mm/vmalloc: remove preempt_disable/enable when do preloading

alloc_vmap_area() is given a gfp_mask for the page allocator. Let's respect that mask and consider it even in the case when doing regular CPU preloading, i.e. where a context can sleep.

Signed-off-by Uladzislau Rezki (Sony) <[email protected]>

---
 mm/vmalloc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

mm/vmalloc.c +4 -4

index b7b443bfdd92..593bf554518d 100644
@@ -1064,9 +1064,9 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
 		return ERR_PTR(-EBUSY);
 
 	might_sleep();
+	gfp_mask = gfp_mask & GFP_RECLAIM_MASK;
 
-	va = kmem_cache_alloc_node(vmap_area_cachep,
-			gfp_mask & GFP_RECLAIM_MASK, node);
+	va = kmem_cache_alloc_node(vmap_area_cachep, gfp_mask, node);
 	if (unlikely(!va))
 		return ERR_PTR(-ENOMEM);
 

@@ -1074,7 +1074,7 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
 	 * Only scan the relevant parts containing pointers to other objects
 	 * to avoid false negatives.
 	 */
-	kmemleak_scan_area(&va->rb_node, SIZE_MAX, gfp_mask & GFP_RECLAIM_MASK);
+	kmemleak_scan_area(&va->rb_node, SIZE_MAX, gfp_mask);
 
 retry:
 	/*

@@ -1100,7 +1100,7 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
 		 * Just proceed as it is. If needed "overflow" path
 		 * will refill the cache we allocate from.
 		 */
-		pva = kmem_cache_alloc_node(vmap_area_cachep, GFP_KERNEL, node);
+		pva = kmem_cache_alloc_node(vmap_area_cachep, gfp_mask, node);
 
 	spin_lock(&vmap_area_lock);
 

Signature · 1 line

2.20.1

Quoted 3 lines

alloc_vmap_area() is given a gfp_mask for the page allocator. Let's respect that mask and consider it even in the case when doing regular CPU preloading, i.e. where a context can sleep.

This is explaining what but it doesn't say why. I would go with "
Allocation functions should comply with the given gfp_mask as much as possible. The preallocation code in alloc_vmap_area doesn't follow that pattern and it is using a hardcoded GFP_KERNEL. Although this doesn't really make much difference because vmalloc is not GFP_NOWAIT compliant in general (e.g. page table allocations are GFP_KERNEL) there is no reason to spread that bad habit and it is good to fix the antipattern. "

Quoted 4 lines

Quoted 3 lines

alloc_vmap_area() is given a gfp_mask for the page allocator. Let's respect that mask and consider it even in the case when doing regular CPU preloading, i.e. where a context can sleep.

Quoted 9 lines

This is explaining what but it doesn't say why. I would go with "
Allocation functions should comply with the given gfp_mask as much as possible. The preallocation code in alloc_vmap_area doesn't follow that pattern and it is using a hardcoded GFP_KERNEL. Although this doesn't really make much difference because vmalloc is not GFP_NOWAIT compliant in general (e.g. page table allocations are GFP_KERNEL) there is no reason to spread that bad habit and it is good to fix the antipattern. "

I can go with that, agree. I am not sure if i need to update the patch and send v4. Or maybe Andrew can directly update it in his tree.

Andrew, should i send or can update?

Quoted 10 lines

Quoted 10 lines

This is explaining what but it doesn't say why. I would go with "
Allocation functions should comply with the given gfp_mask as much as possible. The preallocation code in alloc_vmap_area doesn't follow that pattern and it is using a hardcoded GFP_KERNEL. Although this doesn't really make much difference because vmalloc is not GFP_NOWAIT compliant in general (e.g. page table allocations are GFP_KERNEL) there is no reason to spread that bad habit and it is good to fix the antipattern. "

Quoted 4 lines

I can go with that, agree. I am not sure if i need to update the patch and send v4. Or maybe Andrew can directly update it in his tree.

Andrew, should i send or can update?

I updated the changelog with Michal's words prior to committing. You were cc'ed :)

When fit type is NE_FIT_TYPE there is a need in one extra object. Usually the "ne_fit_preload_node" per-CPU variable has it and there is no need in GFP_NOWAIT allocation, but there are exceptions.

This commit just adds more explanations, as a result giving answers on questions like when it can occur, how often, under which conditions and what happens if GFP_NOWAIT gets failed.

Signed-off-by Uladzislau Rezki (Sony) <[email protected]>

---
 mm/vmalloc.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

mm/vmalloc.c +13 -0

index 593bf554518d..2290a0d270e4 100644
@@ -969,6 +969,19 @@ adjust_va_to_fit_type(struct vmap_area *va,
 			 * There are a few exceptions though, as an example it is
 			 * a first allocation (early boot up) when we have "one"
 			 * big free space that has to be split.
+			 *
+			 * Also we can hit this path in case of regular "vmap"
+			 * allocations, if "this" current CPU was not preloaded.
+			 * See the comment in alloc_vmap_area() why. If so, then
+			 * GFP_NOWAIT is used instead to get an extra object for
+			 * split purpose. That is rare and most time does not
+			 * occur.
+			 *
+			 * What happens if an allocation gets failed. Basically,
+			 * an "overflow" path is triggered to purge lazily freed
+			 * areas to free some memory, then, the "retry" path is
+			 * triggered to repeat one more time. See more details
+			 * in alloc_vmap_area() function.
 			 */
 			lva = kmem_cache_alloc(vmap_area_cachep, GFP_NOWAIT);
 			if (!lva)

Signature · 1 line

2.20.1

Quoted 4 lines · Uladzislau Rezki (Sony)

When fit type is NE_FIT_TYPE there is a need in one extra object. Usually the "ne_fit_preload_node" per-CPU variable has it and there is no need in GFP_NOWAIT allocation, but there are exceptions.

