Command submission via GuC for SKL
Subsystems (2)
cover Command submission via GuC for SKL
From: Alex Dai <[email protected]>
This series of patch is to enable ExecList submission via GuC. Here are some key points related to this series, not in particular order.
*** i915_guc_client ***
We use the term client to avoid confusion with contexts. A i915_guc_client is equivalent to GuC object guc_context_desc. This context descriptor is allocated from a pool of 1024 entries. Kernel driver will allocate doorbell and workqueue for it. Also the process descriptor (guc_process_desc), which is mapped to client space. So the client can write Work Item then ring the doorbell.
To simplify the implementation, we allocate one gem object that contains all pages for doorbell, process descriptor and workqueue.
*** intel_guc ***
Top level structure of guc. It handles firmware loading and manages client pool and doorbells. intel_guc owns a i915_guc_client to do the legacy submission.
** The Scratch registers ***
There are 16 MMIO-based registers start from 0xC180. The kernel driver writes a value to the action register (SOFT_SCRATCH_0) along with any data. It then triggers an interrupt on the GuC via another register write (0xC4C8). Firmware writes a success/fail code back to the action register after processes the request. The kernel driver polls waiting for this update and then proceeds.
Details in intel_guc_action()
*** Work Items ***
There are several types of work items that the host may place into a workqueue, each with its own requirements and limitations. Currently only WQ_TYPE_INORDER is used to support legacy submission via GuC, which represents in-order queue. The kernel driver packs ring tail pointer and an ELSP context descriptor dword into Work Item.
Details in add_workqueue_item()
*** Doorbells ***
Doorbells are interrupts to uKernel. A doorbell is a single cache line (QW) mapped into process space.
Details in ring_doorbell()
*** Firmware versioning ***
The firmware build process will generate a version header file with major and minor version defined. The versions are built into CSS header of firmware too. i915 kernel driver set the minimal firmware version required by each platform. The firmware installation package will install (symbolic link) proper version of firmware.
*** Firmware log ***
Firmware log is enabled by setting i915.guc_log_level to non-negative level. Log data is printed out via reading debugfs i915_guc_log_dump. Reading from i915_guc_load_status will print out firmware loading status and scratch registers value.
TODO: the buffer works like a ring. To not missing any data, driver should handle a GuC2Host interrupt (triggered when log buffer is half full) from GuC.
*** Ring buffer size ***
It is up to 16K (4 pages) per LRC. Not HW limitation but firmware is setup in this way.
*** GuC address space ***
GuC is not expecting any gfx GGTT address that falls into range [0, WOPCM_TOP), which is reserved for Boot ROM, SRAM and WOPCM. Currently this top address is 512K in order to fit in big HuC firmware.
In order to exclude 0-512K address space from GGTT, all gfx objects used by GuC is pinned with PIN_OFFSET_BIAS along with size of WOPCM.
Alex Dai (13):
drm/i915: Add guc firmware interface headers drm/i915: GuC firmware loader
drm/i915: Add firmware version check
drm/i915: Make several execlist helper functions external drm/i915: Add functions to allocate / release gem obj for GuC drm/i915: Functions to support command submission via GuC drm/i915: Integration of GuC client
drm/i915: Interrupt routing for GuC scheduler drm/i915: Enable commands submission via GuC drm/i915: debugfs of GuC status
drm/i915: Enable GuC firmware log
drm/i915: Ring Context allocating for GuC
drm/i915: Notify GuC when RC6 state is changed
Dave Gordon (2):
drm/i915: Unified firmware loading mechanism
drm/i915: Defer default hardware context initialisation until first
open
Michael H. Nguyen (2):
drm/i915: Add i915_gem_object_write() to i915_gem.c drm/i915: Move execlists defines from .c to .h
Sagar Kamble (1):
drm/i915: Taking forcewake during GuC load.
drivers/gpu/drm/i915/Makefile | 8 +- drivers/gpu/drm/i915/i915_debugfs.c | 102 +++++ drivers/gpu/drm/i915/i915_dma.c | 6 + drivers/gpu/drm/i915/i915_drv.h | 17 + drivers/gpu/drm/i915/i915_gem.c | 39 +- drivers/gpu/drm/i915/i915_gem_context.c | 35 +- drivers/gpu/drm/i915/i915_gem_stolen.c | 10 + drivers/gpu/drm/i915/i915_params.c | 9 + drivers/gpu/drm/i915/i915_reg.h | 83 +++- drivers/gpu/drm/i915/intel_guc.h | 186 +++++++++ drivers/gpu/drm/i915/intel_guc_api.h | 218 +++++++++++ drivers/gpu/drm/i915/intel_guc_client.c | 607 +++++++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_guc_loader.c | 466 ++++++++++++++++++++++ drivers/gpu/drm/i915/intel_guc_scheduler.c | 167 ++++++++ drivers/gpu/drm/i915/intel_lrc.c | 112 ++---- drivers/gpu/drm/i915/intel_lrc.h | 3 + drivers/gpu/drm/i915/intel_pm.c | 4 + drivers/gpu/drm/i915/intel_ringbuffer.c | 2 +- drivers/gpu/drm/i915/intel_uc_loader.c | 220 +++++++++++ drivers/gpu/drm/i915/intel_uc_loader.h | 82 ++++ 20 files changed, 2279 insertions(+), 97 deletions(-) create mode 100644 drivers/gpu/drm/i915/intel_guc.h create mode 100644 drivers/gpu/drm/i915/intel_guc_api.h create mode 100644 drivers/gpu/drm/i915/intel_guc_client.c create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c create mode 100644 drivers/gpu/drm/i915/intel_guc_scheduler.c create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.h
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 1/18 drm/i915: Add guc firmware interface headers
From: Alex Dai <[email protected]>
These are the subset of GuC interface that we are going to use to enable command submission through GuC. The definition of these struct and bit setting need to match those in firmware.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/intel_guc_api.h | 217 +++++++++++++++++++++++++++++++++++ 1 file changed, 217 insertions(+) create mode 100644 drivers/gpu/drm/i915/intel_guc_api.h
drivers/gpu/drm/i915/intel_guc_api.h new file +217 -0
@@ -0,0 +1,217 @@ +/* +* Copyright © 2014 Intel Corporation +* +* Permission is hereby granted, free of charge, to any person obtaining a +* copy of this software and associated documentation files (the "Software"), +* to deal in the Software without restriction, including without limitation +* the rights to use, copy, modify, merge, publish, distribute, sublicense, +* and/or sell copies of the Software, and to permit persons to whom the +* Software is furnished to do so, subject to the following conditions: +* +* The above copyright notice and this permission notice (including the next +* paragraph) shall be included in all copies or substantial portions of the +* Software. +* +* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL +* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING +* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS +* IN THE SOFTWARE. +*/ +#ifndef _INTEL_GUC_API_H +#define _INTEL_GUC_API_H + +#define GFXCORE_FAMILY_GEN8 11 +#define GFXCORE_FAMILY_GEN9 12 +#define GFXCORE_FAMILY_FORCE_ULONG 0x7fffffff + +#define GUC_CTX_PRIORITY_CRITICAL 0 +#define GUC_CTX_PRIORITY_HIGH 1 +#define GUC_CTX_PRIORITY_NORMAL 2 +#define GUC_CTX_PRIORITY_LOW 3 + +#define MAX_GUC_GPU_CONTEXTS 1024 + +/* Work queue item header definitions */ +#define WQ_STATUS_ACTIVE 1 +#define WQ_STATUS_SUSPENDED 2 +#define WQ_STATUS_CMD_ERROR 3 +#define WQ_STATUS_ENGINE_ID_NOT_USED 4 +#define WQ_STATUS_SUSPENDED_FROM_RESET 5 +#define WQ_TYPE_SHIFT 0 +#define WQ_TYPE_BATCH_BUF (0x1 << WQ_TYPE_SHIFT) +#define WQ_TYPE_PSEUDO (0x2 << WQ_TYPE_SHIFT) +#define WQ_TYPE_INORDER (0x3 << WQ_TYPE_SHIFT) +#define WQ_TARGET_SHIFT 10 +#define WQ_LEN_SHIFT 16 +#define WQ_NO_WCFLUSH_WAIT (1 << 27) +#define WQ_PRESENT_WORKLOAD (1 << 28) + +#define WQ_RING_TAIL_SHIFT 20 +#define WQ_RING_TAIL_MASK (0x7FF << WQ_RING_TAIL_SHIFT) + +#define GUC_DOORBELL_ENABLED 1 +#define GUC_DOORBELL_DISABLED 0 + +#define GUC_CTX_DESC_ATTR_ACTIVE (1 << 0) +#define GUC_CTX_DESC_ATTR_PENDING_DB (1 << 1) +#define GUC_CTX_DESC_ATTR_KERNEL (1 << 2) +#define GUC_CTX_DESC_ATTR_PREEMPT (1 << 3) +#define GUC_CTX_DESC_ATTR_RESET (1 << 4) +#define GUC_CTX_DESC_ATTR_WQLOCKED (1 << 5) +#define GUC_CTX_DESC_ATTR_PCH (1 << 6) + +/* The guc control data is 10 DWORDs */ +#define GUC_CTL_CTXINFO 0 +#define GUC_CTL_CTXNUM_IN16_SHIFT 0 +#define GUC_CTL_BASE_ADDR_SHIFT 12 +#define GUC_CTL_ARAT_HIGH 1 +#define GUC_CTL_ARAT_LOW 2 +#define GUC_CTL_DEVICE_INFO 3 +#define GUC_CTL_GTTYPE_SHIFT 0 +#define GUC_CTL_COREFAMILY_SHIFT 7 +#define GUC_CTL_LOG_PARAMS 4 +#define GUC_LOG_VALID (1 << 0) +#define GUC_LOG_NOTIFY_ON_HALF_FULL (1 << 1) +#define GUC_LOG_ALLOC_IN_MEGABYTE (1 << 3) +#define GUC_LOG_CRASH_SHIFT 4 +#define GUC_LOG_DPC_SHIFT 6 +#define GUC_LOG_ISR_SHIFT 9 +#define GUC_LOG_BUF_ADDR_SHIFT 12 +#define GUC_CTL_PAGE_FAULT_CONTROL 5 +#define GUC_CTL_WA 6 +#define GUC_CTL_WA_UK_BY_DRIVER (1 << 3) +#define GUC_CTL_FEATURE 7 +#define GUC_CTL_VCS2_ENABLED (1 << 0) +#define GUC_CTL_KERNEL_SUBMISSIONS (1 << 1) +#define GUC_CTL_FEATURE2 (1 << 2) +#define GUC_CTL_POWER_GATING (1 << 3) +#define GUC_CTL_DISABLE_SCHEDULER (1 << 4) +#define GUC_CTL_PREEMPTION_LOG (1 << 5) +#define GUC_CTL_DEBUG 8 +#define GUC_LOG_VERBOSITY_SHIFT 0 +#define GUC_LOG_VERBOSITY_LOW (0 << GUC_LOG_VERBOSITY_SHIFT) +#define GUC_LOG_VERBOSITY_MED (1 << GUC_LOG_VERBOSITY_SHIFT) +#define GUC_LOG_VERBOSITY_HIGH (2 << GUC_LOG_VERBOSITY_SHIFT) +#define GUC_LOG_VERBOSITY_ULTRA (3 << GUC_LOG_VERBOSITY_SHIFT) +/* offset from base of LCRA, in pages */ +#define GUC_CTL_OFFSET 9 +#define GUC_CTL_SHARED_DATA_SHIFT 0 +#define GUC_CTL_RING_BUFFER_SHIFT 16 + +#define GUC_CTL_MAX_DWORDS (GUC_CTL_OFFSET + 1) + +struct guc_doorbell_info { +u32 db_status; +u32 cookie; +u32 reserved[14 ]; +} __packed; + +union guc_doorbell_qw { +struct { +u32 db_status; +u32 cookie; + }; +u64 value_qw; +} __packed; + +struct guc_process_desc { +u32 context_id; +u64 db_base_addr; +u32 head; +u32 tail; +u32 error_offset; +u64 wq_base_addr; +u32 wq_size_bytes; +u32 wq_status; +u32 engine_presence; +u32 priority; +u32 reserved[30 ]; +} __packed; + +/* Work item for submitting workloads into work queue of GuC. */ +struct guc_wq_item { +u32 header; +u32 context_desc; +u32 ring_tail; +u32 fence_id; +} __packed; + +/* The execlist context including software and HW information */ +struct guc_execlist_context { +u32 context_desc; +u32 context_id; +u32 ring_status; +u32 ring_lcra; +u32 ring_begin; +u32 ring_end; +u32 ring_next_free_location; +u32 ring_current_tail_pointer_value; +u8 engine_state_submit_value; +u8 engine_state_wait_value; +u16 pagefault_count; +u16 engine_submit_queue_count; +} __packed; + +/*Context descriptor for communicating between uKernel and Driver*/ +struct guc_context_desc { +u32 sched_common_area; +u32 context_id; +u32 pas_id; +u8 engines_used; +u64 db_trigger_cpu; +u32 db_trigger_uk; +u64 db_trigger_phy; +u16 db_id; + +struct guc_execlist_context lrc[I915_NUM_RINGS ]; + +u8 attribute; + +u32 priority; + +u32 wq_sampled_tail_offset; +u32 wq_total_submit_enqueues; + +u32 process_desc; +u32 wq_addr; +u32 wq_size; + +u32 engine_presence; + +u32 reserved0[1 ]; +u64 reserved1[1 ]; + +u64 desc_private; +} __packed; + +/* This Action will be programmed in C180 - SOFT_SCRATCH_O_REG */ +enum host2guc_action { +HOST2GUC_ACTION_DEFAULT =0x0 , +HOST2GUC_ACTION_ALLOCATE_DOORBELL =0x10 , +HOST2GUC_ACTION_DEALLOCATE_DOORBELL =0x20 , +HOST2GUC_ACTION_LIMIT +}; + +/* +* The GuC sends its response to a command by overwriting the +* command in SS0. The response is distinguishable from a command +* by the fact that all the MASK bits are set. The remaining bits +* give more detail. +*/ +#define GUC2HOST_RESPONSE_MASK 0xF0000000 +#define GUC2HOST_IS_RESPONSE (x) \ + (((x) & GUC2HOST_RESPONSE_MASK) == GUC2HOST_RESPONSE_MASK) +#define GUC2HOST_STATUS (x) (GUC2HOST_RESPONSE_MASK | (x)) + +/* GUC will return status back to SOFT_SCRATCH_O_REG */ +enum guc2host_status { +GUC2HOST_STATUS_SUCCESS =GUC2HOST_STATUS (0x0 ), +GUC2HOST_STATUS_ALLOCATE_DOORBELL_FAIL =GUC2HOST_STATUS (0x10 ), +GUC2HOST_STATUS_DEALLOCATE_DOORBELL_FAIL =GUC2HOST_STATUS (0x20 ), +GUC2HOST_STATUS_GENERIC_FAIL =GUC2HOST_STATUS (0x0000F000 ) +}; + +#endif
