blob: 54ffcfcdd41f9ce3c590f5814fcbe3d3535946ac [file] [log] [blame]
// SPDX-License-Identifier: MIT
/*
* Copyright © 2022 Intel Corporation
*/
#include "xe_guc_submit.h"
#include <linux/bitfield.h>
#include <linux/bitmap.h>
#include <linux/circ_buf.h>
#include <linux/delay.h>
#include <linux/dma-fence-array.h>
#include <drm/drm_managed.h>
#include "abi/guc_actions_abi.h"
#include "abi/guc_klvs_abi.h"
#include "regs/xe_lrc_layout.h"
#include "xe_assert.h"
#include "xe_devcoredump.h"
#include "xe_device.h"
#include "xe_exec_queue.h"
#include "xe_force_wake.h"
#include "xe_gpu_scheduler.h"
#include "xe_gt.h"
#include "xe_guc.h"
#include "xe_guc_ct.h"
#include "xe_guc_exec_queue_types.h"
#include "xe_guc_submit_types.h"
#include "xe_hw_engine.h"
#include "xe_hw_fence.h"
#include "xe_lrc.h"
#include "xe_macros.h"
#include "xe_map.h"
#include "xe_mocs.h"
#include "xe_ring_ops_types.h"
#include "xe_sched_job.h"
#include "xe_trace.h"
#include "xe_vm.h"
static struct xe_guc *
exec_queue_to_guc(struct xe_exec_queue *q)
{
return &q->gt->uc.guc;
}
/*
* Helpers for engine state, using an atomic as some of the bits can transition
* as the same time (e.g. a suspend can be happning at the same time as schedule
* engine done being processed).
*/
#define EXEC_QUEUE_STATE_REGISTERED (1 << 0)
#define ENGINE_STATE_ENABLED (1 << 1)
#define EXEC_QUEUE_STATE_PENDING_ENABLE (1 << 2)
#define EXEC_QUEUE_STATE_PENDING_DISABLE (1 << 3)
#define EXEC_QUEUE_STATE_DESTROYED (1 << 4)
#define ENGINE_STATE_SUSPENDED (1 << 5)
#define EXEC_QUEUE_STATE_RESET (1 << 6)
#define ENGINE_STATE_KILLED (1 << 7)
static bool exec_queue_registered(struct xe_exec_queue *q)
{
return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_REGISTERED;
}
static void set_exec_queue_registered(struct xe_exec_queue *q)
{
atomic_or(EXEC_QUEUE_STATE_REGISTERED, &q->guc->state);
}
static void clear_exec_queue_registered(struct xe_exec_queue *q)
{
atomic_and(~EXEC_QUEUE_STATE_REGISTERED, &q->guc->state);
}
static bool exec_queue_enabled(struct xe_exec_queue *q)
{
return atomic_read(&q->guc->state) & ENGINE_STATE_ENABLED;
}
static void set_exec_queue_enabled(struct xe_exec_queue *q)
{
atomic_or(ENGINE_STATE_ENABLED, &q->guc->state);
}
static void clear_exec_queue_enabled(struct xe_exec_queue *q)
{
atomic_and(~ENGINE_STATE_ENABLED, &q->guc->state);
}
static bool exec_queue_pending_enable(struct xe_exec_queue *q)
{
return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_PENDING_ENABLE;
}
static void set_exec_queue_pending_enable(struct xe_exec_queue *q)
{
atomic_or(EXEC_QUEUE_STATE_PENDING_ENABLE, &q->guc->state);
}
static void clear_exec_queue_pending_enable(struct xe_exec_queue *q)
{
atomic_and(~EXEC_QUEUE_STATE_PENDING_ENABLE, &q->guc->state);
}
static bool exec_queue_pending_disable(struct xe_exec_queue *q)
{
return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_PENDING_DISABLE;
}
static void set_exec_queue_pending_disable(struct xe_exec_queue *q)
{
atomic_or(EXEC_QUEUE_STATE_PENDING_DISABLE, &q->guc->state);
}
static void clear_exec_queue_pending_disable(struct xe_exec_queue *q)
{
atomic_and(~EXEC_QUEUE_STATE_PENDING_DISABLE, &q->guc->state);
}
static bool exec_queue_destroyed(struct xe_exec_queue *q)
{
return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_DESTROYED;
}
static void set_exec_queue_destroyed(struct xe_exec_queue *q)
{
atomic_or(EXEC_QUEUE_STATE_DESTROYED, &q->guc->state);
}
static bool exec_queue_banned(struct xe_exec_queue *q)
{
return (q->flags & EXEC_QUEUE_FLAG_BANNED);
}
static void set_exec_queue_banned(struct xe_exec_queue *q)
{
q->flags |= EXEC_QUEUE_FLAG_BANNED;
}
static bool exec_queue_suspended(struct xe_exec_queue *q)
{
return atomic_read(&q->guc->state) & ENGINE_STATE_SUSPENDED;
}
static void set_exec_queue_suspended(struct xe_exec_queue *q)
{
atomic_or(ENGINE_STATE_SUSPENDED, &q->guc->state);
}
static void clear_exec_queue_suspended(struct xe_exec_queue *q)
{
atomic_and(~ENGINE_STATE_SUSPENDED, &q->guc->state);
}
static bool exec_queue_reset(struct xe_exec_queue *q)
{
return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_RESET;
}
static void set_exec_queue_reset(struct xe_exec_queue *q)
{
atomic_or(EXEC_QUEUE_STATE_RESET, &q->guc->state);
}
static bool exec_queue_killed(struct xe_exec_queue *q)
{
return atomic_read(&q->guc->state) & ENGINE_STATE_KILLED;
}
static void set_exec_queue_killed(struct xe_exec_queue *q)
{
atomic_or(ENGINE_STATE_KILLED, &q->guc->state);
}
static bool exec_queue_killed_or_banned(struct xe_exec_queue *q)
{
return exec_queue_killed(q) || exec_queue_banned(q);
}
#ifdef CONFIG_PROVE_LOCKING
static int alloc_submit_wq(struct xe_guc *guc)
{
int i;
for (i = 0; i < NUM_SUBMIT_WQ; ++i) {
guc->submission_state.submit_wq_pool[i] =
alloc_ordered_workqueue("submit_wq", 0);
if (!guc->submission_state.submit_wq_pool[i])
goto err_free;
}
return 0;
err_free:
while (i)
destroy_workqueue(guc->submission_state.submit_wq_pool[--i]);
return -ENOMEM;
}
static void free_submit_wq(struct xe_guc *guc)
{
int i;
for (i = 0; i < NUM_SUBMIT_WQ; ++i)
destroy_workqueue(guc->submission_state.submit_wq_pool[i]);
}
static struct workqueue_struct *get_submit_wq(struct xe_guc *guc)
{
int idx = guc->submission_state.submit_wq_idx++ % NUM_SUBMIT_WQ;
return guc->submission_state.submit_wq_pool[idx];
}
#else
static int alloc_submit_wq(struct xe_guc *guc)
{
return 0;
}
static void free_submit_wq(struct xe_guc *guc)
{
}
static struct workqueue_struct *get_submit_wq(struct xe_guc *guc)
{
return NULL;
}
#endif
static void guc_submit_fini(struct drm_device *drm, void *arg)
{
struct xe_guc *guc = arg;
xa_destroy(&guc->submission_state.exec_queue_lookup);
ida_destroy(&guc->submission_state.guc_ids);
bitmap_free(guc->submission_state.guc_ids_bitmap);
free_submit_wq(guc);
mutex_destroy(&guc->submission_state.lock);
}
#define GUC_ID_MAX 65535
#define GUC_ID_NUMBER_MLRC 4096
#define GUC_ID_NUMBER_SLRC (GUC_ID_MAX - GUC_ID_NUMBER_MLRC)
#define GUC_ID_START_MLRC GUC_ID_NUMBER_SLRC
static const struct xe_exec_queue_ops guc_exec_queue_ops;
static void primelockdep(struct xe_guc *guc)
{
if (!IS_ENABLED(CONFIG_LOCKDEP))
return;
fs_reclaim_acquire(GFP_KERNEL);
mutex_lock(&guc->submission_state.lock);
might_lock(&guc->submission_state.suspend.lock);
mutex_unlock(&guc->submission_state.lock);
fs_reclaim_release(GFP_KERNEL);
}
int xe_guc_submit_init(struct xe_guc *guc)
{
struct xe_device *xe = guc_to_xe(guc);
struct xe_gt *gt = guc_to_gt(guc);
int err;
guc->submission_state.guc_ids_bitmap =
bitmap_zalloc(GUC_ID_NUMBER_MLRC, GFP_KERNEL);
if (!guc->submission_state.guc_ids_bitmap)
return -ENOMEM;
err = alloc_submit_wq(guc);
if (err) {
bitmap_free(guc->submission_state.guc_ids_bitmap);
return err;
}
gt->exec_queue_ops = &guc_exec_queue_ops;
mutex_init(&guc->submission_state.lock);
xa_init(&guc->submission_state.exec_queue_lookup);
ida_init(&guc->submission_state.guc_ids);
spin_lock_init(&guc->submission_state.suspend.lock);
guc->submission_state.suspend.context = dma_fence_context_alloc(1);
primelockdep(guc);
err = drmm_add_action_or_reset(&xe->drm, guc_submit_fini, guc);
if (err)
return err;
return 0;
}
static void __release_guc_id(struct xe_guc *guc, struct xe_exec_queue *q, u32 xa_count)
{
int i;
lockdep_assert_held(&guc->submission_state.lock);
for (i = 0; i < xa_count; ++i)
xa_erase(&guc->submission_state.exec_queue_lookup, q->guc->id + i);
if (xe_exec_queue_is_parallel(q))
bitmap_release_region(guc->submission_state.guc_ids_bitmap,
q->guc->id - GUC_ID_START_MLRC,
order_base_2(q->width));
else
ida_simple_remove(&guc->submission_state.guc_ids, q->guc->id);
}
static int alloc_guc_id(struct xe_guc *guc, struct xe_exec_queue *q)
{
int ret;
void *ptr;
int i;
/*
* Must use GFP_NOWAIT as this lock is in the dma fence signalling path,
* worse case user gets -ENOMEM on engine create and has to try again.
