Skip to content

Commit d92eabb

Browse files
mbrost05mbrost05
committed
drm/xe: Add SVM debug
Add some useful SVM debug logging fro SVM range which prints the range's state. v2: - Update logging with latest structure layout v3: - Better commit message (Thomas) - New range structure (Thomas) - s/COLLECTOT/s/COLLECTOR (Thomas) v4: - Drop partial evict message (Thomas) - Use %p for pointers print (Thomas) v6: - Cast dma_addr to u64 (CI) - Only compile if CONFIG_DRM_GPUSVM selected (CI, Lucas) Signed-off-by: Matthew Brost <matthew.brost@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20250306012657.3505757-30-matthew.brost@intel.com
1 parent 3ca608d commit d92eabb

3 files changed

Lines changed: 92 additions & 7 deletions

File tree

‎drivers/gpu/drm/xe/xe_pt.c‎

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -654,6 +654,7 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
654654
/* Move this entire thing to xe_svm.c? */
655655
xe_svm_notifier_lock(xe_vma_vm(vma));
656656
if (!xe_svm_range_pages_valid(range)) {
657+
xe_svm_range_debug(range, "BIND PREPARE - RETRY");
657658
xe_svm_notifier_unlock(xe_vma_vm(vma));
658659
return -EAGAIN;
659660
}
@@ -662,6 +663,10 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
662663
range->base.itree.last + 1 - range->base.itree.start,
663664
&curs);
664665
is_devmem = xe_res_is_vram(&curs);
666+
if (is_devmem)
667+
xe_svm_range_debug(range, "BIND PREPARE - DMA VRAM");
668+
else
669+
xe_svm_range_debug(range, "BIND PREPARE - DMA");
665670
} else {
666671
xe_assert(xe, false);
667672
}
@@ -1434,10 +1439,13 @@ static int xe_pt_svm_pre_commit(struct xe_migrate_pt_update *pt_update)
14341439
if (op->subop == XE_VMA_SUBOP_UNMAP_RANGE)
14351440
continue;
14361441

1442+
xe_svm_range_debug(range, "PRE-COMMIT");
1443+
14371444
xe_assert(vm->xe, xe_vma_is_cpu_addr_mirror(op->map_range.vma));
14381445
xe_assert(vm->xe, op->subop == XE_VMA_SUBOP_MAP_RANGE);
14391446

14401447
if (!xe_svm_range_pages_valid(range)) {
1448+
xe_svm_range_debug(range, "PRE-COMMIT - RETRY");
14411449
xe_svm_notifier_unlock(vm);
14421450
return -EAGAIN;
14431451
}

‎drivers/gpu/drm/xe/xe_svm.c‎

Lines changed: 77 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -12,6 +12,18 @@
1212
#include "xe_vm.h"
1313
#include "xe_vm_types.h"
1414

15+
static bool xe_svm_range_in_vram(struct xe_svm_range *range)
16+
{
17+
/* Not reliable without notifier lock */
18+
return range->base.flags.has_devmem_pages;
19+
}
20+
21+
static bool xe_svm_range_has_vram_binding(struct xe_svm_range *range)
22+
{
23+
/* Not reliable without notifier lock */
24+
return xe_svm_range_in_vram(range) && range->tile_present;
25+
}
26+
1527
static struct xe_vm *gpusvm_to_vm(struct drm_gpusvm *gpusvm)
1628
{
1729
return container_of(gpusvm, struct xe_vm, svm.gpusvm);
@@ -37,6 +49,23 @@ static unsigned long xe_svm_range_size(struct xe_svm_range *range)
3749
return drm_gpusvm_range_size(&range->base);
3850
}
3951

52+
#define range_debug(r__, operaton__) \
53+
vm_dbg(&range_to_vm(&(r__)->base)->xe->drm, \
54+
"%s: asid=%u, gpusvm=%p, vram=%d,%d, seqno=%lu, " \
55+
"start=0x%014lx, end=0x%014lx, size=%lu", \
56+
(operaton__), range_to_vm(&(r__)->base)->usm.asid, \
57+
(r__)->base.gpusvm, \
58+
xe_svm_range_in_vram((r__)) ? 1 : 0, \
59+
xe_svm_range_has_vram_binding((r__)) ? 1 : 0, \
60+
(r__)->base.notifier_seq, \
61+
xe_svm_range_start((r__)), xe_svm_range_end((r__)), \
62+
xe_svm_range_size((r__)))
63+
64+
void xe_svm_range_debug(struct xe_svm_range *range, const char *operation)
65+
{
66+
range_debug(range, operation);
67+
}
68+
4069
static void *xe_svm_devm_owner(struct xe_device *xe)
4170
{
4271
return xe;
@@ -74,6 +103,8 @@ xe_svm_garbage_collector_add_range(struct xe_vm *vm, struct xe_svm_range *range,
74103
{
75104
struct xe_device *xe = vm->xe;
76105