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 2/18 drm/i915: Add i915_gem_object_write() to i915_gem.c
In reply to [email protected]From: "Michael H. Nguyen" <[email protected]>
i915_gem_object_write() is a generic function to copy data from user memory to gem object.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/i915_drv.h | 3 +++ drivers/gpu/drm/i915/i915_gem.c | 30 ++++++++++++++++++++++++++++++ 2 files changed, 33 insertions(+)
drivers/gpu/drm/i915/i915_drv.h +3 -0
@@ -2612,6 +2612,9 @@ void i915_init_vm(struct drm_i915_private *dev_priv,void i915_gem_free_object (struct drm_gem_object *obj);void i915_gem_vma_destroy (struct i915_vma *vma); +int i915_gem_object_write (struct drm_i915_gem_object *obj, +const void *data,const size_t size); +#define PIN_MAPPABLE 0x1#define PIN_NONBLOCK 0x2#define PIN_GLOBAL 0x4
drivers/gpu/drm/i915/i915_gem.c +30 -0
@@ -5197,3 +5197,33 @@ bool i915_gem_obj_is_pinned(struct drm_i915_gem_object *obj)return false; } +/* Fill the @obj with the @size amount of @data */ +int i915_gem_object_write (struct drm_i915_gem_object *obj, +const void *data,const size_t size) +{ +struct sg_table *sg; +size_t bytes; +int ret; + + ret =i915_gem_object_get_pages (obj); +if (ret) +return ret; + +i915_gem_object_pin_pages (obj); + + sg = obj->pages; + + bytes =sg_copy_from_buffer (sg->sgl, sg->nents, + (void *)data, (size_t )size); + +i915_gem_object_unpin_pages (obj); + +if (WARN (bytes != size, +"Failed to upload all data (completed %zu bytes out of %zu total" , + bytes, size)) { +i915_gem_object_put_pages (obj); +return -EIO ; + } + +return 0 ; +}
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 3/18 drm/i915: Unified firmware loading mechanism
In reply to [email protected]From: Dave Gordon <[email protected]>
Factor out the common code of loading firmware into a new file, leaving only the uC-specific parts in the GuC loaders.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/Makefile | 3 + drivers/gpu/drm/i915/intel_uc_loader.c | 208 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_uc_loader.h | 78 +++++++++++++ 3 files changed, 289 insertions(+) create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.h
drivers/gpu/drm/i915/Makefile +3 -0
@@ -39,6 +39,9 @@ i915-y += i915_cmd_parser.o \ intel_ringbuffer.o \ intel_uncore.o +# ancilliary microcontroller support +i915-y += intel_uc_loader.o + # autogenerated null render state i915-y += intel_renderstate_gen6.o \ intel_renderstate_gen7.o \
drivers/gpu/drm/i915/intel_uc_loader.c new file +208 -0
208 lines not on this page · raw message
drivers/gpu/drm/i915/intel_uc_loader.h new file +78 -0
78 lines not on this page · raw message
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 4/18 drm/i915: GuC firmware loader
In reply to [email protected]From: Alex Dai <[email protected]>
Add GuC firmware loader. It uses the unified firmware loader to fetch firmware blob first, then load to hw in driver main thread.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/Makefile | 3 +- drivers/gpu/drm/i915/i915_dma.c | 6 + drivers/gpu/drm/i915/i915_drv.h | 6 + drivers/gpu/drm/i915/i915_gem_stolen.c | 10 ++ drivers/gpu/drm/i915/i915_reg.h | 4 +- drivers/gpu/drm/i915/intel_guc.h | 95 +++++++++++ drivers/gpu/drm/i915/intel_guc_loader.c | 271 ++++++++++++++++++++++++++++++++ 7 files changed, 393 insertions(+), 2 deletions(-) create mode 100644 drivers/gpu/drm/i915/intel_guc.h create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c
drivers/gpu/drm/i915/Makefile +2 -1
@@ -40,7 +40,8 @@ i915-y += i915_cmd_parser.o \ intel_uncore.o # ancilliary microcontroller support -i915-y += intel_uc_loader.o +i915-y += intel_uc_loader.o \ + intel_guc_loader.o # autogenerated null render state i915-y += intel_renderstate_gen6.o \
drivers/gpu/drm/i915/i915_dma.c +6 -0
@@ -465,6 +465,7 @@ static int i915_load_modeset_init(struct drm_device *dev) cleanup_gem:mutex_lock (&dev->struct_mutex); +intel_guc_ucode_fini (dev);i915_gem_cleanup_ringbuffer (dev);i915_gem_context_fini (dev);mutex_unlock (&dev->struct_mutex);
@@ -828,6 +829,10 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)intel_uncore_init (dev); +intel_wopcm_init (dev); + +intel_guc_ucode_init (dev); + ret =i915_gem_gtt_init (dev);if (ret) goto out_regs;
@@ -1071,6 +1076,7 @@ int i915_driver_unload(struct drm_device *dev)flush_workqueue (dev_priv->wq);mutex_lock (&dev->struct_mutex); +intel_guc_ucode_fini (dev);i915_gem_cleanup_ringbuffer (dev);i915_gem_batch_pool_fini (&dev_priv->mm.batch_pool);i915_gem_context_fini (dev);
drivers/gpu/drm/i915/i915_drv.h +6 -0
@@ -50,6 +50,7 @@#include <linux/intel-iommu.h> #include <linux/kref.h> #include <linux/pm_qos.h> +#include "intel_guc.h" /* General customization: */
@@ -1576,6 +1577,7 @@ struct drm_i915_private {struct intel_gmbus gmbus[GMBUS_NUM_PORTS ]; +struct intel_guc guc;/** gmbus_mutex protects against concurrent usage of the single hw gmbus * controller on different i2c buses. */
@@ -2405,6 +2407,9 @@ struct drm_i915_cmd_table {#define HAS_RC6 (dev) (INTEL_INFO(dev)->gen >= 6)#define HAS_RC6p (dev) (INTEL_INFO(dev)->gen == 6 || IS_IVYBRIDGE(dev)) +#define HAS_GUC_UCODE (dev) (IS_GEN9(dev) || IS_GEN8(dev)) +#define HAS_GUC_SCHED (dev) (IS_GEN9(dev) || IS_BROADWELL(dev)) +#define INTEL_PCH_DEVICE_ID_MASK 0xff00#define INTEL_PCH_IBX_DEVICE_ID_TYPE 0x3b00#define INTEL_PCH_CPT_DEVICE_ID_TYPE 0x1c00
@@ -2958,6 +2963,7 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_device *dev,u32 stolen_offset, u32 gtt_offset, u32 size); +void intel_wopcm_init (struct drm_device *dev);/* i915_gem_shrinker.c */ unsigned long i915_gem_shrink(struct drm_i915_private *dev_priv,
drivers/gpu/drm/i915/i915_gem_stolen.c +10 -0
@@ -551,3 +551,13 @@ err_out:drm_gem_object_unreference (&obj->base);return NULL ; } + +void intel_wopcm_init (struct drm_device *dev) +{ +struct drm_i915_private *dev_priv = dev->dev_private; + +if (HAS_GUC_UCODE (dev)) { +I915_WRITE (GUC_WOPCM_SIZE ,GUC_WOPCM_SIZE_VALUE ); +I915_WRITE (DMA_GUC_WOPCM_OFFSET ,GUC_WOPCM_OFFSET ); + } +}
drivers/gpu/drm/i915/i915_reg.h +3 -1
@@ -6289,7 +6289,9 @@ enum skl_disp_power_wells {#define GEN9_PGCTL_SSB_EU311_ACK (1 << 14)#define GEN7_MISCCPCTL (0x9424) -#define GEN7_DOP_CLOCK_GATE_ENABLE (1<<0) +#define GEN7_DOP_CLOCK_GATE_ENABLE (1<<0) +#define GEN8_DOP_CLOCK_GATE_CFCLK_ENABLE (1<<2) +#define GEN8_DOP_CLOCK_GATE_GUC_ENABLE (1<<4)/* IVYBRIDGE DPF */ #define GEN7_L3CDERRST1 0xB008/* L3CD Error Status 1 */
drivers/gpu/drm/i915/intel_guc.h new file +95 -0
95 lines not on this page · raw message
drivers/gpu/drm/i915/intel_guc_loader.c new file +271 -0
271 lines not on this page · raw message
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 5/18 drm/i915: Add firmware version check
In reply to [email protected]From: Alex Dai <[email protected]>
Set the firmware version that required by HW. Driver sets required version according to platform. After firmware is loaded but before send to HW, the major.minor version is read from CSS header field, which is 17th DWORD currently. The major version must be same; the minor version must be newer than required.
Currently the GuC fw version is 1.0.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/intel_guc.h | 1 + drivers/gpu/drm/i915/intel_guc_loader.c | 41 +++++++++++++++++++++++++++++++-- drivers/gpu/drm/i915/intel_uc_loader.c | 6 ++++- drivers/gpu/drm/i915/intel_uc_loader.h | 4 ++++ 4 files changed, 49 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_guc.h +1 -0
@@ -50,6 +50,7 @@ struct intel_guc {#define UOS_RSA_SCRATCH_0 0xc200#define UOS_RSA_SIG_SIZE 0x100 +#define UOS_VERSION_OFFSET 0x11#define DMA_ADDR_0_LOW 0xc300#define DMA_ADDR_0_HIGH 0xc304#define DMA_ADDR_1_LOW 0xc308
drivers/gpu/drm/i915/intel_guc_loader.c +39 -2
55 lines not on this page · raw message
drivers/gpu/drm/i915/intel_uc_loader.c +5 -1
@@ -49,10 +49,12 @@ static void uc_fw_finish(struct drm_device *dev, struct intel_uc_fw *uc_fw) uc_fw->uc_name, uc_fw->uc_fw_fetch_status, uc_fw->uc_fw_blob); fw = uc_fw->uc_fw_blob; - uc_fw->uc_fw_blob =NULL ;if (!fw) goto fail; +if (uc_fw->uc_fw_check && !uc_fw->uc_fw_check (uc_fw)) + goto fail; + obj =i915_gem_alloc_object (dev,round_up (fw->size,PAGE_SIZE ));if (!obj) goto fail;
@@ -64,6 +66,7 @@ static void uc_fw_finish(struct drm_device *dev, struct intel_uc_fw *uc_fw) uc_fw->uc_fw_obj = obj; uc_fw->uc_fw_size = fw->size; uc_fw->uc_fw_fetch_status =INTEL_UC_FIRMWARE_SUCCESS ; + uc_fw->uc_fw_blob =NULL ;release_firmware (fw);return ;
@@ -77,6 +80,7 @@ fail:drm_gem_object_unreference (&obj->base); uc_fw->uc_fw_fetch_status =INTEL_UC_FIRMWARE_FAIL ; + uc_fw->uc_fw_blob =NULL ;release_firmware (fw);/* OK even if fw is NULL */ }
drivers/gpu/drm/i915/intel_uc_loader.h +4 -0
@@ -68,11 +68,15 @@ struct intel_uc_fw {struct drm_i915_gem_object * uc_fw_obj;enum intel_uc_fw_status uc_fw_fetch_status;enum intel_uc_fw_status uc_fw_load_status; +uint32_t uc_fw_ver_major; +uint32_t uc_fw_ver_minor; + bool (*uc_fw_check)(struct intel_uc_fw *); };void intel_uc_fw_init (struct drm_device *dev,struct intel_uc_fw *uc_fw,const char *uc_name,const char *fw_path,bool has_ucode);int intel_uc_fw_check (struct drm_device *dev,struct intel_uc_fw *uc_fw);void intel_uc_fw_fini (struct drm_device *dev,struct intel_uc_fw *uc_fw); +bool intel_uc_fw_version_check (struct intel_uc_fw *fw,const char *data);#endif
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 6/18 drm/i915: Defer default hardware context initialisation until first open
In reply to [email protected]From: Dave Gordon <[email protected]>
In order to fully initialise the default contexts, we have to execute batchbuffer commands on the GPU engines. But we can't do that until any required firmware has been loaded, which may not be possible during driver load, because the filesystem(s) containing the firmware may not be mounted until later.
