-----Original Message-----
From: Thomas Hellström <thomas.hellstrom@xxxxxxxxxxxxxxx>
Sent: October 5, 2021 9:48 AM
To: Zeng, Oak <oak.zeng@xxxxxxxxx>; Auld, Matthew
<matthew.auld@xxxxxxxxx>; intel-gfx@xxxxxxxxxxxxxxxxxxxxx
Cc: dri-devel@xxxxxxxxxxxxxxxxxxxxx; Christian König
<christian.koenig@xxxxxxx>
Subject: Re: [PATCH v5 09/13] drm/i915/ttm: add tt shmem
backend
On 10/5/21 04:05, Zeng, Oak wrote:
Hi Matthew/Thomas,
See one question inline
Regards,
Oak
-----Original Message-----
From: Intel-gfx <intel-gfx-bounces@xxxxxxxxxxxxxxxxxxxxx> On Behalf Of
Matthew Auld
Sent: September 27, 2021 7:41 AM
To: intel-gfx@xxxxxxxxxxxxxxxxxxxxx
Cc: dri-devel@xxxxxxxxxxxxxxxxxxxxx; Thomas Hellström
<thomas.hellstrom@xxxxxxxxxxxxxxx>; Christian König
<christian.koenig@xxxxxxx>
Subject: [PATCH v5 09/13] drm/i915/ttm: add tt shmem backend
For cached objects we can allocate our pages directly in shmem. This should
make it possible(in a later patch) to utilise the existing i915-gem shrinker code
for such objects. For now this is still disabled.
v2(Thomas):
- Add optional try_to_writeback hook for objects. Importantly we need
to check if the object is even still shrinkable; in between us
dropping the shrinker LRU lock and acquiring the object lock it could for
example have been moved. Also we need to differentiate between
"lazy" shrinking and the immediate writeback mode. Also later we need
to
handle objects which don't even have mm.pages, so bundling this into
put_pages() would require somehow handling that edge case, hence
just letting the ttm backend handle everything in try_to_writeback
doesn't seem too bad.
v3(Thomas):
- Likely a bad idea to touch the object from the unpopulate hook,
since it's not possible to hold a reference, without also creating
circular dependency, so likely this is too fragile. For now just
ensure we at least mark the pages as dirty/accessed when called from the
shrinker on WILLNEED objects.
- s/try_to_writeback/shrinker_release_pages, since this can do more
than just writeback.
- Get rid of do_backup boolean and just set the SWAPPED flag prior to
calling unpopulate.
- Keep shmem_tt as lowest priority for the TTM LRU bo_swapout walk,
since
these just get skipped anyway. We can try to come up with something
better later.
Signed-off-by: Matthew Auld <matthew.auld@xxxxxxxxx>
Cc: Thomas Hellström <thomas.hellstrom@xxxxxxxxxxxxxxx>
Cc: Christian König <christian.koenig@xxxxxxx>
---
drivers/gpu/drm/i915/gem/i915_gem_object.h | 8 +
.../gpu/drm/i915/gem/i915_gem_object_types.h | 2 +
drivers/gpu/drm/i915/gem/i915_gem_shmem.c | 14 +-
drivers/gpu/drm/i915/gem/i915_gem_shrinker.c | 17 +-
drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 240 ++++++++++++++++-
-
5 files changed, 245 insertions(+), 36 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h
b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 3043fcbd31bd..1c9a1d8d3434 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -601,6 +601,14 @@ int i915_gem_object_wait_migration(struct
drm_i915_gem_object *obj, bool
i915_gem_object_placement_possible(struct drm_i915_gem_object *obj,
enum intel_memory_type type);
+struct sg_table *shmem_alloc_st(struct drm_i915_private *i915,
+ size_t size, struct intel_memory_region *mr,
+ struct address_space *mapping,
+ unsigned int max_segment);
+void shmem_free_st(struct sg_table *st, struct address_space *mapping,
+ bool dirty, bool backup);
+void __shmem_writeback(size_t size, struct address_space *mapping);