*
* FIXME: Have caller pre-alloc or post-alloc /w GFP_KERNEL to prevent
* failure.
*/
lockdep_assert_held(&guc->submission_state.lock);
if (xe_exec_queue_is_parallel(q)) {
void *bitmap = guc->submission_state.guc_ids_bitmap;
ret = bitmap_find_free_region(bitmap, GUC_ID_NUMBER_MLRC,
order_base_2(q->width));
} else {
ret = ida_simple_get(&guc->submission_state.guc_ids, 0,
GUC_ID_NUMBER_SLRC, GFP_NOWAIT);
}
if (ret < 0)
return ret;
q->guc->id = ret;
if (xe_exec_queue_is_parallel(q))
q->guc->id += GUC_ID_START_MLRC;
for (i = 0; i < q->width; ++i) {
ptr = xa_store(&guc->submission_state.exec_queue_lookup,
q->guc->id + i, q, GFP_NOWAIT);
if (IS_ERR(ptr)) {
ret = PTR_ERR(ptr);
goto err_release;
}
}
return 0;
err_release:
__release_guc_id(guc, q, i);
return ret;
}
static void release_guc_id(struct xe_guc *guc, struct xe_exec_queue *q)
{
mutex_lock(&guc->submission_state.lock);
__release_guc_id(guc, q, q->width);
mutex_unlock(&guc->submission_state.lock);
}
struct exec_queue_policy {
u32 count;
struct guc_update_exec_queue_policy h2g;
};
static u32 __guc_exec_queue_policy_action_size(struct exec_queue_policy *policy)
{
size_t bytes = sizeof(policy->h2g.header) +
(sizeof(policy->h2g.klv[0]) * policy->count);
return bytes / sizeof(u32);
}
static void __guc_exec_queue_policy_start_klv(struct exec_queue_policy *policy,
u16 guc_id)
{
policy->h2g.header.action =
XE_GUC_ACTION_HOST2GUC_UPDATE_CONTEXT_POLICIES;
policy->h2g.header.guc_id = guc_id;
policy->count = 0;
}
#define MAKE_EXEC_QUEUE_POLICY_ADD(func, id) \
static void __guc_exec_queue_policy_add_##func(struct exec_queue_policy *policy, \
u32 data) \
{ \
XE_WARN_ON(policy->count >= GUC_CONTEXT_POLICIES_KLV_NUM_IDS); \
\
policy->h2g.klv[policy->count].kl = \
FIELD_PREP(GUC_KLV_0_KEY, \
GUC_CONTEXT_POLICIES_KLV_ID_##id) | \
FIELD_PREP(GUC_KLV_0_LEN, 1); \
policy->h2g.klv[policy->count].value = data; \
policy->count++; \
}
MAKE_EXEC_QUEUE_POLICY_ADD(execution_quantum, EXECUTION_QUANTUM)
MAKE_EXEC_QUEUE_POLICY_ADD(preemption_timeout, PREEMPTION_TIMEOUT)
MAKE_EXEC_QUEUE_POLICY_ADD(priority, SCHEDULING_PRIORITY)
#undef MAKE_EXEC_QUEUE_POLICY_ADD
static const int xe_exec_queue_prio_to_guc[] = {
[XE_EXEC_QUEUE_PRIORITY_LOW] = GUC_CLIENT_PRIORITY_NORMAL,
[XE_EXEC_QUEUE_PRIORITY_NORMAL] = GUC_CLIENT_PRIORITY_KMD_NORMAL,
[XE_EXEC_QUEUE_PRIORITY_HIGH] = GUC_CLIENT_PRIORITY_HIGH,
[XE_EXEC_QUEUE_PRIORITY_KERNEL] = GUC_CLIENT_PRIORITY_KMD_HIGH,
};
static void init_policies(struct xe_guc *guc, struct xe_exec_queue *q)
{
struct exec_queue_policy policy;
struct xe_device *xe = guc_to_xe(guc);
enum xe_exec_queue_priority prio = q->sched_props.priority;
u32 timeslice_us = q->sched_props.timeslice_us;
u32 preempt_timeout_us = q->sched_props.preempt_timeout_us;
xe_assert(xe, exec_queue_registered(q));
__guc_exec_queue_policy_start_klv(&policy, q->guc->id);
__guc_exec_queue_policy_add_priority(&policy, xe_exec_queue_prio_to_guc[prio]);
__guc_exec_queue_policy_add_execution_quantum(&policy, timeslice_us);
__guc_exec_queue_policy_add_preemption_timeout(&policy, preempt_timeout_us);
xe_guc_ct_send(&guc->ct, (u32 *)&policy.h2g,
__guc_exec_queue_policy_action_size(&policy), 0, 0);
}
static void set_min_preemption_timeout(struct xe_guc *guc, struct xe_exec_queue *q)
{
struct exec_queue_policy policy;
__guc_exec_queue_policy_start_klv(&policy, q->guc->id);
__guc_exec_queue_policy_add_preemption_timeout(&policy, 1);
xe_guc_ct_send(&guc->ct, (u32 *)&policy.h2g,
__guc_exec_queue_policy_action_size(&policy), 0, 0);
}
#define parallel_read(xe_, map_, field_) \
xe_map_rd_field(xe_, &map_, 0, struct guc_submit_parallel_scratch, \
field_)
#define parallel_write(xe_, map_, field_, val_) \
xe_map_wr_field(xe_, &map_, 0, struct guc_submit_parallel_scratch, \
field_, val_)
static void __register_mlrc_engine(struct xe_guc *guc,
struct xe_exec_queue *q,
struct guc_ctxt_registration_info *info)
{
#define MAX_MLRC_REG_SIZE (13 + XE_HW_ENGINE_MAX_INSTANCE * 2)
struct xe_device *xe = guc_to_xe(guc);
u32 action[MAX_MLRC_REG_SIZE];
int len = 0;
int i;
xe_assert(xe, xe_exec_queue_is_parallel(q));
action[len++] = XE_GUC_ACTION_REGISTER_CONTEXT_MULTI_LRC;
action[len++] = info->flags;
action[len++] = info->context_idx;
action[len++] = info->engine_class;
action[len++] = info->engine_submit_mask;
action[len++] = info->wq_desc_lo;
action[len++] = info->wq_desc_hi;
action[len++] = info->wq_base_lo;
action[len++] = info->wq_base_hi;
action[len++] = info->wq_size;
action[len++] = q->width;
action[len++] = info->hwlrca_lo;
action[len++] = info->hwlrca_hi;
for (i = 1; i < q->width; ++i) {
struct xe_lrc *lrc = q->lrc + i;
action[len++] = lower_32_bits(xe_lrc_descriptor(lrc));
action[len++] = upper_32_bits(xe_lrc_descriptor(lrc));
}
xe_assert(xe, len <= MAX_MLRC_REG_SIZE);
#undef MAX_MLRC_REG_SIZE
xe_guc_ct_send(&guc->ct, action, len, 0, 0);
}
static void __register_engine(struct xe_guc *guc,
struct guc_ctxt_registration_info *info)
{
u32 action[] = {
XE_GUC_ACTION_REGISTER_CONTEXT,
info->flags,
info->context_idx,
info->engine_class,
info->engine_submit_mask,
info->wq_desc_lo,
info->wq_desc_hi,
info->wq_base_lo,
info->wq_base_hi,
info->wq_size,
info->hwlrca_lo,
info->hwlrca_hi,
};
xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action), 0, 0);
}
static void register_engine(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct xe_lrc *lrc = q->lrc;
struct guc_ctxt_registration_info info;
xe_assert(xe, !exec_queue_registered(q));
memset(&info, 0, sizeof(info));
info.context_idx = q->guc->id;
info.engine_class = xe_engine_class_to_guc_class(q->class);
info.engine_submit_mask = q->logical_mask;
info.hwlrca_lo = lower_32_bits(xe_lrc_descriptor(lrc));
info.hwlrca_hi = upper_32_bits(xe_lrc_descriptor(lrc));
info.flags = CONTEXT_REGISTRATION_FLAG_KMD;
if (xe_exec_queue_is_parallel(q)) {
u32 ggtt_addr = xe_lrc_parallel_ggtt_addr(lrc);
struct iosys_map map = xe_lrc_parallel_map(lrc);
info.wq_desc_lo = lower_32_bits(ggtt_addr +
offsetof(struct guc_submit_parallel_scratch, wq_desc));
info.wq_desc_hi = upper_32_bits(ggtt_addr +
offsetof(struct guc_submit_parallel_scratch, wq_desc));
info.wq_base_lo = lower_32_bits(ggtt_addr +
offsetof(struct guc_submit_parallel_scratch, wq[0]));
info.wq_base_hi = upper_32_bits(ggtt_addr +
offsetof(struct guc_submit_parallel_scratch, wq[0]));
info.wq_size = WQ_SIZE;
q->guc->wqi_head = 0;