Therefore, we now allow the first call to the firmware-loading code to return -EAGAIN to indicate that it's not yet ready, and that it should be retried when the device is first opened from user code, by which time we expect that all required filesystems will have been mounted. The late-retry code will then re-attempt to load the firmware if the early attempt failed.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/i915_gem.c | 9 ++++++++- drivers/gpu/drm/i915/i915_gem_context.c | 35 ++++++++++++++++++++++++++++----- drivers/gpu/drm/i915/intel_guc.h | 2 +- drivers/gpu/drm/i915/intel_guc_loader.c | 22 ++++++++++++++++++--- drivers/gpu/drm/i915/intel_uc_loader.c | 12 +++++++++-- 6 files changed, 69 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +1 -0
@@ -1755,6 +1755,7 @@ struct drm_i915_private {/* hda/i915 audio component */ bool audio_component_registered; +bool contexts_ready;uint32_t hw_context_size;struct list_head context_list;
drivers/gpu/drm/i915/i915_gem.c +8 -1
@@ -4828,8 +4828,15 @@ i915_gem_init_hw(struct drm_device *dev)i915_gem_cleanup_ringbuffer (dev); } +/* We can't enable contexts until all firmware is loaded */ + ret =intel_guc_load_ucode (dev, false); +if (ret == -EAGAIN ) +return 0 ;/* too early */ + ret =i915_gem_context_enable (dev_priv); -if (ret && ret != -EIO ) { +if (ret ==0 ) { + dev_priv->contexts_ready = true; + }else if (ret && ret != -EIO ) {DRM_ERROR ("Context enable failed %d\n" , ret);i915_gem_cleanup_ringbuffer (dev);
drivers/gpu/drm/i915/i915_gem_context.c +30 -5
@@ -445,23 +445,48 @@ static int context_idr_cleanup(int id, void *p, void *data)return 0 ; } +/* Complete any late initialisation here */ +static int i915_gem_context_first_open (struct drm_device *dev) +{ +struct drm_i915_private *dev_priv = dev->dev_private; +int ret; + +/* We can't enable contexts until all firmware is loaded */ + ret =intel_guc_load_ucode (dev, true); +WARN_ON (ret == -EAGAIN ); + + ret =i915_gem_context_enable (dev_priv); +if (ret ==0 ) + dev_priv->contexts_ready = true; +return ret; +} +int i915_gem_context_open (struct drm_device *dev,struct drm_file *file) { +struct drm_i915_private *dev_priv = dev->dev_private;struct drm_i915_file_private *file_priv = file->driver_priv;struct intel_context *ctx; +int ret =0 ;idr_init (&file_priv->context_idr);mutex_lock (&dev->struct_mutex); - ctx =i915_gem_create_context (dev, file_priv); + +if (!dev_priv->contexts_ready) + ret =i915_gem_context_first_open (dev); + +if (ret ==0 ) { + ctx =i915_gem_create_context (dev, file_priv); +if (IS_ERR (ctx)) + ret =PTR_ERR (ctx); + } +mutex_unlock (&dev->struct_mutex); -if (IS_ERR (ctx)) { +if (ret)idr_destroy (&file_priv->context_idr); -return PTR_ERR (ctx); - } -return 0 ; +return ret; }void i915_gem_context_close (struct drm_device *dev,struct drm_file *file)
drivers/gpu/drm/i915/intel_guc.h +1 -1
@@ -89,7 +89,7 @@ struct intel_guc { GUC_ENABLE_READ_CACHE_FOR_WOPCM_DATA)/* intel_guc_loader.c */ -extern int intel_guc_load_ucode(struct drm_device *dev); +extern int intel_guc_load_ucode(struct drm_device *dev, bool wait);extern void intel_guc_ucode_fini (struct drm_device *dev);extern void intel_guc_ucode_init (struct drm_device *dev);
drivers/gpu/drm/i915/intel_guc_loader.c +19 -3
@@ -217,25 +217,41 @@ out: }/* -* Called from gem_init_hw() during driver loading and also after a GPU reset. * Check that the firmware fetching process has succeeded, and if so transfer * the loaded image to the hardware. +* +* However, there are a few checks to do first. The very first call should have +* (wait == FALSE), but the fetch_state will still be PENDING as the firmware may +* not be available that early. Therefore, on this first call, we just return. +* +* The second call should come from the first open of the device (wait == TRUE). +* This is a good time to load the firmware into the device, as by this point it +* must be available. +* +* Any subsequent calls are expected to have wait == FALSE, and indicate that the +* hardware has been reset and so the firmware should be reloaded. */ -int intel_guc_load_ucode (struct drm_device *dev) +int intel_guc_load_ucode (struct drm_device *dev,bool wait) {struct drm_i915_private *dev_priv = dev->dev_private;struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;int err; +DRM_DEBUG_DRIVER ("GuC: wait %d, fetch status %d, load status %d\n" , + wait, guc_fw->uc_fw_fetch_status, guc_fw->uc_fw_load_status); + +if (guc_fw->uc_fw_fetch_status ==INTEL_UC_FIRMWARE_PENDING && !wait) +return -EAGAIN ; + guc_fw->uc_fw_load_status =INTEL_UC_FIRMWARE_NONE ;if (guc_fw->uc_fw_fetch_status ==INTEL_UC_FIRMWARE_NONE )return 0 ; - guc_fw->uc_fw_load_status =INTEL_UC_FIRMWARE_PENDING ; err =intel_uc_fw_check (dev, guc_fw);if (err) goto fail; + guc_fw->uc_fw_load_status =INTEL_UC_FIRMWARE_PENDING ; err =guc_load_ucode (dev);if (err) goto fail;
drivers/gpu/drm/i915/intel_uc_loader.c +10 -2
@@ -49,8 +49,16 @@ static void uc_fw_finish(struct drm_device *dev, struct intel_uc_fw *uc_fw) uc_fw->uc_name, uc_fw->uc_fw_fetch_status, uc_fw->uc_fw_blob); fw = uc_fw->uc_fw_blob; -if (!fw) - goto fail; +if (!fw) { +/* no firmware found; try again in case FS was not mounted */ +DRM_DEBUG_DRIVER ("retry fetching %s fw from <%s>\n" , + uc_fw->uc_name, uc_fw->uc_fw_path); +if (request_firmware (&fw, uc_fw->uc_fw_path, &dev->pdev->dev)) + goto fail; +DRM_DEBUG_DRIVER ("fetch %s fw from <%s> succeeded, fw %p\n" , + uc_fw->uc_name, uc_fw->uc_fw_path, fw); + uc_fw->uc_fw_blob = fw; + }if (uc_fw->uc_fw_check && !uc_fw->uc_fw_check (uc_fw)) goto fail;
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/18] drm/i915: Defer default hardware context initialisation until first open
Quoted 8 lines · On Thu, Mar 26, 2015 at 12:41:13PM -0700, [email protected] wrote:
From: Dave Gordon <[email protected]>
In order to fully initialise the default contexts, we have to execute batchbuffer commands on the GPU engines. But we can't do that until any required firmware has been loaded, which may not be possible during driver load, because the filesystem(s) containing the firmware may not be mounted until later.
Quoted 6 lines
Therefore, we now allow the first call to the firmware-loading code to return
-EAGAINto indicate that it's not yet ready, and that it should be retried when the device is first opened from user code, by which time we expect that all required filesystems will have been mounted. The late-retry code will then re-attempt to load the firmware if the early attempt failed.
We've tried a similar approach a while back and it doesn't work well in conjunction with rps - the hw tends to fall over if the context state isn't properly initialized when going into rc6.
Why exactly can't we load that firmware right at boot-up, or at least stall correctly until it's there?
-Daniel
Quoted 196 lines
Issue: VIZ-4884
Signed-off-by Dave Gordon <[email protected]>Signed-off-by Alex Dai <[email protected]>--- drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/i915_gem.c | 9 ++++++++- drivers/gpu/drm/i915/i915_gem_context.c | 35 ++++++++++++++++++++++++++++----- drivers/gpu/drm/i915/intel_guc.h | 2 +- drivers/gpu/drm/i915/intel_guc_loader.c | 22 ++++++++++++++++++--- drivers/gpu/drm/i915/intel_uc_loader.c | 12 +++++++++-- 6 files changed, 69 insertions(+), 12 deletions(-)drivers/gpu/drm/i915/i915_drv.h +1
-0@@ -1755,6 +1755,7 @@ struct drm_i915_private {/* hda/i915 audio component */ bool audio_component_registered; +bool contexts_ready;uint32_t hw_context_size;struct list_head context_list;drivers/gpu/drm/i915/i915_gem.c +8
-1@@ -4828,8 +4828,15 @@ i915_gem_init_hw(struct drm_device *dev)i915_gem_cleanup_ringbuffer (dev); } +/* We can't enable contexts until all firmware is loaded */ + ret =intel_guc_load_ucode (dev, false); +if (ret == -EAGAIN ) +return 0 ;/* too early */ + ret =i915_gem_context_enable (dev_priv); -if (ret && ret != -EIO ) { +if (ret ==0 ) { + dev_priv->contexts_ready = true; + }else if (ret && ret != -EIO ) {DRM_ERROR ("Context enable failed %d\n" , ret);i915_gem_cleanup_ringbuffer (dev);drivers/gpu/drm/i915/i915_gem_context.c +30
-5@@ -445,23 +445,48 @@ static int context_idr_cleanup(int id, void *p, void *data)return 0 ; } +/* Complete any late initialisation here */ +static int i915_gem_context_first_open (struct drm_device *dev) +{ +struct drm_i915_private *dev_priv = dev->dev_private; +int ret; + +/* We can't enable contexts until all firmware is loaded */ + ret =intel_guc_load_ucode (dev, true); +WARN_ON (ret == -EAGAIN ); + + ret =i915_gem_context_enable (dev_priv); +if (ret ==0 ) + dev_priv->contexts_ready = true; +return ret; +} +int i915_gem_context_open (struct drm_device *dev,struct drm_file *file) { +struct drm_i915_private *dev_priv = dev->dev_private;struct drm_i915_file_private *file_priv = file->driver_priv;struct intel_context *ctx; +int ret =0 ;idr_init (&file_priv->context_idr);mutex_lock (&dev->struct_mutex); - ctx =i915_gem_create_context (dev, file_priv); + +if (!dev_priv->contexts_ready) + ret =i915_gem_context_first_open (dev); + +if (ret ==0 ) { + ctx =i915_gem_create_context (dev, file_priv); +if (IS_ERR (ctx)) + ret =PTR_ERR (ctx); + } +mutex_unlock (&dev->struct_mutex); -if (IS_ERR (ctx)) { +if (ret)idr_destroy (&file_priv->context_idr); -return PTR_ERR (ctx); - } -return 0 ; +return ret; }void i915_gem_context_close (struct drm_device *dev,struct drm_file *file)drivers/gpu/drm/i915/intel_guc.h +1
-1@@ -89,7 +89,7 @@ struct intel_guc { GUC_ENABLE_READ_CACHE_FOR_WOPCM_DATA)/* intel_guc_loader.c */ -extern int intel_guc_load_ucode(struct drm_device *dev); +extern int intel_guc_load_ucode(struct drm_device *dev, bool wait);extern void intel_guc_ucode_fini (struct drm_device *dev);extern void intel_guc_ucode_init (struct drm_device *dev);drivers/gpu/drm/i915/intel_guc_loader.c +19
-3@@ -217,25 +217,41 @@ out: }/* -* Called from gem_init_hw() during driver loading and also after a GPU reset. * Check that the firmware fetching process has succeeded, and if so transfer * the loaded image to the hardware. +* +* However, there are a few checks to do first. The very first call should have +* (wait == FALSE), but the fetch_state will still be PENDING as the firmware may +* not be available that early. Therefore, on this first call, we just return. +* +* The second call should come from the first open of the device (wait == TRUE). +* This is a good time to load the firmware into the device, as by this point it +* must be available. +* +* Any subsequent calls are expected to have wait == FALSE, and indicate that the +* hardware has been reset and so the firmware should be reloaded. */ -int intel_guc_load_ucode (struct drm_device *dev) +int intel_guc_load_ucode (struct drm_device *dev,bool wait) {struct drm_i915_private *dev_priv = dev->dev_private;struct intel_uc_fw *guc_fw = &dev_priv->guc.guc_fw;int err; +DRM_DEBUG_DRIVER ("GuC: wait %d, fetch status %d, load status %d\n" , + wait, guc_fw->uc_fw_fetch_status, guc_fw->uc_fw_load_status); + +if (guc_fw->uc_fw_fetch_status ==INTEL_UC_FIRMWARE_PENDING && !wait) +return -EAGAIN ; + guc_fw->uc_fw_load_status =INTEL_UC_FIRMWARE_NONE ;if (guc_fw->uc_fw_fetch_status ==INTEL_UC_FIRMWARE_NONE )return 0 ; - guc_fw->uc_fw_load_status =INTEL_UC_FIRMWARE_PENDING ; err =intel_uc_fw_check (dev, guc_fw);if (err) goto fail; + guc_fw->uc_fw_load_status =INTEL_UC_FIRMWARE_PENDING ; err =guc_load_ucode (dev);if (err) goto fail;drivers/gpu/drm/i915/intel_uc_loader.c +10
-2@@ -49,8 +49,16 @@ static void uc_fw_finish(struct drm_device *dev, struct intel_uc_fw *uc_fw) uc_fw->uc_name, uc_fw->uc_fw_fetch_status, uc_fw->uc_fw_blob); fw = uc_fw->uc_fw_blob; -if (!fw) - goto fail; +if (!fw) { +/* no firmware found; try again in case FS was not mounted */ +DRM_DEBUG_DRIVER ("retry fetching %s fw from <%s>\n" , + uc_fw->uc_name, uc_fw->uc_fw_path); +if (request_firmware (&fw, uc_fw->uc_fw_path, &dev->pdev->dev)) + goto fail; +DRM_DEBUG_DRIVER ("fetch %s fw from <%s> succeeded, fw %p\n" , + uc_fw->uc_name, uc_fw->uc_fw_path, fw); + uc_fw->uc_fw_blob = fw; + }if (uc_fw->uc_fw_check && !uc_fw->uc_fw_check (uc_fw)) goto fail;Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/18] drm/i915: Defer default hardware context initialisation until first open
Quoted 16 lines · Daniel Vetter
Quoted 8 lines · On Thu, Mar 26, 2015 at 12:41:13PM -0700, [email protected] wrote:
From: Dave Gordon <[email protected]>
In order to fully initialise the default contexts, we have to execute batchbuffer commands on the GPU engines. But we can't do that until any required firmware has been loaded, which may not be possible during driver load, because the filesystem(s) containing the firmware may not be mounted until later.
Quoted 6 lines
Therefore, we now allow the first call to the firmware-loading code to return
-EAGAINto indicate that it's not yet ready, and that it should be retried when the device is first opened from user code, by which time we expect that all required filesystems will have been mounted. The late-retry code will then re-attempt to load the firmware if the early attempt failed.
