Skip to contentSkip to conversation

drm/ttm: Temporarily disable the huge_fault() callback

PATCH In review Thomas Hellström (VMware) 2 participants 5 messages Apr 8, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/ttm: Temporarily disable the huge_fault() callback

From: "Thomas Hellstrom (VMware)" <[email protected]>

With amdgpu and CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y, there are errors like:

BUG: non-zero pgtables_bytes on freeing mm
and:
BUG: Bad rss-counter state

with TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable the huge_fault() callback, eliminating transhuge page-table entries.

---
 drivers/gpu/drm/ttm/ttm_bo_vm.c | 63 ---------------------------------
 1 file changed, 63 deletions(-)
drivers/gpu/drm/ttm/ttm_bo_vm.c +0 -63
index 6ee3b96f0d13..0ad30b112982 100644
@@ -442,66 +442,6 @@ vm_fault_t ttm_bo_vm_fault(struct vm_fault *vmf)
 }
 EXPORT_SYMBOL(ttm_bo_vm_fault);
 
-#ifdef CONFIG_TRANSPARENT_HUGEPAGE
-/**
- * ttm_pgprot_is_wrprotecting - Is a page protection value write-protecting?
- * @prot: The page protection value
- *
- * Return: true if @prot is write-protecting. false otherwise.
- */
-static bool ttm_pgprot_is_wrprotecting(pgprot_t prot)
-{
-	/*
-	 * This is meant to say "pgprot_wrprotect(prot) == prot" in a generic
-	 * way. Unfortunately there is no generic pgprot_wrprotect.
-	 */
-	return pte_val(pte_wrprotect(__pte(pgprot_val(prot)))) ==
-		pgprot_val(prot);
-}
-
-static vm_fault_t ttm_bo_vm_huge_fault(struct vm_fault *vmf,
-				       enum page_entry_size pe_size)
-{
-	struct vm_area_struct *vma = vmf->vma;
-	pgprot_t prot;
-	struct ttm_buffer_object *bo = vma->vm_private_data;
-	vm_fault_t ret;
-	pgoff_t fault_page_size = 0;
-	bool write = vmf->flags & FAULT_FLAG_WRITE;
-
-	switch (pe_size) {
-	case PE_SIZE_PMD:
-		fault_page_size = HPAGE_PMD_SIZE >> PAGE_SHIFT;
-		break;
-#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
-	case PE_SIZE_PUD:
-		fault_page_size = HPAGE_PUD_SIZE >> PAGE_SHIFT;
-		break;
-#endif
-	default:
-		WARN_ON_ONCE(1);
-		return VM_FAULT_FALLBACK;
-	}
-
-	/* Fallback on write dirty-tracking or COW */
-	if (write && ttm_pgprot_is_wrprotecting(vma->vm_page_prot))
-		return VM_FAULT_FALLBACK;
-
-	ret = ttm_bo_vm_reserve(bo, vmf);
-	if (ret)
-		return ret;
-
-	prot = vm_get_page_prot(vma->vm_flags);
-	ret = ttm_bo_vm_fault_reserved(vmf, prot, 1, fault_page_size);
-	if (ret == VM_FAULT_RETRY && !(vmf->flags & FAULT_FLAG_RETRY_NOWAIT))
-		return ret;
-
-	dma_resv_unlock(bo->base.resv);
-
-	return ret;
-}
-#endif
-
 void ttm_bo_vm_open(struct vm_area_struct *vma)
 {
 	struct ttm_buffer_object *bo = vma->vm_private_data;
@@ -604,9 +544,6 @@ static const struct vm_operations_struct ttm_bo_vm_ops = {
 	.open = ttm_bo_vm_open,
 	.close = ttm_bo_vm_close,
 	.access = ttm_bo_vm_access,
-#ifdef CONFIG_TRANSPARENT_HUGEPAGE
-	.huge_fault = ttm_bo_vm_huge_fault,
-#endif
 };
 
 static struct ttm_buffer_object *ttm_bo_vm_lookup(struct ttm_bo_device *bdev,
Signature · 6 lines
2.21.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
#1 lore raw

Hi, Christian,

Quoted 12 lines · Thomas Hellström (VMware)

From: "Thomas Hellstrom (VMware)" <[email protected]>

With amdgpu and CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y, there are errors like:

BUG: non-zero pgtables_bytes on freeing mm
and:
BUG: Bad rss-counter state

with TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable the huge_fault() callback, eliminating transhuge page-table entries.

