Linaro-mm-sig

linaro-mm-sig@lists.linaro.org

2 participants
2906 discussions

[PATCH] RFC: dma-fence: Document recoverable page fault implications

by Daniel Vetter

Recently there was a fairly long thread about recoreable hardware page faults, how they can deadlock, and what to do about that. While the discussion is still fresh I figured good time to try and document the conclusions a bit. References: https://lore.kernel.org/dri-devel/20210107030127.20393-1-Felix.Kuehling@amd… Cc: Maarten Lankhorst <maarten.lankhorst(a)linux.intel.com> Cc: Thomas Hellström <thomas.hellstrom(a)intel.com> Cc: "Christian König" <christian.koenig(a)amd.com> Cc: Jerome Glisse <jglisse(a)redhat.com> Cc: Felix Kuehling <felix.kuehling(a)amd.com> Signed-off-by: Daniel Vetter <daniel.vetter(a)intel.com> Cc: Sumit Semwal <sumit.semwal(a)linaro.org> Cc: linux-media(a)vger.kernel.org Cc: linaro-mm-sig(a)lists.linaro.org -- I'll be away next week, but figured I'll type this up quickly for some comments and to check whether I got this all roughly right. Critique very much wanted on this, so that we can make sure hw which can't preempt (with pagefaults pending) like gfx10 has a clear path to support page faults in upstream. So anything I missed, got wrong or like that would be good. -Daniel --- Documentation/driver-api/dma-buf.rst | 66 ++++++++++++++++++++++++++++ 1 file changed, 66 insertions(+) diff --git a/Documentation/driver-api/dma-buf.rst b/Documentation/driver-api/dma-buf.rst index a2133d69872c..e924c1e4f7a3 100644 --- a/Documentation/driver-api/dma-buf.rst +++ b/Documentation/driver-api/dma-buf.rst @@ -257,3 +257,69 @@ fences in the kernel. This means: userspace is allowed to use userspace fencing or long running compute workloads. This also means no implicit fencing for shared buffers in these cases. + +Recoverable Hardware Page Faults Implications +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Modern hardware supports recoverable page faults, which has a lot of +implications for DMA fences. + +First, a pending page fault obviously holds up the work that's running on the +accelerator and a memory allocation is usually required to resolve the fault. +But memory allocations are not allowed to gate completion of DMA fences, which +means any workload using recoverable page faults cannot use DMA fences for +synchronization. Synchronization fences controlled by userspace must be used +instead. + +On GPUs this poses a problem, because current desktop compositor protocols on +Linus rely on DMA fences, which means without an entirely new userspace stack +built on top of userspace fences, they cannot benefit from recoverable page +faults. The exception is when page faults are only used as migration hints and +never to on-demand fill a memory request. For now this means recoverable page +faults on GPUs are limited to pure compute workloads. + +Furthermore GPUs usually have shared resources between the 3D rendering and +compute side, like compute units or command submission engines. If both a 3D +job with a DMA fence and a compute workload using recoverable page faults are +pending they could deadlock: + +- The 3D workload might need to wait for the compute job to finish and release + hardware resources first. + +- The compute workload might be stuck in a page fault, because the memory + allocation is waiting for the DMA fence of the 3D workload to complete. + +There are a few ways to prevent this problem: + +- Compute workloads can always be preempted, even when a page fault is pending + and not yet repaired. Not all hardware supports this. + +- DMA fence workloads and workloads which need page fault handling have + independent hardware resources to guarantee forward progress. This could be + achieved through e.g. through dedicated engines and minimal compute unit + reservations for DMA fence workloads. + +- The reservation approach could be further refined by only reserving the + hardware resources for DMA fence workloads when they are in-flight. This must + cover the time from when the DMA fence is visible to other threads up to + moment when fence is completed through dma_fence_signal(). + +- As a last resort, if the hardware provides no useful reservation mechanics, + all workloads must be flushed from the GPU when switching between jobs + requiring DMA fences or jobs requiring page fault handling: This means all DMA + fences must complete before a compute job with page fault handling can be + inserted into the scheduler queue. And vice versa, before a DMA fence can be + made visible anywhere in the system, all compute workloads must be preempted + to guarantee all pending GPU page faults are flushed. + +Note that workloads that run on independent hardware like copy engines or other +GPUs do not have any impact. This allows us to keep using DMA fences internally +in the kernel even for resolving hardware page faults, e.g. by using copy +engines to clear or copy memory needed to resolve the page fault. + +In some ways this page fault problem is a special case of the `Infinite DMA +Fences` discussions: Infinite fences from compute workloads are allowed to +depend on DMA fences, but not the other way around. And not even the page fault +problem is new, because some other CPU thread in userspace might +hit a page fault which holds up a userspace fence - supporting page faults on +GPUs doesn't anything fundamentally new. -- 2.30.0

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH v4 04/13] drm/shmem-helper: Provide a vmap function for short-term mappings

by Daniel Vetter

On Wed, Jan 27, 2021 at 01:08:05PM +0100, Thomas Zimmermann wrote: > Hi > > Am 11.01.21 um 17:50 schrieb Daniel Vetter: > > On Fri, Jan 08, 2021 at 10:43:31AM +0100, Thomas Zimmermann wrote: > > > Implementations of the vmap/vunmap GEM callbacks may perform pinning > > > of the BO and may acquire the associated reservation object's lock. > > > Callers that only require a mapping of the contained memory can thus > > > interfere with other tasks that require exact pinning, such as scanout. > > > This is less of an issue with private SHMEM buffers, but may happen > > > with imported ones. > > > > > > Therefore provide the new interfaces drm_gem_shmem_vmap_local() and > > > drm_gem_shmem_vunmap_local(), which only perform the vmap/vunmap > > > operations. Callers have to hold the reservation lock while the mapping > > > persists. > > > > > > This patch also connects GEM SHMEM helpers to GEM object functions with > > > equivalent functionality. > > > > > > v4: > > > * call dma_buf_{vmap,vunmap}_local() where necessary (Daniel) > > > * move driver changes into separate patches (Daniel) > > > > > > Signed-off-by: Thomas Zimmermann <tzimmermann(a)suse.de> > > > --- > > > drivers/gpu/drm/drm_gem_shmem_helper.c | 90 +++++++++++++++++++++++--- > > > include/drm/drm_gem_shmem_helper.h | 2 + > > > 2 files changed, 84 insertions(+), 8 deletions(-) > > > > > > diff --git a/drivers/gpu/drm/drm_gem_shmem_helper.c b/drivers/gpu/drm/drm_gem_shmem_helper.c > > > index 9825c378dfa6..298832b2b43b 100644 > > > --- a/drivers/gpu/drm/drm_gem_shmem_helper.c > > > +++ b/drivers/gpu/drm/drm_gem_shmem_helper.c > > > @@ -32,6 +32,8 @@ static const struct drm_gem_object_funcs drm_gem_shmem_funcs = { > > > .get_sg_table = drm_gem_shmem_get_sg_table, > > > .vmap = drm_gem_shmem_vmap, > > > .vunmap = drm_gem_shmem_vunmap, > > > + .vmap_local = drm_gem_shmem_vmap_local, > > > + .vunmap_local = drm_gem_shmem_vunmap_local, > > > .mmap = drm_gem_shmem_mmap, > > > }; > > > @@ -261,7 +263,8 @@ void drm_gem_shmem_unpin(struct drm_gem_object *obj) > > > } > > > EXPORT_SYMBOL(drm_gem_shmem_unpin); > > > -static int drm_gem_shmem_vmap_locked(struct drm_gem_shmem_object *shmem, struct dma_buf_map *map) > > > +static int drm_gem_shmem_vmap_locked(struct drm_gem_shmem_object *shmem, struct dma_buf_map *map, > > > + bool local) > > > > This is a bit spaghetti and also has the problem that we're not changing > > shmem->vmap_use_count under different locks, depending upon which path > > we're taking. > > > > I think the cleanest would be if we pull the if (import_attach) case out > > of the _locked() version completely, for all cases, and also outside of > > the shmem->vmap_lock. This means no caching of vmaps in the shmem layer > > anymore for imported buffers, but this is no longer a problem: We cache > > them in the exporters instead (I think at least, if not maybe need to fix > > that where it's expensive). > > There's no vmap refcounting in amdgpu AFAICT. So importing pages from there > into an SHMEM object has the potential of breaking. IIRC same fro radeon and > nouveau. As long as the pinning is refcounted I think it should be fine, it's just that if you have multiple vmaps (e.g. 2 udl devices plugged in) we'll set up 2 vmaps. Which is a point pointless, but not really harmful. At least on 64bit where there's enough virtual address space. > So I'm somewhat reluctant to making this change. I guess I'll look elsewhere > first to fix some of the locking issues (e.g., my recent ast cursor > patches). If this would break for amdgpu/radeon/nouveau then we already have a bug, since 2 udl