Quoted 3 lines
We've tried a similar approach a while back and it doesn't work well in conjunction with rps - the hw tends to fall over if the context state isn't properly initialized when going into rc6.
I believe patch 18/18 of this series is to notify GuC when RC6 is on and off. I don't know much details of implementation inside firmware, but I believe GuC will take care of it properly.
Quoted 2 lines
Why exactly can't we load that firmware right at boot-up, or at least stall correctly until it's there?
Dave G. wrote a very good comment about this. Sorry I lost it during patch squashing. Here is a copy of it. I will amend the comment in next version.
The GuC loader uses an asynchronous thread to fetch the firmware image (aka "binary blob") from a file. This thread has then had to wait for the mainline driver loading code to complete GEM initialisation before it can convert the blob into a GEM object and transfer it to the GuC's memory.
Unfortunately, with this scheme, the GuC loading was occurring*after* the internally-generated batches used to initialise contexts had already been submitted (using direct access to the ELSP, since the GuC wasn't ready).
In addition, the one-way synchronisation mechanism resulted in the firmware image being transfeered to the GuC at an indeterminate time (just "sometime" after the mainline thread releases the device's struct_mutex), with consequent confusion.
Rather than complicate the loader further by adding a second sync point and arranging the handover of the struct_mutex from one thread to the other, this commit reverses the synchronisation so that the mainline thread waits for the asynchronous thread rather than vice versa. The firmware loader now only has to save the reference to the blob and signal a completion; the mainline thread can then continue with the rest of the loading process when it catches up. The result is a much simpler (and fully deterministic) process for loading the GuC firmware.
-Alex
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/18] drm/i915: Defer default hardware context initialisation until first open
Quoted 46 lines · Yu Dai
Quoted 16 lines · Daniel Vetter
Quoted 8 lines · On Thu, Mar 26, 2015 at 12:41:13PM -0700, [email protected] wrote:
From: Dave Gordon <[email protected]>
In order to fully initialise the default contexts, we have to execute batchbuffer commands on the GPU engines. But we can't do that until any required firmware has been loaded, which may not be possible during driver load, because the filesystem(s) containing the firmware may not be mounted until later.
Quoted 6 lines
Therefore, we now allow the first call to the firmware-loading code to return
-EAGAINto indicate that it's not yet ready, and that it should be retried when the device is first opened from user code, by which time we expect that all required filesystems will have been mounted. The late-retry code will then re-attempt to load the firmware if the early attempt failed.Quoted 3 lines
We've tried a similar approach a while back and it doesn't work well in conjunction with rps - the hw tends to fall over if the context state isn't properly initialized when going into rc6.
I believe patch 18/18 of this series is to notify GuC when RC6 is on and off. I don't know much details of implementation inside firmware, but I believe GuC will take care of it properly.
Quoted 2 lines
Why exactly can't we load that firmware right at boot-up, or at least stall correctly until it's there?
Dave G. wrote a very good comment about this. Sorry I lost it during patch squashing. Here is a copy of it. I will amend the comment in next version.
The GuC loader uses an asynchronous thread to fetch the firmware image (aka "binary blob") from a file. This thread has then had to wait for the mainline driver loading code to complete GEM initialisation before it can convert the blob into a GEM object and transfer it to the GuC's memory.
Unfortunately, with this scheme, the GuC loading was occurring*after* the internally-generated batches used to initialise contexts had already been submitted (using direct access to the ELSP, since the GuC wasn't ready).
In addition, the one-way synchronisation mechanism resulted in the firmware image being transfeered to the GuC at an indeterminate time (just "sometime" after the mainline thread releases the device's
struct_mutex), with consequent confusion.
Quoted 8 lines
Rather than complicate the loader further by adding a second sync point and arranging the handover of the
struct_mutexfrom one thread to the other, this commit reverses the synchronisation so that the mainline thread waits for the asynchronous thread rather than vice versa. The firmware loader now only has to save the reference to the blob and signal a completion; the mainline thread can then continue with the rest of the loading process when it catches up. The result is a much simpler (and fully deterministic) process for loading the GuC firmware.
Still sounds rather fragile. Chris suggestion to just not submit anything until guc firmware is loaded and then kick of a submission of all outstanding requests sounds much better. That would naturally insert minimal amounts of sync points everywhere you need them using the normal GEM execution tracking.
The only other change we'd need (I think at least, didn't check details) is to wait for the hw context init asynchronously in the rps worker. -Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 06/18] drm/i915: Defer default hardware context initialisation until first open
In reply to Daniel VetterQuoted 16 lines · Daniel Vetter
Quoted 8 lines · On Thu, Mar 26, 2015 at 12:41:13PM -0700, [email protected] wrote:
From: Dave Gordon <[email protected]>
In order to fully initialise the default contexts, we have to execute batchbuffer commands on the GPU engines. But we can't do that until any required firmware has been loaded, which may not be possible during driver load, because the filesystem(s) containing the firmware may not be mounted until later.
Quoted 6 lines
Therefore, we now allow the first call to the firmware-loading code to return
-EAGAINto indicate that it's not yet ready, and that it should be retried when the device is first opened from user code, by which time we expect that all required filesystems will have been mounted. The late-retry code will then re-attempt to load the firmware if the early attempt failed.
Quoted 6 lines
We've tried a similar approach a while back and it doesn't work well in conjunction with rps - the hw tends to fall over if the context state isn't properly initialized when going into rc6.
Why exactly can't we load that firmware right at boot-up, or at least stall correctly until it's there?
Note that is is very simple to build up a request queue prior to enabling the submission port. Without having to touch core GEM code at all.
-Chris
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 7/18 drm/i915: Move execlists defines from .c to .h
In reply to [email protected]From: "Michael H. Nguyen" <[email protected]>
Move defines from intel_lrc.c to i915_reg.h so they are accessible by the guc files
Issue: VIZ-4884
--- drivers/gpu/drm/i915/i915_reg.h | 68 ++++++++++++++++++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_lrc.c | 67 --------------------------------------- 2 files changed, 68 insertions(+), 67 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +68 -0
@@ -7323,4 +7323,72 @@ enum skl_disp_power_wells {#define _PALETTE_A (dev_priv->info.display_mmio_offset + 0xa000)#define _PALETTE_B (dev_priv->info.display_mmio_offset + 0xa800) +/* Exec Lists */ +#define GEN9_LR_CONTEXT_RENDER_SIZE (22 * PAGE_SIZE) +#define GEN8_LR_CONTEXT_RENDER_SIZE (20 * PAGE_SIZE) +#define GEN8_LR_CONTEXT_OTHER_SIZE (2 * PAGE_SIZE) + +#define RING_EXECLIST_QFULL (1 << 0x2) +#define RING_EXECLIST1_VALID (1 << 0x3) +#define RING_EXECLIST0_VALID (1 << 0x4) +#define RING_EXECLIST_ACTIVE_STATUS (3 << 0xE) +#define RING_EXECLIST1_ACTIVE (1 << 0x11) +#define RING_EXECLIST0_ACTIVE (1 << 0x12) + +#define GEN8_CTX_STATUS_IDLE_ACTIVE (1 << 0) +#define GEN8_CTX_STATUS_PREEMPTED (1 << 1) +#define GEN8_CTX_STATUS_ELEMENT_SWITCH (1 << 2) +#define GEN8_CTX_STATUS_ACTIVE_IDLE (1 << 3) +#define GEN8_CTX_STATUS_COMPLETE (1 << 4) +#define GEN8_CTX_STATUS_LITE_RESTORE (1 << 15) + +#define CTX_LRI_HEADER_0 0x01 +#define CTX_CONTEXT_CONTROL 0x02 +#define CTX_RING_HEAD 0x04 +#define CTX_RING_TAIL 0x06 +#define CTX_RING_BUFFER_START 0x08 +#define CTX_RING_BUFFER_CONTROL 0x0a +#define CTX_BB_HEAD_U 0x0c +#define CTX_BB_HEAD_L 0x0e +#define CTX_BB_STATE 0x10 +#define CTX_SECOND_BB_HEAD_U 0x12 +#define CTX_SECOND_BB_HEAD_L 0x14 +#define CTX_SECOND_BB_STATE 0x16 +#define CTX_BB_PER_CTX_PTR 0x18 +#define CTX_RCS_INDIRECT_CTX 0x1a +#define CTX_RCS_INDIRECT_CTX_OFFSET 0x1c +#define CTX_LRI_HEADER_1 0x21 +#define CTX_CTX_TIMESTAMP 0x22 +#define CTX_PDP3_UDW 0x24 +#define CTX_PDP3_LDW 0x26 +#define CTX_PDP2_UDW 0x28 +#define CTX_PDP2_LDW 0x2a +#define CTX_PDP1_UDW 0x2c +#define CTX_PDP1_LDW 0x2e +#define CTX_PDP0_UDW 0x30 +#define CTX_PDP0_LDW 0x32 +#define CTX_LRI_HEADER_2 0x41 +#define CTX_R_PWR_CLK_STATE 0x42 +#define CTX_GPGPU_CSR_BASE_ADDRESS 0x44 + +#define GEN8_CTX_VALID (1<<0) +#define GEN8_CTX_FORCE_PD_RESTORE (1<<1) +#define GEN8_CTX_FORCE_RESTORE (1<<2) +#define GEN8_CTX_L3LLC_COHERENT (1<<5) +#define GEN8_CTX_PRIVILEGE (1<<8) +enum { +ADVANCED_CONTEXT =0 , +LEGACY_CONTEXT , +ADVANCED_AD_CONTEXT , +LEGACY_64B_CONTEXT +}; +#define GEN8_CTX_MODE_SHIFT 3 +enum { +FAULT_AND_HANG =0 , +FAULT_AND_HALT ,/* Debug only */ +FAULT_AND_STREAM , +FAULT_AND_CONTINUE /* Unsupported */ +}; +#define GEN8_CTX_ID_SHIFT 32 +#endif /* _I915_REG_H_ */
drivers/gpu/drm/i915/intel_lrc.c +0 -67
73 lines not on this page · raw message
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 8/18 drm/i915: Make several execlist helper functions external
In reply to [email protected]From: Alex Dai <[email protected]>
These functions will be used by GuC scheduler.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/intel_lrc.c | 6 +++--- drivers/gpu/drm/i915/intel_lrc.h | 3 +++ 2 files changed, 6 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +3 -3
@@ -187,8 +187,8 @@ u32 intel_execlists_ctx_id(struct drm_i915_gem_object *ctx_obj)return lrca >>12 ; } -static uint64_t execlists_ctx_descriptor (struct intel_engine_cs *ring, -struct drm_i915_gem_object *ctx_obj) +uint64_t execlists_ctx_descriptor (struct intel_engine_cs *ring, +struct drm_i915_gem_object *ctx_obj) {struct drm_device *dev = ring->dev;uint64_t desc;
@@ -1761,7 +1761,7 @@ void intel_lr_context_free(struct intel_context *ctx) } } -static uint32_t get_lr_context_size (struct intel_engine_cs *ring) +uint32_t get_lr_context_size (struct intel_engine_cs *ring) {int ret =0 ;
drivers/gpu/drm/i915/intel_lrc.h +3 -0
@@ -68,6 +68,7 @@ int intel_logical_ring_begin(struct intel_ringbuffer *ringbuf, int num_dwords);/* Logical Ring Contexts */ +uint32_t get_lr_context_size (struct intel_engine_cs *ring);void intel_lr_context_free (struct intel_context *ctx);int intel_lr_context_deferred_create (struct intel_context *ctx,struct intel_engine_cs *ring);
@@ -86,6 +87,8 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,struct drm_i915_gem_object *batch_obj, u64 exec_start, u32 dispatch_flags);u32 intel_execlists_ctx_id (struct drm_i915_gem_object *ctx_obj); +uint64_t execlists_ctx_descriptor (struct intel_engine_cs *ring, +struct drm_i915_gem_object *ctx_obj);void intel_lrc_irq_handler (struct intel_engine_cs *ring);void intel_execlists_retire_requests (struct intel_engine_cs *ring);
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 9/18 drm/i915: Add functions to allocate / release gem obj for GuC
In reply to [email protected]From: Alex Dai <[email protected]>
All gem objects used by GuC are pinned to ggtt space out of range [0, WOPCM size]. In GuC address space mapping, [0, WPOCM size] is used internally for its Boot ROM, SRAM etc. Currently this WPOCM size is 512K. This is done by using of PIN_OFFSET_BIAS.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/intel_guc.h | 3 +++ drivers/gpu/drm/i915/intel_guc_loader.c | 34 +++++++++++++++++++++++++++++++++ 2 files changed, 37 insertions(+)
drivers/gpu/drm/i915/intel_guc.h +3 -0
@@ -92,5 +92,8 @@ struct intel_guc {extern int intel_guc_load_ucode (struct drm_device *dev,bool wait);extern void intel_guc_ucode_fini (struct drm_device *dev);extern void intel_guc_ucode_init (struct drm_device *dev); +struct drm_i915_gem_object * +intel_guc_allocate_gem_obj (struct drm_device *dev,u32 size); +void intel_guc_release_gem_obj (struct drm_i915_gem_object *obj);#endif
drivers/gpu/drm/i915/intel_guc_loader.c +34 -0
@@ -30,6 +30,40 @@MODULE_FIRMWARE (I915_GUC_UCODE_GEN8 );MODULE_FIRMWARE (I915_GUC_UCODE_GEN9 ); +struct drm_i915_gem_object * +intel_guc_allocate_gem_obj (struct drm_device *dev,u32 size) +{ +struct drm_i915_gem_object *obj; + + obj =i915_gem_alloc_object (dev, size); +if (!obj) +return NULL ; + +if (i915_gem_object_get_pages (obj)) { +drm_gem_object_unreference (&obj->base); +return NULL ; + } + +if (i915_gem_obj_ggtt_pin (obj,PAGE_SIZE , +PIN_OFFSET_BIAS |GUC_WOPCM_SIZE_VALUE )) { +drm_gem_object_unreference (&obj->base); +return NULL ; + } + +return obj; +} + +void intel_guc_release_gem_obj (struct drm_i915_gem_object *obj) +{ +if (!obj) +return ; + +if (i915_gem_obj_is_pinned (obj)) +i915_gem_object_ggtt_unpin (obj); + +drm_gem_object_unreference (&obj->base); +} +static int copy_rsa (struct drm_i915_private *dev_priv) {struct sg_table *st = dev_priv->guc.guc_fw.uc_fw_obj->pages;
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 09/18] drm/i915: Add functions to allocate / release gem obj for GuC
Quoted 41 lines · On Thu, Mar 26, 2015 at 12:41:16PM -0700, [email protected] wrote:
From: Alex Dai <[email protected]>
All gem objects used by GuC are pinned to ggtt space out of range [0, WOPCM size]. In GuC address space mapping, [0, WPOCM size] is used internally for its Boot ROM, SRAM etc. Currently this WPOCM size is 512K. This is done by using of
PIN_OFFSET_BIAS.Issue: VIZ-4884
Signed-off-by Alex Dai <[email protected]>--- drivers/gpu/drm/i915/intel_guc.h | 3 +++ drivers/gpu/drm/i915/intel_guc_loader.c | 34 +++++++++++++++++++++++++++++++++ 2 files changed, 37 insertions(+)drivers/gpu/drm/i915/intel_guc.h +3
-0@@ -92,5 +92,8 @@ struct intel_guc {extern int intel_guc_load_ucode (struct drm_device *dev,bool wait);extern void intel_guc_ucode_fini (struct drm_device *dev);extern void intel_guc_ucode_init (struct drm_device *dev); +struct drm_i915_gem_object * +intel_guc_allocate_gem_obj (struct drm_device *dev,u32 size); +void intel_guc_release_gem_obj (struct drm_i915_gem_object *obj);#endif drivers/gpu/drm/i915/intel_guc_loader.c +6
-0@@ -30,6 +30,40 @@MODULE_FIRMWARE (I915_GUC_UCODE_GEN8 );MODULE_FIRMWARE (I915_GUC_UCODE_GEN9 ); +struct drm_i915_gem_object * +intel_guc_allocate_gem_obj (struct drm_device *dev,u32 size) +{ +struct drm_i915_gem_object *obj; + + obj =i915_gem_alloc_object (dev, size);
drivers/gpu/drm/i915/intel_guc_loader.c line 48
Quoted 10 lines
drivers/gpu/drm/i915/intel_guc_loader.c +10
-0+if (!obj) +return NULL ; + +if (i915_gem_object_get_pages (obj)) { +drm_gem_object_unreference (&obj->base); +return NULL ; + } + +if (i915_gem_obj_ggtt_pin (obj,PAGE_SIZE , +PIN_OFFSET_BIAS |GUC_WOPCM_SIZE_VALUE )) {
Please add a comment here about how much pin bias you exactly need and why. Otherwise this will be extremely arcane knowledge lost when someone reworks the pin bias stuff or changes the default offset.
