/* * Copyright 2015 Advanced Micro Devices, Inc. * * 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 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 COPYRIGHT HOLDER(S) OR AUTHOR(S) 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. * * Authors: monk liu <monk.liu@amd.com> */ #include <drm/drm_auth.h>#include <drm/drm_drv.h>#include "amdgpu.h"#include "amdgpu_sched.h"#include "amdgpu_ras.h"#include <linux/nospec.h> #define to_amdgpu_ctx_entity(e)	\	container_of((e), struct amdgpu_ctx_entity, entity) const unsigned int amdgpu_ctx_num_entities[AMDGPU_HW_IP_NUM] = {	[AMDGPU_HW_IP_GFX]	=	1,	[AMDGPU_HW_IP_COMPUTE]	=	4,	[AMDGPU_HW_IP_DMA]	=	2,	[AMDGPU_HW_IP_UVD]	=	1,	[AMDGPU_HW_IP_VCE]	=	1,	[AMDGPU_HW_IP_UVD_ENC]	=	1,	[AMDGPU_HW_IP_VCN_DEC]	=	1,	[AMDGPU_HW_IP_VCN_ENC]	=	1,	[AMDGPU_HW_IP_VCN_JPEG]	=	1,	[AMDGPU_HW_IP_VPE]	=	1,}; bool amdgpu_ctx_priority_is_valid(int32_t ctx_prio){	switch (ctx_prio) {	case AMDGPU_CTX_PRIORITY_VERY_LOW:	case AMDGPU_CTX_PRIORITY_LOW:	case AMDGPU_CTX_PRIORITY_NORMAL:	case AMDGPU_CTX_PRIORITY_HIGH:	case AMDGPU_CTX_PRIORITY_VERY_HIGH:		return true;	default:	case AMDGPU_CTX_PRIORITY_UNSET:		/* UNSET priority is not valid and we don't carry that		 * around, but set it to NORMAL in the only place this		 * function is called, amdgpu_ctx_ioctl().		 */		return false;	}} static enum drm_sched_priorityamdgpu_ctx_to_drm_sched_prio(int32_t ctx_prio){	switch (ctx_prio) {	case AMDGPU_CTX_PRIORITY_UNSET:		pr_warn_once("AMD-->DRM context priority value UNSET-->NORMAL");		return DRM_SCHED_PRIORITY_NORMAL; 	case AMDGPU_CTX_PRIORITY_VERY_LOW:		return DRM_SCHED_PRIORITY_LOW; 	case AMDGPU_CTX_PRIORITY_LOW:		return DRM_SCHED_PRIORITY_LOW; 	case AMDGPU_CTX_PRIORITY_NORMAL:		return DRM_SCHED_PRIORITY_NORMAL; 	case AMDGPU_CTX_PRIORITY_HIGH:		return DRM_SCHED_PRIORITY_HIGH; 	case AMDGPU_CTX_PRIORITY_VERY_HIGH:		return DRM_SCHED_PRIORITY_HIGH; 	/* This should not happen as we sanitized userspace provided priority	 * already, WARN if this happens.	 */	default:		WARN(1, "Invalid context priority %d\n", ctx_prio);		return DRM_SCHED_PRIORITY_NORMAL;	} } static int amdgpu_ctx_priority_permit(struct drm_file *filp,				      int32_t priority){	/* NORMAL and below are accessible by everyone */	if (priority <= AMDGPU_CTX_PRIORITY_NORMAL)		return 0; 	if (capable(CAP_SYS_NICE))		return 0; 	if (drm_is_current_master(filp))		return 0; 	return -EACCES;} static enum amdgpu_gfx_pipe_priority amdgpu_ctx_prio_to_gfx_pipe_prio(int32_t prio){	switch (prio) {	case AMDGPU_CTX_PRIORITY_HIGH:	case AMDGPU_CTX_PRIORITY_VERY_HIGH:		return AMDGPU_GFX_PIPE_PRIO_HIGH;	default:		return AMDGPU_GFX_PIPE_PRIO_NORMAL;	}} static enum amdgpu_ring_priority_level amdgpu_ctx_sched_prio_to_ring_prio(int32_t prio){	switch (prio) {	case AMDGPU_CTX_PRIORITY_HIGH:		return AMDGPU_RING_PRIO_1;	case AMDGPU_CTX_PRIORITY_VERY_HIGH:		return AMDGPU_RING_PRIO_2;	default:		return AMDGPU_RING_PRIO_0;	}} static unsigned int amdgpu_ctx_get_hw_prio(struct amdgpu_ctx *ctx, u32 hw_ip){	struct amdgpu_device *adev = ctx->mgr->adev;	unsigned int hw_prio;	int32_t ctx_prio; 	ctx_prio = (ctx->override_priority == AMDGPU_CTX_PRIORITY_UNSET) ?			