[PATCH 1/3] mm/page_alloc: use might_alloc()

[PATCH 1/3] mm/page_alloc: use might_alloc()

[Daniel Vetter]

... instead of open codding it. Completely equivalent code, just
a notch more meaningful when reading.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-mm@kvack.org
---
 mm/page_alloc.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 2db95780e003..277774d170cb 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -5177,10 +5177,7 @@ static inline bool prepare_alloc_pages(gfp_t gfp_mask, unsigned int order,
 			*alloc_flags |= ALLOC_CPUSET;
 	}
 
-	fs_reclaim_acquire(gfp_mask);
-	fs_reclaim_release(gfp_mask);
-
-	might_sleep_if(gfp_mask & __GFP_DIRECT_RECLAIM);
+	might_alloc(gfp_mask);
 
 	if (should_fail_alloc_page(gfp_mask, order))
 		return false;
-- 
2.36.0

[PATCH 2/3] mm/slab: delete cache_alloc_debugcheck_before()

[Daniel Vetter]

It only does a might_sleep_if(GFP_RECLAIM) check, which is already
covered by the might_alloc() in slab_pre_alloc_hook(). And all callers
of cache_alloc_debugcheck_before() call that beforehand already.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: linux-mm@kvack.org
---
 mm/slab.c | 10 ----------
 1 file changed, 10 deletions(-)

diff --git a/mm/slab.c b/mm/slab.c
index b04e40078bdf..75779ac5f5ba 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -2973,12 +2973,6 @@ static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
 	return ac->entry[--ac->avail];
 }
 
-static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
-						gfp_t flags)
-{
-	might_sleep_if(gfpflags_allow_blocking(flags));
-}
-
 #if DEBUG
 static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
 				gfp_t flags, void *objp, unsigned long caller)
@@ -3219,7 +3213,6 @@ slab_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid, size_t orig_
 	if (unlikely(ptr))
 		goto out_hooks;
 
-	cache_alloc_debugcheck_before(cachep, flags);
 	local_irq_save(save_flags);
 
 	if (nodeid == NUMA_NO_NODE)
@@ -3304,7 +3297,6 @@ slab_alloc(struct kmem_cache *cachep, struct list_lru *lru, gfp_t flags,
 	if (unlikely(objp))
 		goto out;
 
-	cache_alloc_debugcheck_before(cachep, flags);
 	local_irq_save(save_flags);
 	objp = __do_cache_alloc(cachep, flags);
 	local_irq_restore(save_flags);
@@ -3541,8 +3533,6 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
 	if (!s)
 		return 0;
 
-	cache_alloc_debugcheck_before(s, flags);
-
 	local_irq_disable();
 	for (i = 0; i < size; i++) {
 		void *objp = kfence_alloc(s, s->object_size, flags) ?: __do_cache_alloc(s, flags);
-- 
2.36.0

Re: [PATCH 2/3] mm/slab: delete cache_alloc_debugcheck_before()

[David Hildenbrand]

On 05.06.22 17:25, Daniel Vetter wrote:
> It only does a might_sleep_if(GFP_RECLAIM) check, which is already
> covered by the might_alloc() in slab_pre_alloc_hook(). And all callers
> of cache_alloc_debugcheck_before() call that beforehand already.
> 
> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
> Cc: Christoph Lameter <cl@linux.com>
> Cc: Pekka Enberg <penberg@kernel.org>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Vlastimil Babka <vbabka@suse.cz>
> Cc: Roman Gushchin <roman.gushchin@linux.dev>
> Cc: linux-mm@kvack.org
> ---

LGTM

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 
Thanks,

David / dhildenb

Re: [PATCH 2/3] mm/slab: delete cache_alloc_debugcheck_before()

[David Rientjes]

On Sun, 5 Jun 2022, Daniel Vetter wrote:

> It only does a might_sleep_if(GFP_RECLAIM) check, which is already
> covered by the might_alloc() in slab_pre_alloc_hook(). And all callers
> of cache_alloc_debugcheck_before() call that beforehand already.
> 
> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
> Cc: Christoph Lameter <cl@linux.com>
> Cc: Pekka Enberg <penberg@kernel.org>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Vlastimil Babka <vbabka@suse.cz>
> Cc: Roman Gushchin <roman.gushchin@linux.dev>
> Cc: linux-mm@kvack.org

Acked-by: David Rientjes <rientjes@google.com>

Re: [PATCH 2/3] mm/slab: delete cache_alloc_debugcheck_before()

[Muchun Song]

On Sun, Jun 05, 2022 at 05:25:38PM +0200, Daniel Vetter wrote:
> It only does a might_sleep_if(GFP_RECLAIM) check, which is already
> covered by the might_alloc() in slab_pre_alloc_hook(). And all callers
> of cache_alloc_debugcheck_before() call that beforehand already.
> 
> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

Nice cleanup.

Reviewed-by: Muchun Song <songmuchun@bytedance.com>

Thanks.

Re: [PATCH 2/3] mm/slab: delete cache_alloc_debugcheck_before()

[Vlastimil Babka]

On 6/5/22 17:25, Daniel Vetter wrote:
> It only does a might_sleep_if(GFP_RECLAIM) check, which is already
> covered by the might_alloc() in slab_pre_alloc_hook(). And all callers
> of cache_alloc_debugcheck_before() call that beforehand already.
> 
> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
> Cc: Christoph Lameter <cl@linux.com>
> Cc: Pekka Enberg <penberg@kernel.org>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Vlastimil Babka <vbabka@suse.cz>
> Cc: Roman Gushchin <roman.gushchin@linux.dev>
> Cc: linux-mm@kvack.org

Thanks, added to slab/for-5.20/cleanup as it's slab-specific and independent
from 1/3 and 3/3.

> ---
>  mm/slab.c | 10 ----------
>  1 file changed, 10 deletions(-)
> 
> diff --git a/mm/slab.c b/mm/slab.c
> index b04e40078bdf..75779ac5f5ba 100644
> --- a/mm/slab.c
> +++ b/mm/slab.c
> @@ -2973,12 +2973,6 @@ static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
>  	return ac->entry[--ac->avail];
>  }
>  
> -static inline void cache_alloc_debugcheck_before(struct kmem_cache *cachep,
> -						gfp_t flags)
> -{
> -	might_sleep_if(gfpflags_allow_blocking(flags));
> -}
> -
>  #if DEBUG
>  static void *cache_alloc_debugcheck_after(struct kmem_cache *cachep,
>  				gfp_t flags, void *objp, unsigned long caller)
> @@ -3219,7 +3213,6 @@ slab_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid, size_t orig_
>  	if (unlikely(ptr))
>  		goto out_hooks;
>  
> -	cache_alloc_debugcheck_before(cachep, flags);
>  	local_irq_save(save_flags);
>  
>  	if (nodeid == NUMA_NO_NODE)
> @@ -3304,7 +3297,6 @@ slab_alloc(struct kmem_cache *cachep, struct list_lru *lru, gfp_t flags,
>  	if (unlikely(objp))
>  		goto out;
>  
> -	cache_alloc_debugcheck_before(cachep, flags);
>  	local_irq_save(save_flags);
>  	objp = __do_cache_alloc(cachep, flags);
>  	local_irq_restore(save_flags);
> @@ -3541,8 +3533,6 @@ int kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t flags, size_t size,
>  	if (!s)
>  		return 0;
>  
> -	cache_alloc_debugcheck_before(s, flags);
> -
>  	local_irq_disable();
>  	for (i = 0; i < size; i++) {
>  		void *objp = kfence_alloc(s, s->object_size, flags) ?: __do_cache_alloc(s, flags);

[PATCH 3/3] mm/mempool: use might_alloc()

[Daniel Vetter]

mempool are generally used for GFP_NOIO, so this wont benefit all that
much because might_alloc currently only checks GFP_NOFS. But it does
validate against mmu notifier pte zapping, some might catch some
drivers doing really silly things, plus it's a bit more meaningful in
what we're checking for here.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-mm@kvack.org
---
 mm/mempool.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/mempool.c b/mm/mempool.c
index b933d0fc21b8..96488b13a1ef 100644
--- a/mm/mempool.c
+++ b/mm/mempool.c
@@ -379,7 +379,7 @@ void *mempool_alloc(mempool_t *pool, gfp_t gfp_mask)
 	gfp_t gfp_temp;
 
 	VM_WARN_ON_ONCE(gfp_mask & __GFP_ZERO);
-	might_sleep_if(gfp_mask & __GFP_DIRECT_RECLAIM);
+	might_alloc(gfp_mask);
 
 	gfp_mask |= __GFP_NOMEMALLOC;	/* don't allocate emergency reserves */
 	gfp_mask |= __GFP_NORETRY;	/* don't loop in __alloc_pages */
-- 
2.36.0

Re: [PATCH 3/3] mm/mempool: use might_alloc()

[Vlastimil Babka (SUSE)]

On 6/5/22 17:25, Daniel Vetter wrote:
> mempool are generally used for GFP_NOIO, so this wont benefit all that
> much because might_alloc currently only checks GFP_NOFS. But it does
> validate against mmu notifier pte zapping, some might catch some
> drivers doing really silly things, plus it's a bit more meaningful in
> what we're checking for here.
> 
> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: linux-mm@kvack.org

Reviewed-by: Vlastimil Babka <vbabka@suse.cz>

> ---
>  mm/mempool.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/mempool.c b/mm/mempool.c
> index b933d0fc21b8..96488b13a1ef 100644
> --- a/mm/mempool.c
> +++ b/mm/mempool.c
> @@ -379,7 +379,7 @@ void *mempool_alloc(mempool_t *pool, gfp_t gfp_mask)
>  	gfp_t gfp_temp;
>  
>  	VM_WARN_ON_ONCE(gfp_mask & __GFP_ZERO);
> -	might_sleep_if(gfp_mask & __GFP_DIRECT_RECLAIM);
> +	might_alloc(gfp_mask);
>  
>  	gfp_mask |= __GFP_NOMEMALLOC;	/* don't allocate emergency reserves */
>  	gfp_mask |= __GFP_NORETRY;	/* don't loop in __alloc_pages */

Re: [PATCH 1/3] mm/page_alloc: use might_alloc()

[David Hildenbrand]

On 05.06.22 17:25, Daniel Vetter wrote:
> ... instead of open codding it. Completely equivalent code, just
> a notch more meaningful when reading.
> 
> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: linux-mm@kvack.org
> ---
>  mm/page_alloc.c | 5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 2db95780e003..277774d170cb 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -5177,10 +5177,7 @@ static inline bool prepare_alloc_pages(gfp_t gfp_mask, unsigned int order,
>  			*alloc_flags |= ALLOC_CPUSET;
>  	}
>  
> -	fs_reclaim_acquire(gfp_mask);
> -	fs_reclaim_release(gfp_mask);
> -
> -	might_sleep_if(gfp_mask & __GFP_DIRECT_RECLAIM);
> +	might_alloc(gfp_mask);
>  
>  	if (should_fail_alloc_page(gfp_mask, order))
>  		return false;

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 
Thanks,

David / dhildenb

Re: [PATCH 1/3] mm/page_alloc: use might_alloc()

[Vlastimil Babka (SUSE)]

On 6/5/22 17:25, Daniel Vetter wrote:
> ... instead of open codding it. Completely equivalent code, just
> a notch more meaningful when reading.
> 
> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: linux-mm@kvack.org

Reviewed-by: Vlastimil Babka <vbabka@suse.cz>

> ---
>  mm/page_alloc.c | 5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 2db95780e003..277774d170cb 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -5177,10 +5177,7 @@ static inline bool prepare_alloc_pages(gfp_t gfp_mask, unsigned int order,
>  			*alloc_flags |= ALLOC_CPUSET;
>  	}
>  
> -	fs_reclaim_acquire(gfp_mask);
> -	fs_reclaim_release(gfp_mask);
> -
> -	might_sleep_if(gfp_mask & __GFP_DIRECT_RECLAIM);
> +	might_alloc(gfp_mask);
>  
>  	if (should_fail_alloc_page(gfp_mask, order))
>  		return false;