Or well just use the right offset directly, since atm PIN_OFFSET_BIAS is 256k and you claim to need 512k.
-Daniel
Quoted 28 lines
drivers/gpu/drm/i915/intel_guc_loader.c +18
-0+drm_gem_object_unreference (&obj->base); +return NULL ; + } + +return obj; +} + +void intel_guc_release_gem_obj (struct drm_i915_gem_object *obj) +{ +if (!obj) +return ; + +if (i915_gem_obj_is_pinned (obj)) +i915_gem_object_ggtt_unpin (obj); + +drm_gem_object_unreference (&obj->base); +} +static int copy_rsa (struct drm_i915_private *dev_priv) {struct sg_table *st = dev_priv->guc.guc_fw.uc_fw_obj->pages;Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 09/18] drm/i915: Add functions to allocate / release gem obj for GuC
Quoted 57 lines · Daniel Vetter
Quoted 41 lines · On Thu, Mar 26, 2015 at 12:41:16PM -0700, [email protected] wrote:
From: Alex Dai <[email protected]>
All gem objects used by GuC are pinned to ggtt space out of range [0, WOPCM size]. In GuC address space mapping, [0, WPOCM size] is used internally for its Boot ROM, SRAM etc. Currently this WPOCM size is 512K. This is done by using of
PIN_OFFSET_BIAS.Issue: VIZ-4884
Signed-off-by Alex Dai <[email protected]>--- drivers/gpu/drm/i915/intel_guc.h | 3 +++ drivers/gpu/drm/i915/intel_guc_loader.c | 34 +++++++++++++++++++++++++++++++++ 2 files changed, 37 insertions(+)drivers/gpu/drm/i915/intel_guc.h +3
-0@@ -92,5 +92,8 @@ struct intel_guc {extern int intel_guc_load_ucode (struct drm_device *dev,bool wait);extern void intel_guc_ucode_fini (struct drm_device *dev);extern void intel_guc_ucode_init (struct drm_device *dev); +struct drm_i915_gem_object * +intel_guc_allocate_gem_obj (struct drm_device *dev,u32 size); +void intel_guc_release_gem_obj (struct drm_i915_gem_object *obj);#endif drivers/gpu/drm/i915/intel_guc_loader.c +6
-0@@ -30,6 +30,40 @@MODULE_FIRMWARE (I915_GUC_UCODE_GEN8 );MODULE_FIRMWARE (I915_GUC_UCODE_GEN9 ); +struct drm_i915_gem_object * +intel_guc_allocate_gem_obj (struct drm_device *dev,u32 size) +{ +struct drm_i915_gem_object *obj; + + obj =i915_gem_alloc_object (dev, size);drivers/gpu/drm/i915/intel_guc_loader.c line 48
Quoted 10 lines
drivers/gpu/drm/i915/intel_guc_loader.c +10
-0+if (!obj) +return NULL ; + +if (i915_gem_object_get_pages (obj)) { +drm_gem_object_unreference (&obj->base); +return NULL ; + } + +if (i915_gem_obj_ggtt_pin (obj,PAGE_SIZE , +PIN_OFFSET_BIAS |GUC_WOPCM_SIZE_VALUE )) {Please add a comment here about how much pin bias you exactly need and why. Otherwise this will be extremely arcane knowledge lost when someone reworks the pin bias stuff or changes the default offset.
Quoted 2 lines
Or well just use the right offset directly, since atm
PIN_OFFSET_BIASis 256k and you claim to need 512k.
Strike that all, I'm blind and your code is all fine.
Sorry for the mess.
-Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 10/18 drm/i915: Functions to support command submission via GuC
In reply to [email protected]From: Alex Dai <[email protected]>
To enable GuC command submission / scheduling, we need to setup firmware initializaion properly. i915.enable_guc_scheduling is introduced to enable / disable GuC submission.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/Makefile | 3 +- drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/i915_params.c | 4 ++ drivers/gpu/drm/i915/intel_guc.h | 10 ++++ drivers/gpu/drm/i915/intel_guc_loader.c | 49 +++++++++++++++++ drivers/gpu/drm/i915/intel_guc_scheduler.c | 87 ++++++++++++++++++++++++++++++ 6 files changed, 153 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/i915/intel_guc_scheduler.c
drivers/gpu/drm/i915/Makefile +2 -1
@@ -37,7 +37,8 @@ i915-y += i915_cmd_parser.o \ i915_trace_points.o \ intel_lrc.o \ intel_ringbuffer.o \ - intel_uncore.o + intel_uncore.o \ + intel_guc_scheduler.o # ancilliary microcontroller support i915-y += intel_uc_loader.o \
drivers/gpu/drm/i915/i915_drv.h +1 -0
@@ -2471,6 +2471,7 @@ struct i915_params {bool reset;bool disable_display;bool disable_vtd_wa; +bool enable_guc_scheduling;int use_mmio_flip;int mmio_debug;bool verbose_state_checks;
drivers/gpu/drm/i915/i915_params.c +4 -0
@@ -52,6 +52,7 @@ struct i915_params i915 __read_mostly = { .mmio_debug = 0, .verbose_state_checks = 1, .nuclear_pageflip = 0, + .enable_guc_scheduling = false, };module_param_named (modeset, i915.modeset, int,0400 );
@@ -178,3 +179,6 @@ MODULE_PARM_DESC(verbose_state_checks,module_param_named_unsafe (nuclear_pageflip, i915.nuclear_pageflip, bool,0600 );MODULE_PARM_DESC (nuclear_pageflip,"Force atomic modeset functionality; only planes work for now (default: false)." ); + +module_param_named (enable_guc_scheduling, i915.enable_guc_scheduling, bool,0400 ); +MODULE_PARM_DESC (enable_guc_scheduling,"Enable GuC scheduling (default:false)" );
drivers/gpu/drm/i915/intel_guc.h +10 -0
@@ -29,6 +29,9 @@struct intel_guc {struct intel_uc_fw guc_fw; + +/* GuC-specific additions */ +struct drm_i915_gem_object *ctx_pool_obj; };#define GUC_STATUS 0xc000
@@ -96,4 +99,11 @@ struct drm_i915_gem_object * intel_guc_allocate_gem_obj(struct drm_device *dev, u32 size);void intel_guc_release_gem_obj (struct drm_i915_gem_object *obj); +/* intel_guc_scheduler.c */ +int guc_scheduler_init (struct drm_device *dev); +void guc_scheduler_fini (struct drm_device *dev); +int guc_scheduler_enable (struct drm_device *dev); +void guc_scheduler_disable (struct drm_device *dev); +bool sanitize_enable_guc_scheduling (struct drm_device *dev); +#endif
drivers/gpu/drm/i915/intel_guc_loader.c +49 -0
81 lines not on this page · raw message
drivers/gpu/drm/i915/intel_guc_scheduler.c new file +87 -0
87 lines not on this page · raw message
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 11/18 drm/i915: Integration of GuC client
In reply to [email protected]From: Alex Dai <[email protected]>
Implementation of GuC client. A GuC client has its own doorbell and workqueue. It maintains the doorbell cache line, process description object and work queue item.
