_get_fw_msg_buf (struct ipu6_isys_stream *stream) +{ +struct ipu6_isys *isys = stream->isys; +struct device *dev = &isys->adev->auxdev.dev; +struct isys_fw_msgs *msg; +unsigned long flags; +int ret; + +spin_lock_irqsave (&isys->listlock, flags); +if (list_empty (&isys->framebuflist)) { +spin_unlock_irqrestore (&isys->listlock, flags); +dev_dbg (dev,"Frame list empty\n" ); + + ret =alloc_fw_msg_bufs (isys,5 ); +if (ret <0 ) +return NULL ; + +spin_lock_irqsave (&isys->listlock, flags); +if (list_empty (&isys->framebuflist)) { +spin_unlock_irqrestore (&isys->listlock, flags); +dev_err (dev,"Frame list empty\n" ); +return NULL ; + } + } + msg = list_last_entry(&isys->framebuflist, struct isys_fw_msgs, head); +list_move (&msg->head, &isys->framebuflist_fw); +spin_unlock_irqrestore (&isys->listlock, flags); +memset (&msg->fw_msg,0 ,sizeof (msg->fw_msg)); + +return msg; +} + +void ipu6_cleanup_fw_msg_bufs (struct ipu6_isys *isys) +{ +struct isys_fw_msgs *fwmsg, *fwmsg0; +unsigned long flags; + +spin_lock_irqsave (&isys->listlock, flags); +list_for_each_entry_safe (fwmsg, fwmsg0, &isys->framebuflist_fw, head) +list_move (&fwmsg->head, &isys->framebuflist); +spin_unlock_irqrestore (&isys->listlock, flags); +} + +void ipu6_put_fw_msg_buf (struct ipu6_isys *isys,u64 data) +{ +struct isys_fw_msgs *msg; +unsigned long flags; +u64 *ptr = (u64 *)data; + +if (!ptr) +return ; + +spin_lock_irqsave (&isys->listlock, flags); + msg = container_of(ptr, struct isys_fw_msgs, fw_msg.dummy); +list_move (&msg->head, &isys->framebuflist); +spin_unlock_irqrestore (&isys->listlock, flags); +} + +static int isys_probe (struct auxiliary_device *auxdev, +const struct auxiliary_device_id *auxdev_id) +{ +struct ipu6_bus_device *adev =auxdev_to_adev (auxdev); +struct ipu6_device *isp = adev->isp; +const struct firmware *fw; +struct ipu6_isys *isys; +unsigned int i; +int ret; + +if (!isp->bus_ready_to_probe) +return -EPROBE_DEFER ; + + isys =devm_kzalloc (&auxdev->dev,sizeof (*isys),GFP_KERNEL ); +if (!isys) +return -ENOMEM ; + + ret =ipu6_mmu_hw_init (adev->mmu); +if (ret) +return ret; + + adev->auxdrv_data = + (const struct ipu6_auxdrv_data *)auxdev_id->driver_data; + adev->auxdrv =to_auxiliary_drv (auxdev->dev.driver); + isys->adev = adev; + isys->pdata = adev->pdata; + +/* initial sensor type */ + isys->sensor_type = isys->pdata->ipdata->sensor_type_start; + +spin_lock_init (&isys->streams_lock); +spin_lock_init (&isys->power_lock); + isys->power =0 ; + isys->phy_termcal_val =0 ; + +mutex_init (&isys->mutex); +mutex_init (&isys->stream_mutex); + +spin_lock_init (&isys->listlock); +INIT_LIST_HEAD (&isys->framebuflist); +INIT_LIST_HEAD (&isys->framebuflist_fw); + + isys->line_align =IPU6_ISYS_2600_MEM_LINE_ALIGN ; + isys->icache_prefetch =0 ; + +dev_set_drvdata (&auxdev->dev, isys); + +isys_stream_init (isys); + +if (!isp->secure_mode) { + fw = isp->cpd_fw; + ret =ipu6_buttress_map_fw_image (adev, fw, &adev->fw_sgt); +if (ret) + goto release_firmware; + + ret =ipu6_cpd_create_pkg_dir (adev, isp->cpd_fw->data); +if (ret) + goto remove_shared_buffer; + } + +cpu_latency_qos_add_request (&isys->pm_qos,PM_QOS_DEFAULT_VALUE ); + + ret =alloc_fw_msg_bufs (isys,20 ); +if (ret <0 ) + goto out_remove_pkg_dir_shared_buffer; + +isys_iwake_watermark_init (isys); + +if (is_ipu6se (adev->isp->hw_ver)) + isys->phy_set_power = ipu6_isys_jsl_phy_set_power; +else if (is_ipu6ep_mtl (adev->isp->hw_ver)) + isys->phy_set_power = ipu6_isys_dwc_phy_set_power; +else + isys->phy_set_power = ipu6_isys_mcd_phy_set_power; + + ret =isys_register_devices (isys); +if (ret) + goto out_remove_pkg_dir_shared_buffer; + +ipu6_mmu_hw_cleanup (adev->mmu); + +return 0 ; + +out_remove_pkg_dir_shared_buffer: +if (!isp->secure_mode) +ipu6_cpd_free_pkg_dir (adev); +remove_shared_buffer: +if (!isp->secure_mode) +ipu6_buttress_unmap_fw_image (adev, &adev->fw_sgt); +release_firmware: +if (!isp->secure_mode) +release_firmware (adev->fw); + +for (i =0 ; i <IPU6_ISYS_MAX_STREAMS ; i++) +mutex_destroy (&isys->streams[i].mutex); + +mutex_destroy (&isys->mutex); +mutex_destroy (&isys->stream_mutex); + +ipu6_mmu_hw_cleanup (adev->mmu); + +return ret; +} + +struct fwmsg { +int type; +char *msg; +bool valid_ts; +}; + +static const struct fwmsg fw_msg[] = { + {IPU6_FW_ISYS_RESP_TYPE_STREAM_OPEN_DONE ,"STREAM_OPEN_DONE" ,0 }, + {IPU6_FW_ISYS_RESP_TYPE_STREAM_CLOSE_ACK ,"STREAM_CLOSE_ACK" ,0 }, + {IPU6_FW_ISYS_RESP_TYPE_STREAM_START_ACK ,"STREAM_START_ACK" ,0 }, + {IPU6_FW_ISYS_RESP_TYPE_STREAM_START_AND_CAPTURE_ACK , +"STREAM_START_AND_CAPTURE_ACK" ,0 }, + {IPU6_FW_ISYS_RESP_TYPE_STREAM_STOP_ACK ,"STREAM_STOP_ACK" ,0 }, + {IPU6_FW_ISYS_RESP_TYPE_STREAM_FLUSH_ACK ,"STREAM_FLUSH_ACK" ,0 }, + {IPU6_FW_ISYS_RESP_TYPE_PIN_DATA_READY ,"PIN_DATA_READY" ,1 }, + {IPU6_FW_ISYS_RESP_TYPE_STREAM_CAPTURE_ACK ,"STREAM_CAPTURE_ACK" ,0 }, + {IPU6_FW_ISYS_RESP_TYPE_STREAM_START_AND_CAPTURE_DONE , +"STREAM_START_AND_CAPTURE_DONE" ,1 }, + {IPU6_FW_ISYS_RESP_TYPE_STREAM_CAPTURE_DONE ,"STREAM_CAPTURE_DONE" ,1 }, + {IPU6_FW_ISYS_RESP_TYPE_FRAME_SOF ,"FRAME_SOF" ,1 }, + {IPU6_FW_ISYS_RESP_TYPE_FRAME_EOF ,"FRAME_EOF" ,1 }, + {IPU6_FW_ISYS_RESP_TYPE_STATS_DATA_READY ,"STATS_READY" ,1 }, + {-1 ,"UNKNOWN MESSAGE" ,0 } +}; + +static int resp_type_to_index (int type) +{ +unsigned int i; + +for (i =0 ; i <ARRAY_SIZE (fw_msg); i++) +if (fw_msg[i].type == type) +return i; + +return ARRAY_SIZE (fw_msg) -1 ; +} + +static int isys_isr_one (struct ipu6_bus_device *adev) +{ +struct ipu6_isys *isys =ipu6_bus_get_drvdata (adev); +struct ipu6_fw_isys_resp_info_abi *resp; +struct ipu6_isys_stream *stream; +struct ipu6_isys_csi2 *csi2 =NULL ; +u64 ts; + +if (!isys->fwcom) +return 1 ; + + resp =ipu6_fw_isys_get_resp (isys->fwcom,IPU6_BASE_MSG_RECV_QUEUES ); +if (!resp) +return 1 ; + + ts = (u64 )resp->timestamp[1 ] <<32 | resp->timestamp[0 ]; + +if (resp->error_info.error ==IPU6_FW_ISYS_ERROR_STREAM_IN_SUSPENSION ) +/* Suspension is kind of special case: not enough buffers */ +dev_dbg (&adev->auxdev.dev, +"FW error resp %02d %s, stream %u, error SUSPENSION, details %d, timestamp 0x%16.16llx, pin %d\n" , + resp->type, + fw_msg[resp_type_to_index (resp->type)].msg, + resp->stream_handle, + resp->error_info.error_details, + fw_msg[resp_type_to_index (resp->type)].valid_ts ? + ts :0 , resp->pin_id); +else if (resp->error_info.error) +dev_dbg (&adev->auxdev.dev, +"FW error resp %02d %s, stream %u, error %d, details %d, timestamp 0x%16.16llx, pin %d\n" , + resp->type, + fw_msg[resp_type_to_index (resp->type)].msg, + resp->stream_handle, + resp->error_info.error, resp->error_info.error_details, + fw_msg[resp_type_to_index (resp->type)].valid_ts ? + ts :0 , resp->pin_id); +else +dev_dbg (&adev->auxdev.dev, +"FW resp %02d %s, stream %u, timestamp 0x%16.16llx, pin %d\n" , + resp->type, + fw_msg[resp_type_to_index (resp->type)].msg, + resp->stream_handle, + fw_msg[resp_type_to_index (resp->type)].valid_ts ? + ts :0 , resp->pin_id); + +if (resp->stream_handle >=IPU6_ISYS_MAX_STREAMS ) { +dev_err (&adev->auxdev.dev,"bad stream handle %u\n" , + resp->stream_handle); + goto leave; + } + + stream =ipu6_isys_query_stream_by_handle (isys, resp->stream_handle); +if (!stream) { +dev_err (&adev->auxdev.dev,"stream of stream_handle %u is unused\n" , + resp->stream_handle); + goto leave; + } + stream->error = resp->error_info.error; + + csi2 =ipu6_isys_subdev_to_csi2 (stream->asd); + +switch (resp->type) { +case IPU6_FW_ISYS_RESP_TYPE_STREAM_OPEN_DONE : +complete (&stream->stream_open_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_CLOSE_ACK : +complete (&stream->stream_close_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_START_ACK : +complete (&stream->stream_start_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_START_AND_CAPTURE_ACK : +complete (&stream->stream_start_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_STOP_ACK : +complete (&stream->stream_stop_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_FLUSH_ACK : +complete (&stream->stream_stop_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_PIN_DATA_READY : +/* +* firmware only release the capture msg until software +* get pin_data_ready event +*/ +ipu6_put_fw_msg_buf (ipu6_bus_get_drvdata (adev), resp->buf_id); +if (resp->pin_id <IPU6_ISYS_OUTPUT_PINS && + stream->output_pins[resp->pin_id].pin_ready) + stream->output_pins[resp->pin_id].pin_ready (stream, + resp); +else +dev_warn (&adev->auxdev.dev, +"%d:No data pin ready handler for pin id %d\n" , + resp->stream_handle, resp->pin_id); +if (csi2) +ipu6_isys_csi2_error (csi2); + +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_CAPTURE_ACK : +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_START_AND_CAPTURE_DONE : +case IPU6_FW_ISYS_RESP_TYPE_STREAM_CAPTURE_DONE : +break ; +case IPU6_FW_ISYS_RESP_TYPE_FRAME_SOF : + +ipu6_isys_csi2_sof_event_by_stream (stream); + stream->seq[stream->seq_index].sequence = +atomic_read (&stream->sequence) -1 ; + stream->seq[stream->seq_index].timestamp = ts; +dev_dbg (&adev->auxdev.dev, +"sof: handle %d: (index %u), timestamp 0x%16.16llx\n" , + resp->stream_handle, + stream->seq[stream->seq_index].sequence, ts); + stream->seq_index = (stream->seq_index +1 ) + %IPU6_ISYS_MAX_PARALLEL_SOF ; +break ; +case IPU6_FW_ISYS_RESP_TYPE_FRAME_EOF : +ipu6_isys_csi2_eof_event_by_stream (stream); +dev_dbg (&adev->auxdev.dev, +"eof: handle %d: (index %u), timestamp 0x%16.16llx\n" , + resp->stream_handle, + stream->seq[stream->seq_index].sequence, ts); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STATS_DATA_READY : +break ; +default : +dev_err (&adev->auxdev.dev,"%d:unknown response type %u\n" , + resp->stream_handle, resp->type); +break ; + } + +ipu6_isys_put_stream (stream); +leave: +ipu6_fw_isys_put_resp (isys->fwcom,IPU6_BASE_MSG_RECV_QUEUES ); +return 0 ; +} + +static const struct ipu6_auxdrv_data ipu6_isys_auxdrv_data = { + .isr = isys_isr, + .isr_threaded =NULL , + .wake_isr_thread = false, +}; + +static const struct auxiliary_device_id ipu6_isys_id_table[] = { + { + .name ="intel_ipu6.isys" , + .driver_data = (kernel_ulong_t)&ipu6_isys_auxdrv_data, + }, + { } +}; +MODULE_DEVICE_TABLE (auxiliary, ipu6_isys_id_table); + +static struct auxiliary_driver isys_driver = { + .name =IPU6_ISYS_NAME , + .probe = isys_probe, + .remove = isys_remove, + .id_table = ipu6_isys_id_table, + .driver = { + .pm = &isys_pm_ops, + }, +}; + +module_auxiliary_driver (isys_driver); + +MODULE_AUTHOR ("Sakari Ailus <[email protected]>" ); +MODULE_AUTHOR ("Tianshu Qiu <[email protected]>" ); +MODULE_AUTHOR ("Bingbu Cao <[email protected]>" ); +MODULE_AUTHOR ("Yunliang Ding <[email protected]>" ); +MODULE_AUTHOR ("Hongju Wang <[email protected]>" ); +MODULE_LICENSE ("GPL" ); +MODULE_DESCRIPTION ("Intel IPU6 input system driver" ); +MODULE_IMPORT_NS (INTEL_IPU6 );
drivers/media/pci/intel/ipu6/ipu6-isys.h new file +201 -0
201 lines not on this page · raw message
Signature · 1 line
2.42.0
Re: [PATCH v2 10/15] media: intel/ipu6: add input system driver
Hi Bingbu,
For some reason my mail client decided to add double > > below. Hope it is not too confusing.
Quoted 16 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
Quoted 16 lines
From: Bingbu Cao <[email protected]>
Input system driver do basic isys hardware setup and irq handling and work with fwnode and v4l2 to register the ISYS v4l2 devices.
Signed-off-by Bingbu Cao <[email protected]>Signed-off-by Hans de Goede <[email protected]>Reported-by Claus Stovgaard <[email protected]>---
drivers/media/pci/intel/ipu6/ipu6-isys.c | 1345 >
++++++++++++++++++++++
drivers/media/pci/intel/ipu6/ipu6-isys.h | 201 ++++2 files changed, 1546 insertions(+)create mode 100644
drivers/media/pci/intel/ipu6/ipu6-isys.ccreate mode 100644drivers/media/pci/intel/ipu6/ipu6-isys.h
...
Quoted 8 lines
Quoted 8 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +7
-0+static int isys_register_video_devices (struct ipu6_isys *isys) +{ + unsigned int i; + int ret; + + for (i = 0; i < NR_OF_VIDEO_DEVICE; i++) { + snprintf(isys->av[i].vdev.name, > sizeof(isys-Quoted 1 line
av[i].vdev.name),
Quoted 8 lines
Quoted 8 lines
+
IPU6_ISYS_ENTITY_PREFIX" ISYS Capture %u", > i);
+ isys->av[i].isys = isys;
+ isys->av[i].aq.buf_prepare =ipu6_isys_buf_prepare; + isys->av[i].aq.fill_frame_buf_set =
+ ipu6_isys_buf_to_fw_frame_buf_pin; + isys->av[i].aq.link_fmt_validate = >ipu6_isys_link_fmt_validate;
These 3 function pointers are always set to the same value, is there any point in that, or could they be called directly?
Quoted 17 lines
Quoted 17 lines
+ isys->av[i].aq.vbq.buf_struct_size = + sizeof(struct
ipu6_isys_video_buffer); + isys->av[i].pfmt = &ipu6_isys_pfmts[0];drivers/media/pci/intel/ipu6/ipu6-isys.c new file +14
-0+ + ret = ipu6_isys_video_init(&isys->av[i]); + if (ret) + goto fail; + } + + return 0; + +fail: + while (i--) + ipu6_isys_video_cleanup(&isys->av[i]); + + return ret; +}
...
Quoted 34 lines
Quoted 34 lines
+ if (
resp->error_info.error== >IPU6_FW_ISYS_ERROR_STREAM_IN_SUSPENSION)
+ /* Suspension is kind of special case: not enough > buffers */
+ dev_dbg(&adev->auxdev.dev,
+ "FW error resp %02d %s, stream %u, error > SUSPENSION, details %d, timestamp 0x%16.16llx, pin %d\n", + resp->type,
+ fw_msg[resp_type_to_index(resp->type)].msg, + resp->stream_handle,
+ resp->error_info.error_details, + fw_msg[resp_type_to_index(resp-Quoted 1 line
type)].valid_> ts ?
+ ts : 0,
resp->pin_id);
+ else if (resp->error_info.error)
+ dev_dbg(&adev->auxdev.dev,
+ "FW error resp %02d %s, stream %u, error %d, > details %d, timestamp 0x%16.16llx, pin %d\n",
+ resp->type,
+ fw_msg[resp_type_to_index(resp->type)].msg, + resp->stream_handle,
+ resp->error_info.error, > resp-Quoted 1 line
error_info.error_details,
+ fw_msg[resp_type_to_index(resp-
Quoted 1 line
type)].valid_> ts ?
+ ts : 0,
resp->pin_id);
+ else
+ dev_dbg(&adev->auxdev.dev,
+ "FW resp %02d %s, stream %u, timestamp > 0x%16.16llx, pin %d\n",
+ resp->type,
+ fw_msg[resp_type_to_index(resp->type)].msg, + resp->stream_handle,
+ fw_msg[resp_type_to_index(resp-
Quoted 2 lines
Quoted 2 lines
Quoted 1 line
type)].valid_> ts ?
+ ts : 0,
resp->pin_id);
fw_msg[resp_type_to_index(resp->type)] is duplicated 6 times above. It would be a lot cleaner if you did the lookup once before the if/else if/else block.
Quoted 1 line
/Andreas
Re: [PATCH v2 10/15] media: intel/ipu6: add input system driver
In reply to [email protected]Hi Bingbu,
Quoted 16 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Input system driver do basic isys hardware setup and irq handling and work with fwnode and v4l2 to register the ISYS v4l2 devices.
Signed-off-by Bingbu Cao <[email protected]>Signed-off-by Hans de Goede <[email protected]>Reported-by Claus Stovgaard <[email protected]>--- drivers/media/pci/intel/ipu6/ipu6-isys.c | 1345 ++++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-isys.h | 201 ++++ 2 files changed, 1546 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys.hdrivers/media/pci/intel/ipu6/ipu6-isys.c +0
-0
...
