drm/ttm: Temporarily disable the huge_fault() callback
Subsystems (3)
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
@@ -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
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 statewith TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable thehuge_fault()callback, eliminating transhuge page-table entries.
Quoted 5 lines
Cc Christian König <[email protected]>Signed-off-by Thomas Hellstrom (VMware) <[email protected]>Reported-by Alex Xu (Hello71) <[email protected]>Tested-by Alex Xu (Hello71) <[email protected]>---
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
Christian König
Acked-byAm 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 statewith TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable thehuge_fault()callback, eliminating transhuge page-table entries.Cc Christian König <[email protected]>Signed-off-by Thomas Hellstrom (VMware) <[email protected]>Reported-by Alex Xu (Hello71) <[email protected]>Tested-by Alex Xu (Hello71) <[email protected]>
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
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 statewith TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable thehuge_fault()callback, eliminating transhuge page-table entries.Cc Christian König <[email protected]>Signed-off-by Thomas Hellstrom (VMware) <[email protected]>Reported-by Alex Xu (Hello71) <[email protected]>Tested-by Alex Xu (Hello71) <[email protected]>Acked-by Christian König <[email protected]>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
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 statewith TTM transparent huge-pages.
Until we've figured out what other TTM drivers do differently compared to vmwgfx, disable thehuge_fault()callback, eliminating transhuge page-table entries.Cc Christian König <[email protected]>Signed-off-by Thomas Hellstrom (VMware) <[email protected]>Reported-by Alex Xu (Hello71) <[email protected]>Tested-by Alex Xu (Hello71) <[email protected]>Acked-by Christian König <[email protected]>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