A default guc_client is created to do the in-order legacy execlist submission.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/Makefile | 3 +- drivers/gpu/drm/i915/i915_drv.h | 5 + drivers/gpu/drm/i915/intel_guc.h | 52 ++++ drivers/gpu/drm/i915/intel_guc_client.c | 399 +++++++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_guc_loader.c | 14 +- drivers/gpu/drm/i915/intel_guc_scheduler.c | 29 ++- 6 files changed, 492 insertions(+), 10 deletions(-) create mode 100644 drivers/gpu/drm/i915/intel_guc_client.c
drivers/gpu/drm/i915/Makefile +2 -1
@@ -42,7 +42,8 @@ i915-y += i915_cmd_parser.o \ # ancilliary microcontroller support i915-y += intel_uc_loader.o \ - intel_guc_loader.o + intel_guc_loader.o \ + intel_guc_client.o # autogenerated null render state i915-y += intel_renderstate_gen6.o \
drivers/gpu/drm/i915/i915_drv.h +5 -0
@@ -1847,6 +1847,11 @@ static inline struct drm_i915_private *dev_to_i915(struct device *dev)return to_i915 (dev_get_drvdata (dev)); } +static inline struct drm_i915_private *guc_to_i915 (struct intel_guc *guc) +{ + return container_of(guc, struct drm_i915_private, guc); +} +/* Iterate over initialised rings */ #define for_each_ring (ring__, dev_priv__, i__) \ for ((i__) = 0; (i__) < I915_NUM_RINGS; (i__)++) \
drivers/gpu/drm/i915/intel_guc.h +52 -0
73 lines not on this page · raw message
drivers/gpu/drm/i915/intel_guc_client.c new file +399 -0
399 lines not on this page · raw message
drivers/gpu/drm/i915/intel_guc_loader.c +7 -7
@@ -233,12 +233,19 @@ static int guc_load_ucode(struct drm_device *dev)/* Set MMIO/WA for GuC init */ +I915_WRITE (DRBMISC1 ,DOORBELL_ENABLE ); +/* Enable MIA caching. GuC clock gating is disabled. */ I915_WRITE (GUC_SHIM_CONTROL ,GUC_SHIM_CONTROL_VALUE );/* WaC6DisallowByGfxPause*/ I915_WRITE (GEN6_GFXPAUSE ,0x30FFF ); +if (IS_SKYLAKE (dev)) +I915_WRITE (GEN9_GT_PM_CONFIG ,GEN8_GT_DOORBELL_ENABLE ); +else +I915_WRITE (GEN8_GT_PM_CONFIG ,GEN8_GT_DOORBELL_ENABLE ); +if (IS_GEN9 (dev)) {/* DOP Clock Gating Enable for GuC clocks */ I915_WRITE(GEN7_MISCCPCTL, (GEN8_DOP_CLOCK_GATE_GUC_ENABLE |
@@ -255,13 +262,6 @@ static int guc_load_ucode(struct drm_device *dev)set_guc_init_params (dev_priv); -/* FIXME: !UPSTREAM - I don't have real keys, so we need to disable the -* authentication. This can only work if the part is fused in a special -* configuration. Therefore, even if it leaked externally, it won't be -* detrimental to security -*/ -I915_WRITE (0xc068 ,0x3 ); - ret =ucode_dma_xfer_sync (dev_priv);/* We can free the object pages now, and we would, except we might as
drivers/gpu/drm/i915/intel_guc_scheduler.c +27 -2
@@ -36,6 +36,8 @@ void guc_scheduler_fini(struct drm_device *dev)if (ctx_pool) {intel_guc_release_gem_obj (ctx_pool); guc->ctx_pool_obj =NULL ; + +ida_destroy (&guc->ctx_ids); } }
@@ -60,21 +62,44 @@ int guc_scheduler_init(struct drm_device *dev)if (!guc->ctx_pool_obj)return -ENOMEM ; +spin_lock_init (&dev_priv->guc.host2guc_lock); + +ida_init (&guc->ctx_ids); + +memset (guc->doorbell_bitmap,0 ,sizeof (guc->doorbell_bitmap)); + guc->db_cacheline =0 ; +return 0 ; }int guc_scheduler_enable (struct drm_device *dev) { +struct drm_i915_private *dev_priv = dev->dev_private; +struct intel_guc *guc = &dev_priv->guc; +if (!i915.enable_guc_scheduling)return 0 ; -/* TODO: placeholder for guc scheduler enabling */ +/* client for execbuf submission */ + guc->execbuf_client = +i915_guc_client_alloc (dev,GUC_CTX_PRIORITY_NORMAL ); +if (!guc->execbuf_client) { +DRM_ERROR ("Failed to create execbuf guc_client\n" ); +return -ENOMEM ; + } +return 0 ; }void guc_scheduler_disable (struct drm_device *dev) { -/* TODO: placeholder for guc scheduler disabling */ +struct drm_i915_private *dev_priv = dev->dev_private; +struct intel_guc *guc = &dev_priv->guc; + +if (guc->execbuf_client) { +i915_guc_client_free (dev, guc->execbuf_client); + guc->execbuf_client =NULL ; + } }bool sanitize_enable_guc_scheduling (struct drm_device *dev)
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 12/18 drm/i915: Interrupt routing for GuC scheduler
In reply to [email protected]From: Alex Dai <[email protected]>
Turn on interrupt steering to route necessary interrupts to GuC.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/i915_reg.h | 11 ++++-- drivers/gpu/drm/i915/intel_guc.h | 7 ++++ drivers/gpu/drm/i915/intel_guc_scheduler.c | 55 +++++++++++++++++++++++++++++- 3 files changed, 70 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +9 -2
@@ -1372,12 +1372,18 @@ enum skl_disp_power_wells {#define GFX_MODE_GEN7 0x0229c#define RING_MODE_GEN7 (ring) ((ring)->mmio_base+0x29c)#define GFX_RUN_LIST_ENABLE (1<<15) +#define GFX_INTERRUPT_STEERING (1<<14)#define GFX_TLB_INVALIDATE_EXPLICIT (1<<13)#define GFX_SURFACE_FAULT_ENABLE (1<<12)#define GFX_REPLAY_MODE (1<<11)#define GFX_PSMI_GRANULARITY (1<<10)#define GFX_PPGTT_ENABLE (1<<9) +#define GFX_FORWARD_VBLANK_MASK (3<<5) +#define GFX_FORWARD_VBLANK_NEVER (0<<5) +#define GFX_FORWARD_VBLANK_ALWAYS (1<<5) +#define GFX_FORWARD_VBLANK_COND (2<<5) +#define VLV_DISPLAY_BASE 0x180000#define VLV_MIPI_BASE VLV_DISPLAY_BASE
@@ -5196,11 +5202,12 @@ enum skl_disp_power_wells {#define GEN8_GT_IIR (which) (0x44308 + (0x10 * (which)))#define GEN8_GT_IER (which) (0x4430c + (0x10 * (which))) -#define GEN8_BCS_IRQ_SHIFT 16#define GEN8_RCS_IRQ_SHIFT 0 -#define GEN8_VCS2_IRQ_SHIFT 16 +#define GEN8_BCS_IRQ_SHIFT 16#define GEN8_VCS1_IRQ_SHIFT 0 +#define GEN8_VCS2_IRQ_SHIFT 16#define GEN8_VECS_IRQ_SHIFT 0 +#define GEN8_WD_IRQ_SHIFT 16#define GEN8_DE_PIPE_ISR (pipe) (0x44400 + (0x10 * (pipe)))#define GEN8_DE_PIPE_IMR (pipe) (0x44404 + (0x10 * (pipe)))
drivers/gpu/drm/i915/intel_guc.h +7 -0
@@ -134,6 +134,13 @@ struct intel_guc {#define GEN8_GTCR 0x4274#define GEN8_GTCR_INVALIDATE (1<<0) +#define DE_GUCRMR 0x44054 + +#define GUC_BCS_RCS_IER 0xC550 +#define GUC_VCS2_VCS1_IER 0xC554 +#define GUC_WD_VECS_IER 0xC558 +#define GUC_PM_P24C_IER 0xC55C +/* intel_guc_loader.c */ extern int intel_guc_load_ucode (struct drm_device *dev,bool wait);extern void intel_guc_ucode_fini (struct drm_device *dev);
drivers/gpu/drm/i915/intel_guc_scheduler.c +54 -1
80 lines not on this page · raw message
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 13/18 drm/i915: Enable commands submission via GuC
In reply to [email protected]From: Alex Dai <[email protected]>
Add functions to submit work queue item and ring the door bell. GuC TLB needs to be invalided if LRC context changes.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/intel_guc.h | 3 + drivers/gpu/drm/i915/intel_guc_client.c | 175 ++++++++++++++++++++++++++++- drivers/gpu/drm/i915/intel_guc_scheduler.c | 2 +- drivers/gpu/drm/i915/intel_lrc.c | 16 ++- 4 files changed, 193 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_guc.h +3 -0
@@ -31,6 +31,7 @@#define GUC_WQ_SIZE (PAGE_SIZE * 2)struct i915_guc_client { +spinlock_t wq_lock;struct drm_i915_gem_object *client_obj;u32 priority;off_t doorbell_offset;
@@ -39,6 +40,8 @@ struct i915_guc_client {uint16_t doorbell_id;uint32_t ctx_index;uint32_t wq_size; +uint32_t wq_tail; +uint32_t cookie; };#define I915_MAX_DOORBELLS 256
drivers/gpu/drm/i915/intel_guc_client.c +174 -1
195 lines not on this page · raw message
drivers/gpu/drm/i915/intel_guc_scheduler.c +1 -1
@@ -160,6 +160,6 @@ bool sanitize_enable_guc_scheduling(struct drm_device *dev)if (!HAS_GUC_UCODE (dev) || !HAS_GUC_SCHED (dev))return false; -return i915.enable_guc_scheduling; +return i915.enable_execlists && i915.enable_guc_scheduling; }
drivers/gpu/drm/i915/intel_lrc.c +15 -1
@@ -735,13 +735,17 @@ intel_logical_ring_advance_and_submit(struct intel_ringbuffer *ringbuf, struct drm_i915_gem_request *request) {struct intel_engine_cs *ring = ringbuf->ring; +struct drm_i915_private *dev_priv = ring->dev->dev_private;intel_logical_ring_advance (ringbuf);if (intel_ring_stopped (ring))return ; -execlists_context_queue (ring, ctx, ringbuf->tail, request); +if (dev_priv->guc.execbuf_client) +i915_guc_client_submit (dev_priv->guc.execbuf_client, ctx, ring); +else +execlists_context_queue (ring, ctx, ringbuf->tail, request); } static int intel_lr_context_pin(struct intel_engine_cs *ring,
@@ -749,6 +753,7 @@ static int intel_lr_context_pin(struct intel_engine_cs *ring, {struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state;struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf; +struct drm_i915_private *dev_priv = ring->dev->dev_private;int ret =0 ;WARN_ON (!mutex_is_locked (&ring->dev->struct_mutex));
@@ -761,6 +766,10 @@ static int intel_lr_context_pin(struct intel_engine_cs *ring, ret =intel_pin_and_map_ringbuffer_obj (ring->dev, ringbuf);if (ret) goto unpin_ctx_obj; + +/* Invalidate GuC TLB. */ +if (i915.enable_guc_scheduling) +I915_WRITE (GEN8_GTCR ,GEN8_GTCR_INVALIDATE ); }return ret;
@@ -1305,8 +1314,13 @@ out:static int gen8_init_rcs_context (struct intel_engine_cs *ring,struct intel_context *ctx) { +struct drm_i915_private *dev_priv = ring->dev->dev_private;int ret; +/* Invalidate GuC TLB. */ +if (i915.enable_guc_scheduling) +I915_WRITE (GEN8_GTCR ,GEN8_GTCR_INVALIDATE ); + ret =intel_logical_ring_workarounds_emit (ring, ctx);if (ret)return ret;
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 14/18 drm/i915: debugfs of GuC status
In reply to [email protected]From: Alex Dai <[email protected]>
Now print out Bootrom, uKernel and MIA Core status. The scratch reg 0 & 15 are used for communication between driver and firmware. Their status is also printed out.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/i915_debugfs.c | 73 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_guc.h | 13 ++++++ drivers/gpu/drm/i915/intel_guc_client.c | 33 +++++++++++---- 3 files changed, 112 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +73 -0
@@ -2239,6 +2239,77 @@ static int i915_llc(struct seq_file *m, void *data)return 0 ; } +static void i915_uc_load_status_info (struct seq_file *m,struct intel_uc_fw *uc_fw) +{ +seq_printf (m,"%s firmware status:\n\tpath: <%s>\n\tfetch: %d\n\tload: %d\n" , + uc_fw->uc_name, + uc_fw->uc_fw_path, + uc_fw->uc_fw_fetch_status, + uc_fw->uc_fw_load_status); +} + +static int i915_guc_load_status_info (struct seq_file *m,void *data) +{ +struct drm_info_node *node = m->private; +struct drm_i915_private *dev_priv = node->minor->dev->dev_private; +u32 tmp, i; + +i915_uc_load_status_info (m, &dev_priv->guc.guc_fw); + + tmp =I915_READ (GUC_STATUS ); + +seq_puts (m,"\nResponse from GuC:\n" ); +seq_printf (m,"\tBootrom status = 0x%x\n" , + (tmp &GS_BOOTROM_MASK ) >>GS_BOOTROM_SHIFT ); +seq_printf (m,"\tuKernel status = 0x%x\n" , + (tmp &GS_UKERNEL_MASK ) >>GS_UKERNEL_SHIFT ); +seq_printf (m,"\tMIA Core status = 0x%x\n" , + (tmp &GS_MIA_MASK ) >>GS_MIA_SHIFT ); +seq_puts (m,"Scratch registers value:\n" ); +for (i =0 ; i <16 ; i++) +seq_printf (m,"\t%2d: \t0x%x\n" , i,I915_READ (SOFT_SCRATCH (i))); + +return 0 ; +} + +static int i915_guc_info (struct seq_file *m,void *data) +{ +struct drm_info_node *node = m->private; +struct drm_device *dev = node->minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +struct intel_guc guc; +struct i915_guc_client client; + +memset (&client,0 ,sizeof (struct i915_guc_client )); + +/* Take a local copy of the GuC data, so we can dump it at leisure */ +spin_lock (&dev_priv->guc.host2guc_lock); + guc = dev_priv->guc; +if (guc.execbuf_client) { +spin_lock (&guc.execbuf_client->wq_lock); + client = *guc.execbuf_client; +spin_unlock (&guc.execbuf_client->wq_lock); + } +spin_unlock (&dev_priv->guc.host2guc_lock); + +seq_printf (m,"GuC total action count: %llu\n" , guc.action_count); +seq_printf (m,"GuC last action command: 0x%x\n" , guc.action_cmd); +seq_printf (m,"GuC last action status: 0x%x\n" , guc.action_status); + +seq_printf (m,"GuC action failure count: %u\n" , guc.action_fail); +seq_printf (m,"GuC last action error code: %d\n" , guc.action_err); + +seq_printf (m,"GuC execbuf client @ %p:\n" , guc.execbuf_client); +seq_printf (m,"\tTotal submissions: %llu\n" , client.submissions); +seq_printf (m,"\tFailed to queue: %u\n" , client.q_fail); +seq_printf (m,"\tFailed doorbell: %u\n" , client.b_fail); +seq_printf (m,"\tLast submission result: %d\n" , client.retcode); + +/* Add more as required ... */ + +return 0 ; +} +static int i915_edp_psr_status (struct seq_file *m,void *data) {struct drm_info_node *node = m->private;
@@ -4661,6 +4732,8 @@ static const struct drm_info_list i915_debugfs_list[] = { {"i915_gem_hws_bsd", i915_hws_info, 0, (void *)VCS}, {"i915_gem_hws_vebox", i915_hws_info, 0, (void *)VECS}, {"i915_gem_batch_pool", i915_gem_batch_pool_info, 0}, + {"i915_guc_info", i915_guc_info, 0}, + {"i915_guc_load_status", i915_guc_load_status_info, 0}, {"i915_frequency_info", i915_frequency_info, 0}, {"i915_hangcheck_info", i915_hangcheck_info, 0}, {"i915_drpc_info", i915_drpc_info, 0},
drivers/gpu/drm/i915/intel_guc.h +13 -0
@@ -42,6 +42,12 @@ struct i915_guc_client {uint32_t wq_size;uint32_t wq_tail;uint32_t cookie; + +/* GuC submission statistics & status */ +uint64_t submissions; +uint32_t q_fail; +uint32_t b_fail; +int retcode; };#define I915_MAX_DOORBELLS 256