Quoted 74 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c +74
-0+static void isys_notifier_cleanup (struct ipu6_isys *isys) +{ +v4l2_async_nf_unregister (&isys->notifier); +v4l2_async_nf_cleanup (&isys->notifier); +} + +static int isys_register_devices (struct ipu6_isys *isys) +{ +struct device *dev = &isys->adev->auxdev.dev; +struct pci_dev *pdev = isys->adev->isp->pdev; +int ret; + + isys->media_dev.dev = dev; +media_device_pci_init (&isys->media_dev, + pdev,IPU6_MEDIA_DEV_MODEL_NAME ); + +strscpy (isys->v4l2_dev.name, isys->media_dev.model, +sizeof (isys->v4l2_dev.name)); + + ret =media_device_register (&isys->media_dev); +if (ret <0 ) + goto out_media_device_unregister; + + isys->v4l2_dev.mdev = &isys->media_dev; + isys->v4l2_dev.ctrl_handler =NULL ; + + ret =v4l2_device_register (dev->parent, &isys->v4l2_dev); +if (ret <0 ) + goto out_media_device_unregister; + + ret =isys_register_video_devices (isys); +if (ret) + goto out_v4l2_device_unregister; + + ret =isys_csi2_register_subdevices (isys); +if (ret) + goto out_isys_unregister_video_device; + + ret =isys_csi2_create_media_links (isys); +if (ret) + goto out_isys_unregister_subdevices; + + ret =isys_notifier_init (isys); +if (ret) + goto out_isys_unregister_subdevices; + +return 0 ; + +out_isys_unregister_subdevices: +isys_csi2_unregister_subdevices (isys); + +out_isys_unregister_video_device: +isys_unregister_video_devices (isys); + +out_v4l2_device_unregister: +v4l2_device_unregister (&isys->v4l2_dev); + +out_media_device_unregister: +media_device_unregister (&isys->media_dev); +media_device_cleanup (&isys->media_dev); + +dev_err (dev,"failed to register isys devices\n" ); + +return ret; +} + +static void isys_unregister_devices (struct ipu6_isys *isys) +{ +isys_unregister_video_devices (isys); +isys_csi2_unregister_subdevices (isys); +v4l2_device_unregister (&isys->v4l2_dev); +media_device_unregister (&isys->media_dev); +media_device_cleanup (&isys->media_dev); +}
...
drivers/media/pci/intel/ipu6/ipu6-isys.c line 912
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c +10
-0+static void isys_remove (struct auxiliary_device *auxdev) +{ +struct ipu6_bus_device *adev =auxdev_to_adev (auxdev); +struct ipu6_isys *isys =dev_get_drvdata (&auxdev->dev); +struct ipu6_device *isp = adev->isp; +struct isys_fw_msgs *fwmsg, *safe; +unsigned int i; + +for (i =0 ; i <IPU6_ISYS_MAX_STREAMS ; i++) +mutex_destroy (&isys->streams[i].mutex);
In my testing with IPU4, I had to move these mutex_destroy's to the end of isys_remove. If we're streaming, they are needed, presumably until isys_unregister_devices is called.
drivers/media/pci/intel/ipu6/ipu6-isys.c line 924
Quoted 12 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c +12
-0+ +list_for_each_entry_safe (fwmsg, safe, &isys->framebuflist, head) +dma_free_attrs (&auxdev->dev,sizeof (struct isys_fw_msgs ), + fwmsg, fwmsg->dma_addr,0 ); + +list_for_each_entry_safe (fwmsg, safe, &isys->framebuflist_fw, head) +dma_free_attrs (&auxdev->dev,sizeof (struct isys_fw_msgs ), + fwmsg, fwmsg->dma_addr,0 ); + +isys_iwake_watermark_cleanup (isys); +isys_notifier_cleanup (isys); +isys_unregister_devices (isys);
Again in IPU4 testing:
I get crashing in stop_streaming when unbinding during streaming. If I move isys_unregister_devices before isys_notifier_cleanup, this no longer happens.
drivers/media/pci/intel/ipu6/ipu6-isys.c line 936
Quoted 12 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c +12
-0+ +cpu_latency_qos_remove_request (&isys->pm_qos); + +if (!isp->secure_mode) { +ipu6_cpd_free_pkg_dir (adev); +ipu6_buttress_unmap_fw_image (adev, &adev->fw_sgt); +release_firmware (adev->fw); + } + +mutex_destroy (&isys->stream_mutex); +mutex_destroy (&isys->mutex); +}
/Andreas
Re: [PATCH v2 10/15] media: intel/ipu6: add input system driver
In reply to [email protected]Hi Bingbu,
Quoted 20 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Input system driver do basic isys hardware setup and irq handling and work with fwnode and v4l2 to register the ISYS v4l2 devices.
Signed-off-by Bingbu Cao <[email protected]>Signed-off-by Hans de Goede <[email protected]>Reported-by Claus Stovgaard <[email protected]>--- drivers/media/pci/intel/ipu6/ipu6-isys.c | 1345 ++++++++++++++++++++++ drivers/media/pci/intel/ipu6/ipu6-isys.h | 201 ++++ 2 files changed, 1546 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys.hdrivers/media/pci/intel/ipu6/ipu6-isys.c new file +0
-0
...
Quoted 44 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +44
-0+struct isys_fw_msgs *ipu6_get_fw_msg_buf (struct ipu6_isys_stream *stream) +{ +struct ipu6_isys *isys = stream->isys; +struct device *dev = &isys->adev->auxdev.dev; +struct isys_fw_msgs *msg; +unsigned long flags; +int ret; + +spin_lock_irqsave (&isys->listlock, flags); +if (list_empty (&isys->framebuflist)) { +spin_unlock_irqrestore (&isys->listlock, flags); +dev_dbg (dev,"Frame list empty\n" ); + + ret =alloc_fw_msg_bufs (isys,5 ); +if (ret <0 ) +return NULL ; + +spin_lock_irqsave (&isys->listlock, flags); +if (list_empty (&isys->framebuflist)) { +spin_unlock_irqrestore (&isys->listlock, flags); +dev_err (dev,"Frame list empty\n" ); +return NULL ; + } + } + msg = list_last_entry(&isys->framebuflist, struct isys_fw_msgs, head); +list_move (&msg->head, &isys->framebuflist_fw); +spin_unlock_irqrestore (&isys->listlock, flags); +memset (&msg->fw_msg,0 ,sizeof (msg->fw_msg)); + +return msg; +} + +void ipu6_cleanup_fw_msg_bufs (struct ipu6_isys *isys) +{ +struct isys_fw_msgs *fwmsg, *fwmsg0; +unsigned long flags; + +spin_lock_irqsave (&isys->listlock, flags); +list_for_each_entry_safe (fwmsg, fwmsg0, &isys->framebuflist_fw, head) +list_move (&fwmsg->head, &isys->framebuflist); +spin_unlock_irqrestore (&isys->listlock, flags); +} + +void ipu6_put_fw_msg_buf (struct ipu6_isys *isys,u64 data)
drivers/media/pci/intel/ipu6/ipu6-isys.c line 1029
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +10
-0+{ +struct isys_fw_msgs *msg; +unsigned long flags; +u64 *ptr = (u64 *)data; + +if (!ptr) +return ; + +spin_lock_irqsave (&isys->listlock, flags); + msg = container_of(ptr, struct isys_fw_msgs, fw_msg.dummy);
The dummy field is only there to do this container_of, but ptr and msg point to the same address, since the union dummy is in is the first field. So the container_of is exactly the same as simply casting the pointer.
Quoted 3 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +3
-0+list_move (&msg->head, &isys->framebuflist); +spin_unlock_irqrestore (&isys->listlock, flags); +}
...
Quoted 76 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +76
-0+static int isys_isr_one (struct ipu6_bus_device *adev) +{ +struct ipu6_isys *isys =ipu6_bus_get_drvdata (adev); +struct ipu6_fw_isys_resp_info_abi *resp; +struct ipu6_isys_stream *stream; +struct ipu6_isys_csi2 *csi2 =NULL ; +u64 ts; + +if (!isys->fwcom) +return 1 ; + + resp =ipu6_fw_isys_get_resp (isys->fwcom,IPU6_BASE_MSG_RECV_QUEUES ); +if (!resp) +return 1 ; + + ts = (u64 )resp->timestamp[1 ] <<32 | resp->timestamp[0 ]; + +if (resp->error_info.error ==IPU6_FW_ISYS_ERROR_STREAM_IN_SUSPENSION ) +/* Suspension is kind of special case: not enough buffers */ +dev_dbg (&adev->auxdev.dev, +"FW error resp %02d %s, stream %u, error SUSPENSION, details %d, timestamp 0x%16.16llx, pin %d\n" , + resp->type, + fw_msg[resp_type_to_index (resp->type)].msg, + resp->stream_handle, + resp->error_info.error_details, + fw_msg[resp_type_to_index (resp->type)].valid_ts ? + ts :0 , resp->pin_id); +else if (resp->error_info.error) +dev_dbg (&adev->auxdev.dev, +"FW error resp %02d %s, stream %u, error %d, details %d, timestamp 0x%16.16llx, pin %d\n" , + resp->type, + fw_msg[resp_type_to_index (resp->type)].msg, + resp->stream_handle, + resp->error_info.error, resp->error_info.error_details, + fw_msg[resp_type_to_index (resp->type)].valid_ts ? + ts :0 , resp->pin_id); +else +dev_dbg (&adev->auxdev.dev, +"FW resp %02d %s, stream %u, timestamp 0x%16.16llx, pin %d\n" , + resp->type, + fw_msg[resp_type_to_index (resp->type)].msg, + resp->stream_handle, + fw_msg[resp_type_to_index (resp->type)].valid_ts ? + ts :0 , resp->pin_id); + +if (resp->stream_handle >=IPU6_ISYS_MAX_STREAMS ) { +dev_err (&adev->auxdev.dev,"bad stream handle %u\n" , + resp->stream_handle); + goto leave; + } + + stream =ipu6_isys_query_stream_by_handle (isys, resp->stream_handle); +if (!stream) { +dev_err (&adev->auxdev.dev,"stream of stream_handle %u is unused\n" , + resp->stream_handle); + goto leave; + } + stream->error = resp->error_info.error; + + csi2 =ipu6_isys_subdev_to_csi2 (stream->asd); + +switch (resp->type) { +case IPU6_FW_ISYS_RESP_TYPE_STREAM_OPEN_DONE : +complete (&stream->stream_open_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_CLOSE_ACK : +complete (&stream->stream_close_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_START_ACK : +complete (&stream->stream_start_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_START_AND_CAPTURE_ACK : +complete (&stream->stream_start_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_STOP_ACK : +complete (&stream->stream_stop_completion);
drivers/media/pci/intel/ipu6/ipu6-isys.c line 1259
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +10
-0+break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_FLUSH_ACK : +complete (&stream->stream_stop_completion); +break ; +case IPU6_FW_ISYS_RESP_TYPE_PIN_DATA_READY : +/* +* firmware only release the capture msg until software +* get pin_data_ready event +*/ +ipu6_put_fw_msg_buf (ipu6_bus_get_drvdata (adev), resp->buf_id);
Is there a difference between IPU4 and IPU6 here? buf_id is always 0 on IPU4, so this doesn't release anything.
The PIN_DATA_READY response is sent multiple times per call to ipu6_get_fw_msg_buf, so if I'm not mistaken, this can only be correct if the firmware only sends the buf_id in one of the PIN_DATA_READY responses.
The correct pointer to release seems to be passed in the various _ACK responses on IPU4. I think it would make a lot more sense to release the buffer there. But of course, if there is a bug in the firmware we might have to do something else.
Quoted 50 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +50
-0+if (resp->pin_id <IPU6_ISYS_OUTPUT_PINS && + stream->output_pins[resp->pin_id].pin_ready) + stream->output_pins[resp->pin_id].pin_ready (stream, + resp); +else +dev_warn (&adev->auxdev.dev, +"%d:No data pin ready handler for pin id %d\n" , + resp->stream_handle, resp->pin_id); +if (csi2) +ipu6_isys_csi2_error (csi2); + +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_CAPTURE_ACK : +break ; +case IPU6_FW_ISYS_RESP_TYPE_STREAM_START_AND_CAPTURE_DONE : +case IPU6_FW_ISYS_RESP_TYPE_STREAM_CAPTURE_DONE : +break ; +case IPU6_FW_ISYS_RESP_TYPE_FRAME_SOF : + +ipu6_isys_csi2_sof_event_by_stream (stream); + stream->seq[stream->seq_index].sequence = +atomic_read (&stream->sequence) -1 ; + stream->seq[stream->seq_index].timestamp = ts; +dev_dbg (&adev->auxdev.dev, +"sof: handle %d: (index %u), timestamp 0x%16.16llx\n" , + resp->stream_handle, + stream->seq[stream->seq_index].sequence, ts); + stream->seq_index = (stream->seq_index +1 ) + %IPU6_ISYS_MAX_PARALLEL_SOF ; +break ; +case IPU6_FW_ISYS_RESP_TYPE_FRAME_EOF : +ipu6_isys_csi2_eof_event_by_stream (stream); +dev_dbg (&adev->auxdev.dev, +"eof: handle %d: (index %u), timestamp 0x%16.16llx\n" , + resp->stream_handle, + stream->seq[stream->seq_index].sequence, ts); +break ; +case IPU6_FW_ISYS_RESP_TYPE_STATS_DATA_READY : +break ; +default : +dev_err (&adev->auxdev.dev,"%d:unknown response type %u\n" , + resp->stream_handle, resp->type); +break ; + } + +ipu6_isys_put_stream (stream); +leave: +ipu6_fw_isys_put_resp (isys->fwcom,IPU6_BASE_MSG_RECV_QUEUES ); +return 0 ; +}
...
Quoted 1 line
+++
b/drivers/media/pci/intel/ipu6/ipu6-isys.h
...
drivers/media/pci/intel/ipu6/ipu6-isys.c
Quoted 12 lines
drivers/media/pci/intel/ipu6/ipu6-isys.c new file +12
-0+struct isys_fw_msgs { +union { +u64 dummy; +struct ipu6_fw_isys_frame_buff_set_abi frame; +struct ipu6_fw_isys_stream_cfg_data_abi stream; + } fw_msg; +struct list_head head; +dma_addr_t dma_addr; +}; + +struct isys_fw_msgs *ipu6_get_fw_msg_buf (struct ipu6_isys_stream *stream); +void ipu6_put_fw_msg_buf (struct ipu6_isys *isys,u64 data);
I it would be easier to understand these functions if the ipu6_put_fw_msg_buf function, took the same type that the get function returns.
/Andreas
PATCH v2 11/15 media: intel/ipu6: input system video capture nodes
In reply to [email protected]From: Bingbu Cao <[email protected]>
Register v4l2 video device and setup the vb2 queue to support basic video capture. Video streaming callback will trigger the input system driver to construct a input system stream configuration for firmware based on data type and stream ID and then queue buffers to firmware to do capture.
--- .../media/pci/intel/ipu6/ipu6-isys-queue.c | 846 +++++++++++ .../media/pci/intel/ipu6/ipu6-isys-queue.h | 94 ++ .../media/pci/intel/ipu6/ipu6-isys-video.c | 1239 +++++++++++++++++ .../media/pci/intel/ipu6/ipu6-isys-video.h | 129 ++ 4 files changed, 2308 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-queue.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-queue.h create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-video.c create mode 100644 drivers/media/pci/intel/ipu6/ipu6-isys-video.h
drivers/media/pci/intel/ipu6/ipu6-isys-queue.c new file +846 -0
@@ -0,0 +1,846 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* +* Copyright (C) 2013 - 2023 Intel Corporation +*/ + +#include <media/videobuf2-dma-contig.h> + +#include "ipu6-isys.h" + +static int queue_setup (struct vb2_queue *q,unsigned int *num_buffers, +unsigned int *num_planes,unsigned int sizes[], +struct device *alloc_devs[]) +{ +struct ipu6_isys_queue *aq =vb2_queue_to_isys_queue (q); +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct device *dev = &av->isys->adev->auxdev.dev; +bool use_fmt = false; +unsigned int i; +u32 size; + +/* num_planes == 0: we're being called through VIDIOC_REQBUFS */ +if (!*num_planes) { + use_fmt = true; + *num_planes = av->mpix.num_planes; + } + +for (i =0 ; i < *num_planes; i++) { + size = av->mpix.plane_fmt[i].sizeimage; +if (use_fmt) { + sizes[i] = size; + }else if (sizes[i] < size) { +dev_err (dev,"%s: queue setup: plane %d size %u < %u\n" , + av->vdev.name, i, sizes[i], size); +return -EINVAL ; + } + + alloc_devs[i] = aq->dev; + } + +return 0 ; +} + +int ipu6_isys_buf_prepare (struct vb2_buffer *vb) +{ +struct ipu6_isys_queue *aq =vb2_queue_to_isys_queue (vb->vb2_queue); +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct device *dev = &av->isys->adev->auxdev.dev; + +dev_dbg (dev,"buffer: %s: configured size %u, buffer size %lu\n" , + av->vdev.name, av->mpix.plane_fmt[0 ].sizeimage, +vb2_plane_size (vb,0 )); + +if (av->mpix.plane_fmt[0 ].sizeimage >vb2_plane_size (vb,0 )) +return -EINVAL ; + +vb2_set_plane_payload (vb,0 , av->mpix.plane_fmt[0 ].bytesperline * + av->mpix.height); + vb->planes[0 ].data_offset =0 ; + +return 0 ; +} + +static int buf_prepare (struct vb2_buffer *vb) +{ +struct ipu6_isys_queue *aq =vb2_queue_to_isys_queue (vb->vb2_queue); + +return aq->buf_prepare (vb); +} + +/* +* Queue a buffer list back to incoming or active queues. The buffers +* are removed from the buffer list. +*/ +void ipu6_isys_buffer_list_queue (struct ipu6_isys_buffer_list *bl, +unsigned long op_flags, +enum vb2_buffer_state state) +{ +struct ipu6_isys_buffer *ib, *ib_safe; +unsigned long flags; +bool first = true; + +if (!bl) +return ; + +WARN_ON_ONCE (!bl->nbufs); +WARN_ON_ONCE (op_flags &IPU6_ISYS_BUFFER_LIST_FL_ACTIVE && + op_flags &IPU6_ISYS_BUFFER_LIST_FL_INCOMING ); + +list_for_each_entry_safe (ib, ib_safe, &bl->head, head) { +struct ipu6_isys_video *av; +struct vb2_buffer *vb =ipu6_isys_buffer_to_vb2_buffer (ib); +struct ipu6_isys_queue *aq = +vb2_queue_to_isys_queue (vb->vb2_queue); +struct device *dev; + +if (WARN_ON_ONCE (ib->type !=IPU6_ISYS_VIDEO_BUFFER )) +continue ; + + av =ipu6_isys_queue_to_video (aq); + dev = &av->isys->adev->auxdev.dev; +spin_lock_irqsave (&aq->lock, flags); +list_del (&ib->head); +if (op_flags &IPU6_ISYS_BUFFER_LIST_FL_ACTIVE ) +list_add (&ib->head, &aq->active); +else if (op_flags &IPU6_ISYS_BUFFER_LIST_FL_INCOMING ) +list_add_tail (&ib->head, &aq->incoming); +spin_unlock_irqrestore (&aq->lock, flags); + +if (op_flags &IPU6_ISYS_BUFFER_LIST_FL_SET_STATE ) +vb2_buffer_done (vb, state); + +if (first) { +dev_dbg (dev, +"queue buf list %p flags %lx, s %d, %d bufs\n" , + bl, op_flags, state, bl->nbufs); + first = false; + } + + bl->nbufs--; + } + +WARN_ON (bl->nbufs); +} + +/* +* flush_firmware_streamon_fail() - Flush in cases where requests may +* have been queued to firmware and the *firmware streamon fails for a +* reason or another. +*/ +static void flush_firmware_streamon_fail (struct ipu6_isys_stream *stream) +{ +struct device *dev = &stream->isys->adev->auxdev.dev; +struct ipu6_isys_queue *aq; +unsigned long flags; + +lockdep_assert_held (&stream->mutex); + +list_for_each_entry (aq, &stream->queues, node) { +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct ipu6_isys_buffer *ib, *ib_safe; + +spin_lock_irqsave (&aq->lock, flags); +list_for_each_entry_safe (ib, ib_safe, &aq->active, head) { +struct vb2_buffer *vb = +ipu6_isys_buffer_to_vb2_buffer (ib); + +list_del (&ib->head); +if (av->streaming) { +dev_dbg (dev, +"%s: queue buffer %u back to incoming\n" , + av->vdev.name, vb->index); +/* Queue already streaming, return to driver. */ +list_add (&ib->head, &aq->incoming); +continue ; + } +/* Queue not yet streaming, return to user. */ +dev_dbg (dev,"%s: return %u back to videobuf2\n" , + av->vdev.name, vb->index); +vb2_buffer_done (ipu6_isys_buffer_to_vb2_buffer (ib), +VB2_BUF_STATE_QUEUED ); + } +spin_unlock_irqrestore (&aq->lock, flags); + } +} + +/* +* Attempt obtaining a buffer list from the incoming queues, a list of buffers +* that contains one entry from each video buffer queue. If a buffer can't be +* obtained from every queue, the buffers are returned back to the queue. +*/ +static int buffer_list_get (struct ipu6_isys_stream *stream, +struct ipu6_isys_buffer_list *bl) +{ +struct device *dev = &stream->isys->adev->auxdev.dev; +struct ipu6_isys_queue *aq; +unsigned long flags; +unsigned long buf_flag =IPU6_ISYS_BUFFER_LIST_FL_INCOMING ; + + bl->nbufs =0 ; +INIT_LIST_HEAD (&bl->head); + +list_for_each_entry (aq, &stream->queues, node) { +struct ipu6_isys_buffer *ib; + +spin_lock_irqsave (&aq->lock, flags); +if (list_empty (&aq->incoming)) { +spin_unlock_irqrestore (&aq->lock, flags); +if (!list_empty (&bl->head)) +ipu6_isys_buffer_list_queue (bl, buf_flag,0 ); +return -ENODATA ; + } + + ib =list_last_entry (&aq->incoming, + struct ipu6_isys_buffer, head); + +dev_dbg (dev,"buffer: %s: buffer %u\n" , +ipu6_isys_queue_to_video (aq)->vdev.name, +ipu6_isys_buffer_to_vb2_buffer (ib)->index); +list_del (&ib->head); +list_add (&ib->head, &bl->head); +spin_unlock_irqrestore (&aq->lock, flags); + + bl->nbufs++; + } + +dev_dbg (dev,"get buffer list %p, %u buffers\n" , bl, bl->nbufs); + +return 0 ; +} + +void +ipu6_isys_buf_to_fw_frame_buf_pin (struct vb2_buffer *vb, +struct ipu6_fw_isys_frame_buff_set_abi *set) +{ +struct ipu6_isys_queue *aq =vb2_queue_to_isys_queue (vb->vb2_queue); + + set->output_pins[aq->fw_output].addr = +vb2_dma_contig_plane_dma_addr (vb,0 ); + set->output_pins[aq->fw_output].out_buf_id = vb->index +1 ; +} + +/* +* Convert a buffer list to a isys fw ABI framebuffer set. The +* buffer list is not modified. +*/ +#define IPU6_ISYS_FRAME_NUM_THRESHOLD (30) +void +ipu6_isys_buf_to_fw_frame_buf (struct ipu6_fw_isys_frame_buff_set_abi *set, +struct ipu6_isys_stream *stream, +struct ipu6_isys_buffer_list *bl) +{ +struct ipu6_isys_buffer *ib; + +WARN_ON (!bl->nbufs); + + set->send_irq_sof =1 ; + set->send_resp_sof =1 ; + set->send_irq_eof =0 ; + set->send_resp_eof =0 ; + +if (stream->streaming) + set->send_irq_capture_ack =0 ; +else + set->send_irq_capture_ack =1 ; + set->send_irq_capture_done =0 ; + + set->send_resp_capture_ack =1 ; + set->send_resp_capture_done =1 ; +if (atomic_read (&stream->sequence) >=IPU6_ISYS_FRAME_NUM_THRESHOLD ) { + set->send_resp_capture_ack =0 ; + set->send_resp_capture_done =0 ; + } + +list_for_each_entry (ib, &bl->head, head) { +struct vb2_buffer *vb =ipu6_isys_buffer_to_vb2_buffer (ib); +struct ipu6_isys_queue *aq = +vb2_queue_to_isys_queue (vb->vb2_queue); + +if (WARN_ON_ONCE (ib->type !