Switchtec Fixes and Gen4 Support
Subsystems (2)
Logan Gunthorpe
maintainercover Switchtec Fixes and Gen4 Support
Hi,
Please find a bunch of patches for the switchtec driver collected over the last few months.
The first 2 patches fix a couple of minor bugs. Patch 3 adds support for a new event that is available in specific firmware versions. Patches 4 and 5 are some code cleanup changes to simplify the logic. And the last 6 patches implement support for the new Gen4 hardware.
This patchset is based on v5.5-rc5 and a git branch is available here:
https://github.com/sbates130272/linux-p2pmem switchtec-next
Thanks,
Logan
--
Kelvin Cao (3):
PCI/switchtec: Add gen4 support in struct flash_info_regs PCI/switchtec: Add permission check for the GAS access MRPC commands PCI/switchtec: Introduce gen4 variant IDS in the device ID table
Logan Gunthorpe (6):
PCI/switchtec: Fix vep_vector_number ioread width PCI/switchtec: Add support for new events
PCI/switchtec: Introduce Generation Variable PCI/switchtec: Separate out gen3 specific fields in the sys_info_regs
structure
PCI/switchtec: Add gen4 support in struct sys_info_regs
PCI: Apply switchtec DMA aliasing quirk to GEN4 devices
Wesley Sheng (3):
PCI/switchtec: Use dma_set_mask_and_coherent() PCI/switchtec: Remove redundant valid PFF number count PCI/switchtec: Move check event id from mask_event() to
switchtec_event_isr()
drivers/pci/quirks.c | 18 ++ drivers/pci/switch/switchtec.c | 365 ++++++++++++++++++++------- include/linux/switchtec.h | 160 ++++++++++-- include/uapi/linux/switchtec_ioctl.h | 13 +- 4 files changed, 450 insertions(+), 106 deletions(-)
--
2.20.1
Logan Gunthorpe
maintainerPATCH 1/12 PCI/switchtec: Use dma_set_mask_and_coherent()
From: Wesley Sheng <[email protected]>
Use dma_set_mask_and_coherent() instead of dma_set_coherent_mask() as the switchtec hardware fully supports 64bit addressing and we should set both the streaming and coherent masks the same.
--- drivers/pci/switch/switchtec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pci/switch/switchtec.c +1 -1
@@ -1349,7 +1349,7 @@ static int switchtec_init_pci(struct switchtec_dev *stdev,if (rc)return rc; - rc =dma_set_coherent_mask (&pdev->dev,DMA_BIT_MASK (64 )); + rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc)return rc;
Signature · 1 line
2.20.1
Logan Gunthorpe
maintainerPATCH 2/12 PCI/switchtec: Fix vep_vector_number ioread width
In reply to Logan Gunthorpevep_vector_number is actually a 16 bit register which should be read with ioread16() instead of ioread32().
--- drivers/pci/switch/switchtec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pci/switch/switchtec.c +1 -1
@@ -1276,7 +1276,7 @@ static int switchtec_init_isr(struct switchtec_dev *stdev)if (nvecs <0 )return nvecs; - event_irq =ioread32 (&stdev->mmio_part_cfg->vep_vector_number); + event_irq =ioread16 (&stdev->mmio_part_cfg->vep_vector_number);if (event_irq <0 || event_irq >= nvecs)return -EFAULT ;
Signature · 1 line
2.20.1
Logan Gunthorpe
maintainerPATCH 3/12 PCI/switchtec: Add support for new events
In reply to Logan GunthorpeThe intercomm notify event was added for PAX variants of switchtec hardware and the UEC Port was added for the MR1 release of GEN3 firmware.
--- drivers/pci/switch/switchtec.c | 3 +++ include/linux/switchtec.h | 7 +++++-- include/uapi/linux/switchtec_ioctl.h | 4 +++- 3 files changed, 11 insertions(+), 3 deletions(-)
drivers/pci/switch/switchtec.c +3 -0
@@ -751,10 +751,13 @@ static const struct event_reg {EV_PAR (SWITCHTEC_IOCTL_EVENT_MRPC_COMP , mrpc_comp_hdr),EV_PAR (SWITCHTEC_IOCTL_EVENT_MRPC_COMP_ASYNC , mrpc_comp_async_hdr),EV_PAR (SWITCHTEC_IOCTL_EVENT_DYN_PART_BIND_COMP , dyn_binding_hdr), +EV_PAR (SWITCHTEC_IOCTL_EVENT_INTERCOMM_REQ_NOTIFY , + intercomm_notify_hdr),EV_PFF (SWITCHTEC_IOCTL_EVENT_AER_IN_P2P , aer_in_p2p_hdr),EV_PFF (SWITCHTEC_IOCTL_EVENT_AER_IN_VEP , aer_in_vep_hdr),EV_PFF (SWITCHTEC_IOCTL_EVENT_DPC , dpc_hdr),EV_PFF (SWITCHTEC_IOCTL_EVENT_CTS , cts_hdr), +EV_PFF (SWITCHTEC_IOCTL_EVENT_UEC , uec_hdr),EV_PFF (SWITCHTEC_IOCTL_EVENT_HOTPLUG , hotplug_hdr),EV_PFF (SWITCHTEC_IOCTL_EVENT_IER , ier_hdr), EV_PFF(SWITCHTEC_IOCTL_EVENT_THRESH, threshold_hdr),
include/linux/switchtec.h +5 -2
@@ -196,7 +196,9 @@ struct part_cfg_regs {u32 mrpc_comp_async_data[5 ];u32 dyn_binding_hdr;u32 dyn_binding_data[5 ]; -u32 reserved4[159 ]; +u32 intercomm_notify_hdr; +u32 intercomm_notify_data[5 ]; +u32 reserved4[153 ]; } __packed;enum {
@@ -320,7 +322,8 @@ struct pff_csr_regs {u32 dpc_data[5 ];u32 cts_hdr;u32 cts_data[5 ]; -u32 reserved3[6 ]; +u32 uec_hdr; +u32 uec_data[5 ];u32 hotplug_hdr;u32 hotplug_data[5 ];u32 ier_hdr;
include/uapi/linux/switchtec_ioctl.h +3 -1
@@ -98,7 +98,9 @@ struct switchtec_ioctl_event_summary {#define SWITCHTEC_IOCTL_EVENT_CREDIT_TIMEOUT 27#define SWITCHTEC_IOCTL_EVENT_LINK_STATE 28#define SWITCHTEC_IOCTL_EVENT_GFMS 29 -#define SWITCHTEC_IOCTL_MAX_EVENTS 30 +#define SWITCHTEC_IOCTL_EVENT_INTERCOMM_REQ_NOTIFY 30 +#define SWITCHTEC_IOCTL_EVENT_UEC 31 +#define SWITCHTEC_IOCTL_MAX_EVENTS 32#define SWITCHTEC_IOCTL_EVENT_LOCAL_PART_IDX -1#define SWITCHTEC_IOCTL_EVENT_IDX_ALL -2
Signature · 1 line
2.20.1
Re: [PATCH 03/12] PCI/switchtec: Add support for new events
Quoted 2 lines · Logan Gunthorpe
The intercomm notify event was added for PAX variants of switchtec hardware and the UEC Port was added for the MR1 release of GEN3 firmware.
Do they actually spell it "intercomm" in the datasheet? Seems like the most common English spelling is "intercom".
Is there some meaningful expansion of "UEC"?
Quoted 17 lines
Signed-off-by Logan Gunthorpe <[email protected]>--- drivers/pci/switch/switchtec.c | 3 +++ include/linux/switchtec.h | 7 +++++-- include/uapi/linux/switchtec_ioctl.h | 4 +++- 3 files changed, 11 insertions(+), 3 deletions(-)drivers/pci/switch/switchtec.c +2
-0@@ -751,10 +751,13 @@ static const struct event_reg {EV_PAR (SWITCHTEC_IOCTL_EVENT_MRPC_COMP , mrpc_comp_hdr),EV_PAR (SWITCHTEC_IOCTL_EVENT_MRPC_COMP_ASYNC , mrpc_comp_async_hdr),EV_PAR (SWITCHTEC_IOCTL_EVENT_DYN_PART_BIND_COMP , dyn_binding_hdr), +EV_PAR (SWITCHTEC_IOCTL_EVENT_INTERCOMM_REQ_NOTIFY , + intercomm_notify_hdr),
Logan Gunthorpe
maintainerRe: [PATCH 03/12] PCI/switchtec: Add support for new events
Quoted 6 lines · Bjorn Helgaas
Quoted 2 lines · Logan Gunthorpe
The intercomm notify event was added for PAX variants of switchtec hardware and the UEC Port was added for the MR1 release of GEN3 firmware.
Do they actually spell it "intercomm" in the datasheet? Seems like the most common English spelling is "intercom".
Hmm, not sure I don't have this event in my datasheet which is a little out of date. Kelvin, whose CC'd, would know if we can change the spelling hopefully he can weigh in on this. As far as I know the define isn't used anywhere yet.
Quoted 1 line
Is there some meaningful expansion of "UEC"?
Upstream Error Containment -- I'll mention that in the commit message.
Logan
Quoted 17 lines
Quoted 17 lines
Signed-off-by Logan Gunthorpe <[email protected]>--- drivers/pci/switch/switchtec.c | 3 +++ include/linux/switchtec.h | 7 +++++-- include/uapi/linux/switchtec_ioctl.h | 4 +++- 3 files changed, 11 insertions(+), 3 deletions(-)drivers/pci/switch/switchtec.c +2
-0@@ -751,10 +751,13 @@ static const struct event_reg {EV_PAR (SWITCHTEC_IOCTL_EVENT_MRPC_COMP , mrpc_comp_hdr),EV_PAR (SWITCHTEC_IOCTL_EVENT_MRPC_COMP_ASYNC , mrpc_comp_async_hdr),EV_PAR (SWITCHTEC_IOCTL_EVENT_DYN_PART_BIND_COMP , dyn_binding_hdr), +EV_PAR (SWITCHTEC_IOCTL_EVENT_INTERCOMM_REQ_NOTIFY , + intercomm_notify_hdr),
RE: [PATCH 03/12] PCI/switchtec: Add support for new events
'InterComm' is the spelling in the latest datasheet, it's short for "Inter Fabric Manager Communication". Thanks.
Kelvin
-----Original Message-----
From: Logan Gunthorpe [mailto:[email protected]]
Sent: Thursday, January 9, 2020 5:48 AM
To: Bjorn Helgaas <[email protected]>
Cc: [email protected]; [email protected]; Kelvin Cao - A31060 <[email protected]>; Eric Pilmore <[email protected]>; Doug Meyer <[email protected]>
Subject: Re: [PATCH 03/12] PCI/switchtec: Add support for new events
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
Quoted 6 lines · Bjorn Helgaas
Quoted 2 lines · Logan Gunthorpe
The intercomm notify event was added for PAX variants of switchtec hardware and the UEC Port was added for the MR1 release of GEN3 firmware.
Do they actually spell it "intercomm" in the datasheet? Seems like the most common English spelling is "intercom".
Hmm, not sure I don't have this event in my datasheet which is a little out of date. Kelvin, whose CC'd, would know if we can change the spelling hopefully he can weigh in on this. As far as I know the define isn't used anywhere yet.
Quoted 1 line
Is there some meaningful expansion of "UEC"?
Upstream Error Containment -- I'll mention that in the commit message.
Logan
Quoted 19 lines
Quoted 19 lines
Signed-off-by Logan Gunthorpe <[email protected]>--- drivers/pci/switch/switchtec.c | 3 +++ include/linux/switchtec.h | 7 +++++-- include/uapi/linux/switchtec_ioctl.h | 4 +++- 3 files changed, 11 insertions(+), 3 deletions(-)(quoted diff) +0
-0
b/drivers/pci/switch/switchtec.cindex9c3ad09d3022..218e67428cf9100644drivers/pci/switch/switchtec.c +2
-0@@ -751,10 +751,13 @@ static const struct event_reg {EV_PAR (SWITCHTEC_IOCTL_EVENT_MRPC_COMP , mrpc_comp_hdr),EV_PAR (SWITCHTEC_IOCTL_EVENT_MRPC_COMP_ASYNC , mrpc_comp_async_hdr),EV_PAR (SWITCHTEC_IOCTL_EVENT_DYN_PART_BIND_COMP , dyn_binding_hdr), +EV_PAR (SWITCHTEC_IOCTL_EVENT_INTERCOMM_REQ_NOTIFY , + intercomm_notify_hdr),
Logan Gunthorpe
maintainerRe: [PATCH 03/12] PCI/switchtec: Add support for new events
Bjorn,
Quoted 1 line · On 2020-01-13 7:07 p.m., [email protected] wrote:
'InterComm' is the spelling in the latest datasheet, it's short for "Inter Fabric Manager Communication". Thanks.
I noticed you fixed up the spelling in your branch but the IOCTL define is still spelled INTERCOM, which is arguably the most important seeing it's a userspace interface.
Thanks,
Logan
(quoted diff) +0 -0
b/include/uapi/linux/switchtec_ioctl.h
index c912b5a678e4..902fffa9e0f5 100644
include/uapi/linux/switchtec_ioctl.h +3 -1
@@ -98,7 +98,9 @@ struct switchtec_ioctl_event_summary {#define SWITCHTEC_IOCTL_EVENT_CREDIT_TIMEOUT 27#define SWITCHTEC_IOCTL_EVENT_LINK_STATE 28#define SWITCHTEC_IOCTL_EVENT_GFMS 29 -#define SWITCHTEC_IOCTL_MAX_EVENTS 30 +#define SWITCHTEC_IOCTL_EVENT_INTERCOM_REQ_NOTIFY 30 +#define SWITCHTEC_IOCTL_EVENT_UEC 31 +#define SWITCHTEC_IOCTL_MAX_EVENTS 32#define SWITCHTEC_IOCTL_EVENT_LOCAL_PART_IDX -1#define SWITCHTEC_IOCTL_EVENT_IDX_ALL -2
Re: [PATCH 03/12] PCI/switchtec: Add support for new events
Quoted 8 lines · Logan Gunthorpe
Bjorn,
Quoted 1 line · On 2020-01-13 7:07 p.m., [email protected] wrote:
'InterComm' is the spelling in the latest datasheet, it's short for "Inter Fabric Manager Communication". Thanks.
I noticed you fixed up the spelling in your branch but the IOCTL define is still spelled INTERCOM, which is arguably the most important seeing it's a userspace interface.
Thanks, I missed that one. Fixed now.
Logan Gunthorpe
maintainerPATCH 4/12 PCI/switchtec: Remove redundant valid PFF number count
In reply to Logan GunthorpeFrom: Wesley Sheng <[email protected]>
init_pff() function has already counted the valid PFF number, so there is no requirement to count it again in ioctl_event_summary().
--- drivers/pci/switch/switchtec.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-)
drivers/pci/switch/switchtec.c +4 -8
@@ -683,11 +683,7 @@ static int ioctl_event_summary(struct switchtec_dev *stdev, s->part[i] = reg; } -for (i =0 ; i <SWITCHTEC_MAX_PFF_CSR ; i++) { - reg =ioread16 (&stdev->mmio_pff_csr[i].vendor_id); -if (reg !=PCI_VENDOR_ID_MICROSEMI ) -break ; - +for (i =0 ; i < stdev->pff_csr_count; i++) { reg =ioread32 (&stdev->mmio_pff_csr[i].pff_event_summary); s->pff[i] = reg; }
@@ -1327,16 +1323,16 @@ static void init_pff(struct switchtec_dev *stdev) stdev->pff_csr_count = i; reg =ioread32 (&pcfg->usp_pff_inst_id); -if (reg <SWITCHTEC_MAX_PFF_CSR ) +if (reg < stdev->pff_csr_count) stdev->pff_local[reg] =1 ; reg =ioread32 (&pcfg->vep_pff_inst_id); -if (reg <SWITCHTEC_MAX_PFF_CSR ) +if (reg < stdev->pff_csr_count) stdev->pff_local[reg] =1 ;for (i =0 ; i <ARRAY_SIZE (pcfg->dsp_pff_inst_id); i++) { reg =ioread32 (&pcfg->dsp_pff_inst_id[i]); -if (reg <SWITCHTEC_MAX_PFF_CSR ) +if (reg < stdev->pff_csr_count) stdev->pff_local[reg] =1 ; } }
Signature · 1 line
2.20.1
Logan Gunthorpe
maintainerPATCH 5/12 PCI/switchtec: Move check event id from mask_event() to switchtec_event_isr()
In reply to Logan GunthorpeFrom: Wesley Sheng <[email protected]>
Event id check doesn't depend on anything in the mask_all_events() to mask_event() path, doing it in switchtec_event_isr() would avoid the CSR read in mask_event().
--- drivers/pci/switch/switchtec.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-)
drivers/pci/switch/switchtec.c +6 -5
@@ -1180,10 +1180,6 @@ static int mask_event(struct switchtec_dev *stdev, int eid, int idx)if (!(hdr &SWITCHTEC_EVENT_OCCURRED && hdr &SWITCHTEC_EVENT_EN_IRQ ))return 0 ; -if (eid ==SWITCHTEC_IOCTL_EVENT_LINK_STATE || - eid ==SWITCHTEC_IOCTL_EVENT_MRPC_COMP ) -return 0 ; -dev_dbg (&stdev->dev,"%s: %d %d %x\n" , __func__, eid, idx, hdr); hdr &= ~(SWITCHTEC_EVENT_EN_IRQ |SWITCHTEC_EVENT_OCCURRED );iowrite32 (hdr, hdr_reg);
@@ -1230,8 +1226,13 @@ static irqreturn_t switchtec_event_isr(int irq, void *dev)check_link_state_events (stdev); -for (eid =0 ; eid <SWITCHTEC_IOCTL_MAX_EVENTS ; eid++) +for (eid =0 ; eid <SWITCHTEC_IOCTL_MAX_EVENTS ; eid++) { +if (eid ==SWITCHTEC_IOCTL_EVENT_LINK_STATE || + eid ==SWITCHTEC_IOCTL_EVENT_MRPC_COMP ) +continue ; + event_count +=mask_all_events (stdev, eid); + }if (event_count) {atomic_inc (&stdev->event_cnt);
Signature · 1 line
2.20.1
Logan Gunthorpe
maintainerPATCH 6/12 PCI/switchtec: Introduce Generation Variable
In reply to Logan GunthorpeAdd a generation variable passed through the device id table.