106+
range_debug(range, "GARBAGE COLLECTOR ADD");
107+
77108
drm_gpusvm_range_set_unmapped(&range->base, mmu_range);
78109

79110
spin_lock(&vm->svm.garbage_collector.lock);
@@ -99,10 +130,14 @@ xe_svm_range_notifier_event_begin(struct xe_vm *vm, struct drm_gpusvm_range *r,
99130

100131
xe_svm_assert_in_notifier(vm);
101132

133+
range_debug(range, "NOTIFIER");
134+
102135
/* Skip if already unmapped or if no binding exist */
103136
if (range->base.flags.unmapped || !range->tile_present)
104137
return 0;
105138

139+
range_debug(range, "NOTIFIER - EXECUTE");
140+
106141
/* Adjust invalidation to range boundaries */
107142
*adj_start = min(xe_svm_range_start(range), mmu_range->start);
108143
*adj_end = max(xe_svm_range_end(range), mmu_range->end);
@@ -153,6 +188,11 @@ static void xe_svm_invalidate(struct drm_gpusvm *gpusvm,
153188

154189
xe_svm_assert_in_notifier(vm);
155190

191+
vm_dbg(&gpusvm_to_vm(gpusvm)->xe->drm,
192+
"INVALIDATE: asid=%u, gpusvm=%p, seqno=%lu, start=0x%016lx, end=0x%016lx, event=%d",
193+
vm->usm.asid, gpusvm, notifier->notifier.invalidate_seq,
194+
mmu_range->start, mmu_range->end, mmu_range->event);
195+
156196
/* Adjust invalidation to notifier boundaries */
157197
adj_start = max(drm_gpusvm_notifier_start(notifier), adj_start);
158198
adj_end = min(drm_gpusvm_notifier_end(notifier), adj_end);
@@ -237,6 +277,8 @@ static int __xe_svm_garbage_collector(struct xe_vm *vm,
237277
{
238278
struct dma_fence *fence;
239279

280+
range_debug(range, "GARBAGE COLLECTOR");
281+
240282
xe_vm_lock(vm, false);
241283
fence = xe_vm_range_unbind(vm, range);
242284
xe_vm_unlock(vm);
@@ -396,16 +438,23 @@ static int xe_svm_copy(struct page **pages, dma_addr_t *dma_addr,
396438
int incr = (match && last) ? 1 : 0;
397439

398440
if (vram_addr != XE_VRAM_ADDR_INVALID) {
399-
if (sram)
441+
if (sram) {
442+
vm_dbg(&tile->xe->drm,
443+
"COPY TO SRAM - 0x%016llx -> 0x%016llx, NPAGES=%ld",
444+
vram_addr, (u64)dma_addr[pos], i - pos + incr);
400445
__fence = xe_migrate_from_vram(tile->migrate,
401446
i - pos + incr,
402447
vram_addr,
403448
dma_addr + pos);
404-
else
449+
} else {
450+
vm_dbg(&tile->xe->drm,
451+
"COPY TO VRAM - 0x%016llx -> 0x%016llx, NPAGES=%ld",
452+
(u64)dma_addr[pos], vram_addr, i - pos + incr);
405453
__fence = xe_migrate_to_vram(tile->migrate,
406454
i - pos + incr,
407455
dma_addr + pos,
408456
vram_addr);
457+
}
409458
if (IS_ERR(__fence)) {
410459
err = PTR_ERR(__fence);
411460
goto err_out;
@@ -425,14 +474,21 @@ static int xe_svm_copy(struct page **pages, dma_addr_t *dma_addr,
425474