=IPU6_ISYS_VIDEO_BUFFER )) +continue ; + +if (aq->fill_frame_buf_set) + aq->fill_frame_buf_set (vb, set); + } +} + +/* Start streaming for real. The buffer list must be available. */ +static int ipu6_isys_stream_start (struct ipu6_isys_video *av, +struct ipu6_isys_buffer_list *bl,bool error) +{ +struct ipu6_isys_stream *stream = av->stream; +struct device *dev = &stream->isys->adev->auxdev.dev; +struct ipu6_isys_buffer_list __bl; +int ret; + +mutex_lock (&stream->isys->stream_mutex); + ret =ipu6_isys_video_set_streaming (av,1 , bl); +mutex_unlock (&stream->isys->stream_mutex); +if (ret) + goto out_requeue; + + stream->streaming =1 ; + + bl = &__bl; + +do { +struct ipu6_fw_isys_frame_buff_set_abi *buf =NULL ; +struct isys_fw_msgs *msg; +u16 send_type =IPU6_FW_ISYS_SEND_TYPE_STREAM_CAPTURE ; + + ret =buffer_list_get (stream, bl); +if (ret <0 ) +break ; + + msg =ipu6_get_fw_msg_buf (stream); +if (!msg) +return -ENOMEM ; + + buf = &msg->fw_msg.frame; +ipu6_isys_buf_to_fw_frame_buf (buf, stream, bl); +ipu6_fw_isys_dump_frame_buff_set (dev, buf, + stream->nr_output_pins); +ipu6_isys_buffer_list_queue (bl,IPU6_ISYS_BUFFER_LIST_FL_ACTIVE , +0 ); + + ret =ipu6_fw_isys_complex_cmd (stream->isys, + stream->stream_handle, buf, + msg->dma_addr,sizeof (*buf), + send_type); + }while (!WARN_ON (ret)); + +return 0 ; + +out_requeue: +if (bl && bl->nbufs) +ipu6_isys_buffer_list_queue (bl, + (IPU6_ISYS_BUFFER_LIST_FL_INCOMING | + error) ? +IPU6_ISYS_BUFFER_LIST_FL_SET_STATE : +0 , error ?VB2_BUF_STATE_ERROR : +VB2_BUF_STATE_QUEUED ); +flush_firmware_streamon_fail (stream); + +return ret; +} + +static void buf_queue (struct vb2_buffer *vb) +{ +struct ipu6_isys_queue *aq =vb2_queue_to_isys_queue (vb->vb2_queue); +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct ipu6_isys_buffer *ib =vb2_buffer_to_ipu6_isys_buffer (vb); +struct device *dev = &av->isys->adev->auxdev.dev; +struct media_pipeline *media_pipe = +media_entity_pipeline (&av->vdev.entity); +struct ipu6_fw_isys_frame_buff_set_abi *buf =NULL ; +struct ipu6_isys_stream *stream = av->stream; +struct ipu6_isys_buffer_list bl; +struct isys_fw_msgs *msg; +unsigned long flags; +dma_addr_t dma; +unsigned int i; +int ret; + +dev_dbg (dev,"queue buffer %u for %s\n" , vb->index, av->vdev.name); + +for (i =0 ; i < vb->num_planes; i++) { + dma =vb2_dma_contig_plane_dma_addr (vb, i); +dev_dbg (dev,"iova: plane %u iova %pad\n" , i, &dma); + } + +spin_lock_irqsave (&aq->lock, flags); +list_add (&ib->head, &aq->incoming); +spin_unlock_irqrestore (&aq->lock, flags); + +if (!media_pipe || !vb->vb2_queue->start_streaming_called) { +dev_dbg (dev,"media pipeline is not ready for %s\n" , + av->vdev.name); +return ; + } + +mutex_lock (&stream->mutex); + +if (stream->nr_streaming != stream->nr_queues) { +dev_dbg (dev,"not streaming yet, adding to incoming\n" ); + goto out; + } + +/* +* We just put one buffer to the incoming list of this queue +* (above). Let's see whether all queues in the pipeline would +* have a buffer. +*/ + ret =buffer_list_get (stream, &bl); +if (ret <0 ) { +dev_warn (dev,"No buffers available\n" ); + goto out; + } + + msg =ipu6_get_fw_msg_buf (stream); +if (!msg) { + ret = -ENOMEM ; + goto out; + } + + buf = &msg->fw_msg.frame; + +ipu6_isys_buf_to_fw_frame_buf (buf, stream, &bl); + +ipu6_fw_isys_dump_frame_buff_set (dev, buf, stream->nr_output_pins); + +if (!stream->streaming) { + ret =ipu6_isys_stream_start (av, &bl, true); +if (ret) +dev_err (dev,"stream start failed.\n" ); + goto out; + } + +/* +* We must queue the buffers in the buffer list to the +* appropriate video buffer queues BEFORE passing them to the +* firmware since we could get a buffer event back before we +* have queued them ourselves to the active queue. +*/ +ipu6_isys_buffer_list_queue (&bl,IPU6_ISYS_BUFFER_LIST_FL_ACTIVE ,0 ); + + ret =ipu6_fw_isys_complex_cmd (stream->isys, stream->stream_handle, + buf, msg->dma_addr,sizeof (*buf), +IPU6_FW_ISYS_SEND_TYPE_STREAM_CAPTURE ); +if (ret <0 ) +dev_err (dev,"send stream capture failed\n" ); + +out: +mutex_unlock (&stream->mutex); +} + +int ipu6_isys_link_fmt_validate (struct ipu6_isys_queue *aq) +{ +struct v4l2_mbus_framefmt format; +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct device *dev = &av->isys->adev->auxdev.dev; +struct media_pad *remote_pad = +media_pad_remote_pad_first (av->vdev.entity.pads); +struct v4l2_subdev *sd; +u32 r_stream; +int ret; + +if (!remote_pad) +return -ENOTCONN ; + + sd =media_entity_to_v4l2_subdev (remote_pad->entity); + r_stream =ipu6_isys_get_src_stream_by_src_pad (sd, remote_pad->index); + + ret =ipu6_isys_get_stream_pad_fmt (sd, remote_pad->index, r_stream, + &format); + +if (ret) { +dev_dbg (dev,"failed to get %s: pad %d, stream:%d format\n" , + sd->entity.name, remote_pad->index, r_stream); +return ret; + } + +if (format.width != av->mpix.width || + format.height != av->mpix.height) { +dev_dbg (dev,"wrong width or height %ux%u (%ux%u expected)\n" , + av->mpix.width, av->mpix.height, + format.width, format.height); +return -EINVAL ; + } + +if (format.field != av->mpix.field) { +dev_dbg (dev,"wrong field value 0x%8.8x (0x%8.8x expected)\n" , + av->mpix.field, format.field); +return -EINVAL ; + } + +if (format.code != av->pfmt->code) { +dev_dbg (dev,"wrong mbus code 0x%8.8x (0x%8.8x expected)\n" , + av->pfmt->code, format.code); +return -EINVAL ; + } + +return 0 ; +} + +static void return_buffers (struct ipu6_isys_queue *aq, +enum vb2_buffer_state state) +{ +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct device *dev = &av->isys->adev->auxdev.dev; +struct ipu6_isys_buffer *ib; +bool need_reset = false; +unsigned long flags; + +spin_lock_irqsave (&aq->lock, flags); +while (!list_empty (&aq->incoming)) { +struct vb2_buffer *vb; + + ib = list_first_entry(&aq->incoming, struct ipu6_isys_buffer, + head); + vb =ipu6_isys_buffer_to_vb2_buffer (ib); +list_del (&ib->head); +spin_unlock_irqrestore (&aq->lock, flags); + +vb2_buffer_done (vb, state); + +dev_dbg (dev,"%s: stop_streaming incoming %u\n" , +ipu6_isys_queue_to_video (vb2_queue_to_isys_queue + (vb->vb2_queue))->vdev.name, + vb->index); + +spin_lock_irqsave (&aq->lock, flags); + } + +/* +* Something went wrong (FW crash / HW hang / not all buffers +* returned from isys) if there are still buffers queued in active +* queue. We have to clean up places a bit. +*/ +while (!list_empty (&aq->active)) { +struct vb2_buffer *vb; + + ib = list_first_entry(&aq->active, struct ipu6_isys_buffer, + head); + vb =ipu6_isys_buffer_to_vb2_buffer (ib); + +list_del (&ib->head); +spin_unlock_irqrestore (&aq->lock, flags); + +vb2_buffer_done (vb, state); + +dev_warn (dev,"%s: cleaning active queue %u\n" , +ipu6_isys_queue_to_video (vb2_queue_to_isys_queue + (vb->vb2_queue))->vdev.name, + vb->index); + +spin_lock_irqsave (&aq->lock, flags); + need_reset = true; + } + +spin_unlock_irqrestore (&aq->lock, flags); + +if (need_reset) { +mutex_lock (&av->isys->mutex); + av->isys->need_reset = true; +mutex_unlock (&av->isys->mutex); + } +} + +static void ipu6_isys_stream_cleanup (struct ipu6_isys_video *av) +{ +video_device_pipeline_stop (&av->vdev); +ipu6_isys_put_stream (av->stream); + av->stream =NULL ; +} + +static int start_streaming (struct vb2_queue *q,unsigned int count) +{ +struct ipu6_isys_queue *aq =vb2_queue_to_isys_queue (q); +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct device *dev = &av->isys->adev->auxdev.dev; +struct ipu6_isys_buffer_list __bl, *bl =NULL ; +struct ipu6_isys_stream *stream; +struct media_entity *source_entity =NULL ; +int nr_queues, ret; + +dev_dbg (dev,"stream: %s: width %u, height %u, css pixelformat %u\n" , + av->vdev.name, av->mpix.width, av->mpix.height, + av->pfmt->css_pixelformat); + + ret =ipu6_isys_setup_video (av, &source_entity, &nr_queues); +if (ret <0 ) { +dev_err (dev,"failed to setup video\n" ); + goto out_return_buffers; + } + + ret = aq->link_fmt_validate (aq); +if (ret) { +dev_err (dev, +"%s: link format validation failed (%d)\n" , + av->vdev.name, ret); + goto out_pipeline_stop; + } + + ret =ipu6_isys_fw_open (av->isys); +if (ret) + goto out_pipeline_stop; + + stream = av->stream; +mutex_lock (&stream->mutex); +if (!stream->nr_streaming) { + ret =ipu6_isys_video_prepare_stream (av, source_entity, + nr_queues); +if (ret) + goto out_fw_close; + } + + stream->nr_streaming++; +dev_dbg (dev,"queue %u of %u\n" , stream->nr_streaming, + stream->nr_queues); +list_add (&aq->node, &stream->queues); + +ipu6_isys_set_csi2_streams_status (av, true); + +ipu6_isys_configure_stream_watermark (av, true); +ipu6_isys_update_stream_watermark (av, true); + +if (stream->nr_streaming != stream->nr_queues) + goto out; + + bl = &__bl; + ret =buffer_list_get (stream, bl); +if (ret <0 ) { +dev_dbg (dev, +"no buffer available, postponing streamon\n" ); + goto out; + } + + ret =ipu6_isys_stream_start (av, bl, false); +if (ret) + goto out_stream_start; + +out: +mutex_unlock (&stream->mutex); + +return 0 ; + +out_stream_start: +list_del (&aq->node); + stream->nr_streaming--; + +out_fw_close: +mutex_unlock (&stream->mutex); +ipu6_isys_fw_close (av->isys); + +out_pipeline_stop: +ipu6_isys_stream_cleanup (av); + +out_return_buffers: +return_buffers (aq,VB2_BUF_STATE_QUEUED ); + +return ret; +} + +static void stop_streaming (struct vb2_queue *q) +{ +struct ipu6_isys_queue *aq =vb2_queue_to_isys_queue (q); +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct ipu6_isys_stream *stream = av->stream; + +ipu6_isys_set_csi2_streams_status (av, false); + +mutex_lock (&stream->mutex); + +ipu6_isys_update_stream_watermark (av, false); + +mutex_lock (&av->isys->stream_mutex); +if (stream->nr_streaming == stream->nr_queues && stream->streaming) +ipu6_isys_video_set_streaming (av,0 ,NULL ); +mutex_unlock (&av->isys->stream_mutex); + + stream->nr_streaming--; +list_del (&aq->node); + stream->streaming =0 ; +mutex_unlock (&stream->mutex); + +ipu6_isys_stream_cleanup (av); + +return_buffers (aq,VB2_BUF_STATE_ERROR ); + +ipu6_isys_fw_close (av->isys); +} + +static unsigned int +get_sof_sequence_by_timestamp (struct ipu6_isys_stream *stream, +struct ipu6_fw_isys_resp_info_abi *info) +{ +u64 time = (u64 )info->timestamp[1 ] <<32 | info->timestamp[0 ]; +struct ipu6_isys *isys = stream->isys; +struct device *dev = &isys->adev->auxdev.dev; +unsigned int i; + +/* +* The timestamp is invalid as no TSC in some FPGA platform, +* so get the sequence from pipeline directly in this case. +*/ +if (time ==0 ) +return atomic_read (&stream->sequence) -1 ; + +for (i =0 ; i <IPU6_ISYS_MAX_PARALLEL_SOF ; i++) +if (time == stream->seq[i].timestamp) { +dev_dbg (dev,"sof: using seq nr %u for ts %llu\n" , + stream->seq[i].sequence, time); +return stream->seq[i].sequence; + } + +for (i =0 ; i <IPU6_ISYS_MAX_PARALLEL_SOF ; i++) +dev_dbg (dev,"sof: sequence %u, timestamp value %llu\n" , + stream->seq[i].sequence, stream->seq[i].timestamp); + +return 0 ; +} + +static u64 get_sof_ns_delta (struct ipu6_isys_video *av, +struct ipu6_fw_isys_resp_info_abi *info) +{ +struct ipu6_bus_device *adev = av->isys->adev; +struct ipu6_device *isp = adev->isp; +u64 delta, tsc_now; + +ipu6_buttress_tsc_read (isp, &tsc_now); +if (!tsc_now) +return 0 ; + + delta = tsc_now - ((u64 )info->timestamp[1 ] <<32 | info->timestamp[0 ]); + +return ipu6_buttress_tsc_ticks_to_ns (delta, isp); +} + +void ipu6_isys_buf_calc_sequence_time (struct ipu6_isys_buffer *ib, +struct ipu6_fw_isys_resp_info_abi *info) +{ +struct vb2_buffer *vb =ipu6_isys_buffer_to_vb2_buffer (ib); +struct vb2_v4l2_buffer *vbuf =to_vb2_v4l2_buffer (vb); +struct ipu6_isys_queue *aq =vb2_queue_to_isys_queue (vb->vb2_queue); +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +struct device *dev = &av->isys->adev->auxdev.dev; +struct ipu6_isys_stream *stream = av->stream; +u64 ns; +u32 sequence; + + ns =ktime_get_ns () -get_sof_ns_delta (av, info); + sequence =get_sof_sequence_by_timestamp (stream, info); + + vbuf->vb2_buf.timestamp = ns; + vbuf->sequence = sequence; + +dev_dbg (dev,"buf: %s: buffer done, CPU-timestamp:%lld, sequence:%d\n" , + av->vdev.name,ktime_get_ns (), sequence); +dev_dbg (dev,"index:%d, vbuf timestamp:%lld\n" , vb->index, + vbuf->vb2_buf.timestamp); +} + +void ipu6_isys_queue_buf_done (struct ipu6_isys_buffer *ib) +{ +struct vb2_buffer *vb =ipu6_isys_buffer_to_vb2_buffer (ib); + +if (atomic_read (&ib->str2mmio_flag)) { +vb2_buffer_done (vb,VB2_BUF_STATE_ERROR ); +/* +* Operation on buffer is ended with error and will be reported +* to the userspace when it is de-queued +*/ +atomic_set (&ib->str2mmio_flag,0 ); + }else { +vb2_buffer_done (vb,VB2_BUF_STATE_DONE ); + } +} + +void ipu6_isys_queue_buf_ready (struct ipu6_isys_stream *stream, +struct ipu6_fw_isys_resp_info_abi *info) +{ +struct ipu6_isys_queue *aq = stream->output_pins[info->pin_id].aq; +struct ipu6_isys *isys = stream->isys; +struct device *dev = &isys->adev->auxdev.dev; +struct ipu6_isys_buffer *ib; +struct vb2_buffer *vb; +unsigned long flags; +bool first = true; +struct vb2_v4l2_buffer *buf; + +dev_dbg (dev,"buffer: %s: received buffer %8.8x\n" , +ipu6_isys_queue_to_video (aq)->vdev.name, info->pin.addr); + +spin_lock_irqsave (&aq->lock, flags); +if (list_empty (&aq->active)) { +spin_unlock_irqrestore (&aq->lock, flags); +dev_err (dev,"active queue empty\n" ); +return ; + } + +list_for_each_entry_reverse (ib, &aq->active, head) { +dma_addr_t addr; + + vb =ipu6_isys_buffer_to_vb2_buffer (ib); + addr =vb2_dma_contig_plane_dma_addr (vb,0 ); + +if (info->pin.addr != addr) { +if (first) +dev_err (dev,"Unexpected buffer address %pad\n" , + &addr); + first = false; +continue ; + } + +if (info->error_info.error == +IPU6_FW_ISYS_ERROR_HW_REPORTED_STR2MMIO ) { +/* +* Check for error message: +* 'IPU6_FW_ISYS_ERROR_HW_REPORTED_STR2MMIO' +*/ +atomic_set (&ib->str2mmio_flag,1 ); + } +dev_dbg (dev,"buffer: found buffer %pad\n" , &addr); + + buf =to_vb2_v4l2_buffer (vb); + buf->field =V4L2_FIELD_NONE ; + +list_del (&ib->head); +spin_unlock_irqrestore (&aq->lock, flags); + +ipu6_isys_buf_calc_sequence_time (ib, info); + +ipu6_isys_queue_buf_done (ib); + +return ; + } + +dev_err (dev,"Failed to find a matching video buffer" ); + +spin_unlock_irqrestore (&aq->lock, flags); +} + +static const struct vb2_ops ipu6_isys_queue_ops = { + .queue_setup = queue_setup, + .wait_prepare = vb2_ops_wait_prepare, + .wait_finish = vb2_ops_wait_finish, + .buf_prepare = buf_prepare, + .start_streaming = start_streaming, + .stop_streaming = stop_streaming, + .buf_queue = buf_queue, +}; + +int ipu6_isys_queue_init (struct ipu6_isys_queue *aq) +{ +struct ipu6_isys *isys =ipu6_isys_queue_to_video (aq)->isys; +struct ipu6_isys_video *av =ipu6_isys_queue_to_video (aq); +int ret; + +/* no support for userptr */ +if (!aq->vbq.io_modes) + aq->vbq.io_modes =VB2_MMAP |VB2_DMABUF ; + + aq->vbq.drv_priv = aq; + aq->vbq.ops = &ipu6_isys_queue_ops; + aq->vbq.lock = &av->mutex; + aq->vbq.mem_ops = &vb2_dma_contig_memops; + aq->vbq.type =V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE ; + aq->vbq.timestamp_flags =V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC ; + + ret =vb2_queue_init (&aq->vbq); +if (ret) +return ret; + + aq->dev = &isys->adev->auxdev.dev; + aq->vbq.dev = &isys->adev->auxdev.dev; +spin_lock_init (&aq->lock); +INIT_LIST_HEAD (&aq->active); +INIT_LIST_HEAD (&aq->incoming); + +return 0 ; +} + +void ipu6_isys_queue_cleanup (struct ipu6_isys_queue *aq) +{ +vb2_queue_release (&aq->vbq); +}
drivers/media/pci/intel/ipu6/ipu6-isys-queue.h new file +94 -0
94 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6-isys-video.c new file +1239 -0
1,239 lines not on this page · raw message
drivers/media/pci/intel/ipu6/ipu6-isys-video.h new file +129 -0
129 lines not on this page · raw message
Signature · 1 line
2.42.0
Re: [PATCH v2 11/15] media: intel/ipu6: input system video capture nodes
Hi Bingbu,
Quoted 10 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Register v4l2 video device and setup the vb2 queue to support basic video capture. Video streaming callback will trigger the input system driver to construct a input system stream configuration for firmware based on data type and stream ID and then queue buffers to firmware to do capture.