q->guc->wqi_tail = 0;
xe_map_memset(xe, &map, 0, 0, PARALLEL_SCRATCH_SIZE - WQ_SIZE);
parallel_write(xe, map, wq_desc.wq_status, WQ_STATUS_ACTIVE);
}
/*
* We must keep a reference for LR engines if engine is registered with
* the GuC as jobs signal immediately and can't destroy an engine if the
* GuC has a reference to it.
*/
if (xe_exec_queue_is_lr(q))
xe_exec_queue_get(q);
set_exec_queue_registered(q);
trace_xe_exec_queue_register(q);
if (xe_exec_queue_is_parallel(q))
__register_mlrc_engine(guc, q, &info);
else
__register_engine(guc, &info);
init_policies(guc, q);
}
static u32 wq_space_until_wrap(struct xe_exec_queue *q)
{
return (WQ_SIZE - q->guc->wqi_tail);
}
static int wq_wait_for_space(struct xe_exec_queue *q, u32 wqi_size)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct iosys_map map = xe_lrc_parallel_map(q->lrc);
unsigned int sleep_period_ms = 1;
#define AVAILABLE_SPACE \
CIRC_SPACE(q->guc->wqi_tail, q->guc->wqi_head, WQ_SIZE)
if (wqi_size > AVAILABLE_SPACE) {
try_again:
q->guc->wqi_head = parallel_read(xe, map, wq_desc.head);
if (wqi_size > AVAILABLE_SPACE) {
if (sleep_period_ms == 1024) {
xe_gt_reset_async(q->gt);
return -ENODEV;
}
msleep(sleep_period_ms);
sleep_period_ms <<= 1;
goto try_again;
}
}
#undef AVAILABLE_SPACE
return 0;
}
static int wq_noop_append(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct iosys_map map = xe_lrc_parallel_map(q->lrc);
u32 len_dw = wq_space_until_wrap(q) / sizeof(u32) - 1;
if (wq_wait_for_space(q, wq_space_until_wrap(q)))
return -ENODEV;
xe_assert(xe, FIELD_FIT(WQ_LEN_MASK, len_dw));
parallel_write(xe, map, wq[q->guc->wqi_tail / sizeof(u32)],
FIELD_PREP(WQ_TYPE_MASK, WQ_TYPE_NOOP) |
FIELD_PREP(WQ_LEN_MASK, len_dw));
q->guc->wqi_tail = 0;
return 0;
}
static void wq_item_append(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct iosys_map map = xe_lrc_parallel_map(q->lrc);
#define WQ_HEADER_SIZE 4 /* Includes 1 LRC address too */
u32 wqi[XE_HW_ENGINE_MAX_INSTANCE + (WQ_HEADER_SIZE - 1)];
u32 wqi_size = (q->width + (WQ_HEADER_SIZE - 1)) * sizeof(u32);
u32 len_dw = (wqi_size / sizeof(u32)) - 1;
int i = 0, j;
if (wqi_size > wq_space_until_wrap(q)) {
if (wq_noop_append(q))
return;
}
if (wq_wait_for_space(q, wqi_size))
return;
wqi[i++] = FIELD_PREP(WQ_TYPE_MASK, WQ_TYPE_MULTI_LRC) |
FIELD_PREP(WQ_LEN_MASK, len_dw);
wqi[i++] = xe_lrc_descriptor(q->lrc);
wqi[i++] = FIELD_PREP(WQ_GUC_ID_MASK, q->guc->id) |
FIELD_PREP(WQ_RING_TAIL_MASK, q->lrc->ring.tail / sizeof(u64));
wqi[i++] = 0;
for (j = 1; j < q->width; ++j) {
struct xe_lrc *lrc = q->lrc + j;
wqi[i++] = lrc->ring.tail / sizeof(u64);
}
xe_assert(xe, i == wqi_size / sizeof(u32));
iosys_map_incr(&map, offsetof(struct guc_submit_parallel_scratch,
wq[q->guc->wqi_tail / sizeof(u32)]));
xe_map_memcpy_to(xe, &map, 0, wqi, wqi_size);
q->guc->wqi_tail += wqi_size;
xe_assert(xe, q->guc->wqi_tail <= WQ_SIZE);
xe_device_wmb(xe);
map = xe_lrc_parallel_map(q->lrc);
parallel_write(xe, map, wq_desc.tail, q->guc->wqi_tail);
}
#define RESUME_PENDING ~0x0ull
static void submit_exec_queue(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct xe_lrc *lrc = q->lrc;
u32 action[3];
u32 g2h_len = 0;
u32 num_g2h = 0;
int len = 0;
bool extra_submit = false;
xe_assert(xe, exec_queue_registered(q));
if (xe_exec_queue_is_parallel(q))
wq_item_append(q);
else
xe_lrc_write_ctx_reg(lrc, CTX_RING_TAIL, lrc->ring.tail);
if (exec_queue_suspended(q) && !xe_exec_queue_is_parallel(q))
return;
if (!exec_queue_enabled(q) && !exec_queue_suspended(q)) {
action[len++] = XE_GUC_ACTION_SCHED_CONTEXT_MODE_SET;
action[len++] = q->guc->id;
action[len++] = GUC_CONTEXT_ENABLE;
g2h_len = G2H_LEN_DW_SCHED_CONTEXT_MODE_SET;
num_g2h = 1;
if (xe_exec_queue_is_parallel(q))
extra_submit = true;
q->guc->resume_time = RESUME_PENDING;
set_exec_queue_pending_enable(q);
set_exec_queue_enabled(q);
trace_xe_exec_queue_scheduling_enable(q);
} else {
action[len++] = XE_GUC_ACTION_SCHED_CONTEXT;
action[len++] = q->guc->id;
trace_xe_exec_queue_submit(q);
}
xe_guc_ct_send(&guc->ct, action, len, g2h_len, num_g2h);
if (extra_submit) {
len = 0;
action[len++] = XE_GUC_ACTION_SCHED_CONTEXT;
action[len++] = q->guc->id;
trace_xe_exec_queue_submit(q);
xe_guc_ct_send(&guc->ct, action, len, 0, 0);
}
}
static struct dma_fence *
guc_exec_queue_run_job(struct drm_sched_job *drm_job)
{
struct xe_sched_job *job = to_xe_sched_job(drm_job);
struct xe_exec_queue *q = job->q;
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
bool lr = xe_exec_queue_is_lr(q);
xe_assert(xe, !(exec_queue_destroyed(q) || exec_queue_pending_disable(q)) ||
exec_queue_banned(q) || exec_queue_suspended(q));
trace_xe_sched_job_run(job);
if (!exec_queue_killed_or_banned(q) && !xe_sched_job_is_error(job)) {
if (!exec_queue_registered(q))
register_engine(q);
if (!lr) /* LR jobs are emitted in the exec IOCTL */
q->ring_ops->emit_job(job);
submit_exec_queue(q);
}
if (lr) {
xe_sched_job_set_error(job, -EOPNOTSUPP);
return NULL;
} else if (test_and_set_bit(JOB_FLAG_SUBMIT, &job->fence->flags)) {
return job->fence;
} else {
return dma_fence_get(job->fence);
}
}
static void guc_exec_queue_free_job(struct drm_sched_job *drm_job)
{
struct xe_sched_job *job = to_xe_sched_job(drm_job);
trace_xe_sched_job_free(job);
xe_sched_job_put(job);
}
static int guc_read_stopped(struct xe_guc *guc)
{
return atomic_read(&guc->submission_state.stopped);
}
#define MAKE_SCHED_CONTEXT_ACTION(q, enable_disable) \
u32 action[] = { \
XE_GUC_ACTION_SCHED_CONTEXT_MODE_SET, \
q->guc->id, \
GUC_CONTEXT_##enable_disable, \
}
static void disable_scheduling_deregister(struct xe_guc *guc,
struct xe_exec_queue *q)
{
MAKE_SCHED_CONTEXT_ACTION(q, DISABLE);
struct xe_device *xe = guc_to_xe(guc);
int ret;
set_min_preemption_timeout(guc, q);
smp_rmb();
ret = wait_event_timeout(guc->ct.wq, !exec_queue_pending_enable(q) ||
guc_read_stopped(guc), HZ * 5);
if (!ret) {
struct xe_gpu_scheduler *sched = &q->guc->sched;
drm_warn(&xe->drm, "Pending enable failed to respond");
xe_sched_submission_start(sched);
xe_gt_reset_async(q->gt);
xe_sched_tdr_queue_imm(sched);
return;
}
clear_exec_queue_enabled(q);
set_exec_queue_pending_disable(q);
set_exec_queue_destroyed(q);
trace_xe_exec_queue_scheduling_disable(q);
/*
* Reserve space for both G2H here as the 2nd G2H is sent from a G2H
* handler and we are not allowed to reserved G2H space in handlers.