ctx->init_priority : ctx->override_priority; 	switch (hw_ip) {	case AMDGPU_HW_IP_GFX:	case AMDGPU_HW_IP_COMPUTE:		hw_prio = amdgpu_ctx_prio_to_gfx_pipe_prio(ctx_prio);		break;	case AMDGPU_HW_IP_VCE:	case AMDGPU_HW_IP_VCN_ENC:		hw_prio = amdgpu_ctx_sched_prio_to_ring_prio(ctx_prio);		break;	default:		hw_prio = AMDGPU_RING_PRIO_DEFAULT;		break;	} 	hw_ip = array_index_nospec(hw_ip, AMDGPU_HW_IP_NUM);	if (adev->gpu_sched[hw_ip][hw_prio].num_scheds == 0)		hw_prio = AMDGPU_RING_PRIO_DEFAULT; 	return hw_prio;} /* Calculate the time spend on the hw */static ktime_t amdgpu_ctx_fence_time(struct dma_fence *fence){	struct drm_sched_fence *s_fence; 	if (!fence)		return ns_to_ktime(0); 	/* When the fence is not even scheduled it can't have spend time */	s_fence = to_drm_sched_fence(fence);	if (!test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &s_fence->scheduled.flags))		return ns_to_ktime(0); 	/* When it is still running account how much already spend */	if (!test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &s_fence->finished.flags))		return ktime_sub(ktime_get(), s_fence->scheduled.timestamp); 	return ktime_sub(s_fence->finished.timestamp,			 s_fence->scheduled.timestamp);} static ktime_t amdgpu_ctx_entity_time(struct amdgpu_ctx *ctx,				      struct amdgpu_ctx_entity *centity){	ktime_t res = ns_to_ktime(0);	uint32_t i; 	spin_lock(&ctx->ring_lock);	for (i = 0; i < amdgpu_sched_jobs; i++) {		res = ktime_add(res, amdgpu_ctx_fence_time(centity->fences[i]));	}	spin_unlock(&ctx->ring_lock);	return res;} static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, u32 hw_ip,				  const u32 ring){	struct drm_gpu_scheduler **scheds = NULL, *sched = NULL;	struct amdgpu_device *adev = ctx->mgr->adev;	struct amdgpu_ctx_entity *entity;	enum drm_sched_priority drm_prio;	unsigned int hw_prio, num_scheds;	int32_t ctx_prio;	int r; 	entity = kzalloc_flex(*entity, fences, amdgpu_sched_jobs);	if (!entity)		return  -ENOMEM; 	ctx_prio = (ctx->override_priority == AMDGPU_CTX_PRIORITY_UNSET) ?			ctx->init_priority : ctx->override_priority;	entity->hw_ip = hw_ip;	entity->sequence = 1;	hw_prio = amdgpu_ctx_get_hw_prio(ctx, hw_ip);	drm_prio = amdgpu_ctx_to_drm_sched_prio(ctx_prio); 	hw_ip = array_index_nospec(hw_ip, AMDGPU_HW_IP_NUM); 	if (!(adev)->xcp_mgr) {		scheds = adev->gpu_sched[hw_ip][hw_prio].sched;		num_scheds = adev->gpu_sched[hw_ip][hw_prio].num_scheds;	} else {		struct amdgpu_fpriv *fpriv; 		/* TODO: Stop using fpriv here, we only need the xcp_id. */		fpriv = container_of(ctx->mgr, struct amdgpu_fpriv, ctx_mgr);		r = amdgpu_xcp_select_scheds(adev, hw_ip, hw_prio, fpriv,						&num_scheds, &scheds);		if (r)			goto error_free_entity;	} 	if (num_scheds == 0) {		r = -EINVAL;		goto error_free_entity;	} 	/* disable load balance if the hw engine retains context among dependent jobs */	if (hw_ip == AMDGPU_HW_IP_VCN_ENC ||	    hw_ip == AMDGPU_HW_IP_VCN_DEC ||	    hw_ip == AMDGPU_HW_IP_UVD_ENC ||	    hw_ip == AMDGPU_HW_IP_UVD) {		sched = drm_sched_pick_best(scheds, num_scheds);		scheds = &sched;		num_scheds = 1;	} 	r = drm_sched_entity_init(&entity->entity, drm_prio, scheds, num_scheds,				  NULL);	if (r)		goto error_free_entity; 	/* It's not an error if we fail to install the new entity */	if (cmpxchg(&ctx->entities[hw_ip][ring], NULL, entity))		goto cleanup_entity; 	return 0; cleanup_entity:	drm_sched_entity_fini(&entity->entity); error_free_entity:	kfree(entity); 	return r;} static ktime_t amdgpu_ctx_fini_entity(struct amdgpu_device *adev,				  struct amdgpu_ctx_entity *entity){	ktime_t res = ns_to_ktime(0);	int i; 	if (!entity)		return res; 	drm_sched_entity_destroy(&entity->entity); 	for (i = 0; i < amdgpu_sched_jobs; ++i) {		res = ktime_add(res, amdgpu_ctx_fence_time(entity->fences[i]));		dma_fence_put(entity->fences[i]);	} 	amdgpu_xcp_release_sched(adev, entity); 	kfree(entity);	return res;} static u32 amdgpu_get_stable_pstate(struct amdgpu_device *adev){	switch (amdgpu_dpm_get_performance_level(adev)) {	case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD:		return AMDGPU_CTX_STABLE_PSTATE_STANDARD;	case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK:		return AMDGPU_CTX_STABLE_PSTATE_MIN_SCLK;	case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK:		return AMDGPU_CTX_STABLE_PSTATE_MIN_MCLK;	case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK:		return AMDGPU_CTX_STABLE_PSTATE_PEAK;	default:		return AMDGPU_CTX_STABLE_PSTATE_NONE;	}} static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,			   struct drm_file *filp, struct amdgpu_ctx *ctx){	struct amdgpu_fpriv *fpriv = filp->driver_priv;	int r; 	r = amdgpu_ctx_priority_permit(filp, priority);	if (r)		return r; 	memset(ctx, 0, sizeof(*ctx)); 	kref_init(&ctx->refcount);	ctx->mgr = mgr;	spin_lock_init(&ctx->ring_lock); 	ctx->reset_counter = atomic_read(&mgr->adev->gpu_reset_counter);	ctx->reset_counter_query = ctx->reset_counter;	ctx->generation = amdgpu_vm_generation(mgr->adev, &fpriv->vm);	ctx->init_priority = priority;	ctx->override_priority = AMDGPU_CTX_PRIORITY_UNSET;	ctx->stable_pstate = AMDGPU_CTX_STABLE_PSTATE_NONE; 	return 0;} static int __amdgpu_ctx_set_stable_pstate(struct amdgpu_ctx *ctx,					  u32 stable_pstate){	struct amdgpu_device *adev = ctx->mgr->adev;	enum amd_dpm_forced_level level;	struct amdgpu_ctx *current_ctx;	u32 current_stable_pstate;	int r = 0; 	lockdep_assert_held(&adev->pm.stable_pstate_ctx_lock); 	switch (stable_pstate) {	case AMDGPU_CTX_STABLE_PSTATE_NONE:		level = AMD_DPM_FORCED_LEVEL_AUTO;		break;	case AMDGPU_CTX_STABLE_PSTATE_STANDARD:		level = AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD;		break;	case AMDGPU_CTX_STABLE_PSTATE_MIN_SCLK:		level = AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK;		break;	case AMDGPU_CTX_STABLE_PSTATE_MIN_MCLK:		level = AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK;		break;	case AMDGPU_CTX_STABLE_PSTATE_PEAK:		level = AMD_DPM_FORCED_LEVEL_PROFILE_PEAK;		