Signed-off-by Bingbu Cao <[email protected]>
While testing this new version I found a small issue in the enum_fmt implementation caused by the mbus_code filtering support I suggested.
Attached is a single-line patch which fixes this, can you please fold this into the next version ?
Regards,
Hans
From 4c3b1445f51e8123ac3fb3fbf76b2c429a1c77fa Mon Sep 17 00:00:00 2001
From: Hans de Goede <[email protected]>
Date: Sun, 5 Nov 2023 17:46:52 +0100
Subject: [PATCH] media: ipu6-isys-video: Do not set mbus_code from
vidioc_enum_fmt_vid_cap
When enumerating userspace will call vidioc_enum_fmt_vid_cap repeatedly with higher v4l2_fmtdesc.index values until it returns -EINVAL to indicate all formats have been enumerated.
Some userspace code such as e.g. v4l2-ctl only fills the v4l2_fmtdesc struct once and then only increases v4l2_fmtdesc.index for subsequent calls.
Setting v4l2_fmtdesc.mbus_code when it was 0 changes subsequent vidioc_enum_fmt_vid_cap from unfiltered mode to mbus-code filtered mode breaking unfiltered enumeration.
Do not set v4l2_fmtdesc.mbus_code in unfiltered mode to fix this.
--- drivers/media/pci/intel/ipu6/ipu6-isys-video.c | 1 - 1 file changed, 1 deletion(-)
drivers/media/pci/intel/ipu6/ipu6-isys-video.c +0 -1
@@ -129,7 +129,6 @@ int ipu6_isys_vidioc_enum_fmt(struct file *file, void *fh,if (!f->mbus_code) { f->flags =0 ; f->pixelformat = ipu6_isys_pfmts[f->index].pixelformat; - f->mbus_code = ipu6_isys_pfmts[f->index].code;return 0 ; }
Signature · 1 line
2.41.0
Re: [PATCH v2 11/15] media: intel/ipu6: input system video capture nodes
Hans,
Quoted 12 lines · Hans de Goede
Hi Bingbu,
Quoted 9 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Register v4l2 video device and setup the vb2 queue to support basic video capture. Video streaming callback will trigger the input system driver to construct a input system stream configuration for firmware based on data type and stream ID and then queue buffers to firmware to do capture.
Quoted 5 lines
Quoted 1 line
Signed-off-by Bingbu Cao <[email protected]>While testing this new version I found a small issue in the
enum_fmtimplementation caused by thembus_codefiltering support I suggested.Attached is a single-line patch which fixes this, can you please fold this into the next version ?
Yes, sure.
Quoted 5 lines
Regards,
Hans
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 11/15] media: intel/ipu6: input system video capture nodes
Quoted 1 line · Bingbu Cao
On 11/6/23 12:59 AM, Hans de Goede wrote:
...
Quoted 1 line
Yes, sure.
It's likely that I meant this series (it's under development, right?) to be Cc'ed to me in its next version.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 11/15] media: intel/ipu6: input system video capture nodes
Hans,
Thanks for your patch, I will squash them into v3.
Andy and Hans, thank you for your review and comments on v2.
Please give me some time to send v2, we have metadata capture support change which depends on some v4l2 APIs update in kernel. I will send v3 ASAP.
Quoted 10 lines · Andy Shevchenko
Quoted 1 line · Bingbu Cao
On 11/6/23 12:59 AM, Hans de Goede wrote:
...
Quoted 1 line
Yes, sure.
It's likely that I meant this series (it's under development, right?) to be Cc'ed to me in its next version.
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 11/15] media: intel/ipu6: input system video capture nodes
In reply to [email protected]Hi Bingbu,
Quoted 11 lines · On Tue, 2023-10-24 at 19:29 +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Register v4l2 video device and setup the vb2 queue to support basic video capture. Video streaming callback will trigger the input system driver to construct a input system stream configuration for firmware based on data type and stream ID and then queue buffers to firmware to do capture.
Signed-off-by Bingbu Cao <[email protected]>---
...
Quoted 1 line
+++
b/drivers/media/pci/intel/ipu6/ipu6-isys-video.c
...
Quoted 4 lines
drivers/media/pci/intel/ipu6/ipu6-isys-video.c new file +4
-0+static int video_release (struct file *file) +{ +return vb2_fop_release (file); +}
...
Quoted 62 lines
drivers/media/pci/intel/ipu6/ipu6-isys-video.c new file +62
-0+/* +* Do everything that's needed to initialise things related to video +* buffer queue, video node, and the related media entity. The caller +* is expected to assign isys field and set the name of the video +* device. +*/ +int ipu6_isys_video_init (struct ipu6_isys_video *av) +{ +struct v4l2_format format = { + .type =V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE , + .fmt.pix_mp = { + .width =1920 , + .height =1080 , + }, + }; +int ret; + +mutex_init (&av->mutex); + av->vdev.device_caps =V4L2_CAP_STREAMING |V4L2_CAP_IO_MC | +V4L2_CAP_VIDEO_CAPTURE_MPLANE ; + av->vdev.vfl_dir =VFL_DIR_RX ; + + ret =ipu6_isys_queue_init (&av->aq); +if (ret) + goto out_free_watermark; + + av->pad.flags =MEDIA_PAD_FL_SINK |MEDIA_PAD_FL_MUST_CONNECT ; + ret =media_entity_pads_init (&av->vdev.entity,1 , &av->pad); +if (ret) + goto out_ipu6_isys_queue_cleanup; + + av->vdev.entity.ops = &entity_ops; + av->vdev.release = video_device_release_empty; + av->vdev.fops = &isys_fops; + av->vdev.v4l2_dev = &av->isys->v4l2_dev; +if (!av->vdev.ioctl_ops) + av->vdev.ioctl_ops = &ioctl_ops_mplane; + av->vdev.queue = &av->aq.vbq; + av->vdev.lock = &av->mutex; + +ipu6_isys_video_try_fmt_vid_mplane (av, &format.fmt.pix_mp); + av->mpix = format.fmt.pix_mp; + +set_bit (V4L2_FL_USES_V4L2_FH , &av->vdev.flags); +video_set_drvdata (&av->vdev, av); + + ret =video_register_device (&av->vdev,VFL_TYPE_VIDEO ,-1 ); +if (ret) + goto out_media_entity_cleanup; + +return ret; + +out_media_entity_cleanup: +video_unregister_device (&av->vdev); +media_entity_cleanup (&av->vdev.entity); + +out_ipu6_isys_queue_cleanup: +ipu6_isys_queue_cleanup (&av->aq); + +out_free_watermark: +mutex_destroy (&av->mutex); +
drivers/media/pci/intel/ipu6/ipu6-isys-video.c line 1239
Quoted 10 lines
drivers/media/pci/intel/ipu6/ipu6-isys-video.c new file +10
-0+return ret; +} + +void ipu6_isys_video_cleanup (struct ipu6_isys_video *av) +{ +video_unregister_device (&av->vdev); +media_entity_cleanup (&av->vdev.entity); +mutex_destroy (&av->mutex); +ipu6_isys_queue_cleanup (&av->aq); +}
I believe you should use vb2_video_unregister_device here instead of video_unregister_device in both ipu6_isys_video_init and ipu6_isys_video_cleanup, and remove the call to ipu6_isys_queue_cleanup.
Commit f729ef5796d8 from Hans Verkuil that added vb2_video_unregister_device states: "If a driver calls (_)vb2_fop_release(), then such a driver should also call vb2_video_unregister_device() instead of video_unregister_device(). This helper will call vb2_queue_release() if a filehandle is marked as
owner of the queue. This ensures that at unregister time any streaming is cancelled and all buffers are returned to userspace."
In my testing, using IPU4, this prevents some crashes when unbinding the PCI device while streaming.
/Andreas
PATCH v2 12/15 media: add Kconfig and Makefile for IPU6
In reply to [email protected]From: Bingbu Cao <[email protected]>
Add IPU6 support in Kconfig and Makefile, with this patch you can build the Intel IPU6 and input system modules by select the CONFIG_VIDEO_INTEL_IPU6 in config.
--- drivers/media/pci/intel/Kconfig | 1 + drivers/media/pci/intel/Makefile | 1 + drivers/media/pci/intel/ipu6/Kconfig | 17 +++++++++++++++++ drivers/media/pci/intel/ipu6/Makefile | 23 +++++++++++++++++++++++ 4 files changed, 42 insertions(+) create mode 100644 drivers/media/pci/intel/ipu6/Kconfig create mode 100644 drivers/media/pci/intel/ipu6/Makefile
drivers/media/pci/intel/Kconfig +1 -0
@@ -1,6 +1,7 @@ # SPDX-License-Identifier: GPL-2.0-only source "drivers/media/pci/intel/ipu3/Kconfig" +source "drivers/media/pci/intel/ipu6/Kconfig" source "drivers/media/pci/intel/ivsc/Kconfig" config IPU_BRIDGE
drivers/media/pci/intel/Makefile +1 -0
@@ -5,3 +5,4 @@ obj-$(CONFIG_IPU_BRIDGE) += ipu-bridge.o obj-y += ipu3/ obj-y += ivsc/ +obj-$(CONFIG_VIDEO_INTEL_IPU6) += ipu6/
drivers/media/pci/intel/ipu6/Kconfig new file +17 -0
@@ -0,0 +1,17 @@ +config VIDEO_INTEL_IPU6 + tristate "Intel IPU6 driver" + depends on ACPI || COMPILE_TEST + depends on MEDIA_SUPPORT + depends on MEDIA_PCI_SUPPORT + depends on X86 && X86_64 + select IOMMU_IOVA + select VIDEO_V4L2_SUBDEV_API + select VIDEOBUF2_DMA_CONTIG + select V4L2_FWNODE + select IPU_BRIDGE + help + This is the 6th Gen Intel Image Processing Unit, found in Intel SoCs + and used for capturing images and video from camera sensors. + + To compile this driver, say Y here! It contains 2 modules - + intel_ipu6 and intel_ipu6_isys.
drivers/media/pci/intel/ipu6/Makefile new file +23 -0
@@ -0,0 +1,23 @@ +# SPDX-License-Identifier: GPL-2.0-only + +intel-ipu6-objs += ipu6.o \ + ipu6-bus.o \ + ipu6-dma.o \ + ipu6-mmu.o \ + ipu6-buttress.o \ + ipu6-cpd.o \ + ipu6-fw-com.o + +obj-$(CONFIG_VIDEO_INTEL_IPU6) += intel-ipu6.o + +intel-ipu6-isys-objs += ipu6-isys.o \ + ipu6-isys-csi2.o \ + ipu6-fw-isys.o \ + ipu6-isys-video.o \ + ipu6-isys-queue.o \ + ipu6-isys-subdev.o \ + ipu6-isys-mcd-phy.o \ + ipu6-isys-jsl-phy.o \ + ipu6-isys-dwc-phy.o + +obj-$(CONFIG_VIDEO_INTEL_IPU6) += intel-ipu6-isys.o
Signature · 1 line
2.42.0
Re: [PATCH v2 12/15] media: add Kconfig and Makefile for IPU6
Quoted 5 lines · On Tue, Oct 24, 2023 at 07:29:21PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Add IPU6 support in
KconfigandMakefile, with this patch you can build the Intel IPU6 and input system modules by select theCONFIG_VIDEO_INTEL_IPU6in config.
...
drivers/media/pci/intel/ipu6/Kconfig line 6
Quoted 1 line
drivers/media/pci/intel/ipu6/Kconfig new file +1
-0+ depends on X86 && X86_64
How is that? One is a superset of the other IIUC.
...
Quoted 1 line
+intel-ipu6-objs += ipu6.o \
+= doesn't seem right.
Besides that 'objs' is for user space, in kernel we use 'y'.
drivers/media/pci/intel/ipu6/Makefile line 13
Quoted 10 lines
drivers/media/pci/intel/ipu6/Makefile new file +10
-0+ ipu6-bus.o \ + ipu6-dma.o \ + ipu6-mmu.o \ + ipu6-buttress.o \ + ipu6-cpd.o \ + ipu6-fw-com.o + +obj-$(CONFIG_VIDEO_INTEL_IPU6) += intel-ipu6.o + +intel-ipu6-isys-objs += ipu6-isys.o \
Ditto.
Quoted 10 lines
drivers/media/pci/intel/ipu6/Makefile new file +10
-0+ ipu6-isys-csi2.o \ + ipu6-fw-isys.o \ + ipu6-isys-video.o \ + ipu6-isys-queue.o \ + ipu6-isys-subdev.o \ + ipu6-isys-mcd-phy.o \ + ipu6-isys-jsl-phy.o \ + ipu6-isys-dwc-phy.o + +obj-$(CONFIG_VIDEO_INTEL_IPU6) += intel-ipu6-isys.o
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 12/15] media: add Kconfig and Makefile for IPU6
Andy,
Thanks for the review.
Quoted 7 lines · Andy Shevchenko
Quoted 5 lines · On Tue, Oct 24, 2023 at 07:29:21PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Add IPU6 support in
KconfigandMakefile, with this patch you can build the Intel IPU6 and input system modules by select theCONFIG_VIDEO_INTEL_IPU6in config.
Quoted 5 lines
...
drivers/media/pci/intel/ipu6/Kconfig line 6
Quoted 1 line
drivers/media/pci/intel/ipu6/Kconfig new file +1
-0+ depends on X86 && X86_64How is that? One is a superset of the other IIUC.
It's trying to prevent build IPU6 with other arch such as 'UM'.
# make W=1 O=build_dir ARCH=um olddefconfig ipu6-dma.c:... error: implicit declaration of function 'clflush_cache_range';
X86_64 looks like not an actual architecture.
Quoted 8 lines
...
Quoted 1 line
+intel-ipu6-objs += ipu6.o \
+= doesn't seem right.
Besides that 'objs' is for user space, in kernel we use 'y'.
Quoted 12 lines
drivers/media/pci/intel/ipu6/Makefile line 13
Quoted 10 lines
drivers/media/pci/intel/ipu6/Makefile new file +10
-0+ ipu6-bus.o \ + ipu6-dma.o \ + ipu6-mmu.o \ + ipu6-buttress.o \ + ipu6-cpd.o \ + ipu6-fw-com.o + +obj-$(CONFIG_VIDEO_INTEL_IPU6) += intel-ipu6.o + +intel-ipu6-isys-objs += ipu6-isys.o \Ditto.
Ack.
Quoted 12 lines
Quoted 11 lines
drivers/media/pci/intel/ipu6/Makefile new file +10
-0+ ipu6-isys-csi2.o \ + ipu6-fw-isys.o \ + ipu6-isys-video.o \ + ipu6-isys-queue.o \ + ipu6-isys-subdev.o \ + ipu6-isys-mcd-phy.o \ + ipu6-isys-jsl-phy.o \ + ipu6-isys-dwc-phy.o + +obj-$(CONFIG_VIDEO_INTEL_IPU6) += intel-ipu6-isys.o
Signature · 2 lines
Best regards, Bingbu Cao
Re: [PATCH v2 12/15] media: add Kconfig and Makefile for IPU6
Quoted 2 lines · Bingbu Cao
Quoted 1 line · Andy Shevchenko
On Tue, Oct 24, 2023 at 07:29:21PM +0800, [email protected] wrote:
...
Quoted 9 lines
Quoted 4 lines
drivers/media/pci/intel/ipu6/Kconfig line 6
Quoted 2 lines
drivers/media/pci/intel/ipu6/Kconfig new file +1
-0+ depends on X86 && X86_64How is that? One is a superset of the other IIUC.
It's trying to prevent build IPU6 with other arch such as 'UM'.
# make W=1 O=build_dir ARCH=um olddefconfig ipu6-dma.c:... error: implicit declaration of function 'clflush_cache_range';
Is UM ever supported 64-bit builds?
Quoted 1 line
X86_64looks like not an actual architecture.
How comes?
Signature · 2 lines
With Best Regards, Andy Shevchenko
Sakari Ailus
maintainerRe: [PATCH v2 12/15] media: add Kconfig and Makefile for IPU6
In reply to Andy ShevchenkoHi Andy,
Quoted 13 lines · Andy Shevchenko
Quoted 5 lines · On Tue, Oct 24, 2023 at 07:29:21PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Add IPU6 support in
KconfigandMakefile, with this patch you can build the Intel IPU6 and input system modules by select theCONFIG_VIDEO_INTEL_IPU6in config....
drivers/media/pci/intel/ipu6/Kconfig line 6
Quoted 1 line
drivers/media/pci/intel/ipu6/Kconfig new file +1
-0+ depends on X86 && X86_64How is that? One is a superset of the other IIUC.
Quoted 6 lines
...
Quoted 1 line
+intel-ipu6-objs += ipu6.o \
+= doesn't seem right.
Besides that 'objs' is for user space, in kernel we use 'y'.
This is a common pattern used in other Makefiles, too... I guess ":=" would work, too.
How otherwise would you do this?
It'd be nice to align the object names starting after the column after "+= ".
Quoted 25 lines
drivers/media/pci/intel/ipu6/Makefile line 13
Quoted 11 lines
drivers/media/pci/intel/ipu6/Makefile new file +10
-0+ ipu6-bus.o \ + ipu6-dma.o \ + ipu6-mmu.o \ + ipu6-buttress.o \ + ipu6-cpd.o \ + ipu6-fw-com.o + +obj-$(CONFIG_VIDEO_INTEL_IPU6) += intel-ipu6.o + +intel-ipu6-isys-objs += ipu6-isys.o \Ditto.
Quoted 10 lines
drivers/media/pci/intel/ipu6/Makefile new file +10
-0+ ipu6-isys-csi2.o \ + ipu6-fw-isys.o \ + ipu6-isys-video.o \ + ipu6-isys-queue.o \ + ipu6-isys-subdev.o \ + ipu6-isys-mcd-phy.o \ + ipu6-isys-jsl-phy.o \ + ipu6-isys-dwc-phy.o + +obj-$(CONFIG_VIDEO_INTEL_IPU6) += intel-ipu6-isys.o
Signature · 3 lines
Regards, Sakari Ailus
Re: [PATCH v2 12/15] media: add Kconfig and Makefile for IPU6
Quoted 2 lines · Sakari Ailus
Quoted 1 line · Andy Shevchenko
On Tue, Oct 24, 2023 at 07:29:21PM +0800, [email protected] wrote:
...