*/
xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
G2H_LEN_DW_SCHED_CONTEXT_MODE_SET +
G2H_LEN_DW_DEREGISTER_CONTEXT, 2);
}
static void guc_exec_queue_print(struct xe_exec_queue *q, struct drm_printer *p);
#if IS_ENABLED(CONFIG_DRM_XE_SIMPLE_ERROR_CAPTURE)
static void simple_error_capture(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct drm_printer p = drm_err_printer("");
struct xe_hw_engine *hwe;
enum xe_hw_engine_id id;
u32 adj_logical_mask = q->logical_mask;
u32 width_mask = (0x1 << q->width) - 1;
int i;
bool cookie;
if (q->vm && !q->vm->error_capture.capture_once) {
q->vm->error_capture.capture_once = true;
cookie = dma_fence_begin_signalling();
for (i = 0; q->width > 1 && i < XE_HW_ENGINE_MAX_INSTANCE;) {
if (adj_logical_mask & BIT(i)) {
adj_logical_mask |= width_mask << i;
i += q->width;
} else {
++i;
}
}
xe_force_wake_get(gt_to_fw(guc_to_gt(guc)), XE_FORCEWAKE_ALL);
xe_guc_ct_print(&guc->ct, &p, true);
guc_exec_queue_print(q, &p);
for_each_hw_engine(hwe, guc_to_gt(guc), id) {
if (hwe->class != q->hwe->class ||
!(BIT(hwe->logical_instance) & adj_logical_mask))
continue;
xe_hw_engine_print(hwe, &p);
}
xe_analyze_vm(&p, q->vm, q->gt->info.id);
xe_force_wake_put(gt_to_fw(guc_to_gt(guc)), XE_FORCEWAKE_ALL);
dma_fence_end_signalling(cookie);
}
}
#else
static void simple_error_capture(struct xe_exec_queue *q)
{
}
#endif
static void xe_guc_exec_queue_trigger_cleanup(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
/** to wakeup xe_wait_user_fence ioctl if exec queue is reset */
wake_up_all(&xe->ufence_wq);
if (xe_exec_queue_is_lr(q))
queue_work(guc_to_gt(guc)->ordered_wq, &q->guc->lr_tdr);
else
xe_sched_tdr_queue_imm(&q->guc->sched);
}
static void xe_guc_exec_queue_lr_cleanup(struct work_struct *w)
{
struct xe_guc_exec_queue *ge =
container_of(w, struct xe_guc_exec_queue, lr_tdr);
struct xe_exec_queue *q = ge->q;
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct xe_gpu_scheduler *sched = &ge->sched;
xe_assert(xe, xe_exec_queue_is_lr(q));
trace_xe_exec_queue_lr_cleanup(q);
/* Kill the run_job / process_msg entry points */
xe_sched_submission_stop(sched);
/*
* Engine state now mostly stable, disable scheduling / deregister if
* needed. This cleanup routine might be called multiple times, where
* the actual async engine deregister drops the final engine ref.
* Calling disable_scheduling_deregister will mark the engine as
* destroyed and fire off the CT requests to disable scheduling /
* deregister, which we only want to do once. We also don't want to mark
* the engine as pending_disable again as this may race with the
* xe_guc_deregister_done_handler() which treats it as an unexpected
* state.
*/
if (exec_queue_registered(q) && !exec_queue_destroyed(q)) {
struct xe_guc *guc = exec_queue_to_guc(q);
int ret;
set_exec_queue_banned(q);
disable_scheduling_deregister(guc, q);
/*
* Must wait for scheduling to be disabled before signalling
* any fences, if GT broken the GT reset code should signal us.
*/
ret = wait_event_timeout(guc->ct.wq,
!exec_queue_pending_disable(q) ||
guc_read_stopped(guc), HZ * 5);
if (!ret) {
drm_warn(&xe->drm, "Schedule disable failed to respond");
xe_sched_submission_start(sched);
xe_gt_reset_async(q->gt);
return;
}
}
xe_sched_submission_start(sched);
}
static enum drm_gpu_sched_stat
guc_exec_queue_timedout_job(struct drm_sched_job *drm_job)
{
struct xe_sched_job *job = to_xe_sched_job(drm_job);
struct xe_sched_job *tmp_job;
struct xe_exec_queue *q = job->q;
struct xe_gpu_scheduler *sched = &q->guc->sched;
struct xe_device *xe = guc_to_xe(exec_queue_to_guc(q));
int err = -ETIME;
int i = 0;
if (!test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &job->fence->flags)) {
xe_assert(xe, !(q->flags & EXEC_QUEUE_FLAG_KERNEL));
xe_assert(xe, !(q->flags & EXEC_QUEUE_FLAG_VM && !exec_queue_killed(q)));
drm_notice(&xe->drm, "Timedout job: seqno=%u, guc_id=%d, flags=0x%lx",
xe_sched_job_seqno(job), q->guc->id, q->flags);
simple_error_capture(q);
xe_devcoredump(q);
} else {
drm_dbg(&xe->drm, "Timedout signaled job: seqno=%u, guc_id=%d, flags=0x%lx",
xe_sched_job_seqno(job), q->guc->id, q->flags);
}
trace_xe_sched_job_timedout(job);
/* Kill the run_job entry point */
xe_sched_submission_stop(sched);
/*
* Kernel jobs should never fail, nor should VM jobs if they do
* somethings has gone wrong and the GT needs a reset
*/
if (q->flags & EXEC_QUEUE_FLAG_KERNEL ||
(q->flags & EXEC_QUEUE_FLAG_VM && !exec_queue_killed(q))) {
if (!xe_sched_invalidate_job(job, 2)) {
xe_sched_add_pending_job(sched, job);
xe_sched_submission_start(sched);
xe_gt_reset_async(q->gt);
goto out;
}
}
/* Engine state now stable, disable scheduling if needed */
if (exec_queue_registered(q)) {
struct xe_guc *guc = exec_queue_to_guc(q);
int ret;
if (exec_queue_reset(q))
err = -EIO;
set_exec_queue_banned(q);
if (!exec_queue_destroyed(q)) {
xe_exec_queue_get(q);
disable_scheduling_deregister(guc, q);
}
/*
* Must wait for scheduling to be disabled before signalling
* any fences, if GT broken the GT reset code should signal us.
*
* FIXME: Tests can generate a ton of 0x6000 (IOMMU CAT fault
* error) messages which can cause the schedule disable to get
* lost. If this occurs, trigger a GT reset to recover.
*/
smp_rmb();
ret = wait_event_timeout(guc->ct.wq,
!exec_queue_pending_disable(q) ||
guc_read_stopped(guc), HZ * 5);
if (!ret || guc_read_stopped(guc)) {
drm_warn(&xe->drm, "Schedule disable failed to respond");
xe_sched_add_pending_job(sched, job);
xe_sched_submission_start(sched);
xe_gt_reset_async(q->gt);
xe_sched_tdr_queue_imm(sched);
goto out;
}
}
/* Stop fence signaling */
xe_hw_fence_irq_stop(q->fence_irq);
/*
* Fence state now stable, stop / start scheduler which cleans up any
* fences that are complete
*/
xe_sched_add_pending_job(sched, job);
xe_sched_submission_start(sched);
xe_guc_exec_queue_trigger_cleanup(q);
/* Mark all outstanding jobs as bad, thus completing them */
spin_lock(&sched->base.job_list_lock);
list_for_each_entry(tmp_job, &sched->base.pending_list, drm.list)
xe_sched_job_set_error(tmp_job, !i++ ? err : -ECANCELED);
spin_unlock(&sched->base.job_list_lock);
/* Start fence signaling */
xe_hw_fence_irq_start(q->fence_irq);
out:
return DRM_GPU_SCHED_STAT_NOMINAL;
}
static void __guc_exec_queue_fini_async(struct work_struct *w)
{
struct xe_guc_exec_queue *ge =
container_of(w, struct xe_guc_exec_queue, fini_async);
struct xe_exec_queue *q = ge->q;
struct xe_guc *guc = exec_queue_to_guc(q);
trace_xe_exec_queue_destroy(q);
if (xe_exec_queue_is_lr(q))
cancel_work_sync(&ge->lr_tdr);
if (q->flags & EXEC_QUEUE_FLAG_PERSISTENT)
xe_device_remove_persistent_exec_queues(gt_to_xe(q->gt), q);
release_guc_id(guc, q);
xe_sched_entity_fini(&ge->entity);
xe_sched_fini(&ge->sched);
kfree(ge);
xe_exec_queue_fini(q);
}
static void guc_exec_queue_fini_async(struct xe_exec_queue *q)
{
INIT_WORK(&q->guc->fini_async, __guc_exec_queue_fini_async);
/* We must block on kernel engines so slabs are empty on driver unload */
if (q->flags & EXEC_QUEUE_FLAG_PERMANENT)
__guc_exec_queue_fini_async(&q->guc->fini_async);
else
queue_work(system_wq, &q->guc->fini_async);
}
static void __guc_exec_queue_fini(struct xe_guc *guc, struct xe_exec_queue *q)
{
/*
* Might be done from within the GPU scheduler, need to do async as we
* fini the scheduler when the engine is fini'd, the scheduler can't
* complete fini within itself (circular dependency). Async resolves
* this we and don't really care when everything is fini'd, just that it
* is.
*/
guc_exec_queue_fini_async(q);
}
static void __guc_exec_queue_process_msg_cleanup(struct xe_sched_msg *msg)
{
struct xe_exec_queue *q = msg->private_data;
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
xe_assert(xe, !(q->flags & EXEC_QUEUE_FLAG_PERMANENT));
trace_xe_exec_queue_cleanup_entity(q);
if (exec_queue_registered(q))
disable_scheduling_deregister(guc, q);
else
__guc_exec_queue_fini(guc, q);
}
static bool guc_exec_queue_allowed_to_change_state(struct xe_exec_queue *q)
{
return !exec_queue_killed_or_banned(q) && exec_queue_registered(q);
}
static void __guc_exec_queue_process_msg_set_sched_props(struct xe_sched_msg *msg)
{
struct xe_exec_queue *q = msg->private_data;
struct xe_guc *guc = exec_queue_to_guc(q);
if (guc_exec_queue_allowed_to_change_state(q))
init_policies(guc, q);
kfree(msg);
}
static void suspend_fence_signal(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
xe_assert(xe, exec_queue_suspended(q) || exec_queue_killed(q) ||
guc_read_stopped(guc));
xe_assert(xe, q->guc->suspend_pending);
q->guc->suspend_pending = false;
smp_wmb();
wake_up(&q->guc->suspend_wait);
}
static void __guc_exec_queue_process_msg_suspend(struct xe_sched_msg *msg)
{
struct xe_exec_queue *q = msg->private_data;
struct xe_guc *guc = exec_queue_to_guc(q);
if (guc_exec_queue_allowed_to_change_state(q) && !exec_queue_suspended(q) &&
exec_queue_enabled(q)) {
wait_event(guc->ct.wq, q->guc->resume_time != RESUME_PENDING ||
guc_read_stopped(guc));
if (!guc_read_stopped(guc)) {
MAKE_SCHED_CONTEXT_ACTION(q, DISABLE);
s64 since_resume_ms =
ktime_ms_delta(ktime_get(),
q->guc->resume_time);
s64 wait_ms = q->vm->preempt.min_run_period_ms -
since_resume_ms;
if (wait_ms > 0 && q->guc->resume_time)
msleep(wait_ms);
set_exec_queue_suspended(q);
clear_exec_queue_enabled(q);
set_exec_queue_pending_disable(q);
trace_xe_exec_queue_scheduling_disable(q);
xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
G2H_LEN_DW_SCHED_CONTEXT_MODE_SET, 1);
}
} else if (q->guc->suspend_pending) {
set_exec_queue_suspended(q);
suspend_fence_signal(q);
}
}
static void __guc_exec_queue_process_msg_resume(struct xe_sched_msg *msg)
{
struct xe_exec_queue *q = msg->private_data;
struct xe_guc *guc = exec_queue_to_guc(q);
if (guc_exec_queue_allowed_to_change_state(q)) {
MAKE_SCHED_CONTEXT_ACTION(q, ENABLE);
q->guc->resume_time = RESUME_PENDING;
clear_exec_queue_suspended(q);
set_exec_queue_pending_enable(q);
set_exec_queue_enabled(q);
trace_xe_exec_queue_scheduling_enable(q);
xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action),
G2H_LEN_DW_SCHED_CONTEXT_MODE_SET, 1);
} else {
clear_exec_queue_suspended(q);
}
}
#define CLEANUP 1 /* Non-zero values to catch uninitialized msg */
#define SET_SCHED_PROPS 2
#define SUSPEND 3
#define RESUME 4
static void guc_exec_queue_process_msg(struct xe_sched_msg *msg)
{
trace_xe_sched_msg_recv(msg);
switch (msg->opcode) {
case CLEANUP:
__guc_exec_queue_process_msg_cleanup(msg);
break;
case SET_SCHED_PROPS:
__guc_exec_queue_process_msg_set_sched_props(msg);
break;
case SUSPEND:
__guc_exec_queue_process_msg_suspend(msg);
break;
case RESUME:
__guc_exec_queue_process_msg_resume(msg);
break;
default:
XE_WARN_ON("Unknown message type");
}
}
static const struct drm_sched_backend_ops drm_sched_ops = {
.run_job = guc_exec_queue_run_job,
.free_job = guc_exec_queue_free_job,
.timedout_job = guc_exec_queue_timedout_job,
};
static const struct xe_sched_backend_ops xe_sched_ops = {
.process_msg = guc_exec_queue_process_msg,
};
static int guc_exec_queue_init(struct xe_exec_queue *q)
{
struct xe_gpu_scheduler *sched;
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct xe_guc_exec_queue *ge;
long timeout;
int err;
xe_assert(xe, xe_device_uc_enabled(guc_to_xe(guc)));
ge = kzalloc(sizeof(*ge), GFP_KERNEL);
if (!ge)
return -ENOMEM;
q->guc = ge;
ge->q = q;
init_waitqueue_head(&ge->suspend_wait);
timeout = (q->vm && xe_vm_in_lr_mode(q->vm)) ? MAX_SCHEDULE_TIMEOUT :
q->hwe->eclass->sched_props.job_timeout_ms;
err = xe_sched_init(&ge->sched, &drm_sched_ops, &xe_sched_ops,
get_submit_wq(guc),
q->lrc[0].ring.size / MAX_JOB_SIZE_BYTES, 64,
timeout, guc_to_gt(guc)->ordered_wq, NULL,
q->name, gt_to_xe(q->gt)->drm.dev);
if (err)
goto err_free;
sched = &ge->sched;
err = xe_sched_entity_init(&ge->entity, sched);
if (err)
goto err_sched;
if (xe_exec_queue_is_lr(q))
INIT_WORK(&q->guc->lr_tdr, xe_guc_exec_queue_lr_cleanup);
mutex_lock(&guc->submission_state.lock);
err = alloc_guc_id(guc, q);
if (err)
goto err_entity;
q->entity = &ge->entity;
if (guc_read_stopped(guc))
xe_sched_stop(sched);
mutex_unlock(&guc->submission_state.lock);
xe_exec_queue_assign_name(q, q->guc->id);
trace_xe_exec_queue_create(q);
return 0;
err_entity:
xe_sched_entity_fini(&ge->entity);
err_sched:
xe_sched_fini(&ge->sched);
err_free:
kfree(ge);
return err;
}
static void guc_exec_queue_kill(struct xe_exec_queue *q)
{
trace_xe_exec_queue_kill(q);
set_exec_queue_killed(q);
xe_guc_exec_queue_trigger_cleanup(q);
}
static void guc_exec_queue_add_msg(struct xe_exec_queue *q, struct xe_sched_msg *msg,
u32 opcode)
{
INIT_LIST_HEAD(&msg->link);
msg->opcode = opcode;
msg->private_data = q;
trace_xe_sched_msg_add(msg);
xe_sched_add_msg(&q->guc->sched, msg);
}
#define STATIC_MSG_CLEANUP 0
#define STATIC_MSG_SUSPEND 1
#define STATIC_MSG_RESUME 2
static void guc_exec_queue_fini(struct xe_exec_queue *q)
{
struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_CLEANUP;
if (!(q->flags & EXEC_QUEUE_FLAG_PERMANENT))
guc_exec_queue_add_msg(q, msg, CLEANUP);
else
__guc_exec_queue_fini(exec_queue_to_guc(q), q);
}
static int guc_exec_queue_set_priority(struct xe_exec_queue *q,
enum xe_exec_queue_priority priority)
{
struct xe_sched_msg *msg;
if (q->sched_props.priority == priority || exec_queue_killed_or_banned(q))
return 0;
msg = kmalloc(sizeof(*msg), GFP_KERNEL);
if (!msg)
return -ENOMEM;
q->sched_props.priority = priority;
guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
return 0;
}
static int guc_exec_queue_set_timeslice(struct xe_exec_queue *q, u32 timeslice_us)
{
struct xe_sched_msg *msg;
if (q->sched_props.timeslice_us == timeslice_us ||
exec_queue_killed_or_banned(q))
return 0;
msg = kmalloc(sizeof(*msg), GFP_KERNEL);
if (!msg)
return -ENOMEM;
q->sched_props.timeslice_us = timeslice_us;
guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
return 0;
}
static int guc_exec_queue_set_preempt_timeout(struct xe_exec_queue *q,
u32 preempt_timeout_us)
{
struct xe_sched_msg *msg;
if (q->sched_props.preempt_timeout_us == preempt_timeout_us ||
exec_queue_killed_or_banned(q))
return 0;
msg = kmalloc(sizeof(*msg), GFP_KERNEL);
if (!msg)
return -ENOMEM;
q->sched_props.preempt_timeout_us = preempt_timeout_us;
guc_exec_queue_add_msg(q, msg, SET_SCHED_PROPS);
return 0;
}
static int guc_exec_queue_set_job_timeout(struct xe_exec_queue *q, u32 job_timeout_ms)
{
struct xe_gpu_scheduler *sched = &q->guc->sched;
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
xe_assert(xe, !exec_queue_registered(q));
xe_assert(xe, !exec_queue_banned(q));
xe_assert(xe, !exec_queue_killed(q));
sched->base.timeout = job_timeout_ms;
return 0;
}
static int guc_exec_queue_suspend(struct xe_exec_queue *q)
{
struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_SUSPEND;
if (exec_queue_killed_or_banned(q) || q->guc->suspend_pending)
return -EINVAL;
q->guc->suspend_pending = true;
guc_exec_queue_add_msg(q, msg, SUSPEND);
return 0;
}
static void guc_exec_queue_suspend_wait(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
wait_event(q->guc->suspend_wait, !q->guc->suspend_pending ||
guc_read_stopped(guc));
}
static void guc_exec_queue_resume(struct xe_exec_queue *q)
{
struct xe_sched_msg *msg = q->guc->static_msgs + STATIC_MSG_RESUME;
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
xe_assert(xe, !q->guc->suspend_pending);
guc_exec_queue_add_msg(q, msg, RESUME);
}
static bool guc_exec_queue_reset_status(struct xe_exec_queue *q)
{
return exec_queue_reset(q);
}
/*
* All of these functions are an abstraction layer which other parts of XE can
* use to trap into the GuC backend. All of these functions, aside from init,
* really shouldn't do much other than trap into the DRM scheduler which
* synchronizes these operations.
*/
static const struct xe_exec_queue_ops guc_exec_queue_ops = {
.init = guc_exec_queue_init,
.kill = guc_exec_queue_kill,
.fini = guc_exec_queue_fini,
.set_priority = guc_exec_queue_set_priority,
.set_timeslice = guc_exec_queue_set_timeslice,
.set_preempt_timeout = guc_exec_queue_set_preempt_timeout,
.set_job_timeout = guc_exec_queue_set_job_timeout,
.suspend = guc_exec_queue_suspend,
.suspend_wait = guc_exec_queue_suspend_wait,
.resume = guc_exec_queue_resume,
.reset_status = guc_exec_queue_reset_status,
};
static void guc_exec_queue_stop(struct xe_guc *guc, struct xe_exec_queue *q)
{
struct xe_gpu_scheduler *sched = &q->guc->sched;
/* Stop scheduling + flush any DRM scheduler operations */
xe_sched_submission_stop(sched);
/* Clean up lost G2H + reset engine state */
if (exec_queue_registered(q)) {
if ((exec_queue_banned(q) && exec_queue_destroyed(q)) ||
xe_exec_queue_is_lr(q))
xe_exec_queue_put(q);
else if (exec_queue_destroyed(q))
__guc_exec_queue_fini(guc, q);
}
if (q->guc->suspend_pending) {
set_exec_queue_suspended(q);
suspend_fence_signal(q);
}
atomic_and(EXEC_QUEUE_STATE_DESTROYED | ENGINE_STATE_SUSPENDED,
&q->guc->state);
q->guc->resume_time = 0;
trace_xe_exec_queue_stop(q);
/*
* Ban any engine (aside from kernel and engines used for VM ops) with a
* started but not complete job or if a job has gone through a GT reset
* more than twice.
*/
if (!(q->flags & (EXEC_QUEUE_FLAG_KERNEL | EXEC_QUEUE_FLAG_VM))) {
struct xe_sched_job *job = xe_sched_first_pending_job(sched);
if (job) {
if ((xe_sched_job_started(job) &&
!xe_sched_job_completed(job)) ||
xe_sched_invalidate_job(job, 2)) {
trace_xe_sched_job_ban(job);
xe_sched_tdr_queue_imm(&q->guc->sched);
set_exec_queue_banned(q);
}
}
}
}
int xe_guc_submit_reset_prepare(struct xe_guc *guc)
{
int ret;
/*
* Using an atomic here rather than submission_state.lock as this
* function can be called while holding the CT lock (engine reset
* failure). submission_state.lock needs the CT lock to resubmit jobs.
* Atomic is not ideal, but it works to prevent against concurrent reset
* and releasing any TDRs waiting on guc->submission_state.stopped.
*/
ret = atomic_fetch_or(1, &guc->submission_state.stopped);
smp_wmb();
wake_up_all(&guc->ct.wq);
return ret;
}
void xe_guc_submit_reset_wait(struct xe_guc *guc)
{
wait_event(guc->ct.wq, !guc_read_stopped(guc));
}
int xe_guc_submit_stop(struct xe_guc *guc)
{
struct xe_exec_queue *q;
unsigned long index;
struct xe_device *xe = guc_to_xe(guc);
xe_assert(xe, guc_read_stopped(guc) == 1);
mutex_lock(&guc->submission_state.lock);
xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
guc_exec_queue_stop(guc, q);
mutex_unlock(&guc->submission_state.lock);
/*
* No one can enter the backend at this point, aside from new engine
* creation which is protected by guc->submission_state.lock.