Quoted 5 lines

This commit just adds more explanations, as a result giving answers on questions like when it can occur, how often, under which conditions and what happens if GFP_NOWAIT gets failed.

Signed-off-by Uladzislau Rezki (Sony) <[email protected]>

Acked-by Michal Hocko <[email protected]>

Some background. The preemption was disabled before to guarantee that a preloaded object is available for a CPU, it was stored for.

The aim was to not allocate in atomic context when spinlock is taken later, for regular vmap allocations. But that approach conflicts with CONFIG_PREEMPT_RT philosophy. It means that calling spin_lock() with disabled preemption is forbidden in the CONFIG_PREEMPT_RT kernel.

Therefore, get rid of preempt_disable() and preempt_enable() when the preload is done for splitting purpose. As a result we do not guarantee now that a CPU is preloaded, instead we minimize the case when it is not, with this change.

For example i run the special test case that follows the preload pattern and path. 20 "unbind" threads run it and each does 1000000 allocations. Only 3.5 times among 1000000 a CPU was not preloaded. So it can happen but the number is negligible.

V2 - > V3:
- update the commit message

V1 -> V2:
- move __this_cpu_cmpxchg check when spin_lock is taken, as proposed by Andrew Morton
- add more explanation in regard of preloading
- adjust and move some comments

Fixes 82dd23e84be3 ("mm/vmalloc.c: preload a CPU with one object for split purpose")

Reviewed-by Steven Rostedt (VMware) <[email protected]>

Acked-by Sebastian Andrzej Siewior <[email protected]>

Acked-by Daniel Wagner <[email protected]>

Signed-off-by Uladzislau Rezki (Sony) <[email protected]>

---
 mm/vmalloc.c | 37 ++++++++++++++++++++-----------------
 1 file changed, 20 insertions(+), 17 deletions(-)

mm/vmalloc.c +20 -17

index e92ff5f7dd8b..b7b443bfdd92 100644
@@ -1078,31 +1078,34 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
 
 retry:
 	/*
-	 * Preload this CPU with one extra vmap_area object to ensure
-	 * that we have it available when fit type of free area is
-	 * NE_FIT_TYPE.
+	 * Preload this CPU with one extra vmap_area object. It is used
+	 * when fit type of free area is NE_FIT_TYPE. Please note, it
+	 * does not guarantee that an allocation occurs on a CPU that
+	 * is preloaded, instead we minimize the case when it is not.
+	 * It can happen because of cpu migration, because there is a
+	 * race until the below spinlock is taken.
 	 *
 	 * The preload is done in non-atomic context, thus it allows us
 	 * to use more permissive allocation masks to be more stable under
-	 * low memory condition and high memory pressure.
+	 * low memory condition and high memory pressure. In rare case,
+	 * if not preloaded, GFP_NOWAIT is used.
 	 *
-	 * Even if it fails we do not really care about that. Just proceed
-	 * as it is. "overflow" path will refill the cache we allocate from.
+	 * Set "pva" to NULL here, because of "retry" path.
 	 */
-	preempt_disable();
-	if (!__this_cpu_read(ne_fit_preload_node)) {
-		preempt_enable();
-		pva = kmem_cache_alloc_node(vmap_area_cachep, GFP_KERNEL, node);
-		preempt_disable();
+	pva = NULL;
 
-		if (__this_cpu_cmpxchg(ne_fit_preload_node, NULL, pva)) {
-			if (pva)
-				kmem_cache_free(vmap_area_cachep, pva);
-		}
-	}
+	if (!this_cpu_read(ne_fit_preload_node))
+		/*
+		 * Even if it fails we do not really care about that.
+		 * Just proceed as it is. If needed "overflow" path
+		 * will refill the cache we allocate from.
+		 */
+		pva = kmem_cache_alloc_node(vmap_area_cachep, GFP_KERNEL, node);
 
 	spin_lock(&vmap_area_lock);
-	preempt_enable();
+
+	if (pva && __this_cpu_cmpxchg(ne_fit_preload_node, NULL, pva))
+		kmem_cache_free(vmap_area_cachep, pva);
 
 	/*
 	 * If an allocation fails, the "vend" address is

Signature · 1 line

2.20.1

Quoted 2 lines

Some background. The preemption was disabled before to guarantee that a preloaded object is available for a CPU, it was stored for.

Probably good to be explicit that this has been achieved by combining the disabling the preemption and taking the spin lock while the ne_fit_preload_node is checked resp. repopulated, right?

Quoted 10 lines

The aim was to not allocate in atomic context when spinlock is taken later, for regular vmap allocations. But that approach conflicts with CONFIG_PREEMPT_RT philosophy. It means that calling spin_lock() with disabled preemption is forbidden in the CONFIG_PREEMPT_RT kernel.

Therefore, get rid of preempt_disable() and preempt_enable() when the preload is done for splitting purpose. As a result we do not guarantee now that a CPU is preloaded, instead we minimize the case when it is not, with this change.

by populating the per cpu preload pointer under the vmap_area_lock. This implies that at least each caller which has done the preallocation will not fallback to an atomic allocation later. It is possible that the preallocation would be pointless or that no preallocation is done because of the race but your data shows that this is really rare.