+
#ifdef CONFIG_MMU_NOTIFIER
static inline bool
i915_gem_object_is_userptr(struct drm_i915_gem_object *obj) diff --git
a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index fa2ba9e2a4d0..f0fb17be2f7a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -56,6 +56,8 @@ struct drm_i915_gem_object_ops {
struct sg_table *pages);
void (*truncate)(struct drm_i915_gem_object *obj);
void (*writeback)(struct drm_i915_gem_object *obj);
+ int (*shrinker_release_pages)(struct drm_i915_gem_object *obj,
+ bool should_writeback);
int (*pread)(struct drm_i915_gem_object *obj,
const struct drm_i915_gem_pread *arg); diff --git
a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
index 36b711ae9e28..19e55cc29a15 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
@@ -25,8 +25,8 @@ static void check_release_pagevec(struct pagevec
*pvec)
cond_resched();
}
-static void shmem_free_st(struct sg_table *st, struct address_space
*mapping,
- bool dirty, bool backup)
+void shmem_free_st(struct sg_table *st, struct address_space *mapping,
+ bool dirty, bool backup)
{
struct sgt_iter sgt_iter;
struct pagevec pvec;
@@ -52,10 +52,10 @@ static void shmem_free_st(struct sg_table *st, struct
address_space *mapping,
kfree(st);
}
-static struct sg_table *shmem_alloc_st(struct drm_i915_private *i915,
- size_t size, struct intel_memory_region
*mr,
- struct address_space *mapping,
- unsigned int max_segment)
+struct sg_table *shmem_alloc_st(struct drm_i915_private *i915,
+ size_t size, struct intel_memory_region *mr,
+ struct address_space *mapping,
+ unsigned int max_segment)
{
const unsigned long page_count = size / PAGE_SIZE;
unsigned long i;
@@ -300,7 +300,7 @@ shmem_truncate(struct drm_i915_gem_object *obj)
obj->mm.pages = ERR_PTR(-EFAULT);
}
-static void __shmem_writeback(size_t size, struct address_space
*mapping)
+void __shmem_writeback(size_t size, struct address_space *mapping)
{
struct writeback_control wbc = {
.sync_mode = WB_SYNC_NONE,
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
index e382b7f2353b..cc80bd23d323 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
@@ -56,19 +56,24 @@ static bool unsafe_drop_pages(struct
drm_i915_gem_object *obj,
return false;
}
-static void try_to_writeback(struct drm_i915_gem_object *obj,
- unsigned int flags)
+static int try_to_writeback(struct drm_i915_gem_object *obj, unsigned
+int flags)
{
+ if (obj->ops->shrinker_release_pages)
+ return obj->ops->shrinker_release_pages(obj,
+ flags &
I915_SHRINK_WRITEBACK);
+
switch (obj->mm.madv) {
case I915_MADV_DONTNEED:
i915_gem_object_truncate(obj);
- return;
+ return 0;
case __I915_MADV_PURGED:
- return;
+ return 0;
}
if (flags & I915_SHRINK_WRITEBACK)
i915_gem_object_writeback(obj);
+
+ return 0;
}
/**
@@ -222,8 +227,8 @@ i915_gem_shrink(struct i915_gem_ww_ctx *ww,
}
if (!__i915_gem_object_put_pages(obj)) {
- try_to_writeback(obj, shrink);
- count += obj->base.size >>
PAGE_SHIFT;
+ if (!try_to_writeback(obj, shrink))
+ count += obj->base.size >>
PAGE_SHIFT;
}
if (!ww)
i915_gem_object_unlock(obj);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
index a77e90f300fe..c7402995a8f9 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
@@ -35,6 +35,8 @@
* @ttm: The base TTM page vector.
* @dev: The struct device used for dma mapping and unmapping.
* @cached_st: The cached scatter-gather table.
+ * @is_shmem: Set if using shmem.
+ * @filp: The shmem file, if using shmem backend.
