similar to: [PATCH drm-misc-next] nouveau/dmem: fix copy-paste error in nouveau_dmem_migrate_chunk()

nouveau_fence_emit() can fail before and after initializing the dma-fence and hence before and after initializing the dma-fence' kref. In order to avoid nouveau_fence_emit() potentially failing before dma-fence initialization pass the channel to nouveau_fence_new() already and perform the required check before even allocating the fence. While at it, restore the original behavior of

The kcalloc() in nouveau_dmem_evict_chunk() will return null if the physical memory has run out. As a result, if we dereference src_pfns, dst_pfns or dma_addrs, the null pointer dereference bugs will happen. Moreover, the GPU is going away. If the kcalloc() fails, we could not evict all pages mapping a chunk. So this patch adds a __GFP_NOFAIL flag in kcalloc(). Finally, as there is no need to

Originally patch 4 was targeted for Jason's rdma tree since other HMM related changes were queued there. Now that those have been merged, these patches just contain changes to nouveau so they could go through any tree. I guess Ben Skeggs' tree would be appropriate. Changes since v2: Added patches 1-3 to fix some minor issues. Eliminated nouveau_find_svmm() since it is easily found.

When memory is migrated to the GPU it is likely to be accessed by GPU code soon afterwards. Instead of waiting for a GPU fault, map the migrated memory into the GPU page tables with the same access permissions as the source CPU page table entries. This preserves copy on write semantics. Signed-off-by: Ralph Campbell <rcampbell at nvidia.com> Cc: Christoph Hellwig <hch at lst.de> Cc:

The hmm_devmem_add() function doesn't return NULL, it returns error pointers. Fixes: 5be73b690875 ("drm/nouveau/dmem: device memory helpers for SVM") Signed-off-by: Dan Carpenter <dan.carpenter at oracle.com> --- drivers/gpu/drm/nouveau/nouveau_dmem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/nouveau/nouveau_dmem.c

Fixes gcc '-Wunused-but-set-variable' warning: drivers/gpu/drm/nouveau/nouveau_dmem.c: In function 'nouveau_dmem_free': drivers/gpu/drm/nouveau/nouveau_dmem.c:103:22: warning: variable 'drm' set but not used [-Wunused-but-set-variable] struct nouveau_drm *drm; ^ Signed-off-by: YueHaibing <yuehaibing at huawei.com> ---

Hi Jérôme, Ben and Jason, below is a series against the hmm tree which starts revamping the migrate_vma functionality. The prime idea is to export three slightly lower level functions and thus avoid the need for migrate_vma_ops callbacks. Diffstat: 5 files changed, 281 insertions(+), 607 deletions(-) A git tree is also available at: git://git.infradead.org/users/hch/misc.git

In nouveau_dmem_pages_alloc(), the drm->dmem->mutex is unlocked before calling nouveau_dmem_chunk_alloc(). Reacquire the lock before continuing to the next page. Signed-off-by: Ralph Campbell <rcampbell at nvidia.com> --- I found this while testing Jason Gunthorpe's hmm tree but this is independant of those changes. I guess it could go through David Airlie's tree for nouveau

Factor out the end of fencing logic from the two migration routines. Signed-off-by: Christoph Hellwig <hch at lst.de> --- drivers/gpu/drm/nouveau/nouveau_dmem.c | 33 ++++++++++++-------------- 1 file changed, 15 insertions(+), 18 deletions(-) diff --git a/drivers/gpu/drm/nouveau/nouveau_dmem.c b/drivers/gpu/drm/nouveau/nouveau_dmem.c index d469bc334438..21052a4aaf69 100644 ---

Hi Jérôme, Ben and Jason, below is a series against the hmm tree which starts revamping the migrate_vma functionality. The prime idea is to export three slightly lower level functions and thus avoid the need for migrate_vma_ops callbacks. Diffstat: 4 files changed, 285 insertions(+), 602 deletions(-) A git tree is also available at: git://git.infradead.org/users/hch/misc.git