*/
return 0;
}
static void guc_exec_queue_start(struct xe_exec_queue *q)
{
struct xe_gpu_scheduler *sched = &q->guc->sched;
if (!exec_queue_killed_or_banned(q)) {
int i;
trace_xe_exec_queue_resubmit(q);
for (i = 0; i < q->width; ++i)
xe_lrc_set_ring_head(q->lrc + i, q->lrc[i].ring.tail);
xe_sched_resubmit_jobs(sched);
}
xe_sched_submission_start(sched);
}
int xe_guc_submit_start(struct xe_guc *guc)
{
struct xe_exec_queue *q;
unsigned long index;
struct xe_device *xe = guc_to_xe(guc);
xe_assert(xe, guc_read_stopped(guc) == 1);
mutex_lock(&guc->submission_state.lock);
atomic_dec(&guc->submission_state.stopped);
xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
guc_exec_queue_start(q);
mutex_unlock(&guc->submission_state.lock);
wake_up_all(&guc->ct.wq);
return 0;
}
static struct xe_exec_queue *
g2h_exec_queue_lookup(struct xe_guc *guc, u32 guc_id)
{
struct xe_device *xe = guc_to_xe(guc);
struct xe_exec_queue *q;
if (unlikely(guc_id >= GUC_ID_MAX)) {
drm_err(&xe->drm, "Invalid guc_id %u", guc_id);
return NULL;
}
q = xa_load(&guc->submission_state.exec_queue_lookup, guc_id);
if (unlikely(!q)) {
drm_err(&xe->drm, "Not engine present for guc_id %u", guc_id);
return NULL;
}
xe_assert(xe, guc_id >= q->guc->id);
xe_assert(xe, guc_id < (q->guc->id + q->width));
return q;
}
static void deregister_exec_queue(struct xe_guc *guc, struct xe_exec_queue *q)
{
u32 action[] = {
XE_GUC_ACTION_DEREGISTER_CONTEXT,
q->guc->id,
};
trace_xe_exec_queue_deregister(q);
xe_guc_ct_send_g2h_handler(&guc->ct, action, ARRAY_SIZE(action));
}
int xe_guc_sched_done_handler(struct xe_guc *guc, u32 *msg, u32 len)
{
struct xe_device *xe = guc_to_xe(guc);
struct xe_exec_queue *q;
u32 guc_id = msg[0];
if (unlikely(len < 2)) {
drm_err(&xe->drm, "Invalid length %u", len);
return -EPROTO;
}
q = g2h_exec_queue_lookup(guc, guc_id);
if (unlikely(!q))
return -EPROTO;
if (unlikely(!exec_queue_pending_enable(q) &&
!exec_queue_pending_disable(q))) {
drm_err(&xe->drm, "Unexpected engine state 0x%04x",
atomic_read(&q->guc->state));
return -EPROTO;
}
trace_xe_exec_queue_scheduling_done(q);
if (exec_queue_pending_enable(q)) {
q->guc->resume_time = ktime_get();
clear_exec_queue_pending_enable(q);
smp_wmb();
wake_up_all(&guc->ct.wq);
} else {
clear_exec_queue_pending_disable(q);
if (q->guc->suspend_pending) {
suspend_fence_signal(q);
} else {
if (exec_queue_banned(q)) {
smp_wmb();
wake_up_all(&guc->ct.wq);
}
deregister_exec_queue(guc, q);
}
}
return 0;
}
int xe_guc_deregister_done_handler(struct xe_guc *guc, u32 *msg, u32 len)
{
struct xe_device *xe = guc_to_xe(guc);
struct xe_exec_queue *q;
u32 guc_id = msg[0];
if (unlikely(len < 1)) {
drm_err(&xe->drm, "Invalid length %u", len);
return -EPROTO;
}
q = g2h_exec_queue_lookup(guc, guc_id);
if (unlikely(!q))
return -EPROTO;
if (!exec_queue_destroyed(q) || exec_queue_pending_disable(q) ||
exec_queue_pending_enable(q) || exec_queue_enabled(q)) {
drm_err(&xe->drm, "Unexpected engine state 0x%04x",
atomic_read(&q->guc->state));
return -EPROTO;
}
trace_xe_exec_queue_deregister_done(q);
clear_exec_queue_registered(q);
if (exec_queue_banned(q) || xe_exec_queue_is_lr(q))
xe_exec_queue_put(q);
else
__guc_exec_queue_fini(guc, q);
return 0;
}
int xe_guc_exec_queue_reset_handler(struct xe_guc *guc, u32 *msg, u32 len)
{
struct xe_device *xe = guc_to_xe(guc);
struct xe_exec_queue *q;
u32 guc_id = msg[0];
if (unlikely(len < 1)) {
drm_err(&xe->drm, "Invalid length %u", len);
return -EPROTO;
}
q = g2h_exec_queue_lookup(guc, guc_id);
if (unlikely(!q))
return -EPROTO;
drm_info(&xe->drm, "Engine reset: guc_id=%d", guc_id);
/* FIXME: Do error capture, most likely async */
trace_xe_exec_queue_reset(q);
/*
* A banned engine is a NOP at this point (came from
* guc_exec_queue_timedout_job). Otherwise, kick drm scheduler to cancel
* jobs by setting timeout of the job to the minimum value kicking
* guc_exec_queue_timedout_job.
*/
set_exec_queue_reset(q);
if (!exec_queue_banned(q))
xe_guc_exec_queue_trigger_cleanup(q);
return 0;
}
int xe_guc_exec_queue_memory_cat_error_handler(struct xe_guc *guc, u32 *msg,
u32 len)
{
struct xe_device *xe = guc_to_xe(guc);
struct xe_exec_queue *q;
u32 guc_id = msg[0];
if (unlikely(len < 1)) {
drm_err(&xe->drm, "Invalid length %u", len);
return -EPROTO;
}
q = g2h_exec_queue_lookup(guc, guc_id);
if (unlikely(!q))
return -EPROTO;
drm_dbg(&xe->drm, "Engine memory cat error: guc_id=%d", guc_id);
trace_xe_exec_queue_memory_cat_error(q);
/* Treat the same as engine reset */
set_exec_queue_reset(q);
if (!exec_queue_banned(q))
xe_guc_exec_queue_trigger_cleanup(q);
return 0;
}
int xe_guc_exec_queue_reset_failure_handler(struct xe_guc *guc, u32 *msg, u32 len)
{
struct xe_device *xe = guc_to_xe(guc);
u8 guc_class, instance;
u32 reason;
if (unlikely(len != 3)) {
drm_err(&xe->drm, "Invalid length %u", len);
return -EPROTO;
}
guc_class = msg[0];
instance = msg[1];
reason = msg[2];
/* Unexpected failure of a hardware feature, log an actual error */
drm_err(&xe->drm, "GuC engine reset request failed on %d:%d because 0x%08X",
guc_class, instance, reason);
xe_gt_reset_async(guc_to_gt(guc));
return 0;
}
static void
guc_exec_queue_wq_snapshot_capture(struct xe_exec_queue *q,
struct xe_guc_submit_exec_queue_snapshot *snapshot)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct iosys_map map = xe_lrc_parallel_map(q->lrc);
int i;
snapshot->guc.wqi_head = q->guc->wqi_head;
snapshot->guc.wqi_tail = q->guc->wqi_tail;
snapshot->parallel.wq_desc.head = parallel_read(xe, map, wq_desc.head);
snapshot->parallel.wq_desc.tail = parallel_read(xe, map, wq_desc.tail);
snapshot->parallel.wq_desc.status = parallel_read(xe, map,
wq_desc.wq_status);
if (snapshot->parallel.wq_desc.head !=
snapshot->parallel.wq_desc.tail) {
for (i = snapshot->parallel.wq_desc.head;
i != snapshot->parallel.wq_desc.tail;
i = (i + sizeof(u32)) % WQ_SIZE)
snapshot->parallel.wq[i / sizeof(u32)] =
parallel_read(xe, map, wq[i / sizeof(u32)]);
}
}
static void
guc_exec_queue_wq_snapshot_print(struct xe_guc_submit_exec_queue_snapshot *snapshot,
struct drm_printer *p)
{
int i;
drm_printf(p, "\tWQ head: %u (internal), %d (memory)\n",
snapshot->guc.wqi_head, snapshot->parallel.wq_desc.head);
drm_printf(p, "\tWQ tail: %u (internal), %d (memory)\n",
snapshot->guc.wqi_tail, snapshot->parallel.wq_desc.tail);
drm_printf(p, "\tWQ status: %u\n", snapshot->parallel.wq_desc.status);
if (snapshot->parallel.wq_desc.head !=
snapshot->parallel.wq_desc.tail) {
for (i = snapshot->parallel.wq_desc.head;
i != snapshot->parallel.wq_desc.tail;
i = (i + sizeof(u32)) % WQ_SIZE)
drm_printf(p, "\tWQ[%zu]: 0x%08x\n", i / sizeof(u32),
snapshot->parallel.wq[i / sizeof(u32)]);
}
}
/**
* xe_guc_exec_queue_snapshot_capture - Take a quick snapshot of the GuC Engine.