@@ -62,6 +68,13 @@ struct intel_guc {struct ida ctx_ids;int db_cacheline;DECLARE_BITMAP (doorbell_bitmap,I915_MAX_DOORBELLS ); + +/* Action status & statistics */ +uint64_t action_count;/* Total commands issued */ +uint32_t action_cmd;/* Last command word */ +uint32_t action_status;/* Last return status */ +uint32_t action_fail;/* Total number of failures */ +int32_t action_err;/* Last error code */ };#define GUC_STATUS 0xc000
drivers/gpu/drm/i915/intel_guc_client.c +26 -7
@@ -50,6 +50,9 @@ static int intel_guc_action(struct intel_guc *guc, u32 *data, u32 len)spin_lock (&dev_priv->guc.host2guc_lock); + dev_priv->guc.action_count +=1 ; + dev_priv->guc.action_cmd = data[0 ]; +for (i =0 ; i < len; i++)I915_WRITE (SOFT_SCRATCH (i), data[i]);
@@ -68,7 +71,11 @@ static int intel_guc_action(struct intel_guc *guc, u32 *data, u32 len)"status=0x%08X response=0x%08X\n" , data[0 ], ret, status, I915_READ(SOFT_SCRATCH(15))); + + dev_priv->guc.action_fail +=1 ; + dev_priv->guc.action_err = ret; } + dev_priv->guc.action_status = status;spin_unlock (&dev_priv->guc.host2guc_lock);
@@ -560,13 +567,25 @@ int i915_guc_client_submit(struct i915_guc_client *client,struct intel_context *ctx, struct intel_engine_cs *ring) { - int ret; - - ret =add_workqueue_item (client, ctx, ring); -if (ret) -return ret; + int q_ret, b_ret; +unsigned long flags; - ret =ring_doorbell (client); + q_ret =add_workqueue_item (client, ctx, ring); +if (q_ret ==0 ) + b_ret =ring_doorbell (client); + +spin_lock_irqsave (&client->wq_lock, flags); + client->submissions +=1 ; +if (q_ret) { + client->q_fail +=1 ; + client->retcode = q_ret; + }else if (b_ret) { + client->b_fail +=1 ; + client->retcode = q_ret = b_ret; + }else { + client->retcode =0 ; + } +spin_unlock_irqrestore (&client->wq_lock, flags); -return ret; +return q_ret; }
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 15/18 drm/i915: Enable GuC firmware log
In reply to [email protected]From: Alex Dai <[email protected]>
Allocate a gem obj to hold GuC log data. Also a debugfs interface (i915_guc_log_dump) is provided to print out the log content.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/i915_debugfs.c | 29 +++++++++++++++ drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/i915_params.c | 5 +++ drivers/gpu/drm/i915/intel_guc.h | 3 +- drivers/gpu/drm/i915/intel_guc_loader.c | 60 +++++++++++++++++++++++++++++- drivers/gpu/drm/i915/intel_guc_scheduler.c | 9 ++--- 6 files changed, 99 insertions(+), 8 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +29 -0
@@ -2310,6 +2310,34 @@ static int i915_guc_info(struct seq_file *m, void *data)return 0 ; } +static int i915_guc_log_dump (struct seq_file *m,void *data) +{ +struct drm_info_node *node = m->private; +struct drm_device *dev = node->minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +struct drm_i915_gem_object *log_obj = dev_priv->guc.log_obj; +u32 *log; +int i =0 , pg; + +if (!log_obj) +return 0 ; + +for (pg =0 ; pg < log_obj->base.size /PAGE_SIZE ; pg++) { + log =kmap_atomic (i915_gem_object_get_page (log_obj, pg)); + +for (i =0 ; i <PAGE_SIZE /sizeof (u32); i +=4 ) +seq_printf (m,"0x%08x 0x%08x 0x%08x 0x%08x\n" , + *(log + i), *(log + i +1 ), + *(log + i +2 ), *(log + i +3 )); + +kunmap_atomic (log); + } + +seq_putc (m,'\n' ); + +return 0 ; +} +static int i915_edp_psr_status (struct seq_file *m,void *data) {struct drm_info_node *node = m->private;
@@ -4734,6 +4762,7 @@ static const struct drm_info_list i915_debugfs_list[] = { {"i915_gem_batch_pool", i915_gem_batch_pool_info, 0}, {"i915_guc_info", i915_guc_info, 0}, {"i915_guc_load_status", i915_guc_load_status_info, 0}, + {"i915_guc_log_dump", i915_guc_log_dump, 0}, {"i915_frequency_info", i915_frequency_info, 0}, {"i915_hangcheck_info", i915_hangcheck_info, 0}, {"i915_drpc_info", i915_drpc_info, 0},
drivers/gpu/drm/i915/i915_drv.h +1 -0
@@ -2477,6 +2477,7 @@ struct i915_params {bool disable_display;bool disable_vtd_wa;bool enable_guc_scheduling; +unsigned int guc_log_level;int use_mmio_flip;int mmio_debug;bool verbose_state_checks;
drivers/gpu/drm/i915/i915_params.c +5 -0
@@ -53,6 +53,7 @@ struct i915_params i915 __read_mostly = { .verbose_state_checks = 1, .nuclear_pageflip = 0, .enable_guc_scheduling = false, + .guc_log_level = 0, };module_param_named (modeset, i915.modeset, int,0400 );
@@ -182,3 +183,7 @@ MODULE_PARM_DESC(nuclear_pageflip,module_param_named (enable_guc_scheduling, i915.enable_guc_scheduling, bool,0400 );MODULE_PARM_DESC (enable_guc_scheduling,"Enable GuC scheduling (default:false)" ); + +module_param_named (guc_log_level, i915.guc_log_level, int,0400 ); +MODULE_PARM_DESC (guc_log_level, +"GuC firmware logging level (0:disabled, 1~4:enabled)" );
drivers/gpu/drm/i915/intel_guc.h +2 -1
@@ -62,6 +62,7 @@ struct intel_guc {spinlock_t host2guc_lock;struct drm_i915_gem_object *ctx_pool_obj; +struct drm_i915_gem_object *log_obj;struct i915_guc_client *execbuf_client;
@@ -170,7 +171,7 @@ int guc_scheduler_init(struct drm_device *dev);void guc_scheduler_fini (struct drm_device *dev);int guc_scheduler_enable (struct drm_device *dev);void guc_scheduler_disable (struct drm_device *dev); -bool sanitize_enable_guc_scheduling (struct drm_device *dev); +void sanitize_enable_guc_scheduling (struct drm_device *dev);/* intel_guc_client.c */ struct i915_guc_client*
drivers/gpu/drm/i915/intel_guc_loader.c +58 -2
90 lines not on this page · raw message
drivers/gpu/drm/i915/intel_guc_scheduler.c +4 -5
@@ -155,11 +155,10 @@ void guc_scheduler_disable(struct drm_device *dev) } } -bool sanitize_enable_guc_scheduling (struct drm_device *dev) +void sanitize_enable_guc_scheduling (struct drm_device *dev) { -if (!HAS_GUC_UCODE (dev) || !HAS_GUC_SCHED (dev)) -return false; - -return i915.enable_execlists && i915.enable_guc_scheduling; +if (!HAS_GUC_UCODE (dev) || !HAS_GUC_SCHED (dev) || + !i915.enable_execlists) + i915.enable_guc_scheduling = false; }
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 16/18 drm/i915: Ring Context allocating for GuC
In reply to [email protected]From: Alex Dai <[email protected]>
GuC firmware uses the one page after Ring Context as shared data. However, GuC uses same offset to address this page for all rings. So we have to allocate same size of lrc context for all rings.
Also, reduce ring buffer size to 4 pages. In GuC, work queue tail is referenced by 11 bits (WQ_RING_TAIL_MASK). It is in QW, so total 14 bits (4 pages).
Issue: VIZ-4884
--- drivers/gpu/drm/i915/intel_lrc.c | 23 ++++++++--------------- drivers/gpu/drm/i915/intel_ringbuffer.c | 2 +- 2 files changed, 9 insertions(+), 16 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +8 -15
@@ -1781,20 +1781,10 @@ uint32_t get_lr_context_size(struct intel_engine_cs *ring)WARN_ON (INTEL_INFO (ring->dev)->gen <8 ); -switch (ring->id) { -case RCS : -if (INTEL_INFO (ring->dev)->gen >=9 ) - ret =GEN9_LR_CONTEXT_RENDER_SIZE ; -else - ret =GEN8_LR_CONTEXT_RENDER_SIZE ; -break ; -case VCS : -case BCS : -case VECS : -case VCS2 : - ret =GEN8_LR_CONTEXT_OTHER_SIZE ; -break ; - } +if (INTEL_INFO (ring->dev)->gen >=9 ) + ret =GEN9_LR_CONTEXT_RENDER_SIZE ; +else + ret =GEN8_LR_CONTEXT_RENDER_SIZE ;return ret; }
@@ -1843,6 +1833,9 @@ int intel_lr_context_deferred_create(struct intel_context *ctx,WARN_ON (ctx->engine[ring->id].state); context_size =round_up (get_lr_context_size (ring),4096 ); +/* One extra page as the sharing data between driver and GuC */ +if (i915.enable_guc_scheduling) + context_size +=PAGE_SIZE ; ctx_obj =i915_gem_alloc_context_obj (dev, context_size);if (IS_ERR (ctx_obj)) {
@@ -1871,7 +1864,7 @@ int intel_lr_context_deferred_create(struct intel_context *ctx, ringbuf->ring = ring; - ringbuf->size =32 *PAGE_SIZE ; + ringbuf->size =4 *PAGE_SIZE ; ringbuf->effective_size = ringbuf->size; ringbuf->head =0 ; ringbuf->tail =0 ;
drivers/gpu/drm/i915/intel_ringbuffer.c +1 -1
@@ -1972,7 +1972,7 @@ static int intel_init_ring_buffer(struct drm_device *dev,INIT_LIST_HEAD (&ring->active_list);INIT_LIST_HEAD (&ring->request_list);INIT_LIST_HEAD (&ring->execlist_queue); - ringbuf->size =32 *PAGE_SIZE ; + ringbuf->size =4 *PAGE_SIZE ; ringbuf->ring = ring;memset (ring->semaphore.sync_seqno,0 ,sizeof (ring->semaphore.sync_seqno));
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 17/18 drm/i915: Taking forcewake during GuC load.
In reply to [email protected]From: Sagar Kamble <[email protected]>
Need to take forcewake before GuC loading.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/intel_guc_loader.c | 3 +++ 1 file changed, 3 insertions(+)
drivers/gpu/drm/i915/intel_guc_loader.c +3 -0
@@ -273,6 +273,7 @@ static int guc_load_ucode(struct drm_device *dev)int ret;WARN_ON (!mutex_is_locked (&dev->struct_mutex)); +intel_uncore_forcewake_get (dev_priv,FORCEWAKE_ALL ); ret =i915_gem_obj_ggtt_pin (guc_fw->uc_fw_obj,0 ,0 );if (ret)
@@ -326,6 +327,8 @@ out:if (pinned)i915_gem_object_ggtt_unpin (guc_fw->uc_fw_obj); +intel_uncore_forcewake_put (dev_priv,FORCEWAKE_ALL ); +return ret; }
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 17/18] drm/i915: Taking forcewake during GuC load.
Quoted 3 lines · On Thu, Mar 26, 2015 at 12:41:24PM -0700, [email protected] wrote:
From: Sagar Kamble <[email protected]>
Need to take forcewake before GuC loading.
Please explain why and how things fall over if we're not doing this. -Daniel
Quoted 36 lines
Issue: VIZ-4884
Change-Id Ie422fc1e122933b161ff63cab23622197e6bba54Signed-off-by Sagar Kamble <[email protected]>--- drivers/gpu/drm/i915/intel_guc_loader.c | 3 +++ 1 file changed, 3 insertions(+)drivers/gpu/drm/i915/intel_guc_loader.c +3
-0@@ -273,6 +273,7 @@ static int guc_load_ucode(struct drm_device *dev)int ret;WARN_ON (!mutex_is_locked (&dev->struct_mutex)); +intel_uncore_forcewake_get (dev_priv,FORCEWAKE_ALL ); ret =i915_gem_obj_ggtt_pin (guc_fw->uc_fw_obj,0 ,0 );if (ret)@@ -326,6 +327,8 @@ out:if (pinned)i915_gem_object_ggtt_unpin (guc_fw->uc_fw_obj); +intel_uncore_forcewake_put (dev_priv,FORCEWAKE_ALL ); +return ret; }Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
PATCH 18/18 drm/i915: Notify GuC when RC6 state is changed
In reply to [email protected]From: Alex Dai <[email protected]>
Whenever RC6 state (0xA210) is changed, driver needs to notify GuC via guc_action.
Issue: VIZ-4884
--- drivers/gpu/drm/i915/intel_guc.h | 1 + drivers/gpu/drm/i915/intel_guc_api.h | 1 + drivers/gpu/drm/i915/intel_guc_client.c | 16 ++++++++++++++++ drivers/gpu/drm/i915/intel_guc_scheduler.c | 3 +++ drivers/gpu/drm/i915/intel_pm.c | 4 ++++ 5 files changed, 25 insertions(+)
drivers/gpu/drm/i915/intel_guc.h +1 -0
@@ -181,5 +181,6 @@ void i915_guc_client_free(struct drm_device *dev,int i915_guc_client_submit (struct i915_guc_client *client,struct intel_context *ctx,struct intel_engine_cs *ring); +int host2guc_sample_forcewake (struct intel_guc *guc,bool enable);#endif
drivers/gpu/drm/i915/intel_guc_api.h +1 -0
@@ -190,6 +190,7 @@ struct guc_context_desc {/* This Action will be programmed in C180 - SOFT_SCRATCH_O_REG */ enum host2guc_action {HOST2GUC_ACTION_DEFAULT =0x0 , +HOST2GUC_ACTION_SAMPLE_FORCEWAKE =0x6 ,HOST2GUC_ACTION_ALLOCATE_DOORBELL =0x10 ,HOST2GUC_ACTION_DEALLOCATE_DOORBELL =0x20 ,HOST2GUC_ACTION_LIMIT
drivers/gpu/drm/i915/intel_guc_client.c +16 -0
@@ -262,6 +262,22 @@ static int host2guc_release_doorbell(struct intel_guc *guc,return intel_guc_action (guc, data,2 ); } +int host2guc_sample_forcewake (struct intel_guc *guc,bool enable) +{ +u32 data[2 ]; + +if (guc->guc_fw.uc_fw_load_status !=INTEL_UC_FIRMWARE_SUCCESS ) +return -ENODEV ; + + data[0 ] =HOST2GUC_ACTION_SAMPLE_FORCEWAKE ; +if (enable) + data[1 ] =3 ;/* Bit0: Render, Bit1: Media */ +else + data[1 ] =0 ; + +return intel_guc_action (guc, data,2 ); +} +static void init_doorbell (struct intel_guc *guc,struct i915_guc_client *client) {
drivers/gpu/drm/i915/intel_guc_scheduler.c +3 -0
@@ -139,6 +139,9 @@ int guc_scheduler_enable(struct drm_device *dev)direct_interrupts_to_guc (dev_priv); +flush_delayed_work (&dev_priv->rps.delayed_resume_work); +host2guc_sample_forcewake (guc, dev_priv->rps.enabled); +return 0 ; }
drivers/gpu/drm/i915/intel_pm.c +4 -0
@@ -4113,6 +4113,8 @@ static void gen9_disable_rps(struct drm_device *dev)I915_WRITE (GEN6_RC_CONTROL ,0 );I915_WRITE (GEN9_PG_ENABLE ,0 ); + +host2guc_sample_forcewake (&dev_priv->guc,0 ); }static void gen6_disable_rps (struct drm_device *dev)
@@ -4330,6 +4332,8 @@ static void gen9_enable_rc6(struct drm_device *dev)intel_uncore_forcewake_put (dev_priv,FORCEWAKE_ALL ); +host2guc_sample_forcewake (&dev_priv->guc, + rc6_mask &GEN6_RC_CTL_RC6_ENABLE ); }static void gen8_enable_rps (struct drm_device *dev)
Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 18/18] drm/i915: Notify GuC when RC6 state is changed
Task id: 6065 -------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV -2 276/276 274/276 ILK -1 303/303 302/303 SNB -1 304/304 303/304 IVB -1 339/339 338/339 BYT -1 287/287 286/287 HSW -2 362/362 360/362 BDW -1 310/310 309/310
-------------------------------------Detailed-------------------------------------
Platform Test drm-intel-nightly Series Applied
PNV igt@gem_userptr_blits@minor-normal-sync DMESG_WARN(1)PASS(3) DMESG_WARN(1)PASS(1)
(dmesg patch applied)WARNING:at_drivers/gpu/drm/i915/i915_gem_evict.c:#i915_gem_evict_vm[i915]()@WARNING:.* at .* i915_gem_evict_vm+0x
*PNV igt@drv_debugfs_reader PASS(2) DMESG_WARN(2)
(dmesg patch applied)INFO:trying_to_register_non-static_key@INFO: trying to register non-static key.