break;	default:		return -EINVAL;	} 	current_ctx = adev->pm.stable_pstate_ctx;	if (current_ctx && current_ctx != ctx)		return -EBUSY; 	current_stable_pstate = amdgpu_get_stable_pstate(adev);	if (current_stable_pstate == stable_pstate)		return 0; 	r = amdgpu_dpm_force_performance_level(adev, level);	if (r)		return r; 	if (!current_ctx) {		adev->pm.stable_pstate_ctx = ctx;		/*		 * Serialized by context taking ownership for the first time		 * while holding adev->pm.stable_pstate_ctx_lock).		 */		WRITE_ONCE(ctx->stable_pstate, current_stable_pstate);	} 	return 0;} static int amdgpu_ctx_set_stable_pstate(struct amdgpu_ctx *ctx,					u32 stable_pstate){	struct amdgpu_device *adev = ctx->mgr->adev;	int r; 	mutex_lock(&adev->pm.stable_pstate_ctx_lock);	r = __amdgpu_ctx_set_stable_pstate(ctx, stable_pstate);	mutex_unlock(&adev->pm.stable_pstate_ctx_lock); 	return r;} void amdgpu_ctx_fini(struct kref *ref){	struct amdgpu_ctx *ctx = container_of(ref, struct amdgpu_ctx, refcount);	struct amdgpu_ctx_mgr *mgr = ctx->mgr;	struct amdgpu_device *adev = mgr->adev;	unsigned i, j, idx; 	if (!adev)		return; 	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {		for (j = 0; j < AMDGPU_MAX_ENTITY_NUM; ++j) {			ktime_t spend; 			spend = amdgpu_ctx_fini_entity(adev, ctx->entities[i][j]);			atomic64_add(ktime_to_ns(spend), &mgr->time_spend[i]);		}	} 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {		mutex_lock(&adev->pm.stable_pstate_ctx_lock);		if (adev->pm.stable_pstate_ctx == ctx) {			__amdgpu_ctx_set_stable_pstate(ctx, ctx->stable_pstate);			adev->pm.stable_pstate_ctx = NULL;		}		mutex_unlock(&adev->pm.stable_pstate_ctx_lock);		drm_dev_exit(idx);	} 	kfree(ctx);} int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance,			  u32 ring, struct drm_sched_entity **entity){	int r;	struct drm_sched_entity *ctx_entity; 	if (hw_ip >= AMDGPU_HW_IP_NUM) {		drm_err(adev_to_drm(ctx->mgr->adev),			"unknown HW IP type: %d\n", hw_ip);		return -EINVAL;	} 	/* Right now all IPs have only one instance - multiple rings. */	if (instance != 0) {		drm_dbg(adev_to_drm(ctx->mgr->adev),			"invalid ip instance: %d\n", instance);		return -EINVAL;	} 	if (ring >= amdgpu_ctx_num_entities[hw_ip]) {		drm_dbg(adev_to_drm(ctx->mgr->adev),			"invalid ring: %d %d\n", hw_ip, ring);		return -EINVAL;	} 	if (ctx->entities[hw_ip][ring] == NULL) {		r = amdgpu_ctx_init_entity(ctx, hw_ip, ring);		if (r)			return r;	} 	ctx_entity = &ctx->entities[hw_ip][ring]->entity;	r = drm_sched_entity_error(ctx_entity);	if (r) {		DRM_DEBUG("error entity %p\n", ctx_entity);		return r;	} 	*entity = ctx_entity;	return 0;} static int amdgpu_ctx_alloc(struct amdgpu_device *adev,			    struct amdgpu_fpriv *fpriv,			    struct drm_file *filp,			    int32_t priority,			    uint32_t *id){	struct amdgpu_ctx_mgr *mgr = &fpriv->ctx_mgr;	struct amdgpu_ctx *ctx;	int r; 	ctx = kmalloc_obj(*ctx);	if (!ctx)		return -ENOMEM; 	r = amdgpu_ctx_init(mgr, priority, filp, ctx);	if (r) {		kfree(ctx);		return r;

Lines 1–500 of 949. Load more