On 3/6/24 06:01, Duoming Zhou wrote: > The kcalloc() in nouveau_dmem_evict_chunk() will return null if > the physical memory has run out. As a result, if we dereference > src_pfns, dst_pfns or dma_addrs, the null pointer dereference bugs > will happen. > > Moreover, the GPU is going away. If the kcalloc() fails, we could not > evict all pages mapping a chunk. So this patch

The kcalloc() in nouveau_dmem_evict_chunk() will return null if the physical memory has run out. As a result, if we dereference src_pfns, dst_pfns or dma_addrs, the null pointer dereference bugs will happen. This patch uses stack variables to replace the kcalloc(). Fixes: 249881232e14 ("nouveau/dmem: evict device private memory during release") Signed-off-by: Duoming Zhou <duoming

Factor the main copy page to vram routine out into a helper that acts on a single page and which doesn't require the nouveau_dmem_migrate structure for argument passing. As an added benefit the new version only allocates the dma address array once and reuses it for each subsequent chunk of work. Signed-off-by: Christoph Hellwig <hch at lst.de> ---

When memory is migrated to the GPU, it is likely to be accessed by GPU code soon afterwards. Instead of waiting for a GPU fault, map the migrated memory into the GPU page tables with the same access permissions as the source CPU page table entries. This preserves copy on write semantics. Signed-off-by: Ralph Campbell <rcampbell at nvidia.com> Cc: Christoph Hellwig <hch at lst.de> Cc:

Hi Jérôme, Ben and Jason, below is a series against the hmm tree which starts revamping the migrate_vma functionality. The prime idea is to export three slightly lower level functions and thus avoid the need for migrate_vma_ops callbacks. Diffstat: 7 files changed, 282 insertions(+), 614 deletions(-) A git tree is also available at: git://git.infradead.org/users/hch/misc.git

Hi, just for your information and maybe for some help: with 5.1rc1 and SVM enabled i see the following backtrace [1] when the nouveau card (reverse prime) goes to sleep, for now i have papered over with [2] which leaves me with userspace hangs. Any pointers where to look for the actual culprit? PS: Card is: nouveau 0000:01:00.0: NVIDIA GP106 (136000a1) Greetings, Tobias [1]: BUG: unable

On Wed, Aug 07, 2019 at 08:02:14AM -0700, Ralph Campbell wrote: > When memory is migrated to the GPU it is likely to be accessed by GPU > code soon afterwards. Instead of waiting for a GPU fault, map the > migrated memory into the GPU page tables with the same access permissions > as the source CPU page table entries. This preserves copy on write > semantics. > >

In nouveau_dmem_init(), a number of struct nouveau_dmem_chunk are allocated and put on the dmem->chunk_empty list. Then in nouveau_dmem_pages_alloc(), a nouveau_dmem_chunk is removed from the list and GPU memory is allocated. However, the nouveau_dmem_chunk is never removed from the chunk_empty list nor placed on the chunk_free or chunk_full lists. This results in only one chunk ever being

On Thu, Mar 21, 2019 at 04:59:14PM +0100, Tobias Klausmann wrote: > Hi, > > just for your information and maybe for some help: with 5.1rc1 and SVM > enabled i see the following backtrace [1] when the nouveau card (reverse > prime) goes to sleep, for now i have papered over with [2] which leaves me > with userspace hangs. Any pointers where to look for the actual culprit? >

On 6/13/19 5:11 PM, Ralph Campbell wrote: > In nouveau_dmem_pages_alloc(), the drm->dmem->mutex is unlocked before > calling nouveau_dmem_chunk_alloc(). > Reacquire the lock before continuing to the next page. > > Signed-off-by: Ralph Campbell <rcampbell at nvidia.com> > --- > > I found this while testing Jason Gunthorpe's hmm tree but this is >