devices can provoke this issue already as-is. So I don't think this should be a blocker. -Daniel > > Best regards > Thomas > > > > > Other option would be to unly pull it out for the _vmap_local case, but > > that's a bit ugly because no longer symmetrical in the various paths. > > > > > { > > > struct drm_gem_object *obj = &shmem->base; > > > int ret = 0; > > > @@ -272,7 +275,10 @@ static int drm_gem_shmem_vmap_locked(struct drm_gem_shmem_object *shmem, struct > > > } > > > if (obj->import_attach) { > > > - ret = dma_buf_vmap(obj->import_attach->dmabuf, map); > > > + if (local) > > > + ret = dma_buf_vmap_local(obj->import_attach->dmabuf, map); > > > + else > > > + ret = dma_buf_vmap(obj->import_attach->dmabuf, map); > > > if (!ret) { > > > if (WARN_ON(map->is_iomem)) { > > > ret = -EIO; > > > @@ -313,7 +319,7 @@ static int drm_gem_shmem_vmap_locked(struct drm_gem_shmem_object *shmem, struct > > > return ret; > > > } > > > -/* > > > +/** > > > * drm_gem_shmem_vmap - Create a virtual mapping for a shmem GEM object > > > * @shmem: shmem GEM object > > > * @map: Returns the kernel virtual address of the SHMEM GEM object's backing > > > @@ -339,15 +345,53 @@ int drm_gem_shmem_vmap(struct drm_gem_object *obj, struct dma_buf_map *map) > > > ret = mutex_lock_interruptible(&shmem->vmap_lock); > > > if (ret) > > > return ret; > > > - ret = drm_gem_shmem_vmap_locked(shmem, map); > > > + ret = drm_gem_shmem_vmap_locked(shmem, map, false); > > > mutex_unlock(&shmem->vmap_lock); > > > return ret; > > > } > > > EXPORT_SYMBOL(drm_gem_shmem_vmap); > > > +/** > > > + * drm_gem_shmem_vmap_local - Create a virtual mapping for a shmem GEM object > > > + * @shmem: shmem GEM object > > > + * @map: Returns the kernel virtual address of the SHMEM GEM object's backing > > > + * store. > > > + * > > > + * This function makes sure that a contiguous kernel virtual address mapping > > > + * exists for the buffer backing the shmem GEM object. > > > + * > > > + * The function is called with the BO's reservation object locked. Callers must > > > + * hold the lock until after unmapping the buffer. > > > + * > > > + * This function can be used to implement &drm_gem_object_funcs.vmap_local. But > > > + * it can also be called by drivers directly, in which case it will hide the > > > + * differences between dma-buf imported and natively allocated objects. > > > > So for the other callbacks I tried to make sure we have different entry > > points for this, since it's not really the same thing and because of the > > locking mess we have with dma_resv_lock vs various pre-existing local > > locking scheme, it's easy to get a mess. > > > > I think the super clean version here would be to also export just the > > internal stuff for the ->v(un)map_local hooks, but that's maybe a bit too > > much boilerplate for no real gain. > > -Daniel > > > > > + * > > > + * Acquired mappings should be cleaned up by calling drm_gem_shmem_vunmap_local(). > > > + * > > > + * Returns: > > > + * 0 on success or a negative error code on failure. > > > + */ > > > +int drm_gem_shmem_vmap_local(struct drm_gem_object *obj, struct dma_buf_map *map) > > > +{ > > > + struct drm_gem_shmem_object *shmem = to_drm_gem_shmem_obj(obj); > > > + int ret; > > > + > > > + dma_resv_assert_held(obj->resv); > > > + > > > + ret = mutex_lock_interruptible(&shmem->vmap_lock); > > > + if (ret) > > > + return ret; > > > + ret = drm_gem_shmem_vmap_locked(shmem, map, true); > > > + mutex_unlock(&shmem->vmap_lock); > > > + > > > + return ret; > > > +} > > > +EXPORT_SYMBOL(drm_gem_shmem_vmap_local); > > > + > > > static void drm_gem_shmem_vunmap_locked(struct drm_gem_shmem_object *shmem, > > > - struct dma_buf_map *map) > > > + struct dma_buf_map *map, bool local) > > > { > > > struct