*ILK igt@drv_suspend@debugfs-reader PASS(2) DMESG_WARN(2)
(dmesg patch applied)INFO:trying_to_register_non-static_key@INFO: trying to register non-static key.
*SNB igt@drv_debugfs_reader PASS(2) DMESG_WARN(2)
(dmesg patch applied)INFO:trying_to_register_non-static_key@INFO: trying to register non-static key.
*IVB igt@drv_debugfs_reader PASS(2) DMESG_WARN(2)
(dmesg patch applied)INFO:trying_to_register_non-static_key@INFO: trying to register non-static key.
*BYT igt@drv_debugfs_reader PASS(2) DMESG_WARN(2)
(dmesg patch applied)INFO:trying_to_register_non-static_key@INFO: trying to register non-static key.
*HSW igt@gem_storedw_loop_bsd PASS(2) DMESG_WARN(1)PASS(1)
(dmesg patch applied)drm:i915_hangcheck_elapsed[i915]]*ERROR*Hangcheck_timer_elapsed...bsd_ring_idle@Hangcheck timer elapsed... bsd ring idle
*HSW igt@pm_rpm@debugfs-read PASS(2) DMESG_WARN(2)
(dmesg patch applied)WARNING:at_drivers/gpu/drm/i915/intel_uncore.c:#assert_device_not_suspended[i915]()@WARNING:.* at .* assert_device_not_suspended+0x drm:fw_domains_get[i915]]*ERROR*render:timed_out_waiting_for_forcewake_ack_to_clear@render: timed out waiting for forcewake ack to clear. drm:fw_domains_get_with_thread_status[i915]]*ERROR*GT_thread_status_wait_timed_out@GT thread status wait timed out INFO:trying_to_register_non-static_key@INFO: trying to register non-static key.
*BDW igt@gem_userptr_blits@forked-sync-swapping-mempressure-normal PASS(2) DMESG_WARN(2)
(dmesg patch applied)drm:intel_uc_fw_check[i915]]*ERROR*Failed_to_fetch_GuC_firmware_from<i915/guc_gen8.bin>@Failed to fetch GuC firmware from INFO:possible_circular_locking_dependency_detected@INFO: possible circular locking dependency detected
Note: You need to pay more attention to line start with '*'
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [PATCH 18/18] drm/i915: Notify GuC when RC6 state is changed
In reply to [email protected]Quoted 8 lines · On Thu, Mar 26, 2015 at 12:41:25PM -0700, [email protected] wrote:
From: Alex Dai <[email protected]>
Whenever RC6 state (
0xA210) is changed, driver needs to notify GuC viaguc_action.Issue: VIZ-4884
Change-Id I15c661a915c670691d020471ecaccb00f7afb624Signed-off-by Alex Dai <[email protected]>
What exactly is the interaction here? Please elaborate in the commit message, since as-is this looks extremely racy: If the guc can fall over if we change rc6 behind it's back it can still fall over right between when we change rc6 mode and when we tell the guc.
Also note that rc6 setup is delayed in resume/driver load (because it takes forever) and we expect the gpu to be fully working while we enable rc6. If the guc falls over on resume because of this there's a problem. We might need specific igt testcases which race resume against gpu submissions.
-Daniel
Quoted 101 lines
--- drivers/gpu/drm/i915/intel_guc.h | 1 + drivers/gpu/drm/i915/intel_guc_api.h | 1 + drivers/gpu/drm/i915/intel_guc_client.c | 16 ++++++++++++++++ drivers/gpu/drm/i915/intel_guc_scheduler.c | 3 +++ drivers/gpu/drm/i915/intel_pm.c | 4 ++++ 5 files changed, 25 insertions(+)drivers/gpu/drm/i915/intel_guc.h +1
-0@@ -181,5 +181,6 @@ void i915_guc_client_free(struct drm_device *dev,int i915_guc_client_submit (struct i915_guc_client *client,struct intel_context *ctx,struct intel_engine_cs *ring); +int host2guc_sample_forcewake (struct intel_guc *guc,bool enable);#endif drivers/gpu/drm/i915/intel_guc_api.h +1
-0@@ -190,6 +190,7 @@ struct guc_context_desc {/* This Action will be programmed in C180 - SOFT_SCRATCH_O_REG */ enum host2guc_action {HOST2GUC_ACTION_DEFAULT =0x0 , +HOST2GUC_ACTION_SAMPLE_FORCEWAKE =0x6 ,HOST2GUC_ACTION_ALLOCATE_DOORBELL =0x10 ,HOST2GUC_ACTION_DEALLOCATE_DOORBELL =0x20 ,HOST2GUC_ACTION_LIMIT drivers/gpu/drm/i915/intel_guc_client.c +16
-0@@ -262,6 +262,22 @@ static int host2guc_release_doorbell(struct intel_guc *guc,return intel_guc_action (guc, data,2 ); } +int host2guc_sample_forcewake (struct intel_guc *guc,bool enable) +{ +u32 data[2 ]; + +if (guc->guc_fw.uc_fw_load_status !=INTEL_UC_FIRMWARE_SUCCESS ) +return -ENODEV ; + + data[0 ] =HOST2GUC_ACTION_SAMPLE_FORCEWAKE ; +if (enable) + data[1 ] =3 ;/* Bit0: Render, Bit1: Media */ +else + data[1 ] =0 ; + +return intel_guc_action (guc, data,2 ); +} +static void init_doorbell (struct intel_guc *guc,struct i915_guc_client *client) {drivers/gpu/drm/i915/intel_guc_scheduler.c +3
-0@@ -139,6 +139,9 @@ int guc_scheduler_enable(struct drm_device *dev)direct_interrupts_to_guc (dev_priv); +flush_delayed_work (&dev_priv->rps.delayed_resume_work); +host2guc_sample_forcewake (guc, dev_priv->rps.enabled); +return 0 ; }drivers/gpu/drm/i915/intel_pm.c +4
-0@@ -4113,6 +4113,8 @@ static void gen9_disable_rps(struct drm_device *dev)I915_WRITE (GEN6_RC_CONTROL ,0 );I915_WRITE (GEN9_PG_ENABLE ,0 ); + +host2guc_sample_forcewake (&dev_priv->guc,0 ); }static void gen6_disable_rps (struct drm_device *dev)@@ -4330,6 +4332,8 @@ static void gen9_enable_rc6(struct drm_device *dev)intel_uncore_forcewake_put (dev_priv,FORCEWAKE_ALL ); +host2guc_sample_forcewake (&dev_priv->guc, + rc6_mask &GEN6_RC_CTL_RC6_ENABLE ); }static void gen8_enable_rps (struct drm_device *dev)Signature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoted 63 lines · On Thu, Mar 26, 2015 at 12:41:07PM -0700, [email protected] wrote:
From: Alex Dai <[email protected]>
This series of patch is to enable ExecList submission via GuC. Here are some key points related to this series, not in particular order.
***
i915_guc_client***
We use the term client to avoid confusion with contexts. Ai915_guc_clientis equivalent to GuC objectguc_context_desc. This context descriptor is allocated from a pool of 1024 entries. Kernel driver will allocate doorbell and workqueue for it. Also the process descriptor (guc_process_desc), which is mapped to client space. So the client can write Work Item then ring the doorbell.To simplify the implementation, we allocate one gem object that contains all pages for doorbell, process descriptor and workqueue.
***
intel_guc***
Top level structure of guc. It handles firmware loading and manages client pool and doorbells.intel_gucowns ai915_guc_clientto do the legacy submission.** The Scratch registers ***
There are 16 MMIO-based registers start from0xC180. The kernel driver writes a value to the action register (SOFT_SCRATCH_0) along with any data. It then triggers an interrupt on the GuC via another register write (0xC4C8). Firmware writes a success/fail code back to the action register after processes the request. The kernel driver polls waiting for this update and then proceeds.Details in
intel_guc_action()*** Work Items ***
There are several types of work items that the host may place into a workqueue, each with its own requirements and limitations. Currently onlyWQ_TYPE_INORDERis used to support legacy submission via GuC, which represents in-order queue. The kernel driver packs ring tail pointer and an ELSP context descriptor dword into Work Item.Details in
add_workqueue_item()*** Doorbells ***
Doorbells are interrupts to uKernel. A doorbell is a single cache line (QW) mapped into process space.Details in
ring_doorbell()*** Firmware versioning ***
The firmware build process will generate a version header file with major and minor version defined. The versions are built into CSS header of firmware too. i915 kernel driver set the minimal firmware version required by each platform. The firmware installation package will install (symbolic link) proper version of firmware.*** Firmware log ***
Firmware log is enabled by setting i915.guc_log_level to non-negative level. Log data is printed out via reading debugfsi915_guc_log_dump. Reading fromi915_guc_load_statuswill print out firmware loading status and scratch registers value.TODO: the buffer works like a ring. To not missing any data, driver should handle a GuC2Host interrupt (triggered when log buffer is half full) from GuC.
*** Ring buffer size ***
It is up to 16K (4 pages) per LRC. Not HW limitation but firmware is setup in this way.
Quoted 7 lines
*** GuC address space ***
GuC is not expecting any gfx GGTT address that falls into range [0,WOPCM_TOP), which is reserved for Boot ROM, SRAM and WOPCM. Currently this top address is 512K in order to fit in big HuC firmware.In order to exclude 0-512K address space from GGTT, all gfx objects used by GuC is pinned with
PIN_OFFSET_BIASalong with size of WOPCM.
There's lots of great comments both in the code and overview text in this coverletter here, but no kerneldoc. Can you please create a patch on top to document the new guc stuff in an overview section and also document the non-static functions you're adding?
Thanks, Daniel
Quoted 64 lines
Alex Dai (13):
drm/i915: Add guc firmware interface headers drm/i915: GuC firmware loader
drm/i915: Add firmware version check
drm/i915: Make several execlist helper functions external drm/i915: Add functions to allocate / release gem obj for GuC drm/i915: Functions to support command submission via GuC drm/i915: Integration of GuC client
drm/i915: Interrupt routing for GuC scheduler drm/i915: Enable commands submission via GuC drm/i915: debugfs of GuC status
drm/i915: Enable GuC firmware log
drm/i915: Ring Context allocating for GuC
drm/i915: Notify GuC when RC6 state is changedDave Gordon (2):
drm/i915: Unified firmware loading mechanism
drm/i915: Defer default hardware context initialisation until first
openMichael H. Nguyen (2):
drm/i915: Addi915_gem_object_write()toi915_gem.cdrm/i915: Move execlists defines from .c to .hSagar Kamble (1):
drm/i915: Taking forcewake during GuC load.drivers/gpu/drm/i915/Makefile | 8 +- drivers/gpu/drm/i915/i915_debugfs.c | 102 +++++ drivers/gpu/drm/i915/i915_dma.c | 6 + drivers/gpu/drm/i915/i915_drv.h | 17 + drivers/gpu/drm/i915/i915_gem.c | 39 +- drivers/gpu/drm/i915/i915_gem_context.c | 35 +- drivers/gpu/drm/i915/i915_gem_stolen.c | 10 + drivers/gpu/drm/i915/i915_params.c | 9 + drivers/gpu/drm/i915/i915_reg.h | 83 +++- drivers/gpu/drm/i915/intel_guc.h | 186 +++++++++ drivers/gpu/drm/i915/intel_guc_api.h | 218 +++++++++++ drivers/gpu/drm/i915/intel_guc_client.c | 607 +++++++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_guc_loader.c | 466 ++++++++++++++++++++++ drivers/gpu/drm/i915/intel_guc_scheduler.c | 167 ++++++++ drivers/gpu/drm/i915/intel_lrc.c | 112 ++---- drivers/gpu/drm/i915/intel_lrc.h | 3 + drivers/gpu/drm/i915/intel_pm.c | 4 + drivers/gpu/drm/i915/intel_ringbuffer.c | 2 +- drivers/gpu/drm/i915/intel_uc_loader.c | 220 +++++++++++ drivers/gpu/drm/i915/intel_uc_loader.h | 82 ++++ 20 files changed, 2279 insertions(+), 97 deletions(-) create mode 100644 drivers/gpu/drm/i915/intel_guc.h create mode 100644 drivers/gpu/drm/i915/intel_guc_api.h create mode 100644 drivers/gpu/drm/i915/intel_guc_client.c create mode 100644 drivers/gpu/drm/i915/intel_guc_loader.c create mode 100644 drivers/gpu/drm/i915/intel_guc_scheduler.c create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.c create mode 100644 drivers/gpu/drm/i915/intel_uc_loader.hSignature · 6 lines
1.9.1 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx