linux_dsm_epyc7002/drivers/gpu/drm/i915/gt/intel_ring.h
Chris Wilson 32a55a109f drm/i915/gt: Mark up racy read of intel_ring.head
The intel_ring.head is updated as the requests are retired, but is
sampled at any time as we submit requests. Furthermore, it tracks
RING_HEAD which is inherently asynchronous.

[  148.630314] BUG: KCSAN: data-race in execlists_dequeue [i915] / i915_request_retire [i915]
[  148.630349]
[  148.630374] write to 0xffff8881f4e28ddc of 4 bytes by task 90 on cpu 2:
[  148.630752]  i915_request_retire+0xed/0x770 [i915]
[  148.631123]  retire_requests+0x7a/0xd0 [i915]
[  148.631491]  engine_retire+0xa6/0xe0 [i915]
[  148.631523]  process_one_work+0x3af/0x640
[  148.631552]  worker_thread+0x80/0x670
[  148.631581]  kthread+0x19a/0x1e0
[  148.631609]  ret_from_fork+0x1f/0x30
[  148.631629]
[  148.631652] read to 0xffff8881f4e28ddc of 4 bytes by task 14288 on cpu 3:
[  148.632019]  execlists_dequeue+0x1300/0x1680 [i915]
[  148.632384]  __execlists_submission_tasklet+0x48/0x60 [i915]
[  148.632770]  execlists_submit_request+0x38e/0x3c0 [i915]
[  148.633146]  submit_notify+0x8f/0xc0 [i915]
[  148.633512]  __i915_sw_fence_complete+0x5d/0x3e0 [i915]
[  148.633875]  i915_sw_fence_complete+0x58/0x80 [i915]
[  148.634238]  i915_sw_fence_commit+0x16/0x20 [i915]
[  148.634613]  __i915_request_queue+0x60/0x70 [i915]
[  148.634985]  i915_gem_do_execbuffer+0x2de0/0x42b0 [i915]
[  148.635366]  i915_gem_execbuffer2_ioctl+0x2ab/0x580 [i915]
[  148.635400]  drm_ioctl_kernel+0xe9/0x130
[  148.635429]  drm_ioctl+0x27d/0x45e
[  148.635456]  ksys_ioctl+0x89/0xb0
[  148.635482]  __x64_sys_ioctl+0x42/0x60
[  148.635510]  do_syscall_64+0x6e/0x2c0
[  148.635542]  entry_SYSCALL_64_after_hwframe+0x44/0xa9

[  645.071436] BUG: KCSAN: data-race in gen8_emit_fini_breadcrumb [i915] / i915_request_retire [i915]
[  645.071456]
[  645.071467] write to 0xffff8881efe403dc of 4 bytes by task 14668 on cpu 3:
[  645.071647]  i915_request_retire+0xed/0x770 [i915]
[  645.071824]  i915_request_create+0x6c/0x160 [i915]
[  645.072000]  i915_gem_do_execbuffer+0x206d/0x42b0 [i915]
[  645.072177]  i915_gem_execbuffer2_ioctl+0x2ab/0x580 [i915]
[  645.072194]  drm_ioctl_kernel+0xe9/0x130
[  645.072208]  drm_ioctl+0x27d/0x45e
[  645.072222]  ksys_ioctl+0x89/0xb0
[  645.072235]  __x64_sys_ioctl+0x42/0x60
[  645.072248]  do_syscall_64+0x6e/0x2c0
[  645.072263]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  645.072275]
[  645.072285] read to 0xffff8881efe403dc of 4 bytes by interrupt on cpu 2:
[  645.072458]  gen8_emit_fini_breadcrumb+0x158/0x300 [i915]
[  645.072636]  __i915_request_submit+0x204/0x430 [i915]
[  645.072809]  execlists_dequeue+0x8e1/0x1680 [i915]
[  645.072982]  __execlists_submission_tasklet+0x48/0x60 [i915]
[  645.073154]  execlists_submit_request+0x38e/0x3c0 [i915]
[  645.073330]  submit_notify+0x8f/0xc0 [i915]
[  645.073499]  __i915_sw_fence_complete+0x5d/0x3e0 [i915]
[  645.073668]  i915_sw_fence_wake+0xc2/0x130 [i915]
[  645.073836]  __i915_sw_fence_complete+0x2cf/0x3e0 [i915]
[  645.074006]  i915_sw_fence_complete+0x58/0x80 [i915]
[  645.074175]  dma_i915_sw_fence_wake+0x3e/0x80 [i915]
[  645.074344]  signal_irq_work+0x62f/0x710 [i915]
[  645.074360]  irq_work_run_list+0xd7/0x110
[  645.074373]  irq_work_run+0x1d/0x50
[  645.074386]  smp_irq_work_interrupt+0x21/0x30
[  645.074400]  irq_work_interrupt+0xf/0x20
[  645.074414]  _raw_spin_unlock_irqrestore+0x34/0x40
[  645.074585]  execlists_submission_tasklet+0xde/0x170 [i915]
[  645.074602]  tasklet_action_common.isra.0+0x42/0x90
[  645.074617]  __do_softirq+0xc8/0x206
[  645.074629]  irq_exit+0xcd/0xe0
[  645.074642]  do_IRQ+0x44/0xc0
[  645.074654]  ret_from_intr+0x0/0x1c
[  645.074667]  finish_task_switch+0x73/0x230
[  645.074679]  __schedule+0x1c5/0x4c0
[  645.074691]  schedule+0x45/0xb0
[  645.074704]  worker_thread+0x194/0x670
[  645.074716]  kthread+0x19a/0x1e0
[  645.074729]  ret_from_fork+0x1f/0x30