This will allow us to add generation 4 and other devices after they are functionally supported.
--- drivers/pci/switch/switchtec.c | 66 ++++++++++++++++++---------------- include/linux/switchtec.h | 7 ++++ 2 files changed, 42 insertions(+), 31 deletions(-)
drivers/pci/switch/switchtec.c +35 -31
@@ -1420,6 +1420,8 @@ static int switchtec_pci_probe(struct pci_dev *pdev,if (IS_ERR (stdev))return PTR_ERR (stdev); + stdev->gen = id->driver_data; + rc =switchtec_init_pci (stdev, pdev);if (rc) goto err_put;
@@ -1467,7 +1469,7 @@ static void switchtec_pci_remove(struct pci_dev *pdev)put_device (&stdev->dev); } -#define SWITCHTEC_PCI_DEVICE (device_id) \ +#define SWITCHTEC_PCI_DEVICE (device_id, gen) \ { \ .vendor = PCI_VENDOR_ID_MICROSEMI, \ .device = device_id, \
@@ -1475,6 +1477,7 @@ static void switchtec_pci_remove(struct pci_dev *pdev) .subdevice = PCI_ANY_ID, \ .class = (PCI_CLASS_MEMORY_OTHER << 8), \ .class_mask = 0xFFFFFFFF, \ + .driver_data = gen, \ }, \ { \ .vendor = PCI_VENDOR_ID_MICROSEMI, \
@@ -1483,39 +1486,40 @@ static void switchtec_pci_remove(struct pci_dev *pdev) .subdevice = PCI_ANY_ID, \ .class = (PCI_CLASS_BRIDGE_OTHER << 8), \ .class_mask = 0xFFFFFFFF, \ + .driver_data = gen, \ }static const struct pci_device_id switchtec_pci_tbl[] = { -SWITCHTEC_PCI_DEVICE (0x8531 ),//PFX 24xG3 -SWITCHTEC_PCI_DEVICE (0x8532 ),//PFX 32xG3 -SWITCHTEC_PCI_DEVICE (0x8533 ),//PFX 48xG3 -SWITCHTEC_PCI_DEVICE (0x8534 ),//PFX 64xG3 -SWITCHTEC_PCI_DEVICE (0x8535 ),//PFX 80xG3 -SWITCHTEC_PCI_DEVICE (0x8536 ),//PFX 96xG3 -SWITCHTEC_PCI_DEVICE (0x8541 ),//PSX 24xG3 -SWITCHTEC_PCI_DEVICE (0x8542 ),//PSX 32xG3 -SWITCHTEC_PCI_DEVICE (0x8543 ),//PSX 48xG3 -SWITCHTEC_PCI_DEVICE (0x8544 ),//PSX 64xG3 -SWITCHTEC_PCI_DEVICE (0x8545 ),//PSX 80xG3 -SWITCHTEC_PCI_DEVICE (0x8546 ),//PSX 96xG3 -SWITCHTEC_PCI_DEVICE (0x8551 ),//PAX 24XG3 -SWITCHTEC_PCI_DEVICE (0x8552 ),//PAX 32XG3 -SWITCHTEC_PCI_DEVICE (0x8553 ),//PAX 48XG3 -SWITCHTEC_PCI_DEVICE (0x8554 ),//PAX 64XG3 -SWITCHTEC_PCI_DEVICE (0x8555 ),//PAX 80XG3 -SWITCHTEC_PCI_DEVICE (0x8556 ),//PAX 96XG3 -SWITCHTEC_PCI_DEVICE (0x8561 ),//PFXL 24XG3 -SWITCHTEC_PCI_DEVICE (0x8562 ),//PFXL 32XG3 -SWITCHTEC_PCI_DEVICE (0x8563 ),//PFXL 48XG3 -SWITCHTEC_PCI_DEVICE (0x8564 ),//PFXL 64XG3 -SWITCHTEC_PCI_DEVICE (0x8565 ),//PFXL 80XG3 -SWITCHTEC_PCI_DEVICE (0x8566 ),//PFXL 96XG3 -SWITCHTEC_PCI_DEVICE (0x8571 ),//PFXI 24XG3 -SWITCHTEC_PCI_DEVICE (0x8572 ),//PFXI 32XG3 -SWITCHTEC_PCI_DEVICE (0x8573 ),//PFXI 48XG3 -SWITCHTEC_PCI_DEVICE (0x8574 ),//PFXI 64XG3 -SWITCHTEC_PCI_DEVICE (0x8575 ),//PFXI 80XG3 -SWITCHTEC_PCI_DEVICE (0x8576 ),//PFXI 96XG3 +SWITCHTEC_PCI_DEVICE (0x8531 ,SWITCHTEC_GEN3 ),//PFX 24xG3 +SWITCHTEC_PCI_DEVICE (0x8532 ,SWITCHTEC_GEN3 ),//PFX 32xG3 +SWITCHTEC_PCI_DEVICE (0x8533 ,SWITCHTEC_GEN3 ),//PFX 48xG3 +SWITCHTEC_PCI_DEVICE (0x8534 ,SWITCHTEC_GEN3 ),//PFX 64xG3 +SWITCHTEC_PCI_DEVICE (0x8535 ,SWITCHTEC_GEN3 ),//PFX 80xG3 +SWITCHTEC_PCI_DEVICE (0x8536 ,SWITCHTEC_GEN3 ),//PFX 96xG3 +SWITCHTEC_PCI_DEVICE (0x8541 ,SWITCHTEC_GEN3 ),//PSX 24xG3 +SWITCHTEC_PCI_DEVICE (0x8542 ,SWITCHTEC_GEN3 ),//PSX 32xG3 +SWITCHTEC_PCI_DEVICE (0x8543 ,SWITCHTEC_GEN3 ),//PSX 48xG3 +SWITCHTEC_PCI_DEVICE (0x8544 ,SWITCHTEC_GEN3 ),//PSX 64xG3 +SWITCHTEC_PCI_DEVICE (0x8545 ,SWITCHTEC_GEN3 ),//PSX 80xG3 +SWITCHTEC_PCI_DEVICE (0x8546 ,SWITCHTEC_GEN3 ),//PSX 96xG3 +SWITCHTEC_PCI_DEVICE (0x8551 ,SWITCHTEC_GEN3 ),//PAX 24XG3 +SWITCHTEC_PCI_DEVICE (0x8552 ,SWITCHTEC_GEN3 ),//PAX 32XG3 +SWITCHTEC_PCI_DEVICE (0x8553 ,SWITCHTEC_GEN3 ),//PAX 48XG3 +SWITCHTEC_PCI_DEVICE (0x8554 ,SWITCHTEC_GEN3 ),//PAX 64XG3 +SWITCHTEC_PCI_DEVICE (0x8555 ,SWITCHTEC_GEN3 ),//PAX 80XG3 +SWITCHTEC_PCI_DEVICE (0x8556 ,SWITCHTEC_GEN3 ),//PAX 96XG3 +SWITCHTEC_PCI_DEVICE (0x8561 ,SWITCHTEC_GEN3 ),//PFXL 24XG3 +SWITCHTEC_PCI_DEVICE (0x8562 ,SWITCHTEC_GEN3 ),//PFXL 32XG3 +SWITCHTEC_PCI_DEVICE (0x8563 ,SWITCHTEC_GEN3 ),//PFXL 48XG3 +SWITCHTEC_PCI_DEVICE (0x8564 ,SWITCHTEC_GEN3 ),//PFXL 64XG3 +SWITCHTEC_PCI_DEVICE (0x8565 ,SWITCHTEC_GEN3 ),//PFXL 80XG3 +SWITCHTEC_PCI_DEVICE (0x8566 ,SWITCHTEC_GEN3 ),//PFXL 96XG3 +SWITCHTEC_PCI_DEVICE (0x8571 ,SWITCHTEC_GEN3 ),//PFXI 24XG3 +SWITCHTEC_PCI_DEVICE (0x8572 ,SWITCHTEC_GEN3 ),//PFXI 32XG3 +SWITCHTEC_PCI_DEVICE (0x8573 ,SWITCHTEC_GEN3 ),//PFXI 48XG3 +SWITCHTEC_PCI_DEVICE (0x8574 ,SWITCHTEC_GEN3 ),//PFXI 64XG3 +SWITCHTEC_PCI_DEVICE (0x8575 ,SWITCHTEC_GEN3 ),//PFXI 80XG3 +SWITCHTEC_PCI_DEVICE (0x8576 ,SWITCHTEC_GEN3 ),//PFXI 96XG3 {0 } };MODULE_DEVICE_TABLE (pci, switchtec_pci_tbl);
include/linux/switchtec.h +7 -0
@@ -32,6 +32,11 @@ enum { SWITCHTEC_GAS_PFF_CSR_OFFSET = 0x134000, }; +enum switchtec_gen { +SWITCHTEC_GEN3 , +SWITCHTEC_GEN4 , +}; +struct mrpc_regs {u8 input_data[SWITCHTEC_MRPC_PAYLOAD_SIZE ];u8 output_data[SWITCHTEC_MRPC_PAYLOAD_SIZE ];
@@ -358,6 +363,8 @@ struct switchtec_dev {struct device dev;struct cdev cdev; +enum switchtec_gen gen; +int partition;int partition_count;int pff_csr_count;
Signature · 1 line
2.20.1
Logan Gunthorpe
maintainerPATCH 7/12 PCI/switchtec: Separate out gen3 specific fields in the sys_info_regs structure
In reply to Logan GunthorpeSince gen4 system info region in GAS is diverged from gen3, separate oute the gen3 specific stuff in preparation for adding a gen4 variant.
--- drivers/pci/switch/switchtec.c | 18 +++++++++--------- include/linux/switchtec.h | 12 ++++++++---- 2 files changed, 17 insertions(+), 13 deletions(-)
drivers/pci/switch/switchtec.c +9 -9
@@ -317,8 +317,8 @@ static ssize_t field ## _show(struct device *dev, \ struct device_attribute *attr, char *buf) \ { \ struct switchtec_dev *stdev = to_stdev(dev); \ -return io_string_show (buf, &stdev->mmio_sys_info->field, \ -sizeof (stdev->mmio_sys_info->field)); \ +return io_string_show (buf, &stdev->mmio_sys_info->gen3.field, \ +sizeof (stdev->mmio_sys_info->gen3.field)); \ } \ \static DEVICE_ATTR_RO (field)
@@ -332,7 +332,7 @@ static ssize_t component_id_show(struct device *dev, struct device_attribute *attr, char *buf) { struct switchtec_dev *stdev = to_stdev(dev); -int id =ioread16 (&stdev->mmio_sys_info->component_id); +int id =ioread16 (&stdev->mmio_sys_info->gen3.component_id);return sprintf (buf,"PM%04X\n" , id); }
@@ -342,7 +342,7 @@ static ssize_t component_revision_show(struct device *dev, struct device_attribute *attr, char *buf) { struct switchtec_dev *stdev = to_stdev(dev); -int rev =ioread8 (&stdev->mmio_sys_info->component_revision); +int rev =ioread8 (&stdev->mmio_sys_info->gen3.component_revision);return sprintf (buf,"%d\n" , rev); }
@@ -599,25 +599,25 @@ static int ioctl_flash_part_info(struct switchtec_dev *stdev,case SWITCHTEC_IOCTL_PART_CFG0 : active_addr =ioread32 (&fi->active_cfg);set_fw_info_part (&info, &fi->cfg0); -if (ioread16 (&si->cfg_running) ==SWITCHTEC_CFG0_RUNNING ) +if (ioread16 (&si->gen3.cfg_running) ==SWITCHTEC_CFG0_RUNNING ) info.active |=SWITCHTEC_IOCTL_PART_RUNNING ;break ;case SWITCHTEC_IOCTL_PART_CFG1 : active_addr =ioread32 (&fi->active_cfg);set_fw_info_part (&info, &fi->cfg1); -if (ioread16 (&si->cfg_running) ==SWITCHTEC_CFG1_RUNNING ) +if (ioread16 (&si->gen3.cfg_running) ==SWITCHTEC_CFG1_RUNNING ) info.active |=SWITCHTEC_IOCTL_PART_RUNNING ;break ;case SWITCHTEC_IOCTL_PART_IMG0 : active_addr =ioread32 (&fi->active_img);set_fw_info_part (&info, &fi->img0); -if (ioread16 (&si->img_running) ==SWITCHTEC_IMG0_RUNNING ) +if (ioread16 (&si->gen3.img_running) ==SWITCHTEC_IMG0_RUNNING ) info.active |=SWITCHTEC_IOCTL_PART_RUNNING ;break ;case SWITCHTEC_IOCTL_PART_IMG1 : active_addr =ioread32 (&fi->active_img);set_fw_info_part (&info, &fi->img1); -if (ioread16 (&si->img_running) ==SWITCHTEC_IMG1_RUNNING ) +if (ioread16 (&si->gen3.img_running) ==SWITCHTEC_IMG1_RUNNING ) info.active |=SWITCHTEC_IOCTL_PART_RUNNING ;break ;case SWITCHTEC_IOCTL_PART_NVLOG :
@@ -1378,7 +1378,7 @@ static int switchtec_init_pci(struct switchtec_dev *stdev, stdev->mmio_sys_info = stdev->mmio +SWITCHTEC_GAS_SYS_INFO_OFFSET ; stdev->mmio_flash_info = stdev->mmio +SWITCHTEC_GAS_FLASH_INFO_OFFSET ; stdev->mmio_ntb = stdev->mmio +SWITCHTEC_GAS_NTB_OFFSET ; - stdev->partition =ioread8 (&stdev->mmio_sys_info->partition_id); + stdev->partition =ioread8 (&stdev->mmio_sys_info->gen3.partition_id); stdev->partition_count =ioread8 (&stdev->mmio_ntb->partition_count); stdev->mmio_part_cfg_all = stdev->mmio +SWITCHTEC_GAS_PART_CFG_OFFSET ; stdev->mmio_part_cfg = &stdev->mmio_part_cfg_all[stdev->partition];
include/linux/switchtec.h +8 -4
@@ -109,10 +109,7 @@ enum { SWITCHTEC_IMG1_RUNNING = 0x07, }; -struct sys_info_regs { -u32 device_id; -u32 device_version; -u32 firmware_version; +struct sys_info_regs_gen3 {u32 reserved1;u32 vendor_table_revision;u32 table_format_version;
@@ -129,6 +126,13 @@ struct sys_info_regs {u8 component_revision; } __packed; +struct sys_info_regs { +u32 device_id; +u32 device_version; +u32 firmware_version; +struct sys_info_regs_gen3 gen3; +} __packed; +struct flash_info_regs {u32 flash_part_map_upd_idx;
Signature · 1 line
2.20.1
Logan Gunthorpe
maintainerPATCH 8/12 PCI/switchtec: Add gen4 support in struct sys_info_regs
In reply to Logan GunthorpeAdd a union with gen3 and gen4 sys_info structs. Ensure any use of the gen3 struct is gaurded by an if statement on stdev->gen.
--- drivers/pci/switch/switchtec.c | 41 +++++++++++++++++++++++++++--- include/linux/switchtec.h | 46 +++++++++++++++++++++++++++++++++- 2 files changed, 82 insertions(+), 5 deletions(-)
drivers/pci/switch/switchtec.c +37 -4
@@ -317,8 +317,14 @@ static ssize_t field ## _show(struct device *dev, \ struct device_attribute *attr, char *buf) \ { \ struct switchtec_dev *stdev = to_stdev(dev); \ -return io_string_show (buf, &stdev->mmio_sys_info->gen3.field, \ -sizeof (stdev->mmio_sys_info->gen3.field)); \ + struct sys_info_regs __iomem *si = stdev->mmio_sys_info; \ + \ +if (stdev->gen ==SWITCHTEC_GEN4 ) \ +return io_string_show (buf, &si->gen4.field, \ +sizeof (si->gen4.field)); \ +else \ +return io_string_show (buf, &si->gen3.field, \ +sizeof (si->gen3.field)); \ } \ \static DEVICE_ATTR_RO (field)
@@ -326,7 +332,21 @@ static DEVICE_ATTR_RO(field)DEVICE_ATTR_SYS_INFO_STR (vendor_id);DEVICE_ATTR_SYS_INFO_STR (product_id);DEVICE_ATTR_SYS_INFO_STR (product_revision); -DEVICE_ATTR_SYS_INFO_STR (component_vendor); + +static ssize_t component_vendor_show (struct device *dev, +struct device_attribute *attr,char *buf) +{ +struct switchtec_dev *stdev =to_stdev (dev); +struct sys_info_regs __iomem *si = stdev->mmio_sys_info; + +/* component_vendor field not supported after gen4 */ +if (stdev->gen !=SWITCHTEC_GEN3 ) +return sprintf (buf,"none\n" ); + +return io_string_show (buf, &si->gen3.component_vendor, +sizeof (si->gen3.component_vendor)); +} +static DEVICE_ATTR_RO (component_vendor);static ssize_t component_id_show (struct device *dev,struct device_attribute *attr,char *buf)
@@ -334,6 +354,10 @@ static ssize_t component_id_show(struct device *dev,struct switchtec_dev *stdev =to_stdev (dev);int id =ioread16 (&stdev->mmio_sys_info->gen3.component_id); +/* component_id field not supported after gen4 */ +if (stdev->gen !=SWITCHTEC_GEN3 ) +return sprintf (buf,"none\n" ); +return sprintf (buf,"PM%04X\n" , id); }static DEVICE_ATTR_RO (component_id);
@@ -344,6 +368,10 @@ static ssize_t component_revision_show(struct device *dev,struct switchtec_dev *stdev =to_stdev (dev);int rev =ioread8 (&stdev->mmio_sys_info->gen3.component_revision); +/* component_revision field not supported after gen4 */ +if (stdev->gen !=SWITCHTEC_GEN3 ) +return sprintf (buf,"255\n" ); +return sprintf (buf,"%d\n" , rev); }static DEVICE_ATTR_RO (component_revision);
@@ -1344,6 +1372,7 @@ static int switchtec_init_pci(struct switchtec_dev *stdev,int rc;void __iomem *map;unsigned long res_start, res_len; +u32 __iomem *part_id; rc =pcim_enable_device (pdev);if (rc)
@@ -1378,7 +1407,11 @@ static int switchtec_init_pci(struct switchtec_dev *stdev, stdev->mmio_sys_info = stdev->mmio +SWITCHTEC_GAS_SYS_INFO_OFFSET ; stdev->mmio_flash_info = stdev->mmio +SWITCHTEC_GAS_FLASH_INFO_OFFSET ; stdev->mmio_ntb = stdev->mmio +SWITCHTEC_GAS_NTB_OFFSET ; - stdev->partition =ioread8 (&stdev->mmio_sys_info->gen3.partition_id); +if (stdev->gen ==SWITCHTEC_GEN4 ) + part_id = &stdev->mmio_sys_info->gen4.partition_id; +else + part_id = &stdev->mmio_sys_info->gen3.partition_id; + stdev->partition =ioread32 (part_id); stdev->partition_count =ioread8 (&stdev->mmio_ntb->partition_count); stdev->mmio_part_cfg_all = stdev->mmio +SWITCHTEC_GAS_PART_CFG_OFFSET ; stdev->mmio_part_cfg = &stdev->mmio_part_cfg_all[stdev->partition];
include/linux/switchtec.h +45 -1
56 lines not on this page · raw message
Signature · 1 line
2.20.1
Re: [PATCH 08/12] PCI/switchtec: Add gen4 support in struct sys_info_regs
Quoted 2 lines · Logan Gunthorpe
Add a union with gen3 and gen4
sys_infostructs. Ensure any use of the gen3 struct is gaurded by an if statement onstdev->gen.
s/gaurded/guarded/
Quoted 31 lines
Signed-off-by Logan Gunthorpe <[email protected]>--- drivers/pci/switch/switchtec.c | 41 +++++++++++++++++++++++++++--- include/linux/switchtec.h | 46 +++++++++++++++++++++++++++++++++- 2 files changed, 82 insertions(+), 5 deletions(-)drivers/pci/switch/switchtec.c +8
-2@@ -317,8 +317,14 @@ static ssize_t field ## _show(struct device *dev, \ struct device_attribute *attr, char *buf) \ { \ struct switchtec_dev *stdev = to_stdev(dev); \ -return io_string_show (buf, &stdev->mmio_sys_info->gen3.field, \ -sizeof (stdev->mmio_sys_info->gen3.field)); \ + struct sys_info_regs __iomem *si = stdev->mmio_sys_info; \ + \ +if (stdev->gen ==SWITCHTEC_GEN4 ) \ +return io_string_show (buf, &si->gen4.field, \ +sizeof (si->gen4.field)); \ +else \ +return io_string_show (buf, &si->gen3.field, \ +sizeof (si->gen3.field)); \ } \ \static DEVICE_ATTR_RO (field)@@ -326,7 +332,21 @@ static DEVICE_ATTR_RO(field)DEVICE_ATTR_SYS_INFO_STR (vendor_id);DEVICE_ATTR_SYS_INFO_STR (product_id);DEVICE_ATTR_SYS_INFO_STR (product_revision);
drivers/pci/switch/switchtec.c static DEVICE_ATTR_RO(field) · line 343
Quoted 10 lines
drivers/pci/switch/switchtec.c +9
-1-DEVICE_ATTR_SYS_INFO_STR (component_vendor); + +static ssize_t component_vendor_show (struct device *dev, +struct device_attribute *attr,char *buf) +{ +struct switchtec_dev *stdev =to_stdev (dev); +struct sys_info_regs __iomem *si = stdev->mmio_sys_info; + +/* component_vendor field not supported after gen4 */ +if (stdev->gen !=SWITCHTEC_GEN3 )
I assume the comment should say "after gen3"? More occurrences below.
Quoted 28 lines
drivers/pci/switch/switchtec.c +14
-0+return sprintf (buf,"none\n" ); + +return io_string_show (buf, &si->gen3.component_vendor, +sizeof (si->gen3.component_vendor)); +} +static DEVICE_ATTR_RO (component_vendor);static ssize_t component_id_show (struct device *dev,struct device_attribute *attr,char *buf)@@ -334,6 +354,10 @@ static ssize_t component_id_show(struct device *dev,struct switchtec_dev *stdev =to_stdev (dev);int id =ioread16 (&stdev->mmio_sys_info->gen3.component_id); +/* component_id field not supported after gen4 */ +if (stdev->gen !=SWITCHTEC_GEN3 ) +return sprintf (buf,"none\n" ); +return sprintf (buf,"PM%04X\n" , id); }static DEVICE_ATTR_RO (component_id);@@ -344,6 +368,10 @@ static ssize_t component_revision_show(struct device *dev,struct switchtec_dev *stdev =to_stdev (dev);int rev =ioread8 (&stdev->mmio_sys_info->gen3.component_revision); +/* component_revision field not supported after gen4 */ +if (stdev->gen !=SWITCHTEC_GEN3 ) +return sprintf (buf,"255\n" ); +
Logan Gunthorpe
maintainerPATCH 9/12 PCI/switchtec: Add gen4 support in struct flash_info_regs
In reply to Logan GunthorpeFrom: Kelvin Cao <[email protected]>
Add a union with gen3 and gen4 flash_info structs.
--- drivers/pci/switch/switchtec.c | 200 ++++++++++++++++++++++----- include/linux/switchtec.h | 85 ++++++++++-- include/uapi/linux/switchtec_ioctl.h | 9 ++ 3 files changed, 247 insertions(+), 47 deletions(-)
drivers/pci/switch/switchtec.c +166 -34
@@ -596,8 +596,15 @@ static int ioctl_flash_info(struct switchtec_dev *stdev,struct switchtec_ioctl_flash_info info = {0 };struct flash_info_regs __iomem *fi = stdev->mmio_flash_info; - info.flash_length =ioread32 (&fi->flash_length); - info.num_partitions =SWITCHTEC_IOCTL_NUM_PARTITIONS ; +if (stdev->gen ==SWITCHTEC_GEN3 ) { + info.flash_length =ioread32 (&fi->gen3.flash_length); + info.num_partitions =SWITCHTEC_NUM_PARTITIONS_GEN3 ; + }else if (stdev->gen ==SWITCHTEC_GEN4 ) { + info.flash_length =ioread32 (&fi->gen4.flash_length); + info.num_partitions =SWITCHTEC_NUM_PARTITIONS_GEN4 ; + }else { +return -ENOTSUPP ; + }if (copy_to_user (uinfo, &info,sizeof (info)))return -EFAULT ;
@@ -612,75 +619,200 @@ static void set_fw_info_part(struct switchtec_ioctl_flash_part_info *info, info->length =ioread32 (&pi->length); } -static int ioctl_flash_part_info (struct switchtec_dev *stdev, -struct switchtec_ioctl_flash_part_info __user *uinfo) +static int flash_part_info_gen3 (struct switchtec_dev *stdev, +struct switchtec_ioctl_flash_part_info *info) { -struct switchtec_ioctl_flash_part_info info = {0 }; -struct flash_info_regs __iomem *fi = stdev->mmio_flash_info; -struct sys_info_regs __iomem *si = stdev->mmio_sys_info; +struct flash_info_regs_gen3 __iomem *fi = &stdev->mmio_flash_info->gen3; +struct sys_info_regs_gen3 __iomem *si = &stdev->mmio_sys_info->gen3;u32 active_addr =-1 ; -if (copy_from_user (&info, uinfo,sizeof (info))) -return -EFAULT ; - -switch (info.flash_partition) { +switch (info->flash_partition) {case SWITCHTEC_IOCTL_PART_CFG0 : active_addr =ioread32 (&fi->active_cfg); -set_fw_info_part (&info, &fi->cfg0); -if (ioread16 (&si->gen3.cfg_running) ==SWITCHTEC_CFG0_RUNNING ) - info.active |=SWITCHTEC_IOCTL_PART_RUNNING ; +set_fw_info_part (info, &fi->cfg0); +if (ioread16 (&si->cfg_running) ==SWITCHTEC_GEN3_CFG0_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ;break ;case SWITCHTEC_IOCTL_PART_CFG1 : active_addr =ioread32 (&fi->active_cfg); -set_fw_info_part (&info, &fi->cfg1); -if (ioread16 (&si->gen3.cfg_running) ==SWITCHTEC_CFG1_RUNNING ) - info.active |=SWITCHTEC_IOCTL_PART_RUNNING ; +set_fw_info_part (info, &fi->cfg1); +if (ioread16 (&si->cfg_running) ==SWITCHTEC_GEN3_CFG1_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ;break ;case SWITCHTEC_IOCTL_PART_IMG0 : active_addr =ioread32 (&fi->active_img); -set_fw_info_part (&info, &fi->img0); -if (ioread16 (&si->gen3.img_running) ==SWITCHTEC_IMG0_RUNNING ) - info.active |=SWITCHTEC_IOCTL_PART_RUNNING ; +set_fw_info_part (info, &fi->img0); +if (ioread16 (&si->img_running) ==SWITCHTEC_GEN3_IMG0_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ;break ;case SWITCHTEC_IOCTL_PART_IMG1 : active_addr =ioread32 (&fi->active_img); -set_fw_info_part (&info, &fi->img1); -if (ioread16 (&si->gen3.img_running) ==SWITCHTEC_IMG1_RUNNING ) - info.active |=SWITCHTEC_IOCTL_PART_RUNNING ; +set_fw_info_part (info, &fi->img1); +if (ioread16 (&si->img_running) ==SWITCHTEC_GEN3_IMG1_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ;break ;case SWITCHTEC_IOCTL_PART_NVLOG : -set_fw_info_part (&info, &fi->nvlog); +set_fw_info_part (info, &fi->nvlog);break ;case SWITCHTEC_IOCTL_PART_VENDOR0 : -set_fw_info_part (&info, &fi->vendor[0 ]); +set_fw_info_part (info, &fi->vendor[0 ]);break ;case SWITCHTEC_IOCTL_PART_VENDOR1 : -set_fw_info_part (&info, &fi->vendor[1 ]); +set_fw_info_part (info, &fi->vendor[1 ]);break ;case SWITCHTEC_IOCTL_PART_VENDOR2 : -set_fw_info_part (&info, &fi->vendor[2 ]); +set_fw_info_part (info, &fi->vendor[2 ]);break ;case SWITCHTEC_IOCTL_PART_VENDOR3 : -set_fw_info_part (&info, &fi->vendor[3 ]); +set_fw_info_part (info, &fi->vendor[3 ]);break ;case SWITCHTEC_IOCTL_PART_VENDOR4 : -set_fw_info_part (&info, &fi->vendor[4 ]); +set_fw_info_part (info, &fi->vendor[4 ]);break ;case SWITCHTEC_IOCTL_PART_VENDOR5 : -set_fw_info_part (&info, &fi->vendor[5 ]); +set_fw_info_part (info, &fi->vendor[5 ]);break ;case SWITCHTEC_IOCTL_PART_VENDOR6 : -set_fw_info_part (&info, &fi->vendor[6 ]); +set_fw_info_part (info, &fi->vendor[6 ]);break ;case SWITCHTEC_IOCTL_PART_VENDOR7 : -set_fw_info_part (&info, &fi->vendor[7 ]); +set_fw_info_part (info, &fi->vendor[7 ]);break ;default :return -EINVAL ; } -if (info.address == active_addr) - info.active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (info->address == active_addr) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; + +return 0 ; +} + +static int flash_part_info_gen4 (struct switchtec_dev *stdev, +struct switchtec_ioctl_flash_part_info *info) +{ +struct flash_info_regs_gen4 __iomem *fi = &stdev->mmio_flash_info->gen4; +struct sys_info_regs_gen4 __iomem *si = &stdev->mmio_sys_info->gen4; +struct active_partition_info_gen4 __iomem *af = &fi->active_flag; + +switch (info->flash_partition) { +case SWITCHTEC_IOCTL_PART_MAP_0 : +set_fw_info_part (info, &fi->map0); +break ; +case SWITCHTEC_IOCTL_PART_MAP_1 : +set_fw_info_part (info, &fi->map1); +break ; +case SWITCHTEC_IOCTL_PART_KEY_0 : +set_fw_info_part (info, &fi->key0); +if (ioread8 (&af->key) ==SWITCHTEC_GEN4_KEY0_ACTIVE ) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (ioread16 (&si->key_running) ==SWITCHTEC_GEN4_KEY0_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ; +break ; +case SWITCHTEC_IOCTL_PART_KEY_1 : +set_fw_info_part (info, &fi->key1); +if (ioread8 (&af->key) ==SWITCHTEC_GEN4_KEY1_ACTIVE ) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (ioread16 (&si->key_running) ==SWITCHTEC_GEN4_KEY1_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ; +break ; +case SWITCHTEC_IOCTL_PART_BL2_0 : +set_fw_info_part (info, &fi->bl2_0); +if (ioread8 (&af->bl2) ==SWITCHTEC_GEN4_BL2_0_ACTIVE ) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (ioread16 (&si->bl2_running) ==SWITCHTEC_GEN4_BL2_0_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ; +break ; +case SWITCHTEC_IOCTL_PART_BL2_1 : +set_fw_info_part (info, &fi->bl2_1); +if (ioread8 (&af->bl2) ==SWITCHTEC_GEN4_BL2_1_ACTIVE ) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (ioread16 (&si->bl2_running) ==SWITCHTEC_GEN4_BL2_1_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ; +break ; +case SWITCHTEC_IOCTL_PART_CFG0 : +set_fw_info_part (info, &fi->cfg0); +if (ioread8 (&af->cfg) ==SWITCHTEC_GEN4_CFG0_ACTIVE ) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (ioread16 (&si->cfg_running) ==SWITCHTEC_GEN4_CFG0_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ; +break ; +case SWITCHTEC_IOCTL_PART_CFG1 : +set_fw_info_part (info, &fi->cfg1); +if (ioread8 (&af->cfg) ==SWITCHTEC_GEN4_CFG1_ACTIVE ) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (ioread16 (&si->cfg_running) ==SWITCHTEC_GEN4_CFG1_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ; +break ; +case SWITCHTEC_IOCTL_PART_IMG0 : +set_fw_info_part (info, &fi->img0); +if (ioread8 (&af->img) ==SWITCHTEC_GEN4_IMG0_ACTIVE ) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (ioread16 (&si->img_running) ==SWITCHTEC_GEN4_IMG0_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ; +break ; +case SWITCHTEC_IOCTL_PART_IMG1 : +set_fw_info_part (info, &fi->img1); +if (ioread8 (&af->img) ==SWITCHTEC_GEN4_IMG1_ACTIVE ) + info->active |=SWITCHTEC_IOCTL_PART_ACTIVE ; +if (ioread16 (&si->img_running) ==SWITCHTEC_GEN4_IMG1_RUNNING ) + info->active |=SWITCHTEC_IOCTL_PART_RUNNING ; +break ; +case SWITCHTEC_IOCTL_PART_NVLOG : +set_fw_info_part (info, &fi->nvlog); +break ; +case SWITCHTEC_IOCTL_PART_VENDOR0 : +set_fw_info_part (info, &fi->vendor[0 ]); +break ; +case SWITCHTEC_IOCTL_PART_VENDOR1 : +set_fw_info_part (info, &fi->vendor[1 ]); +break ; +case SWITCHTEC_IOCTL_PART_VENDOR2 : +set_fw_info_part (info, &fi->vendor[2 ]); +break ; +case SWITCHTEC_IOCTL_PART_VENDOR3 : +set_fw_info_part (info, &fi->vendor[3 ]); +break ; +case SWITCHTEC_IOCTL_PART_VENDOR4 : +set_fw_info_part (info, &fi->vendor[4 ]); +break ; +case SWITCHTEC_IOCTL_PART_VENDOR5 : +set_fw_info_part (info, &fi->vendor[5 ]); +break ; +case SWITCHTEC_IOCTL_PART_VENDOR6 : +set_fw_info_part (info, &fi->vendor[6 ]); +break ; +case SWITCHTEC_IOCTL_PART_VENDOR7 : +set_fw_info_part (info, &fi->vendor[7 ]); +break ; +default : +return -EINVAL ; + } + +return 0 ; +} + +static int ioctl_flash_part_info (struct switchtec_dev *stdev, +struct switchtec_ioctl_flash_part_info __user *uinfo) +{ +int ret; +struct switchtec_ioctl_flash_part_info info = {0 }; + +if (copy_from_user (&info, uinfo,sizeof (info))) +return -EFAULT ; + +if (stdev->gen ==SWITCHTEC_GEN3 ) { + ret =flash_part_info_gen3 (stdev, &info); +if (ret) +return ret; + }else if (stdev->gen ==SWITCHTEC_GEN4 ) { + ret =flash_part_info_gen4 (stdev, &info); +if (ret) +return ret; + }else { +return -ENOTSUPP ; + } +if (copy_to_user (uinfo, &info,sizeof (info)))return -EFAULT ;
include/linux/switchtec.h +72 -13
114 lines not on this page · raw message
include/uapi/linux/switchtec_ioctl.h +9 -0
@@ -32,7 +32,16 @@#define SWITCHTEC_IOCTL_PART_VENDOR5 10#define SWITCHTEC_IOCTL_PART_VENDOR6 11#define SWITCHTEC_IOCTL_PART_VENDOR7 12 +#define SWITCHTEC_IOCTL_PART_BL2_0 13 +#define SWITCHTEC_IOCTL_PART_BL2_1 14 +#define SWITCHTEC_IOCTL_PART_MAP_0 15 +#define SWITCHTEC_IOCTL_PART_MAP_1 16 +#define SWITCHTEC_IOCTL_PART_KEY_0 17 +#define SWITCHTEC_IOCTL_PART_KEY_1 18 +#define SWITCHTEC_IOCTL_NUM_PARTITIONS 13 +#define SWITCHTEC_NUM_PARTITIONS_GEN3 13 +#define SWITCHTEC_NUM_PARTITIONS_GEN4 19struct switchtec_ioctl_flash_info {__u64 flash_length;
Signature · 1 line
2.20.1
Re: [PATCH 09/12] PCI/switchtec: Add gen4 support in struct flash_info_regs
Quoted 3 lines · Logan Gunthorpe
From: Kelvin Cao <[email protected]>
Add a union with gen3 and gen4
flash_infostructs.
This does a lot more than add a union :)
I think this looks reasonable, but I would like it even better if this and related patches could be split up a little bit differently:
- Rename SWITCHTEC_CFG0_RUNNING to SWITCHTEC_GEN3_CFG0_RUNNING, etc (purely mechanical change, so trivial and obvious).
- Add switchtec_gen and the tests where it's needed, but with only SWITCHTEC_GEN3 cases for now.
- Refactor ioctl_flash_part_info() (still only supports GEN3). Maybe adds struct flash_info_regs and union, but only with gen3.
- Add GEN4 support (patch basically contains only GEN4-related things and doesn't touch GEN3 things at all). Maybe it would still make sense to split the GEN4 support into multiple patches (as in this series), or maybe they could be squashed into a single GEN4 patch?
- It seems like at least the aliasing quirk and the driver device ID update could/should be squashed since they contain the same constants.
Bjorn
Logan Gunthorpe
maintainerRe: [PATCH 09/12] PCI/switchtec: Add gen4 support in struct flash_info_regs
Quoted 25 lines · Bjorn Helgaas
Quoted 3 lines · Logan Gunthorpe
From: Kelvin Cao <[email protected]>
Add a union with gen3 and gen4
flash_infostructs.This does a lot more than add a union :)
I think this looks reasonable, but I would like it even better if this and related patches could be split up a little bit differently:
- Rename
SWITCHTEC_CFG0_RUNNINGtoSWITCHTEC_GEN3_CFG0_RUNNING, etc (purely mechanical change, so trivial and obvious).- Add
switchtec_genand the tests where it's needed, but with onlySWITCHTEC_GEN3cases for now.- Refactor
ioctl_flash_part_info()(still only supports GEN3). Maybe adds structflash_info_regsand union, but only with gen3.- Add GEN4 support (patch basically contains only GEN4-related things and doesn't touch GEN3 things at all). Maybe it would still make sense to split the GEN4 support into multiple patches (as in this series), or maybe they could be squashed into a single GEN4 patch?
Quoted 3 lines
- It seems like at least the aliasing quirk and the driver device ID update could/should be squashed since they contain the same constants.
Thanks for the review. Yes, I should be able to clean this up and submit a v2 in the next week or two.
Logan
Logan Gunthorpe
maintainerPATCH 10/12 PCI/switchtec: Add permission check for the GAS access MRPC commands
In reply to Logan GunthorpeFrom: Kelvin Cao <[email protected]>
GEN4 hardware provides new MRPC commands to read and write from directly from any address in the PCI BAR (which Microsemi refers to as GAS). Seeing accessing BAR registers can be dangerous and break the driver, we don't want unpriviliged users to have this ability.
Therefore, for the local and remote GAS access MRPC commands, the requesting process should need CAP_SYS_ADMIN. Priviligded processes will already have access to the bar through the sysfs resource file so this doesn't give userspace any capabilities it didn't already have.
--- drivers/pci/switch/switchtec.c | 6 ++++++ include/linux/switchtec.h | 5 +++++ 2 files changed, 11 insertions(+)
drivers/pci/switch/switchtec.c +6 -0
@@ -478,6 +478,12 @@ static ssize_t switchtec_dev_write(struct file *filp, const char __user *data, rc = -EFAULT ; goto out; } +if (((MRPC_CMD_ID (stuser->cmd) ==MRPC_GAS_WRITE ) || + (MRPC_CMD_ID (stuser->cmd) ==MRPC_GAS_READ )) && + !capable (CAP_SYS_ADMIN )) { + rc = -EPERM ; + goto out; + } data +=sizeof (stuser->cmd); rc =copy_from_user (&stuser->data, data, size -sizeof (stuser->cmd));
include/linux/switchtec.h +5 -0
@@ -21,6 +21,11 @@#define SWITCHTEC_EVENT_FATAL BIT(4)#define SWITCHTEC_DMA_MRPC_EN BIT(0) + +#define MRPC_GAS_READ 0x29 +#define MRPC_GAS_WRITE 0x87 +#define MRPC_CMD_ID (x) ((x) & 0xffff) +enum {SWITCHTEC_GAS_MRPC_OFFSET =0x0000 ,SWITCHTEC_GAS_TOP_CFG_OFFSET =0x1000 ,
Signature · 1 line
2.20.1
Logan Gunthorpe
maintainerPATCH 11/12 PCI/switchtec: Introduce gen4 variant IDS in the device ID table
In reply to Logan GunthorpeFrom: Kelvin Cao <[email protected]>
Now that GEN4 is properly supported, advertise support in the module's device ID table.
--- drivers/pci/switch/switchtec.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
drivers/pci/switch/switchtec.c +18 -0
@@ -1691,6 +1691,24 @@ static const struct pci_device_id switchtec_pci_tbl[] = {SWITCHTEC_PCI_DEVICE (0x8574 ,SWITCHTEC_GEN3 ),//PFXI 64XG3 SWITCHTEC_PCI_DEVICE (0x8575 ,SWITCHTEC_GEN3 ),//PFXI 80XG3 SWITCHTEC_PCI_DEVICE (0x8576 ,SWITCHTEC_GEN3 ),//PFXI 96XG3 +SWITCHTEC_PCI_DEVICE (0x4000 ,SWITCHTEC_GEN4 ),//PFX 100XG4 +SWITCHTEC_PCI_DEVICE (0x4084 ,SWITCHTEC_GEN4 ),//PFX 84XG4 +SWITCHTEC_PCI_DEVICE (0x4068 ,SWITCHTEC_GEN4 ),//PFX 68XG4 +SWITCHTEC_PCI_DEVICE (0x4052 ,SWITCHTEC_GEN4 ),//PFX 52XG4 +SWITCHTEC_PCI_DEVICE (0x4036 ,SWITCHTEC_GEN4 ),//PFX 36XG4 +SWITCHTEC_PCI_DEVICE (0x4028 ,SWITCHTEC_GEN4 ),//PFX 28XG4 +SWITCHTEC_PCI_DEVICE (0x4100 ,SWITCHTEC_GEN4 ),//PSX 100XG4 +SWITCHTEC_PCI_DEVICE (0x4184 ,SWITCHTEC_GEN4 ),//PSX 84XG4 +SWITCHTEC_PCI_DEVICE (0x4168 ,SWITCHTEC_GEN4 ),//PSX 68XG4 +SWITCHTEC_PCI_DEVICE (0x4152 ,SWITCHTEC_GEN4 ),//PSX 52XG4 +SWITCHTEC_PCI_DEVICE (0x4136 ,SWITCHTEC_GEN4 ),//PSX 36XG4 +SWITCHTEC_PCI_DEVICE (0x4128 ,SWITCHTEC_GEN4 ),//PSX 28XG4 +SWITCHTEC_PCI_DEVICE (0x4200 ,SWITCHTEC_GEN4 ),//PAX 100XG4 +SWITCHTEC_PCI_DEVICE (0x4284 ,SWITCHTEC_GEN4 ),//PAX 84XG4 +SWITCHTEC_PCI_DEVICE (0x4268 ,SWITCHTEC_GEN4 ),//PAX 68XG4 +SWITCHTEC_PCI_DEVICE (0x4252 ,SWITCHTEC_GEN4 ),//PAX 52XG4 +SWITCHTEC_PCI_DEVICE (0x4236 ,SWITCHTEC_GEN4 ),//PAX 36XG4 +SWITCHTEC_PCI_DEVICE (0x4228 ,SWITCHTEC_GEN4 ),//PAX 28XG4 {0} };MODULE_DEVICE_TABLE (pci, switchtec_pci_tbl);
Signature · 1 line
2.20.1
Logan Gunthorpe
maintainerPATCH 12/12 PCI: Apply switchtec DMA aliasing quirk to GEN4 devices
In reply to Logan GunthorpeAdd GEN4 device IDs for quirk_switchtec_ntb_dma_alias().