drm_gem_object *obj = &shmem->base; > > > @@ -358,7 +402,10 @@ static void drm_gem_shmem_vunmap_locked(struct drm_gem_shmem_object *shmem, > > > return; > > > if (obj->import_attach) > > > - dma_buf_vunmap(obj->import_attach->dmabuf, map); > > > + if (local) > > > + dma_buf_vunmap_local(obj->import_attach->dmabuf, map); > > > + else > > > + dma_buf_vunmap(obj->import_attach->dmabuf, map); > > > else > > > vunmap(shmem->vaddr); > > > @@ -366,7 +413,7 @@ static void drm_gem_shmem_vunmap_locked(struct drm_gem_shmem_object *shmem, > > > drm_gem_shmem_put_pages(shmem); > > > } > > > -/* > > > +/** > > > * drm_gem_shmem_vunmap - Unmap a virtual mapping fo a shmem GEM object > > > * @shmem: shmem GEM object > > > * @map: Kernel virtual address where the SHMEM GEM object was mapped > > > @@ -384,11 +431,38 @@ void drm_gem_shmem_vunmap(struct drm_gem_object *obj, struct dma_buf_map *map) > > > struct drm_gem_shmem_object *shmem = to_drm_gem_shmem_obj(obj); > > > mutex_lock(&shmem->vmap_lock); > > > - drm_gem_shmem_vunmap_locked(shmem, map); > > > + drm_gem_shmem_vunmap_locked(shmem, map, false); > > > mutex_unlock(&shmem->vmap_lock); > > > } > > > EXPORT_SYMBOL(drm_gem_shmem_vunmap); > > > +/** > > > + * drm_gem_shmem_vunmap_local - Unmap a virtual mapping fo a shmem GEM object > > > + * @shmem: shmem GEM object > > > + * @map: Kernel virtual address where the SHMEM GEM object was mapped > > > + * > > > + * This function cleans up a kernel virtual address mapping acquired by > > > + * drm_gem_shmem_vmap_local(). The mapping is only removed when the use count > > > + * drops to zero. > > > + * > > > + * The function is called with the BO's reservation object locked. > > > + * > > > + * This function can be used to implement &drm_gem_object_funcs.vmap_local. > > > + * But it can also be called by drivers directly, in which case it will hide > > > + * the differences between dma-buf imported and natively allocated objects. > > > + */ > > > +void drm_gem_shmem_vunmap_local(struct drm_gem_object *obj, struct dma_buf_map *map) > > > +{ > > > + struct drm_gem_shmem_object *shmem = to_drm_gem_shmem_obj(obj); > > > + > > > + dma_resv_assert_held(obj->resv); > > > + > > > + mutex_lock(&shmem->vmap_lock); > > > + drm_gem_shmem_vunmap_locked(shmem, map, true); > > > + mutex_unlock(&shmem->vmap_lock); > > > +} > > > +EXPORT_SYMBOL(drm_gem_shmem_vunmap_local); > > > + > > > struct drm_gem_shmem_object * > > > drm_gem_shmem_create_with_handle(struct drm_file *file_priv, > > > struct drm_device *dev, size_t size, > > > diff --git a/include/drm/drm_gem_shmem_helper.h b/include/drm/drm_gem_shmem_helper.h > > > index 434328d8a0d9..3f59bdf749aa 100644 > > > --- a/include/drm/drm_gem_shmem_helper.h > > > +++ b/include/drm/drm_gem_shmem_helper.h > > > @@ -114,7 +114,9 @@ void drm_gem_shmem_put_pages(struct drm_gem_shmem_object *shmem); > > > int drm_gem_shmem_pin(struct drm_gem_object *obj); > > > void drm_gem_shmem_unpin(struct drm_gem_object *obj); > > > int drm_gem_shmem_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); > > > +int drm_gem_shmem_vmap_local(struct drm_gem_object *obj, struct dma_buf_map *map); > > > void drm_gem_shmem_vunmap(struct drm_gem_object *obj, struct dma_buf_map *map); > > > +void drm_gem_shmem_vunmap_local(struct drm_gem_object *obj, struct dma_buf_map *map); > > > int drm_gem_shmem_madvise(struct drm_gem_object *obj, int madv); > > > -- > > > 2.29.2 > > > > > > > -- > Thomas Zimmermann > Graphics Driver Developer > SUSE Software Solutions Germany GmbH > Maxfeldstr. 5, 90409 Nürnberg, Germany > (HRB 36809, AG Nürnberg) > Geschäftsführer: Felix Imendörffer > -- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH v3] dmabuf: Add the capability to expose DMA-BUF stats in sysfs

by Sumit Semwal

Hi Simon, On Thu, 28 Jan 2021 at 20:01, Simon Ser <contact(a)emersion.fr> wrote: > > On Thursday, January 28th, 2021 at 1:03 PM, Sumit Semwal <sumit.semwal(a)linaro.org> wrote: > > > Since he didn't comment over Hridya's last clarification about the > > tracepoints to track total GPU memory allocations being orthogonal to > > this series, I assumed he agreed with it. > > IIRC he's away this week. (I don't remember when he comes back.) > > > Daniel, do you still have objections around adding this patch in? > > (Adding him explicitly in CC) Thanks for doing this! Best, Sumit.

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH 08/29] dma-buf: Avoid comma separated statements

by Christian König

Am 31.01.21 um 18:39 schrieb Joe Perches: > On Wed, 2021-02-03 at 14:26 +0100, Christian König wrote: >> Am 30.01.21 um 19:47 schrieb Joe Perches: >>> On Mon, 2020-08-24 at 21:56 -0700, Joe Perches wrote: >>>> Use semicolons and braces. >>> Ping? >>>> Signed-off-by: Joe Perches <joe(a)perches.com> >> Reviewed-by: Christian König <christian.koenig(a)amd.com> >> >> Do you have commit rights to drm-misc-next? > No. Pushed. Thanks for the help, Christian. > >>>> --- >>>> drivers/dma-buf/st-dma-fence.c | 7 +++++-- >>>> 1 file changed, 5 insertions(+), 2 deletions(-) >>>> >>>> diff --git a/drivers/dma-buf/st-dma-fence.c b/drivers/dma-buf/st-dma-fence.c >>>> index e593064341c8..c8a12d7ad71a 100644 >>>> --- a/drivers/dma-buf/st-dma-fence.c >>>> +++ b/drivers/dma-buf/st-dma-fence.c >>>> @@ -471,8 +471,11 @@ static int thread_signal_callback(void *arg) >>>> dma_fence_signal(f1); >>>> >>>> smp_store_mb(cb.seen, false); >>>> - if (!f2 || dma_fence_add_callback(f2, &cb.cb, simple_callback)) >>>> - miss++, cb.seen = true; >>>> + if (!f2 || >>>> + dma_fence_add_callback(f2, &cb.cb, simple_callback)) { >>>> + miss++; >>>> + cb.seen = true; >>>> + } >>>> >>>> if (!t->before) >>>> dma_fence_signal(f1); >

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH v4 2/4] mm: failfast mode with __GFP_NORETRY in alloc_contig_range

by Minchan Kim

On Thu, Jan 28, 2021 at 08:53:25AM +0100, Michal Hocko wrote: > On Wed 27-01-21 12:42:45, Minchan Kim wrote: > > On Tue, Jan 26, 2021 at 08:44:49AM +0100, Michal Hocko wrote: > > > On Mon 25-01-21 11:33:36, Minchan Kim wrote: > > > > On Mon, Jan 25, 2021 at 02:12:00PM +0100, Michal Hocko wrote: > > > > > On Thu 21-01-21 09:55:00, Minchan Kim wrote: > > > > > > Contiguous memory allocation can be stalled due to waiting > > > > > > on page writeback and/or page lock which causes unpredictable > > > > > > delay. It's a unavoidable cost for the requestor to get *big* > > > > > > contiguous memory but it's expensive for *small* contiguous > > > > > > memory(e.g., order-4) because caller could retry the request > > > > > > in different range where would have easy migratable pages > > > > > > without stalling. > > > > > > > > > > > > This patch introduce __GFP_NORETRY as compaction gfp_mask in > > > > > > alloc_contig_range so it will fail fast without blocking > > > > > > when it encounters pages needed waiting. > > > > > > > > > > I am not against controling how hard this allocator tries with gfp mask > > > > > but this changelog is rather void on any data and any user. > > > > > > > > > > It is also rather dubious to have retries when then caller says to not > > > > > retry. > > > > > > > > Since max_tries is 1 with ++tries, it shouldn't retry. > > > > > > OK, I have missed that. This is a tricky code. ASYNC mode should be > > > completely orthogonal to the retries count. Those are different things. > > > Page allocator does an explicit bail out based on __GFP_NORETRY. You > > > should be doing the same. > > > > Before sending next revision, let me check this part again. > > > > I want to use __GFP_NORETRY to indicate "opportunistic-easy-to-fail attempt" > > and I want to use ASYNC migrate_mode to help the goal. > > > > Do you see the problem? > > No, as I've said. This is a normal NORETRY policy. And ASYNC migration > is a mere implementation detail you do not have bother your users about. > This is the semantic view. From the implementation POV it should be the > gfp mask to drive decisions rather than a random (ASYNC) flag to control > retries as you did here. Make sense. Let me cook next revision. Thanks for the review, Michal.

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH v4 2/4] mm: failfast mode with __GFP_NORETRY in alloc_contig_range

by Minchan Kim

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH v4 1/4] mm: cma: introduce gfp flag in cma_alloc instead of no_warn