* @q: Xe exec queue.
*
* This can be printed out in a later stage like during dev_coredump
* analysis.
*
* Returns: a GuC Submit Engine snapshot object that must be freed by the
* caller, using `xe_guc_exec_queue_snapshot_free`.
*/
struct xe_guc_submit_exec_queue_snapshot *
xe_guc_exec_queue_snapshot_capture(struct xe_exec_queue *q)
{
struct xe_guc *guc = exec_queue_to_guc(q);
struct xe_device *xe = guc_to_xe(guc);
struct xe_gpu_scheduler *sched = &q->guc->sched;
struct xe_sched_job *job;
struct xe_guc_submit_exec_queue_snapshot *snapshot;
int i;
snapshot = kzalloc(sizeof(*snapshot), GFP_ATOMIC);
if (!snapshot) {
drm_err(&xe->drm, "Skipping GuC Engine snapshot entirely.\n");
return NULL;
}
snapshot->guc.id = q->guc->id;
memcpy(&snapshot->name, &q->name, sizeof(snapshot->name));
snapshot->class = q->class;
snapshot->logical_mask = q->logical_mask;
snapshot->width = q->width;
snapshot->refcount = kref_read(&q->refcount);
snapshot->sched_timeout = sched->base.timeout;
snapshot->sched_props.timeslice_us = q->sched_props.timeslice_us;
snapshot->sched_props.preempt_timeout_us =
q->sched_props.preempt_timeout_us;
snapshot->lrc = kmalloc_array(q->width, sizeof(struct lrc_snapshot),
GFP_ATOMIC);
if (!snapshot->lrc) {
drm_err(&xe->drm, "Skipping GuC Engine LRC snapshot.\n");
} else {
for (i = 0; i < q->width; ++i) {
struct xe_lrc *lrc = q->lrc + i;
snapshot->lrc[i].context_desc =
lower_32_bits(xe_lrc_ggtt_addr(lrc));
snapshot->lrc[i].head = xe_lrc_ring_head(lrc);
snapshot->lrc[i].tail.internal = lrc->ring.tail;
snapshot->lrc[i].tail.memory =
xe_lrc_read_ctx_reg(lrc, CTX_RING_TAIL);
snapshot->lrc[i].start_seqno = xe_lrc_start_seqno(lrc);
snapshot->lrc[i].seqno = xe_lrc_seqno(lrc);
}
}
snapshot->schedule_state = atomic_read(&q->guc->state);
snapshot->exec_queue_flags = q->flags;
snapshot->parallel_execution = xe_exec_queue_is_parallel(q);
if (snapshot->parallel_execution)
guc_exec_queue_wq_snapshot_capture(q, snapshot);
spin_lock(&sched->base.job_list_lock);
snapshot->pending_list_size = list_count_nodes(&sched->base.pending_list);
snapshot->pending_list = kmalloc_array(snapshot->pending_list_size,
sizeof(struct pending_list_snapshot),
GFP_ATOMIC);
if (!snapshot->pending_list) {
drm_err(&xe->drm, "Skipping GuC Engine pending_list snapshot.\n");
} else {
i = 0;
list_for_each_entry(job, &sched->base.pending_list, drm.list) {
snapshot->pending_list[i].seqno =
xe_sched_job_seqno(job);
snapshot->pending_list[i].fence =
dma_fence_is_signaled(job->fence) ? 1 : 0;
snapshot->pending_list[i].finished =
dma_fence_is_signaled(&job->drm.s_fence->finished)
? 1 : 0;
i++;
}
}
spin_unlock(&sched->base.job_list_lock);
return snapshot;
}
/**
* xe_guc_exec_queue_snapshot_print - Print out a given GuC Engine snapshot.
* @snapshot: GuC Submit Engine snapshot object.
* @p: drm_printer where it will be printed out.
*
* This function prints out a given GuC Submit Engine snapshot object.
*/
void
xe_guc_exec_queue_snapshot_print(struct xe_guc_submit_exec_queue_snapshot *snapshot,
struct drm_printer *p)
{
int i;
if (!snapshot)
return;
drm_printf(p, "\nGuC ID: %d\n", snapshot->guc.id);
drm_printf(p, "\tName: %s\n", snapshot->name);
drm_printf(p, "\tClass: %d\n", snapshot->class);
drm_printf(p, "\tLogical mask: 0x%x\n", snapshot->logical_mask);
drm_printf(p, "\tWidth: %d\n", snapshot->width);
drm_printf(p, "\tRef: %d\n", snapshot->refcount);
drm_printf(p, "\tTimeout: %ld (ms)\n", snapshot->sched_timeout);
drm_printf(p, "\tTimeslice: %u (us)\n",
snapshot->sched_props.timeslice_us);
drm_printf(p, "\tPreempt timeout: %u (us)\n",
snapshot->sched_props.preempt_timeout_us);
for (i = 0; snapshot->lrc && i < snapshot->width; ++i) {
drm_printf(p, "\tHW Context Desc: 0x%08x\n",
snapshot->lrc[i].context_desc);
drm_printf(p, "\tLRC Head: (memory) %u\n",
snapshot->lrc[i].head);
drm_printf(p, "\tLRC Tail: (internal) %u, (memory) %u\n",
snapshot->lrc[i].tail.internal,
snapshot->lrc[i].tail.memory);
drm_printf(p, "\tStart seqno: (memory) %d\n",
snapshot->lrc[i].start_seqno);
drm_printf(p, "\tSeqno: (memory) %d\n", snapshot->lrc[i].seqno);
}
drm_printf(p, "\tSchedule State: 0x%x\n", snapshot->schedule_state);
drm_printf(p, "\tFlags: 0x%lx\n", snapshot->exec_queue_flags);
if (snapshot->parallel_execution)
guc_exec_queue_wq_snapshot_print(snapshot, p);
for (i = 0; snapshot->pending_list && i < snapshot->pending_list_size;
i++)
drm_printf(p, "\tJob: seqno=%d, fence=%d, finished=%d\n",
snapshot->pending_list[i].seqno,
snapshot->pending_list[i].fence,
snapshot->pending_list[i].finished);
}
/**
* xe_guc_exec_queue_snapshot_free - Free all allocated objects for a given
* snapshot.
* @snapshot: GuC Submit Engine snapshot object.
*
* This function free all the memory that needed to be allocated at capture
* time.
*/
void xe_guc_exec_queue_snapshot_free(struct xe_guc_submit_exec_queue_snapshot *snapshot)
{
if (!snapshot)
return;
kfree(snapshot->lrc);
kfree(snapshot->pending_list);
kfree(snapshot);
}
static void guc_exec_queue_print(struct xe_exec_queue *q, struct drm_printer *p)
{
struct xe_guc_submit_exec_queue_snapshot *snapshot;
snapshot = xe_guc_exec_queue_snapshot_capture(q);
xe_guc_exec_queue_snapshot_print(snapshot, p);
xe_guc_exec_queue_snapshot_free(snapshot);
}
/**
* xe_guc_submit_print - GuC Submit Print.
* @guc: GuC.
* @p: drm_printer where it will be printed out.
*
* This function capture and prints snapshots of **all** GuC Engines.
*/
void xe_guc_submit_print(struct xe_guc *guc, struct drm_printer *p)
{
struct xe_exec_queue *q;
unsigned long index;
if (!xe_device_uc_enabled(guc_to_xe(guc)))
return;
mutex_lock(&guc->submission_state.lock);
xa_for_each(&guc->submission_state.exec_queue_lookup, index, q)
guc_exec_queue_print(q, p);
mutex_unlock(&guc->submission_state.lock);
}