Quoted 9 lines
Quoted 4 lines
Quoted 1 line
+intel-ipu6-objs += ipu6.o \
+= doesn't seem right.
Besides that 'objs' is for user space, in kernel we use 'y'.This is a common pattern used in other Makefiles, too... I guess ":=" would work, too.
How otherwise would you do this?
:= has different meaning in make language. How you will guarantee somewhere else somebody won't use same target, note that make is recursive.
Signature · 2 lines
With Best Regards, Andy Shevchenko
PATCH v2 13/15 MAINTAINERS: add maintainers for Intel IPU6 input system driver
In reply to [email protected]From: Bingbu Cao <[email protected]>
Update MAINTAINERS file for Intel IPU6 input system driver.
--- MAINTAINERS | 10 ++++++++++ 1 file changed, 10 insertions(+)
MAINTAINERS +10 -0
@@ -10626,6 +10626,16 @@ F: Documentation/admin-guide/media/ipu3_rcb.svg F: Documentation/userspace-api/media/v4l/metafmt-intel-ipu3.rst F: drivers/staging/media/ipu3/ +INTEL IPU6 INPUT SYSTEM DRIVER +M: Sakari Ailus <[email protected]> +M: Bingbu Cao <[email protected]> +R: Tianshu Qiu <[email protected]> +L: [email protected] +S: Maintained +T: git git://linuxtv.org/media_tree.git +F: Documentation/admin-guide/media/ipu6-isys.rst +F: drivers/media/pci/intel/ipu6/ + INTEL ISHTP ECLITE DRIVER M: Sumesh K Naduvalath <[email protected]> L: [email protected]
Signature · 1 line
2.42.0
PATCH v2 14/15 Documentation: add Intel IPU6 ISYS driver admin-guide doc
In reply to [email protected]From: Bingbu Cao <[email protected]>
This document mainly describe the functionality of IPU6 and IPU6 isys driver, and gives an example that how user can do imaging capture with tools.
--- Documentation/admin-guide/media/ipu6-isys.rst | 159 +++ .../admin-guide/media/ipu6_isys_graph.svg | 338 +++++ .../admin-guide/media/ipu6_isys_multi.svg | 1124 +++++++++++++++++ .../admin-guide/media/v4l-drivers.rst | 1 + 4 files changed, 1622 insertions(+) create mode 100644 Documentation/admin-guide/media/ipu6-isys.rst create mode 100644 Documentation/admin-guide/media/ipu6_isys_graph.svg create mode 100644 Documentation/admin-guide/media/ipu6_isys_multi.svg
Documentation/admin-guide/media/ipu6-isys.rst new file +159 -0
@@ -0,0 +1,159 @@ +.. SPDX-License-Identifier: GPL-2.0 + +.. include:: <isonum.txt> + +=============================================================== +Intel Image Processing Unit 6 (IPU6) Input System driver +=============================================================== + +Copyright |copy| 2023 Intel Corporation + +Introduction +============ + +This file documents the Intel IPU6 (6th generation Image Processing Unit) +Input System (MIPI CSI2 receiver) drivers located under +drivers/media/pci/intel/ipu6. + +The Intel IPU6 can be found in certain Intel Chipsets but not in all SKUs: + +* TigerLake +* JasperLake +* AlderLake +* RaptorLake +* MeteorLake + +Intel IPU6 is made up of two components - Input System (ISYS) and Processing +System (PSYS). + +The Input System mainly works as MIPI CSI2 receiver which receives and +processes the imaging data from the sensors and outputs the frames to memory. + +There are 2 driver modules - intel_ipu6 and intel_ipu6_isys. intel_ipu6 is an +IPU6 common driver which does PCI configuration, firmware loading and parsing, +firmware authentication, DMA mapping and IPU-MMU (internal Memory mapping Unit) +configuration. intel_ipu6_isys implements V4L2, Media Controller and V4L2 +sub-device interfaces. The IPU6 ISYS driver supports camera sensors connected +to the IPU6 ISYS through V4L2 sub-device sensor drivers. + +.. Note:: See Documentation/driver-api/media/drivers/ipu6.rst for more + information about the IPU6 hardware. + + +Input system driver +=================== + +The input System driver mainly configures CSI2 DPHY, constructs the firmware +stream configuration, sends commands to firmware, gets response from hardware +and firmware and then returns buffers to user. +The ISYS is represented as several V4L2 sub-devices - 'Intel IPU6 CSI2 $port', +which provide V4L2 subdev interfaces to the user space, there are also several +video nodes for each CSI-2 stream capture - 'Intel IPU6 ISYS capture $num' which +provide interface to user to set formats, queue buffers and streaming. + +.. kernel-figure:: ipu6_isys_graph.svg + :alt: ipu6 isys media graph without multiple streams support + + ipu6 isys media graph without multiple streams support + +.. kernel-figure:: ipu6_isys_multi.svg + :alt: ipu6 isys media graph with multiple streams support + + ipu6 isys media graph with multiple streams support + +Capturing frames by IPU6 ISYS +------------------------------------ + +IPU6 ISYS is used to capture frames from the camera sensors connected to the +CSI2 ports. The supported input formats of ISYS are listed in table below: + +.. tabularcolumns:: |p{0.8cm}|p{4.0cm}|p{4.0cm}| + +.. flat-table:: + :header-rows: 1 + + * - IPU6 ISYS supported input formats + + * - RGB565, RGB888 + + * - UYVY8, YUYV8 + + * - RAW8, RAW10, RAW12 + +Here is an example of IPU6 ISYS raw capture on Dell XPS 9315 laptop. On this +machine, ov01a10 sensor is connected to IPU ISYS CSI2 port 2, which can +generate images at sBGGR10 with resolution 1280x800. + +Using the media controller APIs, we can configure ov01a10 sensor by +media-ctl [#f1]_ and yavta [#f2]_ to transmit frames to IPU6 ISYS. + +.. code-block:: none + + # Example 1 capture frame from ov01a10 camera sensor + # This example assumes /dev/media0 as the IPU ISYS media device + export MDEV=/dev/media0 + + # Establish the link for the media devices using media-ctl + media-ctl -d $MDEV -l "\"ov01a10 3-0036\":0 -> \"Intel IPU6 CSI2 2\":0[1]" + + # Set the format for the media devices + media-ctl -d $MDEV -V "ov01a10:0 [fmt:SBGGR10/1280x800]" + media-ctl -d $MDEV -V "Intel IPU6 CSI2 2:0 [fmt:SBGGR10/1280x800]" + media-ctl -d $MDEV -V "Intel IPU6 CSI2 2:1 [fmt:SBGGR10/1280x800]" + +Once the media pipeline is configured, desired sensor specific settings +(such as exposure and gain settings) can be set, using the yavta tool. + +e.g + +.. code-block:: none + + # and that ov01a10 sensor is connected to i2c bus 3 with address 0x36 + export SDEV=$(media-ctl -d $MDEV -e "ov01a10 3-0036") + + yavta -w 0x009e0903 400 $SDEV + yavta -w 0x009e0913 1000 $SDEV + yavta -w 0x009e0911 2000 $SDEV + +Once the desired sensor settings are set, frame captures can be done as below. + +e.g + +.. code-block:: none + + yavta --data-prefix -u -c10 -n5 -I -s 1280x800 --file=/tmp/frame-#.bin \ + -f SBGGR10 $(media-ctl -d $MDEV -e "Intel IPU6 ISYS Capture 0") + +With the above command, 10 frames are captured at 1280x800 resolution with +sBGGR10 format. The captured frames are available as /tmp/frame-#.bin files. + +Here is another example of IPU6 ISYS RAW and metadata capture from camera +sensor ov2740 on Lenovo X1 Yoga laptop. + +.. code-block:: none + + media-ctl -l "\"ov2740 14-0036\":0 -> \"Intel IPU6 CSI2 1\":0[1]" + media-ctl -l "\"Intel IPU6 CSI2 1\":1 -> \"Intel IPU6 ISYS Capture 0\":0[5]" + media-ctl -l "\"Intel IPU6 CSI2 1\":2 -> \"Intel IPU6 ISYS Capture 1\":0[5]" + + # set routing + media-ctl -v -R "\"Intel IPU6 CSI2 1\" [0/0->1/0[1],0/1->2/1[1]]" + + media-ctl -v "\"Intel IPU6 CSI2 1\":0/0 [fmt:SGRBG10/1932x1092]" + media-ctl -v "\"Intel IPU6 CSI2 1\":0/1 [fmt:GENERIC_8/97x1]" + media-ctl -v "\"Intel IPU6 CSI2 1\":1/0 [fmt:SGRBG10/1932x1092]" + media-ctl -v "\"Intel IPU6 CSI2 1\":2/1 [fmt:GENERIC_8/97x1]" + + CAPTURE_DEV=$(media-ctl -e "Intel IPU6 ISYS Capture 0") + ./yavta --data-prefix -c100 -n5 -I -s1932x1092 --file=/tmp/frame-#.bin \ + -f SGRBG10 ${CAPTURE_DEV} + + CAPTURE_META=$(media-ctl -e "Intel IPU6 ISYS Capture 1") + ./yavta --data-prefix -c100 -n5 -I -s97x1 -B meta-capture \ + --file=/tmp/meta-#.bin -f GENERIC_8 ${CAPTURE_META} + +References +========== + +.. [#f1] https://git.ideasonboard.org/?p=media-ctl.git;a=summary +.. [#f2] https://git.ideasonboard.org/yavta.git
Documentation/admin-guide/media/ipu6_isys_graph.svg new file +338 -0
338 lines not on this page · raw message
Documentation/admin-guide/media/ipu6_isys_multi.svg new file +1124 -0
1,124 lines not on this page · raw message
Documentation/admin-guide/media/v4l-drivers.rst +1 -0
@@ -16,6 +16,7 @@ Video4Linux (V4L) driver-specific documentation imx imx7 ipu3 + ipu6-isys ivtv omap3isp omap4_camera
Signature · 1 line
2.42.0
Sakari Ailus
maintainerRe: [PATCH v2 14/15] Documentation: add Intel IPU6 ISYS driver admin-guide doc
Hi Bingbu,
Quoted 23 lines · On Tue, Oct 24, 2023 at 07:29:23PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This document mainly describe the functionality of IPU6 and IPU6 isys driver, and gives an example that how user can do imaging capture with tools.
Signed-off-by Bingbu Cao <[email protected]>--- Documentation/admin-guide/media/ipu6-isys.rst | 159 +++ .../admin-guide/media/ipu6_isys_graph.svg | 338 +++++ .../admin-guide/media/ipu6_isys_multi.svg | 1124 +++++++++++++++++ .../admin-guide/media/v4l-drivers.rst | 1 + 4 files changed, 1622 insertions(+) create mode 100644 Documentation/admin-guide/media/ipu6-isys.rst create mode 100644 Documentation/admin-guide/media/ipu6_isys_graph.svg create mode 100644 Documentation/admin-guide/media/ipu6_isys_multi.svgDocumentation/admin-guide/media/ipu6-isys.rst new file +0
-0
Documentation/admin-guide/media/ipu6-isys.rst line 7
Quoted 7 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +7
-0@@ -0,0 +1,159 @@ +.. SPDX-License-Identifier: GPL-2.0 + +.. include:: <isonum.txt> + +=============================================================== +Intel Image Processing Unit 6 (IPU6) Input System driver +===============================================================
Too many '='s.
Quoted 23 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +23
-0+ +Copyright |copy| 2023 Intel Corporation + +Introduction +============ + +This file documents the Intel IPU6 (6th generation Image Processing Unit) +Input System (MIPI CSI2 receiver) drivers located under +drivers/media/pci/intel/ipu6. + +The Intel IPU6 can be found in certain Intel Chipsets but not in all SKUs: + +* TigerLake +* JasperLake +* AlderLake +* RaptorLake +* MeteorLake + +Intel IPU6 is made up of two components - Input System (ISYS) and Processing +System (PSYS). + +The Input System mainly works as MIPI CSI2 receiver which receives and +processes the imaging data from the sensors and outputs the frames to memory.
Documentation/admin-guide/media/ipu6-isys.rst line 40
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +There are 2 driver modules - intel_ipu6 and intel_ipu6_isys. intel_ipu6 is an +IPU6 common driver which does PCI configuration, firmware loading and parsing, +firmware authentication, DMA mapping and IPU-MMU (internal Memory mapping Unit) +configuration. intel_ipu6_isys implements V4L2, Media Controller and V4L2 +sub-device interfaces. The IPU6 ISYS driver supports camera sensors connected +to the IPU6 ISYS through V4L2 sub-device sensor drivers. + +.. Note:: See Documentation/driver-api/media/drivers/ipu6.rst for more + information about the IPU6 hardware.
A direct reference would be nice.
Quoted 12 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +12
-0+ + +Input system driver +=================== + +The input System driver mainly configures CSI2 DPHY, constructs the firmware +stream configuration, sends commands to firmware, gets response from hardware +and firmware and then returns buffers to user. +The ISYS is represented as several V4L2 sub-devices - 'Intel IPU6 CSI2 $port', +which provide V4L2 subdev interfaces to the user space, there are also several +video nodes for each CSI-2 stream capture - 'Intel IPU6 ISYS capture $num' which +provide interface to user to set formats, queue buffers and streaming.
Documentation/admin-guide/media/ipu6-isys.rst line 62
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +.. kernel-figure:: ipu6_isys_graph.svg + :alt: ipu6 isys media graph without multiple streams support + + ipu6 isys media graph without multiple streams support + +.. kernel-figure:: ipu6_isys_multi.svg + :alt: ipu6 isys media graph with multiple streams support + + ipu6 isys media graph with multiple streams support
Is there a reason for having links from the CSI-2 receivers' source pads to multiple video nodes? Isn't a single one enough?
Quoted 97 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +97
-0+ +Capturing frames by IPU6 ISYS +------------------------------------ + +IPU6 ISYS is used to capture frames from the camera sensors connected to the +CSI2 ports. The supported input formats of ISYS are listed in table below: + +.. tabularcolumns:: |p{0.8cm}|p{4.0cm}|p{4.0cm}| + +.. flat-table:: + :header-rows: 1 + + * - IPU6 ISYS supported input formats + + * - RGB565, RGB888 + + * - UYVY8, YUYV8 + + * - RAW8, RAW10, RAW12 + +Here is an example of IPU6 ISYS raw capture on Dell XPS 9315 laptop. On this +machine, ov01a10 sensor is connected to IPU ISYS CSI2 port 2, which can +generate images at sBGGR10 with resolution 1280x800. + +Using the media controller APIs, we can configure ov01a10 sensor by +media-ctl [#f1]_ and yavta [#f2]_ to transmit frames to IPU6 ISYS. + +.. code-block:: none + + # Example 1 capture frame from ov01a10 camera sensor + # This example assumes /dev/media0 as the IPU ISYS media device + export MDEV=/dev/media0 + + # Establish the link for the media devices using media-ctl + media-ctl -d $MDEV -l "\"ov01a10 3-0036\":0 -> \"Intel IPU6 CSI2 2\":0[1]" + + # Set the format for the media devices + media-ctl -d $MDEV -V "ov01a10:0 [fmt:SBGGR10/1280x800]" + media-ctl -d $MDEV -V "Intel IPU6 CSI2 2:0 [fmt:SBGGR10/1280x800]" + media-ctl -d $MDEV -V "Intel IPU6 CSI2 2:1 [fmt:SBGGR10/1280x800]" + +Once the media pipeline is configured, desired sensor specific settings +(such as exposure and gain settings) can be set, using the yavta tool. + +e.g + +.. code-block:: none + + # and that ov01a10 sensor is connected to i2c bus 3 with address 0x36 + export SDEV=$(media-ctl -d $MDEV -e "ov01a10 3-0036") + + yavta -w 0x009e0903 400 $SDEV + yavta -w 0x009e0913 1000 $SDEV + yavta -w 0x009e0911 2000 $SDEV + +Once the desired sensor settings are set, frame captures can be done as below. + +e.g + +.. code-block:: none + + yavta --data-prefix -u -c10 -n5 -I -s 1280x800 --file=/tmp/frame-#.bin \ + -f SBGGR10 $(media-ctl -d $MDEV -e "Intel IPU6 ISYS Capture 0") + +With the above command, 10 frames are captured at 1280x800 resolution with +sBGGR10 format. The captured frames are available as /tmp/frame-#.bin files. + +Here is another example of IPU6 ISYS RAW and metadata capture from camera +sensor ov2740 on Lenovo X1 Yoga laptop. + +.. code-block:: none + + media-ctl -l "\"ov2740 14-0036\":0 -> \"Intel IPU6 CSI2 1\":0[1]" + media-ctl -l "\"Intel IPU6 CSI2 1\":1 -> \"Intel IPU6 ISYS Capture 0\":0[5]" + media-ctl -l "\"Intel IPU6 CSI2 1\":2 -> \"Intel IPU6 ISYS Capture 1\":0[5]" + + # set routing + media-ctl -v -R "\"Intel IPU6 CSI2 1\" [0/0->1/0[1],0/1->2/1[1]]" + + media-ctl -v "\"Intel IPU6 CSI2 1\":0/0 [fmt:SGRBG10/1932x1092]" + media-ctl -v "\"Intel IPU6 CSI2 1\":0/1 [fmt:GENERIC_8/97x1]" + media-ctl -v "\"Intel IPU6 CSI2 1\":1/0 [fmt:SGRBG10/1932x1092]" + media-ctl -v "\"Intel IPU6 CSI2 1\":2/1 [fmt:GENERIC_8/97x1]" + + CAPTURE_DEV=$(media-ctl -e "Intel IPU6 ISYS Capture 0") + ./yavta --data-prefix -c100 -n5 -I -s1932x1092 --file=/tmp/frame-#.bin \ + -f SGRBG10 ${CAPTURE_DEV} + + CAPTURE_META=$(media-ctl -e "Intel IPU6 ISYS Capture 1") + ./yavta --data-prefix -c100 -n5 -I -s97x1 -B meta-capture \ + --file=/tmp/meta-#.bin -f GENERIC_8 ${CAPTURE_META} + +References +========== + +.. [#f1] https://git.ideasonboard.org/?p=media-ctl.git;a=summary +.. [#f2] https://git.ideasonboard.org/yavta.git
Signature · 3 lines
Regards, Sakari Ailus
Re: [PATCH v2 14/15] Documentation: add Intel IPU6 ISYS driver admin-guide doc
Sakari,
Quoted 89 lines · Sakari Ailus
Hi Bingbu,
Quoted 23 lines · On Tue, Oct 24, 2023 at 07:29:23PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This document mainly describe the functionality of IPU6 and IPU6 isys driver, and gives an example that how user can do imaging capture with tools.
Signed-off-by Bingbu Cao <[email protected]>--- Documentation/admin-guide/media/ipu6-isys.rst | 159 +++ .../admin-guide/media/ipu6_isys_graph.svg | 338 +++++ .../admin-guide/media/ipu6_isys_multi.svg | 1124 +++++++++++++++++ .../admin-guide/media/v4l-drivers.rst | 1 + 4 files changed, 1622 insertions(+) create mode 100644 Documentation/admin-guide/media/ipu6-isys.rst create mode 100644 Documentation/admin-guide/media/ipu6_isys_graph.svg create mode 100644 Documentation/admin-guide/media/ipu6_isys_multi.svgDocumentation/admin-guide/media/ipu6-isys.rst new file +0
-0Documentation/admin-guide/media/ipu6-isys.rst line 7
Quoted 7 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +7
-0@@ -0,0 +1,159 @@ +.. SPDX-License-Identifier: GPL-2.0 + +.. include:: <isonum.txt> + +=============================================================== +Intel Image Processing Unit 6 (IPU6) Input System driver +===============================================================Too many '='s.
Quoted 23 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +23
-0+ +Copyright |copy| 2023 Intel Corporation + +Introduction +============ + +This file documents the Intel IPU6 (6th generation Image Processing Unit) +Input System (MIPI CSI2 receiver) drivers located under +drivers/media/pci/intel/ipu6. + +The Intel IPU6 can be found in certain Intel Chipsets but not in all SKUs: + +* TigerLake +* JasperLake +* AlderLake +* RaptorLake +* MeteorLake + +Intel IPU6 is made up of two components - Input System (ISYS) and Processing +System (PSYS). + +The Input System mainly works as MIPI CSI2 receiver which receives and +processes the imaging data from the sensors and outputs the frames to memory.Documentation/admin-guide/media/ipu6-isys.rst line 40
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +There are 2 driver modules - intel_ipu6 and intel_ipu6_isys. intel_ipu6 is an +IPU6 common driver which does PCI configuration, firmware loading and parsing, +firmware authentication, DMA mapping and IPU-MMU (internal Memory mapping Unit) +configuration. intel_ipu6_isys implements V4L2, Media Controller and V4L2 +sub-device interfaces. The IPU6 ISYS driver supports camera sensors connected +to the IPU6 ISYS through V4L2 sub-device sensor drivers. + +.. Note:: See Documentation/driver-api/media/drivers/ipu6.rst for more + information about the IPU6 hardware.A direct reference would be nice.
Quoted 17 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +17
-0+ + +Input system driver +=================== + +The input System driver mainly configures CSI2 DPHY, constructs the firmware +stream configuration, sends commands to firmware, gets response from hardware +and firmware and then returns buffers to user. +The ISYS is represented as several V4L2 sub-devices - 'Intel IPU6 CSI2 $port', +which provide V4L2 subdev interfaces to the user space, there are also several +video nodes for each CSI-2 stream capture - 'Intel IPU6 ISYS capture $num' which +provide interface to user to set formats, queue buffers and streaming. + +.. kernel-figure:: ipu6_isys_graph.svg + :alt: ipu6 isys media graph without multiple streams support + + ipu6 isys media graph without multiple streams support
Quoted 9 lines
Documentation/admin-guide/media/ipu6-isys.rst line 62
Quoted 5 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +5
-0+ +.. kernel-figure:: ipu6_isys_multi.svg + :alt: ipu6 isys media graph with multiple streams support + + ipu6 isys media graph with multiple streams supportIs there a reason for having links from the CSI-2 receivers' source pads to multiple video nodes? Isn't a single one enough?
Sorry, I forgot to reply before. Firmware sees each MIPI VC as a single stream, for multiple VCs case, SW need create multiple firmware streams.
<snip>
Signature · 2 lines
Best regards, Bingbu Cao
Sakari Ailus
maintainerRe: [PATCH v2 14/15] Documentation: add Intel IPU6 ISYS driver admin-guide doc
Hi Bingbu,
Quoted 102 lines · Bingbu Cao
Sakari,
Quoted 89 lines · Sakari Ailus
Hi Bingbu,
Quoted 23 lines · On Tue, Oct 24, 2023 at 07:29:23PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This document mainly describe the functionality of IPU6 and IPU6 isys driver, and gives an example that how user can do imaging capture with tools.
Signed-off-by Bingbu Cao <[email protected]>--- Documentation/admin-guide/media/ipu6-isys.rst | 159 +++ .../admin-guide/media/ipu6_isys_graph.svg | 338 +++++ .../admin-guide/media/ipu6_isys_multi.svg | 1124 +++++++++++++++++ .../admin-guide/media/v4l-drivers.rst | 1 + 4 files changed, 1622 insertions(+) create mode 100644 Documentation/admin-guide/media/ipu6-isys.rst create mode 100644 Documentation/admin-guide/media/ipu6_isys_graph.svg create mode 100644 Documentation/admin-guide/media/ipu6_isys_multi.svgDocumentation/admin-guide/media/ipu6-isys.rst new file +0
-0Documentation/admin-guide/media/ipu6-isys.rst line 7
Quoted 7 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +7
-0@@ -0,0 +1,159 @@ +.. SPDX-License-Identifier: GPL-2.0 + +.. include:: <isonum.txt> + +=============================================================== +Intel Image Processing Unit 6 (IPU6) Input System driver +===============================================================Too many '='s.
Quoted 23 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +23
-0+ +Copyright |copy| 2023 Intel Corporation + +Introduction +============ + +This file documents the Intel IPU6 (6th generation Image Processing Unit) +Input System (MIPI CSI2 receiver) drivers located under +drivers/media/pci/intel/ipu6. + +The Intel IPU6 can be found in certain Intel Chipsets but not in all SKUs: + +* TigerLake +* JasperLake +* AlderLake +* RaptorLake +* MeteorLake + +Intel IPU6 is made up of two components - Input System (ISYS) and Processing +System (PSYS). + +The Input System mainly works as MIPI CSI2 receiver which receives and +processes the imaging data from the sensors and outputs the frames to memory.Documentation/admin-guide/media/ipu6-isys.rst line 40
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +There are 2 driver modules - intel_ipu6 and intel_ipu6_isys. intel_ipu6 is an +IPU6 common driver which does PCI configuration, firmware loading and parsing, +firmware authentication, DMA mapping and IPU-MMU (internal Memory mapping Unit) +configuration. intel_ipu6_isys implements V4L2, Media Controller and V4L2 +sub-device interfaces. The IPU6 ISYS driver supports camera sensors connected +to the IPU6 ISYS through V4L2 sub-device sensor drivers. + +.. Note:: See Documentation/driver-api/media/drivers/ipu6.rst for more + information about the IPU6 hardware.A direct reference would be nice.
Quoted 17 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +17
-0+ + +Input system driver +=================== + +The input System driver mainly configures CSI2 DPHY, constructs the firmware +stream configuration, sends commands to firmware, gets response from hardware +and firmware and then returns buffers to user. +The ISYS is represented as several V4L2 sub-devices - 'Intel IPU6 CSI2 $port', +which provide V4L2 subdev interfaces to the user space, there are also several +video nodes for each CSI-2 stream capture - 'Intel IPU6 ISYS capture $num' which +provide interface to user to set formats, queue buffers and streaming. + +.. kernel-figure:: ipu6_isys_graph.svg + :alt: ipu6 isys media graph without multiple streams support + + ipu6 isys media graph without multiple streams supportQuoted 9 lines
Documentation/admin-guide/media/ipu6-isys.rst line 62
Quoted 5 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +5
-0+ +.. kernel-figure:: ipu6_isys_multi.svg + :alt: ipu6 isys media graph with multiple streams support + + ipu6 isys media graph with multiple streams supportIs there a reason for having links from the CSI-2 receivers' source pads to multiple video nodes? Isn't a single one enough?
Quoted 2 lines
Sorry, I forgot to reply before. Firmware sees each MIPI VC as a single stream, for multiple VCs case, SW need create multiple firmware streams.
Yes, I understand that, but this does not explain why you have multiple links between the source pads and video nodes. You can only capture one stream at a time from a video node. The user needs to select the correct video node related to a source pad, not to pick a random video node and then figure out which link to enable.
The media graph does look impressive though. :-)
Signature · 3 lines
Regards, Sakari Ailus
Re: [PATCH v2 14/15] Documentation: add Intel IPU6 ISYS driver admin-guide doc
Sakari,
Quoted 105 lines · Sakari Ailus
Hi Bingbu,
Quoted 102 lines · Bingbu Cao
Sakari,
Quoted 98 lines · Sakari Ailus
Hi Bingbu,
Quoted 23 lines · On Tue, Oct 24, 2023 at 07:29:23PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This document mainly describe the functionality of IPU6 and IPU6 isys driver, and gives an example that how user can do imaging capture with tools.
Signed-off-by Bingbu Cao <[email protected]>--- Documentation/admin-guide/media/ipu6-isys.rst | 159 +++ .../admin-guide/media/ipu6_isys_graph.svg | 338 +++++ .../admin-guide/media/ipu6_isys_multi.svg | 1124 +++++++++++++++++ .../admin-guide/media/v4l-drivers.rst | 1 + 4 files changed, 1622 insertions(+) create mode 100644 Documentation/admin-guide/media/ipu6-isys.rst create mode 100644 Documentation/admin-guide/media/ipu6_isys_graph.svg create mode 100644 Documentation/admin-guide/media/ipu6_isys_multi.svgDocumentation/admin-guide/media/ipu6-isys.rst new file +0
-0Documentation/admin-guide/media/ipu6-isys.rst line 7
Quoted 7 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +7
-0@@ -0,0 +1,159 @@ +.. SPDX-License-Identifier: GPL-2.0 + +.. include:: <isonum.txt> + +=============================================================== +Intel Image Processing Unit 6 (IPU6) Input System driver +===============================================================Too many '='s.
Quoted 23 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +23
-0+ +Copyright |copy| 2023 Intel Corporation + +Introduction +============ + +This file documents the Intel IPU6 (6th generation Image Processing Unit) +Input System (MIPI CSI2 receiver) drivers located under +drivers/media/pci/intel/ipu6. + +The Intel IPU6 can be found in certain Intel Chipsets but not in all SKUs: + +* TigerLake +* JasperLake +* AlderLake +* RaptorLake +* MeteorLake + +Intel IPU6 is made up of two components - Input System (ISYS) and Processing +System (PSYS). + +The Input System mainly works as MIPI CSI2 receiver which receives and +processes the imaging data from the sensors and outputs the frames to memory.Documentation/admin-guide/media/ipu6-isys.rst line 40
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +There are 2 driver modules - intel_ipu6 and intel_ipu6_isys. intel_ipu6 is an +IPU6 common driver which does PCI configuration, firmware loading and parsing, +firmware authentication, DMA mapping and IPU-MMU (internal Memory mapping Unit) +configuration. intel_ipu6_isys implements V4L2, Media Controller and V4L2 +sub-device interfaces. The IPU6 ISYS driver supports camera sensors connected +to the IPU6 ISYS through V4L2 sub-device sensor drivers. + +.. Note:: See Documentation/driver-api/media/drivers/ipu6.rst for more + information about the IPU6 hardware.A direct reference would be nice.
Quoted 12 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +12
-0+ + +Input system driver +=================== + +The input System driver mainly configures CSI2 DPHY, constructs the firmware +stream configuration, sends commands to firmware, gets response from hardware +and firmware and then returns buffers to user. +The ISYS is represented as several V4L2 sub-devices - 'Intel IPU6 CSI2 $port', +which provide V4L2 subdev interfaces to the user space, there are also several +video nodes for each CSI-2 stream capture - 'Intel IPU6 ISYS capture $num' which +provide interface to user to set formats, queue buffers and streaming.Documentation/admin-guide/media/ipu6-isys.rst line 62
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +.. kernel-figure:: ipu6_isys_graph.svg + :alt: ipu6 isys media graph without multiple streams support + + ipu6 isys media graph without multiple streams support + +.. kernel-figure:: ipu6_isys_multi.svg + :alt: ipu6 isys media graph with multiple streams support + + ipu6 isys media graph with multiple streams supportIs there a reason for having links from the CSI-2 receivers' source pads to multiple video nodes? Isn't a single one enough?
Quoted 8 lines
Quoted 2 lines
Sorry, I forgot to reply before. Firmware sees each MIPI VC as a single stream, for multiple VCs case, SW need create multiple firmware streams.
Yes, I understand that, but this does not explain why you have multiple links between the source pads and video nodes. You can only capture one stream at a time from a video node. The user needs to select the correct video node related to a source pad, not to pick a random video node and then figure out which link to enable.
The user can select any of the video nodes to one source pad. This figure shows that user have the capability to link any source pad to any of the video node (in dotted line). In reality, only 1 link is active (solid line).
Quoted 3 lines
The media graph does look impressive though. :-)
Signature · 2 lines
Best regards, Bingbu Cao
Sakari Ailus
maintainerRe: [PATCH v2 14/15] Documentation: add Intel IPU6 ISYS driver admin-guide doc
Hi Bingbu,
Quoted 117 lines · Bingbu Cao
Sakari,
Quoted 105 lines · Sakari Ailus
Hi Bingbu,
Quoted 102 lines · Bingbu Cao
Sakari,
Quoted 98 lines · Sakari Ailus
Hi Bingbu,
Quoted 23 lines · On Tue, Oct 24, 2023 at 07:29:23PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This document mainly describe the functionality of IPU6 and IPU6 isys driver, and gives an example that how user can do imaging capture with tools.
Signed-off-by Bingbu Cao <[email protected]>--- Documentation/admin-guide/media/ipu6-isys.rst | 159 +++ .../admin-guide/media/ipu6_isys_graph.svg | 338 +++++ .../admin-guide/media/ipu6_isys_multi.svg | 1124 +++++++++++++++++ .../admin-guide/media/v4l-drivers.rst | 1 + 4 files changed, 1622 insertions(+) create mode 100644 Documentation/admin-guide/media/ipu6-isys.rst create mode 100644 Documentation/admin-guide/media/ipu6_isys_graph.svg create mode 100644 Documentation/admin-guide/media/ipu6_isys_multi.svgDocumentation/admin-guide/media/ipu6-isys.rst new file +0
-0Documentation/admin-guide/media/ipu6-isys.rst line 7
Quoted 7 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +7
-0@@ -0,0 +1,159 @@ +.. SPDX-License-Identifier: GPL-2.0 + +.. include:: <isonum.txt> + +=============================================================== +Intel Image Processing Unit 6 (IPU6) Input System driver +===============================================================Too many '='s.
Quoted 23 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +23
-0+ +Copyright |copy| 2023 Intel Corporation + +Introduction +============ + +This file documents the Intel IPU6 (6th generation Image Processing Unit) +Input System (MIPI CSI2 receiver) drivers located under +drivers/media/pci/intel/ipu6. + +The Intel IPU6 can be found in certain Intel Chipsets but not in all SKUs: + +* TigerLake +* JasperLake +* AlderLake +* RaptorLake +* MeteorLake + +Intel IPU6 is made up of two components - Input System (ISYS) and Processing +System (PSYS). + +The Input System mainly works as MIPI CSI2 receiver which receives and +processes the imaging data from the sensors and outputs the frames to memory.Documentation/admin-guide/media/ipu6-isys.rst line 40
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +There are 2 driver modules - intel_ipu6 and intel_ipu6_isys. intel_ipu6 is an +IPU6 common driver which does PCI configuration, firmware loading and parsing, +firmware authentication, DMA mapping and IPU-MMU (internal Memory mapping Unit) +configuration. intel_ipu6_isys implements V4L2, Media Controller and V4L2 +sub-device interfaces. The IPU6 ISYS driver supports camera sensors connected +to the IPU6 ISYS through V4L2 sub-device sensor drivers. + +.. Note:: See Documentation/driver-api/media/drivers/ipu6.rst for more + information about the IPU6 hardware.A direct reference would be nice.
Quoted 12 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +12
-0+ + +Input system driver +=================== + +The input System driver mainly configures CSI2 DPHY, constructs the firmware +stream configuration, sends commands to firmware, gets response from hardware +and firmware and then returns buffers to user. +The ISYS is represented as several V4L2 sub-devices - 'Intel IPU6 CSI2 $port', +which provide V4L2 subdev interfaces to the user space, there are also several +video nodes for each CSI-2 stream capture - 'Intel IPU6 ISYS capture $num' which +provide interface to user to set formats, queue buffers and streaming.Documentation/admin-guide/media/ipu6-isys.rst line 62
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +.. kernel-figure:: ipu6_isys_graph.svg + :alt: ipu6 isys media graph without multiple streams support + + ipu6 isys media graph without multiple streams support + +.. kernel-figure:: ipu6_isys_multi.svg + :alt: ipu6 isys media graph with multiple streams support + + ipu6 isys media graph with multiple streams supportIs there a reason for having links from the CSI-2 receivers' source pads to multiple video nodes? Isn't a single one enough?
Quoted 8 lines
Quoted 2 lines
Sorry, I forgot to reply before. Firmware sees each MIPI VC as a single stream, for multiple VCs case, SW need create multiple firmware streams.
Yes, I understand that, but this does not explain why you have multiple links between the source pads and video nodes. You can only capture one stream at a time from a video node. The user needs to select the correct video node related to a source pad, not to pick a random video node and then figure out which link to enable.
Quoted 4 lines
The user can select any of the video nodes to one source pad. This figure shows that user have the capability to link any source pad to any of the video node (in dotted line). In reality, only 1 link is active (solid line).
Yes, that's how it's currently implemented but I don't think there's really a reason to do that. Could you remove all but one link to each video node?
Signature · 3 lines
Regards, Sakari Ailus
Re: [PATCH v2 14/15] Documentation: add Intel IPU6 ISYS driver admin-guide doc
Sakari,
Quoted 120 lines · Sakari Ailus
Hi Bingbu,
Quoted 117 lines · Bingbu Cao
Sakari,
Quoted 113 lines · Sakari Ailus
Hi Bingbu,
Quoted 102 lines · Bingbu Cao
Sakari,
Quoted 89 lines · Sakari Ailus
Hi Bingbu,
Quoted 23 lines · On Tue, Oct 24, 2023 at 07:29:23PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This document mainly describe the functionality of IPU6 and IPU6 isys driver, and gives an example that how user can do imaging capture with tools.
Signed-off-by Bingbu Cao <[email protected]>--- Documentation/admin-guide/media/ipu6-isys.rst | 159 +++ .../admin-guide/media/ipu6_isys_graph.svg | 338 +++++ .../admin-guide/media/ipu6_isys_multi.svg | 1124 +++++++++++++++++ .../admin-guide/media/v4l-drivers.rst | 1 + 4 files changed, 1622 insertions(+) create mode 100644 Documentation/admin-guide/media/ipu6-isys.rst create mode 100644 Documentation/admin-guide/media/ipu6_isys_graph.svg create mode 100644 Documentation/admin-guide/media/ipu6_isys_multi.svgDocumentation/admin-guide/media/ipu6-isys.rst new file +0
-0Documentation/admin-guide/media/ipu6-isys.rst line 7
Quoted 7 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +7
-0@@ -0,0 +1,159 @@ +.. SPDX-License-Identifier: GPL-2.0 + +.. include:: <isonum.txt> + +=============================================================== +Intel Image Processing Unit 6 (IPU6) Input System driver +===============================================================Too many '='s.
Quoted 23 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +23
-0+ +Copyright |copy| 2023 Intel Corporation + +Introduction +============ + +This file documents the Intel IPU6 (6th generation Image Processing Unit) +Input System (MIPI CSI2 receiver) drivers located under +drivers/media/pci/intel/ipu6. + +The Intel IPU6 can be found in certain Intel Chipsets but not in all SKUs: + +* TigerLake +* JasperLake +* AlderLake +* RaptorLake +* MeteorLake + +Intel IPU6 is made up of two components - Input System (ISYS) and Processing +System (PSYS). + +The Input System mainly works as MIPI CSI2 receiver which receives and +processes the imaging data from the sensors and outputs the frames to memory.Documentation/admin-guide/media/ipu6-isys.rst line 40
Quoted 10 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +10
-0+ +There are 2 driver modules - intel_ipu6 and intel_ipu6_isys. intel_ipu6 is an +IPU6 common driver which does PCI configuration, firmware loading and parsing, +firmware authentication, DMA mapping and IPU-MMU (internal Memory mapping Unit) +configuration. intel_ipu6_isys implements V4L2, Media Controller and V4L2 +sub-device interfaces. The IPU6 ISYS driver supports camera sensors connected +to the IPU6 ISYS through V4L2 sub-device sensor drivers. + +.. Note:: See Documentation/driver-api/media/drivers/ipu6.rst for more + information about the IPU6 hardware.A direct reference would be nice.
Quoted 17 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +17
-0+ + +Input system driver +=================== + +The input System driver mainly configures CSI2 DPHY, constructs the firmware +stream configuration, sends commands to firmware, gets response from hardware +and firmware and then returns buffers to user. +The ISYS is represented as several V4L2 sub-devices - 'Intel IPU6 CSI2 $port', +which provide V4L2 subdev interfaces to the user space, there are also several +video nodes for each CSI-2 stream capture - 'Intel IPU6 ISYS capture $num' which +provide interface to user to set formats, queue buffers and streaming. + +.. kernel-figure:: ipu6_isys_graph.svg + :alt: ipu6 isys media graph without multiple streams support + + ipu6 isys media graph without multiple streams supportQuoted 9 lines
Documentation/admin-guide/media/ipu6-isys.rst line 62
Quoted 5 lines
Documentation/admin-guide/media/ipu6-isys.rst new file +5
-0+ +.. kernel-figure:: ipu6_isys_multi.svg + :alt: ipu6 isys media graph with multiple streams support + + ipu6 isys media graph with multiple streams supportIs there a reason for having links from the CSI-2 receivers' source pads to multiple video nodes? Isn't a single one enough?
Quoted 2 lines
Sorry, I forgot to reply before. Firmware sees each MIPI VC as a single stream, for multiple VCs case, SW need create multiple firmware streams.
Yes, I understand that, but this does not explain why you have multiple links between the source pads and video nodes. You can only capture one stream at a time from a video node. The user needs to select the correct video node related to a source pad, not to pick a random video node and then figure out which link to enable.