by Minchan Kim

On Tue, Jan 26, 2021 at 08:38:08AM +0100, Michal Hocko wrote: > On Mon 25-01-21 11:42:34, Minchan Kim wrote: > > On Mon, Jan 25, 2021 at 02:07:01PM +0100, Michal Hocko wrote: > > > On Thu 21-01-21 09:54:59, Minchan Kim wrote: > > > > The upcoming patch will introduce __GFP_NORETRY semantic > > > > in alloc_contig_range which is a failfast mode of the API. > > > > Instead of adding a additional parameter for gfp, replace > > > > no_warn with gfp flag. > > > > > > > > To keep old behaviors, it follows the rule below. > > > > > > > > no_warn gfp_flags > > > > > > > > false GFP_KERNEL > > > > true GFP_KERNEL|__GFP_NOWARN > > > > gfp & __GFP_NOWARN GFP_KERNEL | (gfp & __GFP_NOWARN) > > > > > > > > Reviewed-by: Suren Baghdasaryan <surenb(a)google.com> > > > > Signed-off-by: Minchan Kim <minchan(a)kernel.org> > > > [...] > > > > diff --git a/mm/cma.c b/mm/cma.c > > > > index 0ba69cd16aeb..d50627686fec 100644 > > > > --- a/mm/cma.c > > > > +++ b/mm/cma.c > > > > @@ -419,13 +419,13 @@ static inline void cma_debug_show_areas(struct cma *cma) { } > > > > * @cma: Contiguous memory region for which the allocation is performed. > > > > * @count: Requested number of pages. > > > > * @align: Requested alignment of pages (in PAGE_SIZE order). > > > > - * @no_warn: Avoid printing message about failed allocation > > > > + * @gfp_mask: GFP mask to use during the cma allocation. > > > > > > Call out supported gfp flags explicitly. Have a look at kvmalloc_node > > > for a guidance. > > > > How about this? > > > > > > diff --git a/mm/cma.c b/mm/cma.c > > index d50627686fec..b94727b694d6 100644 > > --- a/mm/cma.c > > +++ b/mm/cma.c > > @@ -423,6 +423,10 @@ static inline void cma_debug_show_areas(struct cma *cma) { } > > * > > * This function allocates part of contiguous memory on specific > > * contiguous memory area. > > + * > > + * For gfp_mask, GFP_KERNEL and __GFP_NORETRY are supported. __GFP_NORETRY > > + * will avoid costly functions(e.g., waiting on page_writeback and locking) > > + * at current implementaion during the page migration. > > rather than explicitly mentioning what the flag implies I think it would > be more useful to state the intended usecase. See how kvmalloc_node says > "__GFP_RETRY_MAYFAIL is supported, and it should be used only if kmalloc is > preferable to the vmalloc fallback, due to visible performance > drawbacks. > __GFP_NOWARN is also supported to suppress allocation failure messages." > > This would help people not familiar with internals to see whether this > flag is a good fit for them. > > In this case I woul go with > " > @flags: gfp mask. Must be compatible (superset) with GFP_KERNEL. > [...] > Reclaim modifiers (__GFP_RETRY_MAYFAIL, __GFP_NOFAIL) are not supported. > __GFP_NORETRY is supported, and it should be used for opportunistic > allocation attempts that should rather fail quickly when the caller has > a fallback strategy. > " > > Obviously for this patch you will go with a simple statement that > Reclaim modifiers are not supported at all. After more discussion for gfp_flags in thread of next patch, let me changes a bit more based on it. Thanks for the suggestion, Michal.

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH v4 4/4] dma-buf: heaps: add chunk heap to dmabuf heaps

by Minchan Kim

On Tue, Jan 26, 2021 at 08:46:05AM +0100, Michal Hocko wrote: > On Thu 21-01-21 09:55:02, Minchan Kim wrote: > > From: Hyesoo Yu <hyesoo.yu(a)samsung.com> > > > > This patch supports chunk heap that allocates the buffers that > > arranged into a list a fixed size chunks taken from CMA. > > > > The chunk heap driver is bound directly to a reserved_memory > > node by following Rob Herring's suggestion in [1]. > > > > [1] https://lore.kernel.org/lkml/20191025225009.50305-2-john.stultz@linaro.org/… > > Who is using this allocator in the kernel? Userspace uses the memory via mapping it via dmabuf.

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH v4 2/4] mm: failfast mode with __GFP_NORETRY in alloc_contig_range

by Minchan Kim

On Tue, Jan 26, 2021 at 08:44:49AM +0100, Michal Hocko wrote: > On Mon 25-01-21 11:33:36, Minchan Kim wrote: > > On Mon, Jan 25, 2021 at 02:12:00PM +0100, Michal Hocko wrote: > > > On Thu 21-01-21 09:55:00, Minchan Kim wrote: > > > > Contiguous memory allocation can be stalled due to waiting > > > > on page writeback and/or page lock which causes unpredictable > > > > delay. It's a unavoidable cost for the requestor to get *big* > > > > contiguous memory but it's expensive for *small* contiguous > > > > memory(e.g., order-4) because caller could retry the request > > > > in different range where would have easy migratable pages > > > > without stalling. > > > > > > > > This patch introduce __GFP_NORETRY as compaction gfp_mask in > > > > alloc_contig_range so it will fail fast without blocking > > > > when it encounters pages needed waiting. > > > > > > I am not against controling how hard this allocator tries with gfp mask > > > but this changelog is rather void on any data and any user. > > > > > > It is also rather dubious to have retries when then caller says to not > > > retry. > > > > Since max_tries is 1 with ++tries, it shouldn't retry. > > OK, I have missed that. This is a tricky code. ASYNC mode should be > completely orthogonal to the retries count. Those are different things. > Page allocator does an explicit bail out based on __GFP_NORETRY. You > should be doing the same. A concern with __GFP_NOWAIT is regardless of flags passed to cma_alloc, internal implementation of alloc_contig_range inside will use blockable operation. See __alloc_contig_migrate_range. If we go with __GFP_NOWAIT, we should propagate the gfp_mask inside of __alloc_contig_migrate_range to make cma_alloc consistent with alloc_pages. (IIUC, that's what you want - make gfp_mask consistent between cma_alloc and alloc_pages) but I am worry about the direction will make complicate situation since cma invovles migration context as well as target page allocation context. Sometime, the single gfp flag could be trouble to express both contexts all at once. > > > > > > > Also why didn't you consider GFP_NOWAIT semantic for non blocking mode? > > > > GFP_NOWAIT seems to be low(specific) flags rather than the one I want to > > express. Even though I said only page writeback/lock in the description, > > the goal is to avoid costly operations we might find later so such > > "failfast", I thought GFP_NORETRY would be good fit. > > I suspect you are too focused on implementation details here. Think > about the indended semantic. Callers of this functionality will not > think about those (I hope because if they rely on these details then the > whole thing will become unmaintainable because any change would require > an audit of all existing users). All you should be caring about is to > control how expensive the call can be. GFP_NOWAIT is not really low > level from that POV. It gives you a very lightweight non-sleeping > attempt to allocate. GFP_NORETRY will give you potentially sleeping but > an opportunistic-easy-to-fail attempt. And so on. See how that is > absolutely free of any page writeback or any specific locking. With above reason I mentioned, I wanted to express __GFP_NORETRY as "opportunistic-easy-to-fail attempt" to support cma_alloc as "failfast" for migration context. > -- > Michal Hocko > SUSE Labs

4 years, 5 months

Re: [Linaro-mm-sig] [PATCH v4 1/4] mm: cma: introduce gfp flag in cma_alloc instead of no_warn

by Minchan Kim

On Mon, Jan 25, 2021 at 02:07:01PM +0100, Michal Hocko wrote: > On Thu 21-01-21 09:54:59, Minchan Kim wrote: > > The upcoming patch will introduce __GFP_NORETRY semantic > > in alloc_contig_range which is a failfast mode of the API. > > Instead of adding a additional parameter for gfp, replace > > no_warn with gfp flag. > > > > To keep old behaviors, it follows the rule below. > > > > no_warn gfp_flags > > > > false GFP_KERNEL > > true GFP_KERNEL|__GFP_NOWARN > > gfp & __GFP_NOWARN GFP_KERNEL | (gfp & __GFP_NOWARN) > > > > Reviewed-by: Suren Baghdasaryan <surenb(a)google.com> > > Signed-off-by: Minchan Kim <minchan(a)kernel.org> > [...] > > diff --git a/mm/cma.c b/mm/cma.c > > index 0ba69cd16aeb..d50627686fec 100644 > > --- a/mm/cma.c > > +++ b/mm/cma.c > > @@ -419,13 +419,13 @@ static inline void cma_debug_show_areas(struct cma *cma) { } > > * @cma: Contiguous memory region for which the allocation is performed. > > * @count: Requested number of pages. > > * @align: Requested alignment of pages (in PAGE_SIZE order). > > - * @no_warn: Avoid printing message about failed allocation > > + * @gfp_mask: GFP mask to use during the cma allocation. > > Call out supported gfp flags explicitly. Have a look at kvmalloc_node > for a guidance. How about this? diff --git a/mm/cma.c b/mm/cma.c index d50627686fec..b94727b694d6 100644 --- a/mm/cma.c +++ b/mm/cma.c @@ -423,6 +423,10 @@ static inline void cma_debug_show_areas(struct cma *cma) { } * * This function allocates part of contiguous memory on specific * contiguous memory area. + * + * For gfp_mask, GFP_KERNEL and __GFP_NORETRY are supported. __GFP_NORETRY + * will avoid costly functions(e.g., waiting on page_writeback and locking) + * at current implementaion during the page migration. */ struct page *cma_alloc(struct cma *cma, size_t count, unsigned int align, gfp_t gfp_mask)

4 years, 5 months

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

Linaro-mm-sig