Quoted 5 lines

---

Without being able to test and track this down on amdgpu there's little more than this I can do at the moment. Hopefully I'll be able to test on nouveau/ttm after getting back from vacation to see if I can reproduce.

It looks like some part of the kernel mistakes a huge page-table entry for a page directory, and that would be a path that is not hit with vmwgfx.

/Thomas

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#2 lore raw

Am 08.04.20 um 14:01 schrieb Thomas Hellström (VMware):

Quoted 19 lines

Hi, Christian,

Quoted 16 lines · Thomas Hellström (VMware)

From: "Thomas Hellstrom (VMware)" <[email protected]>

With amdgpu and CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y, there are errors like:

BUG: non-zero pgtables_bytes on freeing mm
and:
BUG: Bad rss-counter state

with TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable the huge_fault() callback, eliminating transhuge page-table entries.

Quoted 5 lines
Quoted 1 line

---

Without being able to test and track this down on amdgpu there's little more than this I can do at the moment. Hopefully I'll be able to test on nouveau/ttm after getting back from vacation to see if I can reproduce.

It looks like some part of the kernel mistakes a huge page-table entry for a page directory, and that would be a path that is not hit with vmwgfx.

Well that looks like an ugly one and I don't know enough about the page table handling to hunt this down either.

BTW: Have you seen the coverity warning about  "WARN_ON_ONCE(ret = VM_FAULT_FALLBACK);"?

Regards,
Christian.

Quoted 4 lines

/Thomas

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#3 lore raw
Quoted 25 lines · Christian König

Am 08.04.20 um 14:01 schrieb Thomas Hellström (VMware):

Quoted 19 lines

Hi, Christian,

Quoted 16 lines · Thomas Hellström (VMware)

From: "Thomas Hellstrom (VMware)" <[email protected]>

With amdgpu and CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y, there are errors like:

BUG: non-zero pgtables_bytes on freeing mm
and:
BUG: Bad rss-counter state

with TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable the huge_fault() callback, eliminating transhuge page-table entries.

Quoted 2 lines
Quoted 1 line

---

Quoted 7 lines
Quoted 3 lines

Without being able to test and track this down on amdgpu there's little more than this I can do at the moment. Hopefully I'll be able to test on nouveau/ttm after getting back from vacation to see if I can reproduce.

It looks like some part of the kernel mistakes a huge page-table entry for a page directory, and that would be a path that is not hit with vmwgfx.

Well that looks like an ugly one and I don't know enough about the page table handling to hunt this down either.

BTW: Have you seen the coverity warning about  "WARN_ON_ONCE(ret = VM_FAULT_FALLBACK);"?

Yes, that's a false warning but it might be that it should be rewritten for clarity like so:

ret = VM_FAULT_FALLBACK;
WARN_ON_ONCE(true);

/Thomas

Quoted 8 lines

Regards,
Christian.

Quoted 4 lines

/Thomas

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#4 lore raw

Am 08.04.20 um 15:49 schrieb Thomas Hellström (VMware):

Quoted 30 lines · Christian König
Quoted 29 lines

Am 08.04.20 um 14:01 schrieb Thomas Hellström (VMware):

Quoted 19 lines

Hi, Christian,

Quoted 16 lines · Thomas Hellström (VMware)

From: "Thomas Hellstrom (VMware)" <[email protected]>

With amdgpu and CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y, there are errors like:

BUG: non-zero pgtables_bytes on freeing mm
and:
BUG: Bad rss-counter state

with TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable the huge_fault() callback, eliminating transhuge page-table entries.

Quoted 5 lines
Quoted 1 line

---

Without being able to test and track this down on amdgpu there's little more than this I can do at the moment. Hopefully I'll be able to test on nouveau/ttm after getting back from vacation to see if I can reproduce.

It looks like some part of the kernel mistakes a huge page-table entry for a page directory, and that would be a path that is not hit with vmwgfx.

Quoted 8 lines
Quoted 3 lines

Well that looks like an ugly one and I don't know enough about the page table handling to hunt this down either.

BTW: Have you seen the coverity warning about  "WARN_ON_ONCE(ret = VM_FAULT_FALLBACK);"?

Yes, that's a false warning but it might be that it should be rewritten for clarity like so:

ret = VM_FAULT_FALLBACK;
WARN_ON_ONCE(true);

Using an additional () also usually works, but I'm not even sure if WARN_ON_ONCE() isn't compiled to a no-op under some circumstances.

So better save than sorry,
Christian.

Quoted 14 lines

/Thomas

Quoted 8 lines

Regards,
Christian.

Quoted 4 lines

/Thomas

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel