To be able to switch to a spinlock and reduce lock contention in the TTM shrinker we don't want to hold a mutex while unmapping and freeing pages from the pool.
But then we somehow need to prevent a race between (for example) the shrinker trying to free pages and hotplug trying to remove the device which those pages belong to.
Taking and releasing the shrinker semaphore on the write side after unmapping and freeing all pages should make sure that no shrinker is running in paralell any more.
Signed-off-by: Christian König christian.koenig@amd.com --- include/linux/shrinker.h | 1 + mm/vmscan.c | 10 ++++++++++ 2 files changed, 11 insertions(+)
diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h index 0f80123650e2..6b75dc372fce 100644 --- a/include/linux/shrinker.h +++ b/include/linux/shrinker.h @@ -92,4 +92,5 @@ extern void register_shrinker_prepared(struct shrinker *shrinker); extern int register_shrinker(struct shrinker *shrinker); extern void unregister_shrinker(struct shrinker *shrinker); extern void free_prealloced_shrinker(struct shrinker *shrinker); +extern void sync_shrinkers(void); #endif diff --git a/mm/vmscan.c b/mm/vmscan.c index 562e87cbd7a1..46cd9c215d73 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -408,6 +408,16 @@ void unregister_shrinker(struct shrinker *shrinker) } EXPORT_SYMBOL(unregister_shrinker);
+/** + * sync_shrinker - Wait for all running shrinkers to complete. + */ +void sync_shrinkers(void) +{ + down_write(&shrinker_rwsem); + up_write(&shrinker_rwsem); +} +EXPORT_SYMBOL(sync_shrinkers); + #define SHRINK_BATCH 128
static unsigned long do_shrink_slab(struct shrink_control *shrinkctl,
Switch back to using a spinlock again by moving the IOMMU unmap outside of the locked region.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_pool.c | 40 +++++++++++++++------------------- 1 file changed, 18 insertions(+), 22 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.c index cb38b1a17b09..a8b4abe687ce 100644 --- a/drivers/gpu/drm/ttm/ttm_pool.c +++ b/drivers/gpu/drm/ttm/ttm_pool.c @@ -70,7 +70,7 @@ static struct ttm_pool_type global_uncached[MAX_ORDER]; static struct ttm_pool_type global_dma32_write_combined[MAX_ORDER]; static struct ttm_pool_type global_dma32_uncached[MAX_ORDER];
-static struct mutex shrinker_lock; +static spinlock_t shrinker_lock; static struct list_head shrinker_list; static struct shrinker mm_shrinker;
@@ -263,9 +263,9 @@ static void ttm_pool_type_init(struct ttm_pool_type *pt, struct ttm_pool *pool, spin_lock_init(&pt->lock); INIT_LIST_HEAD(&pt->pages);
- mutex_lock(&shrinker_lock); + spin_lock(&shrinker_lock); list_add_tail(&pt->shrinker_list, &shrinker_list); - mutex_unlock(&shrinker_lock); + spin_unlock(&shrinker_lock); }
/* Remove a pool_type from the global shrinker list and free all pages */ @@ -273,9 +273,9 @@ static void ttm_pool_type_fini(struct ttm_pool_type *pt) { struct page *p;
- mutex_lock(&shrinker_lock); + spin_lock(&shrinker_lock); list_del(&pt->shrinker_list); - mutex_unlock(&shrinker_lock); + spin_unlock(&shrinker_lock);
while ((p = ttm_pool_type_take(pt))) ttm_pool_free_page(pt->pool, pt->caching, pt->order, p); @@ -313,24 +313,19 @@ static struct ttm_pool_type *ttm_pool_select_type(struct ttm_pool *pool, static unsigned int ttm_pool_shrink(void) { struct ttm_pool_type *pt; - unsigned int num_freed; struct page *p;
- mutex_lock(&shrinker_lock); + spin_lock(&shrinker_lock); pt = list_first_entry(&shrinker_list, typeof(*pt), shrinker_list); + list_move_tail(&pt->shrinker_list, &shrinker_list); + spin_unlock(&shrinker_lock);
p = ttm_pool_type_take(pt); - if (p) { - ttm_pool_free_page(pt->pool, pt->caching, pt->order, p); - num_freed = 1 << pt->order; - } else { - num_freed = 0; - } - - list_move_tail(&pt->shrinker_list, &shrinker_list); - mutex_unlock(&shrinker_lock); + if (!p) + return 0;
- return num_freed; + ttm_pool_free_page(pt->pool, pt->caching, pt->order, p); + return 1 << pt->order; }
/* Return the allocation order based for a page */ @@ -530,6 +525,7 @@ void ttm_pool_fini(struct ttm_pool *pool) for (j = 0; j < MAX_ORDER; ++j) ttm_pool_type_fini(&pool->caching[i].orders[j]); } + sync_shrinkers(); }
/* As long as pages are available make sure to release at least one */ @@ -604,7 +600,7 @@ static int ttm_pool_debugfs_globals_show(struct seq_file *m, void *data) { ttm_pool_debugfs_header(m);
- mutex_lock(&shrinker_lock); + spin_lock(&shrinker_lock); seq_puts(m, "wc\t:"); ttm_pool_debugfs_orders(global_write_combined, m); seq_puts(m, "uc\t:"); @@ -613,7 +609,7 @@ static int ttm_pool_debugfs_globals_show(struct seq_file *m, void *data) ttm_pool_debugfs_orders(global_dma32_write_combined, m); seq_puts(m, "uc 32\t:"); ttm_pool_debugfs_orders(global_dma32_uncached, m); - mutex_unlock(&shrinker_lock); + spin_unlock(&shrinker_lock);
ttm_pool_debugfs_footer(m);
@@ -640,7 +636,7 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m)
ttm_pool_debugfs_header(m);
- mutex_lock(&shrinker_lock); + spin_lock(&shrinker_lock); for (i = 0; i < TTM_NUM_CACHING_TYPES; ++i) { seq_puts(m, "DMA "); switch (i) { @@ -656,7 +652,7 @@ int ttm_pool_debugfs(struct ttm_pool *pool, struct seq_file *m) } ttm_pool_debugfs_orders(pool->caching[i].orders, m); } - mutex_unlock(&shrinker_lock); + spin_unlock(&shrinker_lock);
ttm_pool_debugfs_footer(m); return 0; @@ -693,7 +689,7 @@ int ttm_pool_mgr_init(unsigned long num_pages) if (!page_pool_size) page_pool_size = num_pages;
- mutex_init(&shrinker_lock); + spin_lock_init(&shrinker_lock); INIT_LIST_HEAD(&shrinker_list);
for (i = 0; i < MAX_ORDER; ++i) {
On 4/9/21 9:17 AM, Christian König wrote:
To be able to switch to a spinlock and reduce lock contention in the TTM shrinker we don't want to hold a mutex while unmapping and freeing pages from the pool.
Does using spinlock instead of mutex really reduce lock contention?
But then we somehow need to prevent a race between (for example) the shrinker trying to free pages and hotplug trying to remove the device which those pages belong to.
Taking and releasing the shrinker semaphore on the write side after unmapping and freeing all pages should make sure that no shrinker is running in paralell any more.
So you explain this in this commit log for adding the function, but then the next patch just adds a sync_shrinkers() call without any comment. I would expect there a comment explaining why it's done there - what it protects against, as it's not an obvious pattern IMHO.
Signed-off-by: Christian König christian.koenig@amd.com
include/linux/shrinker.h | 1 + mm/vmscan.c | 10 ++++++++++ 2 files changed, 11 insertions(+)
diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h index 0f80123650e2..6b75dc372fce 100644 --- a/include/linux/shrinker.h +++ b/include/linux/shrinker.h @@ -92,4 +92,5 @@ extern void register_shrinker_prepared(struct shrinker *shrinker); extern int register_shrinker(struct shrinker *shrinker); extern void unregister_shrinker(struct shrinker *shrinker); extern void free_prealloced_shrinker(struct shrinker *shrinker); +extern void sync_shrinkers(void); #endif diff --git a/mm/vmscan.c b/mm/vmscan.c index 562e87cbd7a1..46cd9c215d73 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -408,6 +408,16 @@ void unregister_shrinker(struct shrinker *shrinker) } EXPORT_SYMBOL(unregister_shrinker);
+/**
- sync_shrinker - Wait for all running shrinkers to complete.
- */
+void sync_shrinkers(void) +{
- down_write(&shrinker_rwsem);
- up_write(&shrinker_rwsem);
+} +EXPORT_SYMBOL(sync_shrinkers);
#define SHRINK_BATCH 128
static unsigned long do_shrink_slab(struct shrink_control *shrinkctl,
Am 09.04.21 um 13:00 schrieb Vlastimil Babka:
On 4/9/21 9:17 AM, Christian König wrote:
To be able to switch to a spinlock and reduce lock contention in the TTM shrinker we don't want to hold a mutex while unmapping and freeing pages from the pool.
Does using spinlock instead of mutex really reduce lock contention?
Well using the spinlock instead of the mutex is only the cherry on the cake.
The real improvement for the contention is the fact that we just grab the next pool and drop the lock again instead of doing the whole IOMMU unmap and flushing of the CPU TLB dance while holding the lock.
But then we somehow need to prevent a race between (for example) the shrinker trying to free pages and hotplug trying to remove the device which those pages belong to.
Taking and releasing the shrinker semaphore on the write side after unmapping and freeing all pages should make sure that no shrinker is running in paralell any more.
So you explain this in this commit log for adding the function, but then the next patch just adds a sync_shrinkers() call without any comment. I would expect there a comment explaining why it's done there - what it protects against, as it's not an obvious pattern IMHO.
Good point, going to add a comment.
Thanks, Christian.
Signed-off-by: Christian König christian.koenig@amd.com
include/linux/shrinker.h | 1 + mm/vmscan.c | 10 ++++++++++ 2 files changed, 11 insertions(+)
diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h index 0f80123650e2..6b75dc372fce 100644 --- a/include/linux/shrinker.h +++ b/include/linux/shrinker.h @@ -92,4 +92,5 @@ extern void register_shrinker_prepared(struct shrinker *shrinker); extern int register_shrinker(struct shrinker *shrinker); extern void unregister_shrinker(struct shrinker *shrinker); extern void free_prealloced_shrinker(struct shrinker *shrinker); +extern void sync_shrinkers(void); #endif diff --git a/mm/vmscan.c b/mm/vmscan.c index 562e87cbd7a1..46cd9c215d73 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -408,6 +408,16 @@ void unregister_shrinker(struct shrinker *shrinker) } EXPORT_SYMBOL(unregister_shrinker);
+/**
- sync_shrinker - Wait for all running shrinkers to complete.
- */
+void sync_shrinkers(void) +{
- down_write(&shrinker_rwsem);
- up_write(&shrinker_rwsem);
+} +EXPORT_SYMBOL(sync_shrinkers);
#define SHRINK_BATCH 128
static unsigned long do_shrink_slab(struct shrink_control *shrinkctl,
dri-devel@lists.freedesktop.org
-
Christian König -
Vlastimil Babka