Quoted 8 lines
Quoted 4 lines
The user can select any of the video nodes to one source pad. This figure shows that user have the capability to link any source pad to any of the video node (in dotted line). In reality, only 1 link is active (solid line).
Yes, that's how it's currently implemented but I don't think there's really a reason to do that. Could you remove all but one link to each video node?
Ack. I generated this graph by media-ctl and dot automatically. Let me try to draw a new one.
Thanks!
Signature · 2 lines
Best regards, Bingbu Cao
PATCH v2 15/15 Documentation: add documentation of Intel IPU6 driver and hardware overview
In reply to [email protected]From: Bingbu Cao <[email protected]>
Add a documentation for an overview of IPU6 hardware and describe the main the components of IPU6 driver.
--- .../driver-api/media/drivers/index.rst | 1 + .../driver-api/media/drivers/ipu6.rst | 205 ++++++++++++++++++ 2 files changed, 206 insertions(+) create mode 100644 Documentation/driver-api/media/drivers/ipu6.rst
Documentation/driver-api/media/drivers/index.rst +1 -0
@@ -26,6 +26,7 @@ Video4Linux (V4L) drivers vimc-devel zoran ccs/ccs + ipu6 Digital TV drivers
Documentation/driver-api/media/drivers/ipu6.rst new file +205 -0
205 lines not on this page · raw message
Signature · 1 line
2.42.0
Sakari Ailus
maintainerRe: [PATCH v2 15/15] Documentation: add documentation of Intel IPU6 driver and hardware overview
Hi Bingbu,
Thanks for the patch. It's a nice piece of documentation for the most complex CSI-2 receiver driver. :-)
Quoted 34 lines · On Tue, Oct 24, 2023 at 07:29:24PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Add a documentation for an overview of IPU6 hardware and describe the main the components of IPU6 driver.
Signed-off-by Bingbu Cao <[email protected]>--- .../driver-api/media/drivers/index.rst | 1 + .../driver-api/media/drivers/ipu6.rst | 205 ++++++++++++++++++ 2 files changed, 206 insertions(+) create mode 100644 Documentation/driver-api/media/drivers/ipu6.rstDocumentation/driver-api/media/drivers/index.rst +1
-0@@ -26,6 +26,7 @@ Video4Linux (V4L) drivers vimc-devel zoran ccs/ccs + ipu6 Digital TV driversDocumentation/driver-api/media/drivers/ipu6.rst new file +4
-0@@ -0,0 +1,205 @@ +.. SPDX-License-Identifier: GPL-2.0 + +================== +Intel IPU6 Driver
Documentation/driver-api/media/drivers/ipu6.rst line 14
Quoted 10 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +10
-0+================== + +Author: Bingbu Cao <[email protected]> + +Overview +========= + +Intel IPU6 is the sixth generation of Intel Image Processing Unit used in some +Intel Chipsets such as Tiger Lake, Jasper Lake, Alder Lake, Raptor Lake and +Meteor Lake. IPU6 consists of two major systems - Input System (IS) and
I'd use a semicolon: ^
Quoted 135 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +135
-0+Processing System (PS). IPU6 are visible on the PCI bus as a single device, +it can be found by ``lspci``: + +``0000:00:05.0 Multimedia controller: Intel Corporation Device xxxx (rev xx)`` + +IPU6 has a 16 MB BAR in PCI configuration Space for MMIO registers which is +visible for driver. + +Buttress +========= + +The IPU6 is connecting to the system fabric with ``Buttress`` which is enabling +host driver to control the IPU6, it also allows IPU6 access the system memory to +store and load frame pixel streams and any other metadata. + +``Buttress`` mainly manages several system functionalities - power management, +interrupt handling, firmware authentication and global timer sync. + +IS and PS Power flow +--------------------------- + +IPU6 driver initialize the IS and PS power up or down request by setting the +Buttress frequency control register for IS and PS - +``IPU6_BUTTRESS_REG_IS_FREQ_CTL`` and ``IPU6_BUTTRESS_REG_PS_FREQ_CTL`` in +function: + +.. c:function:: int ipu6_buttress_power(..., bool on) + +Buttress forwards the request to Punit, after Punit execute the power up flow, +buttress indicates driver that IS or PS is powered up by updating the power +status registers. + +.. Note:: IS power up needs take place prior to PS power up, IS power down needs + take place after PS power down due to hardware limitation. + + +Interrupt +------------ + +IPU6 interrupt can be generated as MSI or INTA, interrupt will be triggered +when IS, PS, Buttress event or error happen, driver can get the interrupt +cause by reading the interrupt status register ``BUTTRESS_REG_ISR_STATUS``, +driver firstly clear the irq status and then call specific IS or PS irq handler. + +.. c:function:: irqreturn_t ipu6_buttress_isr(int irq, ...) + +Security and firmware authentication +------------------------------------- +To address the IPU6 firmware security concerns, the IPU6 firmware needs to +undergo an authentication process before it is allowed to executed on the IPU6 +internal processors. Driver will work with Converged Security Engine (CSE) to +complete authentication process. CSE is responsible of authenticating the +IPU6 firmware, the authenticated firmware binary is copied into an isolated +memory region. Firmware authentication process is implemented by CSE following +an IPC handshake with driver. There are some Buttress registers used by CSE and +driver to communicate with each other as IPC messages. + +.. c:function:: int ipu6_buttress_authenticate(...) + +Global timer sync +------------------ +IPU driver initiates a Hammock Harbor synchronization flow each time it starts +camera operation. IPU will synchronizes an internal counter in the Buttress +with a copy of SoC time, this counter keeps the updated time until camera +operation is stopped. Driver can use this time counter to calibrate the +timestamp based on the timestamp in response event from firmware. + +.. c:function:: int ipu6_buttress_start_tsc_sync(...) + + +DMA and MMU +============ + +IPU6 has its own scalar processor where the firmware run at, it has +an internal 32-bits virtual address space. IPU6 has MMU address translation +hardware to allow that scalar process access the internal memory and external +system memory through IPU6 virtual address. The address translation is +based on two levels of page lookup tables stored in system memory which are +maintained by IPU6 driver. IPU6 driver sets the level-1 page table base address +to MMU register and allow MMU to lookup the page table. + +IPU6 driver exports its own DMA operations. Driver will update the page table +entries for each DMA operation and invalidate the MMU TLB after each unmap and +free. + +.. code-block:: none + + const struct dma_map_ops ipu6_dma_ops = { + .alloc = ipu6_dma_alloc, + .free = ipu6_dma_free, + .mmap = ipu6_dma_mmap, + .map_sg = ipu6_dma_map_sg, + .unmap_sg = ipu6_dma_unmap_sg, + ... + }; + +.. Note:: IPU6 MMU works behind IOMMU, so for each IPU6 DMA ops, driver will + call generic PCI DMA ops to ask IOMMU to do the additional mapping + if VT-d enabled. + + +Firmware file format +===================== + +IPU6 release the firmware in Code Partition Directory (CPD) file format. The +CPD firmware contains a CPD header, several CPD entries and CPD components. +CPD component includes 3 entries - manifest, metadata and module data. Manifest +and metadata are defined by CSE and used by CSE for authentication. Module data +is defined by IPU6 which holds the binary data of firmware called package +directory. IPU6 driver (``ipu6-cpd.c``) parses and validates the CPD firmware +file and get the package directory binary data of IPU6 firmware, copy it to +specific DMA buffer and sets its base address to Buttress ``FW_SOURCE_BASE`` +register, CSE will do authentication for this firmware binary. + + +Syscom interface +================ + +IPU6 driver communicates with firmware via syscom ABI. Syscom is an +inter-processor communication mechanism between IPU scalar processor and CPU. +There are a number of resources shared between firmware and software. +A system memory region where the message queues reside, firmware can access the +memory region via IPU MMU. Syscom queues are FIFO fixed depth queues with +configurable elements ``token`` (message). There is also a common IPU MMIO +registers where the queue read and write indices reside. Software and firmware +work as producer and consumer of tokens in queue, and update the write and read +indices separately when sending or receiving each message. + +IPU6 driver must prepare and configure the number of input and output queues, +configure the count of tokens per queue and the size of per token before +initiate and start the communication with firmware, firmware and software must +use same configurations. IPU6 Buttress has a number of firmware boot parameter +registers which can be used to store the address of configuration and initiate +the Syscom state, then driver can request firmware to start and run via setting +the scalar processor control status register.
Documentation/driver-api/media/drivers/ipu6.rst line 159
Quoted 10 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +10
-0+ + +Input System +============== + +IPU6 input system consists of MIPI D-PHY and several CSI receiver controllers, +it can capture image pixel data from camera sensors or other MIPI CSI output +devices. + +DPHYs and CSI2 ports lane mapping
"D-PHY" and "CSI-2"
Quoted 33 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +33
-0+--------------------------------- + +IPU6 integrates different D-PHY IPs on different SoCs, on Tiger Lake and Alder +Lake, IPU6 integrates MCD10 D-PHY, IPU6SE on Jasper Lake integrates JSL D-PHY +and IPU6EP on Meteor Lake integrates a Synopsys DWC D-PHY. There is an adaption +layer between D-PHY and CSI receiver controller which includes port +configuration, PHY wrapper or private test interfaces for D-PHY. There are 3 +D-PHY drivers ``ipu6-isys-mcd-phy.c``, ``ipu6-isys-jsl-phy.c`` and +``ipu6-isys-dwc-phy.c`` program the above 3 D-PHYs in IPU6. + +Different IPU6 version has different DPHY lanes mappings, On Tiger Lake, there +are 12 data lanes and 8 clock lanes, IPU6 support maximum 8 CSI2 ports, see +the ppi mmapping in ``ipu6-isys-mcd-phy.c`` for more information. On Jasper Lake +and Alder Lake, DPHY has 8 data lanes and 4 clock lanes, IPU6 support maximum 4 +CSI2 ports. For Meteor Lake, DPHY has 12 data lanes and 6 clock lanes, IPU6 +support maximum 6 CSI2 ports. + +.. Note:: Each adjacent CSI ports work as a pair and share the data lanes. + For example, for CSI port 0 and 1, CSI port 0 support maximum 4 + data lanes, CSI port 1 support maximum 2 data lanes, CSI port 0 + with 2 data lanes can work together with CSI port 1 with 2 data lanes. + If trying to use CSI port 0 with 4 lanes, CSI port 1 will not be + available as the 4 data lanes are shared by CSI port 0 and 1. Same + scenario is also applied for CSI port 2/3, 4/5 and 7/8. + +IS firmware ABIs +---------------- + +IPU6 firmware define a series of ABIs to software. In general, software firstly +prepare the stream configuration ``struct ipu6_fw_isys_stream_cfg_data_abi`` +and send the configuration to firmware via sending ``STREAM_OPEN`` command. +Stream configuration includes input pins and output pins, input pin +``struct ipu6_fw_isys_input_pin_info_abi`` defines the resolution and data type
Documentation/driver-api/media/drivers/ipu6.rst line 202
Quoted 10 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +10
-0+of input source, output pin ``struct ipu6_fw_isys_output_pin_info_abi`` +defines the output resolution, stride and frame format, etc. Once driver get the +interrupt from firmware that indicates stream open successfully, driver will +send the ``STREAM_START`` and ``STREAM_CAPTURE`` command to request firmware to +start capturing image frames. ``STREAM_CAPTURE`` command queues the buffers to +firmware with ``struct ipu6_fw_isys_frame_buff_set``, software then wait the +interrupt and response from firmware, ``PIN_DATA_READY`` means data ready +on specific output pin and then software return the buffers to user. + +.. Note:: See Documentation/admin-guide/media/ipu6-isys.rst for how to do
Could you add a direct link instead? See e.g. Documentation/driver-api/media/camera-sensor.rst for an example.
Quoted 3 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +3
-0+ capture by IPU6 IS driver. + +
Signature · 3 lines
Regards, Sakari Ailus
Re: [PATCH v2 15/15] Documentation: add documentation of Intel IPU6 driver and hardware overview
Sakari,
Thanks for the review.
Quoted 44 lines · Sakari Ailus
Hi Bingbu,
Thanks for the patch. It's a nice piece of documentation for the most complex CSI-2 receiver driver. :-)
Quoted 38 lines · On Tue, Oct 24, 2023 at 07:29:24PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Add a documentation for an overview of IPU6 hardware and describe the main the components of IPU6 driver.
Signed-off-by Bingbu Cao <[email protected]>--- .../driver-api/media/drivers/index.rst | 1 + .../driver-api/media/drivers/ipu6.rst | 205 ++++++++++++++++++ 2 files changed, 206 insertions(+) create mode 100644 Documentation/driver-api/media/drivers/ipu6.rstDocumentation/driver-api/media/drivers/index.rst +1
-0@@ -26,6 +26,7 @@ Video4Linux (V4L) drivers vimc-devel zoran ccs/ccs + ipu6 Digital TV driversDocumentation/driver-api/media/drivers/ipu6.rst new file +8
-0@@ -0,0 +1,205 @@ +.. SPDX-License-Identifier: GPL-2.0 + +================== +Intel IPU6 Driver +================== + +Author: Bingbu Cao <[email protected]> +
Quoted 8 lines
Documentation/driver-api/media/drivers/ipu6.rst line 14
Quoted 6 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +6
-0+Overview +========= + +Intel IPU6 is the sixth generation of Intel Image Processing Unit used in some +Intel Chipsets such as Tiger Lake, Jasper Lake, Alder Lake, Raptor Lake and +Meteor Lake. IPU6 consists of two major systems - Input System (IS) andI'd use a semicolon: ^
semicolon or colon? :)
Quoted 140 lines
Quoted 140 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +139
-0+Processing System (PS). IPU6 are visible on the PCI bus as a single device, +it can be found by ``lspci``: + +``0000:00:05.0 Multimedia controller: Intel Corporation Device xxxx (rev xx)`` + +IPU6 has a 16 MB BAR in PCI configuration Space for MMIO registers which is +visible for driver. + +Buttress +========= + +The IPU6 is connecting to the system fabric with ``Buttress`` which is enabling +host driver to control the IPU6, it also allows IPU6 access the system memory to +store and load frame pixel streams and any other metadata. + +``Buttress`` mainly manages several system functionalities - power management, +interrupt handling, firmware authentication and global timer sync. + +IS and PS Power flow +--------------------------- + +IPU6 driver initialize the IS and PS power up or down request by setting the +Buttress frequency control register for IS and PS - +``IPU6_BUTTRESS_REG_IS_FREQ_CTL`` and ``IPU6_BUTTRESS_REG_PS_FREQ_CTL`` in +function: + +.. c:function:: int ipu6_buttress_power(..., bool on) + +Buttress forwards the request to Punit, after Punit execute the power up flow, +buttress indicates driver that IS or PS is powered up by updating the power +status registers. + +.. Note:: IS power up needs take place prior to PS power up, IS power down needs + take place after PS power down due to hardware limitation. + + +Interrupt +------------ + +IPU6 interrupt can be generated as MSI or INTA, interrupt will be triggered +when IS, PS, Buttress event or error happen, driver can get the interrupt +cause by reading the interrupt status register ``BUTTRESS_REG_ISR_STATUS``, +driver firstly clear the irq status and then call specific IS or PS irq handler. + +.. c:function:: irqreturn_t ipu6_buttress_isr(int irq, ...) + +Security and firmware authentication +------------------------------------- +To address the IPU6 firmware security concerns, the IPU6 firmware needs to +undergo an authentication process before it is allowed to executed on the IPU6 +internal processors. Driver will work with Converged Security Engine (CSE) to +complete authentication process. CSE is responsible of authenticating the +IPU6 firmware, the authenticated firmware binary is copied into an isolated +memory region. Firmware authentication process is implemented by CSE following +an IPC handshake with driver. There are some Buttress registers used by CSE and +driver to communicate with each other as IPC messages. + +.. c:function:: int ipu6_buttress_authenticate(...) + +Global timer sync +------------------ +IPU driver initiates a Hammock Harbor synchronization flow each time it starts +camera operation. IPU will synchronizes an internal counter in the Buttress +with a copy of SoC time, this counter keeps the updated time until camera +operation is stopped. Driver can use this time counter to calibrate the +timestamp based on the timestamp in response event from firmware. + +.. c:function:: int ipu6_buttress_start_tsc_sync(...) + + +DMA and MMU +============ + +IPU6 has its own scalar processor where the firmware run at, it has +an internal 32-bits virtual address space. IPU6 has MMU address translation +hardware to allow that scalar process access the internal memory and external +system memory through IPU6 virtual address. The address translation is +based on two levels of page lookup tables stored in system memory which are +maintained by IPU6 driver. IPU6 driver sets the level-1 page table base address +to MMU register and allow MMU to lookup the page table. + +IPU6 driver exports its own DMA operations. Driver will update the page table +entries for each DMA operation and invalidate the MMU TLB after each unmap and +free. + +.. code-block:: none + + const struct dma_map_ops ipu6_dma_ops = { + .alloc = ipu6_dma_alloc, + .free = ipu6_dma_free, + .mmap = ipu6_dma_mmap, + .map_sg = ipu6_dma_map_sg, + .unmap_sg = ipu6_dma_unmap_sg, + ... + }; + +.. Note:: IPU6 MMU works behind IOMMU, so for each IPU6 DMA ops, driver will + call generic PCI DMA ops to ask IOMMU to do the additional mapping + if VT-d enabled. + + +Firmware file format +===================== + +IPU6 release the firmware in Code Partition Directory (CPD) file format. The +CPD firmware contains a CPD header, several CPD entries and CPD components. +CPD component includes 3 entries - manifest, metadata and module data. Manifest +and metadata are defined by CSE and used by CSE for authentication. Module data +is defined by IPU6 which holds the binary data of firmware called package +directory. IPU6 driver (``ipu6-cpd.c``) parses and validates the CPD firmware +file and get the package directory binary data of IPU6 firmware, copy it to +specific DMA buffer and sets its base address to Buttress ``FW_SOURCE_BASE`` +register, CSE will do authentication for this firmware binary. + + +Syscom interface +================ + +IPU6 driver communicates with firmware via syscom ABI. Syscom is an +inter-processor communication mechanism between IPU scalar processor and CPU. +There are a number of resources shared between firmware and software. +A system memory region where the message queues reside, firmware can access the +memory region via IPU MMU. Syscom queues are FIFO fixed depth queues with +configurable elements ``token`` (message). There is also a common IPU MMIO +registers where the queue read and write indices reside. Software and firmware +work as producer and consumer of tokens in queue, and update the write and read +indices separately when sending or receiving each message. + +IPU6 driver must prepare and configure the number of input and output queues, +configure the count of tokens per queue and the size of per token before +initiate and start the communication with firmware, firmware and software must +use same configurations. IPU6 Buttress has a number of firmware boot parameter +registers which can be used to store the address of configuration and initiate +the Syscom state, then driver can request firmware to start and run via setting +the scalar processor control status register. + + +Input System +==============
Quoted 8 lines
Documentation/driver-api/media/drivers/ipu6.rst line 159
Quoted 6 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +6
-0+ +IPU6 input system consists of MIPI D-PHY and several CSI receiver controllers, +it can capture image pixel data from camera sensors or other MIPI CSI output +devices. + +DPHYs and CSI2 ports lane mapping"D-PHY" and "CSI-2"
Ack.