426475
/* Extra mismatched device page, copy it */
427476
if (!match && last && vram_addr != XE_VRAM_ADDR_INVALID) {
428-
if (sram)
477+
if (sram) {
478+
vm_dbg(&tile->xe->drm,
479+
"COPY TO SRAM - 0x%016llx -> 0x%016llx, NPAGES=%d",
480+
vram_addr, (u64)dma_addr[pos], 1);
429481
__fence = xe_migrate_from_vram(tile->migrate, 1,
430482
vram_addr,
431483
dma_addr + pos);
432-
else
484+
} else {
485+
vm_dbg(&tile->xe->drm,
486+
"COPY TO VRAM - 0x%016llx -> 0x%016llx, NPAGES=%d",
487+
(u64)dma_addr[pos], vram_addr, 1);
433488
__fence = xe_migrate_to_vram(tile->migrate, 1,
434489
dma_addr + pos,
435490
vram_addr);
491+
}
436492
if (IS_ERR(__fence)) {
437493
err = PTR_ERR(__fence);
438494
goto err_out;
@@ -609,6 +665,8 @@ static int xe_svm_alloc_vram(struct xe_vm *vm, struct xe_tile *tile,
609665
ktime_t end = 0;
610666
int err;
611667

668+
range_debug(range, "ALLOCATE VRAM");
669+
612670
if (!mmget_not_zero(mm))
613671
return -EFAULT;
614672
mmap_read_lock(mm);
@@ -699,6 +757,8 @@ int xe_svm_handle_pagefault(struct xe_vm *vm, struct xe_vma *vma,
699757
if (xe_svm_range_is_valid(range, tile))
700758
return 0;
701759

760+
range_debug(range, "PAGE FAULT");
761+
702762
/* XXX: Add migration policy, for now migrate range once */
703763
if (!range->skip_migrate && range->base.flags.migrate_devmem &&
704764
xe_svm_range_size(range) >= SZ_64K) {
@@ -714,18 +774,26 @@ int xe_svm_handle_pagefault(struct xe_vm *vm, struct xe_vma *vma,
714774
}
715775
}
716776

777+
range_debug(range, "GET PAGES");
717778
err = drm_gpusvm_range_get_pages(&vm->svm.gpusvm, r, &ctx);
718779
/* Corner where CPU mappings have changed */
719780
if (err == -EOPNOTSUPP || err == -EFAULT || err == -EPERM) {
720-
if (err == -EOPNOTSUPP)
781+
if (err == -EOPNOTSUPP) {
782+
range_debug(range, "PAGE FAULT - EVICT PAGES");
721783
drm_gpusvm_range_evict(&vm->svm.gpusvm, &range->base);
784+
}
722785
drm_dbg(&vm->xe->drm,
723786
"Get pages failed, falling back to retrying, asid=%u, gpusvm=%p, errno=%pe\n",
724787
vm->usm.asid, &vm->svm.gpusvm, ERR_PTR(err));
788+
range_debug(range, "PAGE FAULT - RETRY PAGES");
725789
goto retry;
726790
}
727-
if (err)
791+
if (err) {
792+
range_debug(range, "PAGE FAULT - FAIL PAGE COLLECT");
728793
goto err_out;
794+
}
795+
796+
range_debug(range, "PAGE FAULT - BIND");
729797

730798
retry_bind:
731799
drm_exec_init(&exec, 0, 0);
@@ -741,8 +809,10 @@ int xe_svm_handle_pagefault(struct xe_vm *vm, struct xe_vma *vma,
741809
if (IS_ERR(fence)) {
742810
drm_exec_fini(&exec);
743811
err = PTR_ERR(fence);
744-
if (err == -EAGAIN)
812+
if (err == -EAGAIN) {
813+
range_debug(range, "PAGE FAULT - RETRY BIND");
745814
goto retry;
815+
}
746816
if (xe_vm_validate_should_retry(&exec, err, &end))
747817
goto retry_bind;
748818
goto err_out;

‎drivers/gpu/drm/xe/xe_svm.h‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -70,6 +70,8 @@ int xe_svm_handle_pagefault(struct xe_vm *vm, struct xe_vma *vma,
7070
bool xe_svm_has_mapping(struct xe_vm *vm, u64 start, u64 end);
7171

7272
int xe_svm_bo_evict(struct xe_bo *bo);
73+
74+
void xe_svm_range_debug(struct xe_svm_range *range, const char *operation);
7375
#else
7476
static inline bool xe_svm_range_pages_valid(struct xe_svm_range *range)
7577
{
@@ -117,6 +119,11 @@ int xe_svm_bo_evict(struct xe_bo *bo)
117119
{
118120
return 0;
119121
}
122+
123+
static inline
124+
void xe_svm_range_debug(struct xe_svm_range *range, const char *operation)
125+
{
126+
}
120127
#endif
121128

122129
/**

0 commit comments

Comments
 (0)