--- drivers/pci/quirks.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
drivers/pci/quirks.c +18 -0
@@ -5373,6 +5373,24 @@ SWITCHTEC_QUIRK(0x8573); /* PFXI 48XG3 */SWITCHTEC_QUIRK (0x8574 );/* PFXI 64XG3 */ SWITCHTEC_QUIRK (0x8575 );/* PFXI 80XG3 */ SWITCHTEC_QUIRK (0x8576 );/* PFXI 96XG3 */ +SWITCHTEC_QUIRK (0x4000 );/* PFX 100XG4 */ +SWITCHTEC_QUIRK (0x4084 );/* PFX 84XG4 */ +SWITCHTEC_QUIRK (0x4068 );/* PFX 68XG4 */ +SWITCHTEC_QUIRK (0x4052 );/* PFX 52XG4 */ +SWITCHTEC_QUIRK (0x4036 );/* PFX 36XG4 */ +SWITCHTEC_QUIRK (0x4028 );/* PFX 28XG4 */ +SWITCHTEC_QUIRK (0x4100 );/* PSX 100XG4 */ +SWITCHTEC_QUIRK (0x4184 );/* PSX 84XG4 */ +SWITCHTEC_QUIRK (0x4168 );/* PSX 68XG4 */ +SWITCHTEC_QUIRK (0x4152 );/* PSX 52XG4 */ +SWITCHTEC_QUIRK (0x4136 );/* PSX 36XG4 */ +SWITCHTEC_QUIRK (0x4128 );/* PSX 28XG4 */ +SWITCHTEC_QUIRK (0x4200 );/* PAX 100XG4 */ +SWITCHTEC_QUIRK (0x4284 );/* PAX 84XG4 */ +SWITCHTEC_QUIRK (0x4268 );/* PAX 68XG4 */ +SWITCHTEC_QUIRK (0x4252 );/* PAX 52XG4 */ +SWITCHTEC_QUIRK (0x4236 );/* PAX 36XG4 */ +SWITCHTEC_QUIRK (0x4228 );/* PAX 28XG4 */ /* * On Lenovo Thinkpad P50 SKUs with a Nvidia Quadro M1000M, the BIOS does
Signature · 1 line
2.20.1
Quoted 34 lines · Logan Gunthorpe
Hi,
Please find a bunch of patches for the switchtec driver collected over the last few months.
The first 2 patches fix a couple of minor bugs. Patch 3 adds support for a new event that is available in specific firmware versions. Patches 4 and 5 are some code cleanup changes to simplify the logic. And the last 6 patches implement support for the new Gen4 hardware.
This patchset is based on v5.5-rc5 and a git branch is available here:
https://github.com/sbates130272/linux-p2pmem switchtec-next
Thanks,
Logan
--
Kelvin Cao (3):
PCI/switchtec: Add gen4 support in structflash_info_regsPCI/switchtec: Add permission check for the GAS access MRPC commands PCI/switchtec: Introduce gen4 variant IDS in the device ID tableLogan Gunthorpe (6):
PCI/switchtec: Fixvep_vector_numberioread width PCI/switchtec: Add support for new events
PCI/switchtec: Introduce Generation Variable PCI/switchtec: Separate out gen3 specific fields in thesys_info_regs
structure
PCI/switchtec: Add gen4 support in structsys_info_regs
PCI: Apply switchtec DMA aliasing quirk to GEN4 devices
Quoted 5 lines
Wesley Sheng (3):
PCI/switchtec: Usedma_set_mask_and_coherent()PCI/switchtec: Remove redundant valid PFF number count PCI/switchtec: Move check event id frommask_event()to
switchtec_event_isr()
Current order is:
[PATCH 01/12] PCI/switchtec: Use dma_set_mask_and_coherent()
[PATCH 02/12] PCI/switchtec: Fix vep_vector_number ioread width
[PATCH 03/12] PCI/switchtec: Add support for new events
[PATCH 04/12] PCI/switchtec: Remove redundant valid PFF number count
[PATCH 05/12] PCI/switchtec: Move check event id from mask_event() to switchtec_event_isr() [PATCH 06/12] PCI/switchtec: Introduce Generation Variable
[PATCH 07/12] PCI/switchtec: Separate out gen3 specific fields in the sys_info_regs structure [PATCH 08/12] PCI/switchtec: Add gen4 support in struct sys_info_regs
[PATCH 09/12] PCI/switchtec: Add gen4 support in struct flash_info_regs
[PATCH 10/12] PCI/switchtec: Add permission check for the GAS access MRPC commands
10/12 looks lonely in the middle of the gen4 stuff, and it looks like it's unrelated to gen3/gen4? Maybe it could be moved up after 05/12?
[PATCH 11/12] PCI/switchtec: Introduce gen4 variant IDS in the device ID table [PATCH 12/12] PCI: Apply switchtec DMA aliasing quirk to GEN4 devices
I speculatively reordered the permission check patch and applied these to my pci/switchtec branch for v5.6 (reverse order from "git log"):
b96abab6314f ("PCI/switchtec: Add permission check for the GAS access MRPC commands") 5f23367bd4df ("PCI/switchtec: Move check event ID from mask_event() to switchtec_event_isr()") 6722d609bc82 ("PCI/switchtec: Remove redundant valid PFF number count") 3f3a521ecc81 ("PCI/switchtec: Add support for intercom notify and UEC Port") 9375646b4cf0 ("PCI/switchtec: Fix vep_vector_number ioread width") aa82130a22f7 ("PCI/switchtec: Use dma_set_mask_and_coherent()")
If you rework any of the subsequent ones, you can just post those without reposting these first six.
Bjorn
Logan Gunthorpe
maintainerQuoted 54 lines · Bjorn Helgaas
Quoted 34 lines · Logan Gunthorpe
Hi,
Please find a bunch of patches for the switchtec driver collected over the last few months.
The first 2 patches fix a couple of minor bugs. Patch 3 adds support for a new event that is available in specific firmware versions. Patches 4 and 5 are some code cleanup changes to simplify the logic. And the last 6 patches implement support for the new Gen4 hardware.
This patchset is based on v5.5-rc5 and a git branch is available here:
https://github.com/sbates130272/linux-p2pmem switchtec-next
Thanks,
Logan
--
Kelvin Cao (3):
PCI/switchtec: Add gen4 support in structflash_info_regsPCI/switchtec: Add permission check for the GAS access MRPC commands PCI/switchtec: Introduce gen4 variant IDS in the device ID tableLogan Gunthorpe (6):
PCI/switchtec: Fixvep_vector_numberioread width PCI/switchtec: Add support for new events
PCI/switchtec: Introduce Generation Variable PCI/switchtec: Separate out gen3 specific fields in thesys_info_regs
structure
PCI/switchtec: Add gen4 support in structsys_info_regs
PCI: Apply switchtec DMA aliasing quirk to GEN4 devicesQuoted 5 lines
Wesley Sheng (3):
PCI/switchtec: Usedma_set_mask_and_coherent()PCI/switchtec: Remove redundant valid PFF number count PCI/switchtec: Move check event id frommask_event()to
switchtec_event_isr()Current order is:
[PATCH 01/12] PCI/switchtec: Use
dma_set_mask_and_coherent()
[PATCH 02/12] PCI/switchtec: Fixvep_vector_numberioread width
[PATCH 03/12] PCI/switchtec: Add support for new events
[PATCH 04/12] PCI/switchtec: Remove redundant valid PFF number count
[PATCH 05/12] PCI/switchtec: Move check event id frommask_event()toswitchtec_event_isr()[PATCH 06/12] PCI/switchtec: Introduce Generation Variable
[PATCH 07/12] PCI/switchtec: Separate out gen3 specific fields in thesys_info_regsstructure [PATCH 08/12] PCI/switchtec: Add gen4 support in structsys_info_regs
[PATCH 09/12] PCI/switchtec: Add gen4 support in structflash_info_regs
[PATCH 10/12] PCI/switchtec: Add permission check for the GAS access MRPC commands
Quoted 2 lines
10/12 looks lonely in the middle of the gen4 stuff, and it looks like it's unrelated to gen3/gen4? Maybe it could be moved up after 05/12?
Yes, sort of: It's related to GEN4 because the GAS access MRPC command is introduced in GEN4. So it won't really have any effect until after the GEN4 IDs are added. But there is no harm in applying it earlier.
Quoted 10 lines
I speculatively reordered the permission check patch and applied these to my pci/switchtec branch for v5.6 (reverse order from "git log"):
b96abab6314f("PCI/switchtec: Add permission check for the GAS access MRPC commands")5f23367bd4df("PCI/switchtec: Move check event ID frommask_event()toswitchtec_event_isr()")6722d609bc82("PCI/switchtec: Remove redundant valid PFF number count")3f3a521ecc81("PCI/switchtec: Add support for intercom notify and UEC Port")9375646b4cf0("PCI/switchtec: Fixvep_vector_numberioread width")aa82130a22f7("PCI/switchtec: Usedma_set_mask_and_coherent()")
Quoted 2 lines
If you rework any of the subsequent ones, you can just post those without reposting these first six.
Sounds good, thanks!
Logan