*
* Note that DMA may be going on right up to the point where the page-
* vector is unpopulated in delayed destroy. Hence keep the @@ -46,6
+48,9 @@ struct i915_ttm_tt {
struct ttm_tt ttm;
struct device *dev;
struct sg_table *cached_st;
+
+ bool is_shmem;
+ struct file *filp;
};
static const struct ttm_place sys_placement_flags = { @@ -179,12 +184,90
@@ i915_ttm_placement_from_obj(const struct drm_i915_gem_object *obj,
placement->busy_placement = busy;
}
+static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev,
+ struct ttm_tt *ttm,
+ struct ttm_operation_ctx *ctx) {
+ struct drm_i915_private *i915 = container_of(bdev, typeof(*i915),
bdev);
+ struct intel_memory_region *mr = i915-
mm.regions[INTEL_MEMORY_SYSTEM];
+ struct i915_ttm_tt *i915_tt = container_of(ttm, typeof(*i915_tt),
ttm);
+ const unsigned int max_segment = i915_sg_segment_size();
+ const size_t size = ttm->num_pages << PAGE_SHIFT;
+ struct file *filp = i915_tt->filp;
+ struct sgt_iter sgt_iter;
+ struct sg_table *st;
+ struct page *page;
+ unsigned long i;
+ int err;
+
+ if (!filp) {
+ struct address_space *mapping;
+ gfp_t mask;
+
+ filp = shmem_file_setup("i915-shmem-tt", size,
VM_NORESERVE);
+ if (IS_ERR(filp))
+ return PTR_ERR(filp);
+
+ mask = GFP_HIGHUSER | __GFP_RECLAIMABLE;
+
+ mapping = filp->f_mapping;
+ mapping_set_gfp_mask(mapping, mask);
+ GEM_BUG_ON(!(mapping_gfp_mask(mapping) &
__GFP_RECLAIM));
+
+ i915_tt->filp = filp;
+ }
+
+ st = shmem_alloc_st(i915, size, mr, filp->f_mapping, max_segment);
+ if (IS_ERR(st))
+ return PTR_ERR(st);
+
+ err = dma_map_sg_attrs(i915_tt->dev,
+ st->sgl, st->nents,
+ PCI_DMA_BIDIRECTIONAL,
+ DMA_ATTR_SKIP_CPU_SYNC |
+ DMA_ATTR_NO_KERNEL_MAPPING |
+ DMA_ATTR_NO_WARN);
+ if (err <= 0) {
+ err = -EINVAL;
+ goto err_free_st;
+ }
+
+ i = 0;
+ for_each_sgt_page(page, sgt_iter, st)
+ ttm->pages[i++] = page;
+
+ if (ttm->page_flags & TTM_TT_FLAG_SWAPPED)
+ ttm->page_flags &= ~TTM_TT_FLAG_SWAPPED;
+
+ i915_tt->cached_st = st;
+ return 0;
+
+err_free_st:
+ shmem_free_st(st, filp->f_mapping, false, false);
+ return err;
+}
+
+static void i915_ttm_tt_shmem_unpopulate(struct ttm_tt *ttm) {
+ struct i915_ttm_tt *i915_tt = container_of(ttm, typeof(*i915_tt),
ttm);
+ bool backup = ttm->page_flags & TTM_TT_FLAG_SWAPPED;
+
+ dma_unmap_sg(i915_tt->dev, i915_tt->cached_st->sgl,
+ i915_tt->cached_st->nents,
+ PCI_DMA_BIDIRECTIONAL);
+
+ shmem_free_st(fetch_and_zero(&i915_tt->cached_st),
+ file_inode(i915_tt->filp)->i_mapping,
+ backup, backup);
Should we do something to undo the shmem_file_setup operation here?
From its implementation it does take a reference counter of inode and
allocate file:
https://elixir.bootlin.com/linux/latest/source/mm/shmem.c#L4084
Regards,
Oak
Hi, Oak,
That's done in i915_ttm_tt_destroy() afaict.
/Thomas