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200407221832.15465-1-chris@chris-wilson.co.uk
2020-04-08 13:40:07 +01:00

141 lines
4.0 KiB
C

/*
* SPDX-License-Identifier: MIT
*
* Copyright © 2019 Intel Corporation
*/
#ifndef INTEL_RING_H
#define INTEL_RING_H
#include "i915_gem.h" /* GEM_BUG_ON */
#include "i915_request.h"
#include "intel_ring_types.h"
struct intel_engine_cs;
struct intel_ring *
intel_engine_create_ring(struct intel_engine_cs *engine, int size);
u32 *intel_ring_begin(struct i915_request *rq, unsigned int num_dwords);
int intel_ring_cacheline_align(struct i915_request *rq);
unsigned int intel_ring_update_space(struct intel_ring *ring);
int intel_ring_pin(struct intel_ring *ring);
void intel_ring_unpin(struct intel_ring *ring);
void intel_ring_reset(struct intel_ring *ring, u32 tail);
void intel_ring_free(struct kref *ref);
static inline struct intel_ring *intel_ring_get(struct intel_ring *ring)
{
kref_get(&ring->ref);
return ring;
}
static inline void intel_ring_put(struct intel_ring *ring)
{
kref_put(&ring->ref, intel_ring_free);
}
static inline void intel_ring_advance(struct i915_request *rq, u32 *cs)
{
/* Dummy function.
*
* This serves as a placeholder in the code so that the reader
* can compare against the preceding intel_ring_begin() and
* check that the number of dwords emitted matches the space
* reserved for the command packet (i.e. the value passed to
* intel_ring_begin()).
*/
GEM_BUG_ON((rq->ring->vaddr + rq->ring->emit) != cs);
}
static inline u32 intel_ring_wrap(const struct intel_ring *ring, u32 pos)
{
return pos & (ring->size - 1);
}
static inline int intel_ring_direction(const struct intel_ring *ring,
u32 next, u32 prev)
{
typecheck(typeof(ring->size), next);
typecheck(typeof(ring->size), prev);
return (next - prev) << ring->wrap;
}
static inline bool
intel_ring_offset_valid(const struct intel_ring *ring,
unsigned int pos)
{
if (pos & -ring->size) /* must be strictly within the ring */
return false;
if (!IS_ALIGNED(pos, 8)) /* must be qword aligned */
return false;
return true;
}
static inline u32 intel_ring_offset(const struct i915_request *rq, void *addr)
{
/* Don't write ring->size (equivalent to 0) as that hangs some GPUs. */
u32 offset = addr - rq->ring->vaddr;
GEM_BUG_ON(offset > rq->ring->size);
return intel_ring_wrap(rq->ring, offset);
}
static inline void
assert_ring_tail_valid(const struct intel_ring *ring, unsigned int tail)
{
unsigned int head = READ_ONCE(ring->head);
GEM_BUG_ON(!intel_ring_offset_valid(ring, tail));
/*
* "Ring Buffer Use"
* Gen2 BSpec "1. Programming Environment" / 1.4.4.6
* Gen3 BSpec "1c Memory Interface Functions" / 2.3.4.5
* Gen4+ BSpec "1c Memory Interface and Command Stream" / 5.3.4.5
* "If the Ring Buffer Head Pointer and the Tail Pointer are on the
* same cacheline, the Head Pointer must not be greater than the Tail
* Pointer."
*
* We use ring->head as the last known location of the actual RING_HEAD,
* it may have advanced but in the worst case it is equally the same
* as ring->head and so we should never program RING_TAIL to advance
* into the same cacheline as ring->head.
*/
#define cacheline(a) round_down(a, CACHELINE_BYTES)
GEM_BUG_ON(cacheline(tail) == cacheline(head) && tail < head);
#undef cacheline
}
static inline unsigned int
intel_ring_set_tail(struct intel_ring *ring, unsigned int tail)
{
/* Whilst writes to the tail are strictly order, there is no
* serialisation between readers and the writers. The tail may be
* read by i915_request_retire() just as it is being updated
* by execlists, as although the breadcrumb is complete, the context
* switch hasn't been seen.
*/
assert_ring_tail_valid(ring, tail);
ring->tail = tail;
return tail;
}
static inline unsigned int
__intel_ring_space(unsigned int head, unsigned int tail, unsigned int size)
{
/*
* "If the Ring Buffer Head Pointer and the Tail Pointer are on the
* same cacheline, the Head Pointer must not be greater than the Tail
* Pointer."
*/
GEM_BUG_ON(!is_power_of_2(size));
return (head - tail - CACHELINE_BYTES) & (size - 1);
}
#endif /* INTEL_RING_H */