Quoted 39 lines
Quoted 39 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +38
-0+--------------------------------- + +IPU6 integrates different D-PHY IPs on different SoCs, on Tiger Lake and Alder +Lake, IPU6 integrates MCD10 D-PHY, IPU6SE on Jasper Lake integrates JSL D-PHY +and IPU6EP on Meteor Lake integrates a Synopsys DWC D-PHY. There is an adaption +layer between D-PHY and CSI receiver controller which includes port +configuration, PHY wrapper or private test interfaces for D-PHY. There are 3 +D-PHY drivers ``ipu6-isys-mcd-phy.c``, ``ipu6-isys-jsl-phy.c`` and +``ipu6-isys-dwc-phy.c`` program the above 3 D-PHYs in IPU6. + +Different IPU6 version has different DPHY lanes mappings, On Tiger Lake, there +are 12 data lanes and 8 clock lanes, IPU6 support maximum 8 CSI2 ports, see +the ppi mmapping in ``ipu6-isys-mcd-phy.c`` for more information. On Jasper Lake +and Alder Lake, DPHY has 8 data lanes and 4 clock lanes, IPU6 support maximum 4 +CSI2 ports. For Meteor Lake, DPHY has 12 data lanes and 6 clock lanes, IPU6 +support maximum 6 CSI2 ports. + +.. Note:: Each adjacent CSI ports work as a pair and share the data lanes. + For example, for CSI port 0 and 1, CSI port 0 support maximum 4 + data lanes, CSI port 1 support maximum 2 data lanes, CSI port 0 + with 2 data lanes can work together with CSI port 1 with 2 data lanes. + If trying to use CSI port 0 with 4 lanes, CSI port 1 will not be + available as the 4 data lanes are shared by CSI port 0 and 1. Same + scenario is also applied for CSI port 2/3, 4/5 and 7/8. + +IS firmware ABIs +---------------- + +IPU6 firmware define a series of ABIs to software. In general, software firstly +prepare the stream configuration ``struct ipu6_fw_isys_stream_cfg_data_abi`` +and send the configuration to firmware via sending ``STREAM_OPEN`` command. +Stream configuration includes input pins and output pins, input pin +``struct ipu6_fw_isys_input_pin_info_abi`` defines the resolution and data type +of input source, output pin ``struct ipu6_fw_isys_output_pin_info_abi`` +defines the output resolution, stride and frame format, etc. Once driver get the +interrupt from firmware that indicates stream open successfully, driver will +send the ``STREAM_START`` and ``STREAM_CAPTURE`` command to request firmware to +start capturing image frames. ``STREAM_CAPTURE`` command queues the buffers to
Quoted 8 lines
Documentation/driver-api/media/drivers/ipu6.rst line 202
Quoted 5 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +5
-0+firmware with ``struct ipu6_fw_isys_frame_buff_set``, software then wait the +interrupt and response from firmware, ``PIN_DATA_READY`` means data ready +on specific output pin and then software return the buffers to user. + +.. Note:: See Documentation/admin-guide/media/ipu6-isys.rst for how to doCould you add a direct link instead? See e.g.
Documentation/driver-api/media/camera-sensor.rstfor an example.
Good idea, thanks.
Quoted 5 lines
Quoted 4 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +3
-0+ capture by IPU6 IS driver. + +
Signature · 2 lines
Best regards, Bingbu Cao
Sakari Ailus
maintainerRe: [PATCH v2 15/15] Documentation: add documentation of Intel IPU6 driver and hardware overview
Hi Bingbu,
Quoted 58 lines · Bingbu Cao
Sakari,
Thanks for the review.
Quoted 44 lines · Sakari Ailus
Hi Bingbu,
Thanks for the patch. It's a nice piece of documentation for the most complex CSI-2 receiver driver. :-)
Quoted 38 lines · On Tue, Oct 24, 2023 at 07:29:24PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
Add a documentation for an overview of IPU6 hardware and describe the main the components of IPU6 driver.
Signed-off-by Bingbu Cao <[email protected]>--- .../driver-api/media/drivers/index.rst | 1 + .../driver-api/media/drivers/ipu6.rst | 205 ++++++++++++++++++ 2 files changed, 206 insertions(+) create mode 100644 Documentation/driver-api/media/drivers/ipu6.rstDocumentation/driver-api/media/drivers/index.rst +1
-0@@ -26,6 +26,7 @@ Video4Linux (V4L) drivers vimc-devel zoran ccs/ccs + ipu6 Digital TV driversDocumentation/driver-api/media/drivers/ipu6.rst new file +8
-0@@ -0,0 +1,205 @@ +.. SPDX-License-Identifier: GPL-2.0 + +================== +Intel IPU6 Driver +================== + +Author: Bingbu Cao <[email protected]> +Quoted 8 lines
Documentation/driver-api/media/drivers/ipu6.rst line 14
Quoted 6 lines
Documentation/driver-api/media/drivers/ipu6.rst new file +6
-0+Overview +========= + +Intel IPU6 is the sixth generation of Intel Image Processing Unit used in some +Intel Chipsets such as Tiger Lake, Jasper Lake, Alder Lake, Raptor Lake and +Meteor Lake. IPU6 consists of two major systems - Input System (IS) andI'd use a semicolon: ^
Quoted 1 line
semicolon or colon? :)
A colon, yes, indeed.
Signature · 3 lines
Regards, Sakari Ailus
Hi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt: V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.
Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
Regards,
Hans
Hans,
------------------------------------------------------------------------
BRs,
Bingbu Cao
Quoted 35 lines
-----Original Message-----
From: Hans de Goede <[email protected]>
Sent: Wednesday, November 8, 2023 8:00 PM
To: Cao, Bingbu <[email protected]>; [email protected]; [email protected]; [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Qiu, Tian Shu <[email protected]>; Wang, Hongju <[email protected]>
Subject: Re: [PATCH v2 00/15] Intel IPU6 and IPU6 input system driversHi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.
Quoted 7 lines
Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
Thanks for the test.
I remember I verified SGRBG10P format on my ADL device before, but I have not tested that for a while, give me some time to test again and come back.
Quoted 5 lines
Regards,
Hans
Hi Bingbu,
Quoted 45 lines · Bingbu
Hans,
------------------------------------------------------------------------
BRs,
Bingbu CaoQuoted 39 lines
-----Original Message-----
From: Hans de Goede <[email protected]>
Sent: Wednesday, November 8, 2023 8:00 PM
To: Cao, Bingbu <[email protected]>; [email protected]; [email protected]; [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Qiu, Tian Shu <[email protected]>; Wang, Hongju <[email protected]>
Subject: Re: [PATCH v2 00/15] Intel IPU6 and IPU6 input system driversHi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Quoted 8 lines
Quoted 3 lines
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
Thanks for the test.
I remember I verified SGRBG10P format on my ADL device before, but I have not tested that for a while, give me some time to test again and come back.
I can confirm that SGRBG10P works on Alderlake it appears to be broken on Tiger Lake though.
Regards,
Hans
Hi,
Quoted 23 lines · Hans de Goede
Hi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.
Quoted 7 lines
Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
Bingbu, so we've been trying to get software debayering to work on the Dell Latitude 9420 with ov01a1s and I have just learned that this sensor has a non standard bayer grid.
bayer_order="GRGB_IGIG_GBGR_IGIG"
I'm wondering how to interpret these 4 pixel orders? For the OV2740 this just says:
bayer_order="GRBG"
and this single quartet of letters describes a 2x2 block like this:
GR
BG
but now there are 4 quartets so how to interpret this, I guess this is the correct way to interpret this? :
GRGB
IGIG
GBGR
IGIG
Regards,
Hans
Hans,
Quoted 34 lines · Hans de Goede
Hi,
Quoted 23 lines · Hans de Goede
Hi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.Quoted 7 lines
Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
Quoted 7 lines
Bingbu, so we've been trying to get software debayering to work on the Dell Latitude 9420 with ov01a1s and I have just learned that this sensor has a non standard bayer grid.
bayer_order="GRGB_IGIG_GBGR_IGIG"
The IR is not enabled in Linux camera stack yet if I remember correctly. Current IPU6 processing system have a single hardware block extract the RGB and IR separately - "rgb_ir_2_0" in xml.
Quoted 11 lines
I'm wondering how to interpret these 4 pixel orders? For the OV2740 this just says:
bayer_order="GRBG"
and this single quartet of letters describes a 2x2 block like this:
GR
BG
Quoted 7 lines
but now there are 4 quartets so how to interpret this, I guess this is the correct way to interpret this? :
GRGB
IGIG
GBGR
IGIG
I have no idea how to interpret the output of ov01a1s. Maybe you can try to zero(bypass) the IR data and interpret as SGRBG10.
Quoted 7 lines
Regards,
Hans
Signature · 2 lines
Best regards, Bingbu Cao
Hi,
Quoted 23 lines · Hans de Goede
Hi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.
Quoted 7 lines
Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
I have just rebased my personal tree with these patches in it on top of media_staging. I had to make some changes to make this series work on top of media_staging.
I'm attaching these changes here as a patch in case they are useful for someone else. Feel free to squash these into the next version.
Talking about a next version I would think it is about time to post a v3 of this ?
From 93d23e346aaa2e2ba59b4bc95ca029b674e01195 Mon Sep 17 00:00:00 2001
From: Hans de Goede <[email protected]>
Date: Mon, 4 Dec 2023 12:13:02 +0100
Subject: [PATCH] media: ipu6: Add support for kernel 6.8
Add support for the upcoming v4l2-subdev API changes in kernel 6.8.
--- drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c | 8 +++----- .../media/pci/intel/ipu6/ipu6-isys-subdev.c | 19 +++++++++++-------- .../media/pci/intel/ipu6/ipu6-isys-subdev.h | 2 -- .../media/pci/intel/ipu6/ipu6-isys-video.c | 3 +-- 4 files changed, 15 insertions(+), 17 deletions(-)
drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c +3 -5
@@ -386,12 +386,11 @@ static int ipu6_isys_csi2_set_sel(struct v4l2_subdev *sd,if (!sink_ffmt)return -EINVAL ; - src_ffmt =v4l2_subdev_state_get_stream_format (state, sel->pad, - sel->stream); + src_ffmt =v4l2_subdev_state_get_format (state, sel->pad, sel->stream);if (!src_ffmt)return -EINVAL ; - crop =v4l2_subdev_state_get_stream_crop (state, sel->pad, sel->stream); + crop =v4l2_subdev_state_get_crop (state, sel->pad, sel->stream);if (!crop)return -EINVAL ;
@@ -436,7 +435,7 @@ static int ipu6_isys_csi2_get_sel(struct v4l2_subdev *sd,if (!sink_ffmt)return -EINVAL ; - crop =v4l2_subdev_state_get_stream_crop (state, sel->pad, sel->stream); + crop =v4l2_subdev_state_get_crop (state, sel->pad, sel->stream);if (!crop)return -EINVAL ;
@@ -463,7 +462,6 @@ static const struct v4l2_subdev_video_ops csi2_sd_video_ops = { };static const struct v4l2_subdev_pad_ops csi2_sd_pad_ops = { - .init_cfg = ipu6_isys_subdev_init_cfg, .get_fmt = v4l2_subdev_get_fmt, .set_fmt = ipu6_isys_subdev_set_fmt, .get_selection = ipu6_isys_csi2_get_sel,
drivers/media/pci/intel/ipu6/ipu6-isys-subdev.c +11 -8
@@ -143,8 +143,7 @@ int ipu6_isys_subdev_set_fmt(struct v4l2_subdev *sd, format->format.field =V4L2_FIELD_NONE ;/* Store the format and propagate it to the source pad. */ - fmt =v4l2_subdev_state_get_stream_format (state, format->pad, - format->stream); + fmt =v4l2_subdev_state_get_format (state, format->pad, format->stream);if (!fmt)return -EINVAL ;
@@ -169,8 +168,7 @@ int ipu6_isys_subdev_set_fmt(struct v4l2_subdev *sd,if (ret)return -EINVAL ; - crop =v4l2_subdev_state_get_stream_crop (state, other_pad, - other_stream); + crop =v4l2_subdev_state_get_crop (state, other_pad, other_stream);/* reset crop */ crop->left =0 ; crop->top =0 ;
@@ -229,7 +227,7 @@ int ipu6_isys_get_stream_pad_fmt(struct v4l2_subdev *sd, u32 pad, u32 stream,return -EINVAL ; state =v4l2_subdev_lock_and_get_active_state (sd); - fmt =v4l2_subdev_state_get_stream_format (state, pad, stream); + fmt =v4l2_subdev_state_get_format (state, pad, stream);if (fmt) { *format = *fmt; ret =0 ;
@@ -250,7 +248,7 @@ int ipu6_isys_get_stream_pad_crop(struct v4l2_subdev *sd, u32 pad, u32 stream,return -EINVAL ; state =v4l2_subdev_lock_and_get_active_state (sd); - rect =v4l2_subdev_state_get_stream_crop (state, pad, stream); + rect =v4l2_subdev_state_get_crop (state, pad, stream);if (rect) { *crop = *rect; ret =0 ;
@@ -284,8 +282,8 @@ u32 ipu6_isys_get_src_stream_by_src_pad(struct v4l2_subdev *sd, u32 pad)return source_stream; } -int ipu6_isys_subdev_init_cfg (struct v4l2_subdev *sd, -struct v4l2_subdev_state *state) +static int ipu6_isys_subdev_init_state (struct v4l2_subdev *sd, +struct v4l2_subdev_state *state) {struct v4l2_subdev_route route = { .sink_pad =0 ,
@@ -310,6 +308,10 @@ int ipu6_isys_subdev_set_routing(struct v4l2_subdev *sd,return subdev_set_routing (sd, state, routing); } +static const struct v4l2_subdev_internal_ops ipu6_isys_subdev_internal_ops = { + .init_state = ipu6_isys_subdev_init_state, +}; + int ipu6_isys_subdev_init(struct ipu6_isys_subdev *asd, const struct v4l2_subdev_ops *ops, unsigned int nr_ctrls,
@@ -327,6 +329,7 @@ int ipu6_isys_subdev_init(struct ipu6_isys_subdev *asd,V4L2_SUBDEV_FL_STREAMS ; asd->sd.owner =THIS_MODULE ; asd->sd.entity.function =MEDIA_ENT_F_VID_IF_BRIDGE ; + asd->sd.internal_ops = &ipu6_isys_subdev_internal_ops; asd->pad =devm_kcalloc (&asd->isys->adev->auxdev.dev, num_pads,sizeof (*asd->pad),GFP_KERNEL );
drivers/media/pci/intel/ipu6/ipu6-isys-subdev.h +0 -2
@@ -43,8 +43,6 @@ int ipu6_isys_get_stream_pad_fmt(struct v4l2_subdev *sd, u32 pad, u32 stream, struct v4l2_mbus_framefmt *format);int ipu6_isys_get_stream_pad_crop (struct v4l2_subdev *sd,u32 pad,u32 stream,struct v4l2_rect *crop); -int ipu6_isys_subdev_init_cfg (struct v4l2_subdev *sd, -struct v4l2_subdev_state *state); int ipu6_isys_subdev_set_routing(struct v4l2_subdev *sd, struct v4l2_subdev_state *state, enum v4l2_subdev_format_whence which,
drivers/media/pci/intel/ipu6/ipu6-isys-video.c +1 -2
@@ -281,8 +281,7 @@ static int link_validate(struct media_link *link)v4l2_subdev_lock_state (s_state); - s_fmt =v4l2_subdev_state_get_stream_format (s_state, s_pad->index, - s_stream); + s_fmt =v4l2_subdev_state_get_format (s_state, s_pad->index, s_stream);if (!s_fmt) {dev_err (dev,"failed to get source pad format\n" ); goto unlock;
Signature · 1 line
2.41.0
Quoted 1 line · Hans de Goede
On 11/8/23 12:59, Hans de Goede wrote:
...
Quoted 2 lines
Talking about a next version I would think it is about time to post a v3 of this ?
Yes, would be wonderful to see v3 sooner than later.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Hans,
Quoted 23 lines · Hans de Goede
Hi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.
Quoted 7 lines
Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
I just tested the packed bayer on my device (ov01a10) and did not observe the problems like yours, could you share the isys kernel log?
Quoted 6 lines
Regards,
Hans
Signature · 2 lines
Best regards, Bingbu Cao
Hi Bingbu,
Quoted 30 lines · Bingbu Cao
Hans,
Quoted 27 lines · Hans de Goede
Hi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Quoted 6 lines
Quoted 3 lines
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
I just tested the packed bayer on my device (ov01a10) and did not observe the problems like yours, could you share the isys kernel log?
Was this on a device with a Tiger Lake CPU / plain ipu6 (not ipu6ep) ?
Regards,
Hans
Hans,
Quoted 33 lines · Hans de Goede
Hi Bingbu,
Quoted 30 lines · Bingbu Cao
Hans,
Quoted 27 lines · Hans de Goede
Hi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Quoted 8 lines
Quoted 6 lines
Quoted 3 lines
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
I just tested the packed bayer on my device (ov01a10) and did not observe the problems like yours, could you share the isys kernel log?
Was this on a device with a Tiger Lake CPU / plain ipu6 (not ipu6ep) ?
Sorry, it's ipu6ep.
I have no TigerLake so far.
The difference between ipu6 and ipu6ep is the MIPI CSI-2 D-PHY. But it should not have no impact on the packed format.
Could you share the debug log on TigerLake?
Quoted 6 lines
Regards,
Hans
Signature · 2 lines
Best regards, Bingbu Cao
Hi Bingbu,
Quoted 46 lines · Bingbu Cao
Hans,
Quoted 33 lines · Hans de Goede
Hi Bingbu,
Quoted 30 lines · Bingbu Cao
Hans,
Quoted 27 lines · Hans de Goede
Hi Bingbu,
Quoted 7 lines · On 10/24/23 13:29, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I have been testing this on a TigerLake system, a Dell Latitude 9420 with ov01a1s and the packed 10 bit bayer pixel fmt:
V4L2_PIX_FMT_SGRBG10P, which libcamera together with (WIP) software debayer support picks by default does not work. There are many many CSI errors in dmesg and only the first 10 or so lines of the image show.Disabling the packed format by removing it from ipu6_isys_pfmts[], making libcamera pick the unpacked (every 10 bits per pixel data stored in a 16 bit word in output buffer) fixes this.
Quoted 8 lines
Quoted 6 lines
Quoted 3 lines
Are the packed bayer formats supposed to work on Tiger Lake, or is this a known issue which Intel's own userspace stack avoids by always picking the unpacked format ?
I just tested the packed bayer on my device (ov01a10) and did not observe the problems like yours, could you share the isys kernel log?
Was this on a device with a Tiger Lake CPU / plain ipu6 (not ipu6ep) ?
Quoted 7 lines
Sorry, it's ipu6ep.
I have no TigerLake so far.
The difference between ipu6 and ipu6ep is the MIPI CSI-2 D-PHY. But it should not have no impact on the packed format.Could you share the debug log on TigerLake?
I don't have access to the machine atm, but I should get it back in 2 weeks from now. I'll email you debug logs then.
Regards,
Hans
Sakari Ailus
maintainerHi folks,
Quoted 7 lines · On Tue, Oct 24, 2023 at 07:29:09PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I've made some changes on top of this (besides some done by Bingbu). I've pushed the current patches here <URL:https://git.linuxtv.org/sailus/media_tree.git/log/?h=ipu6>. What's changed:
- CSI-2 port address calculation simplification.
- Tested (and fixed) driver probing on a Dell XPS 9315 that has IVSC and Linux camera description that ignores the IVSC. This works if the intel-ipu6 driver is first removed and then re-probed. I'm not quite sure why it doesn't re-probe on its own yet. I haven't tried streaming yet.
- The patches are on top of the metadata set. If you want to try without, just drop the metadata patches (and the last functional driver patch). It might compile. I'll test this soon, too.
So overall there's some progress but work remains to do. I'll post the IVSC and ACPI patches after a little bit more testing.
Signature · 3 lines
Regards, Sakari Ailus
Sakari Ailus
maintainerQuoted 29 lines · Sakari Ailus
Hi folks,
Quoted 7 lines · On Tue, Oct 24, 2023 at 07:29:09PM +0800, [email protected] wrote:
From: Bingbu Cao <[email protected]>
This patch series adds a driver for Intel IPU6 input system. IPU6 is the sixth generation of Imaging Processing Unit, it is a PCI device which can be found in some Intel Client Platforms. User can use IPU6 to capture images from MIPI camera sensors.
Quoted 6 lines
IPU6 has its own firmware which exposes ABIs to driver, and communicates with CSE to do firmware authentication. IPU6 has its MMU hardware, so the driver sets up a page table to allow IPU6 DMA to access the system memory.
IPU6 input system driver uses MC and V4L2 sub-device APIs besides V4L2.
I've made some changes on top of this (besides some done by Bingbu). I've pushed the current patches here <URL:https://git.linuxtv.org/sailus/media_tree.git/log/?h=ipu6>. What's changed:
- CSI-2 port address calculation simplification.
- Tested (and fixed) driver probing on a Dell XPS 9315 that has IVSC and Linux camera description that ignores the IVSC. This works if the intel-ipu6 driver is first removed and then re-probed. I'm not quite sure why it doesn't re-probe on its own yet. I haven't tried streaming yet.
Quoted 6 lines
- The patches are on top of the metadata set. If you want to try without, just drop the metadata patches (and the last functional driver patch). It might compile. I'll test this soon, too.
So overall there's some progress but work remains to do. I'll post the IVSC and ACPI patches after a little bit more testing.
I've added further patches to the branch and dropped some... Capturing works now on Dell XPS 9315 with ov01a10 without issues (AFAICT), and that pipeline includes IVSC. I'm upstreaming the un-media patches to the respective trees so some of this is going to take a while to merged back to the media tree.
I've also removed the massive number of extra links from the CSI-2 receiver subdevs and video nodes. This part remains untested though, the ov01a10 sensor doesn't have support for embedded data (!).
Just FYI.
Signature · 1 line
Sakari Ailus