Skip to contentSkip to conversation

apparmor: global buffers spin lock may get contended

PATCH In review Sergey Senozhatsky 4 participants 21 messages Jul 13, 2021 – Oct 26, 2023 2 messages not in the archive

Subsystems (2)
Security (LSM) Apparmor security module
lore
Share & tools
lore

Sergey Senozhatsky

author
#0 lore raw

Hi,

We've notices that apparmor has switched from using per-CPU buffer pool and per-CPU spin_lock to a global spin_lock in df323337e507a0009d3db1ea.

This seems to be causing some contention on our build machines (with quite a bit of cores). Because that global spin lock is a part of the stat() sys call (and perhaps some other)

E.g.

- 9.29% 0.00% clang++ [kernel.vmlinux]
- 9.28% entry_SYSCALL_64_after_hwframe
- 8.98% do_syscall_64
- 7.43% __do_sys_newlstat
- 7.43% vfs_statx
- 7.18% security_inode_getattr
- 7.15% apparmor_inode_getattr
- aa_path_perm
- 3.53% aa_get_buffer
- 3.47% _raw_spin_lock
3.44% native_queued_spin_lock_slowpath
- 3.49% aa_put_buffer.part.0
- 3.45% _raw_spin_lock
3.43% native_queued_spin_lock_slowpath

Can we fix this contention?

maintainer
#1 lore raw
Quoted 26 lines · Sergey Senozhatsky

Hi,

We've notices that apparmor has switched from using per-CPU buffer pool and per-CPU spin_lock to a global spin_lock in df323337e507a0009d3db1ea.

This seems to be causing some contention on our build machines (with quite a bit of cores). Because that global spin lock is a part of the stat() sys call (and perhaps some other)

E.g.

- 9.29% 0.00% clang++ [kernel.vmlinux]
- 9.28% entry_SYSCALL_64_after_hwframe
- 8.98% do_syscall_64
- 7.43% __do_sys_newlstat
- 7.43% vfs_statx
- 7.18% security_inode_getattr
- 7.15% apparmor_inode_getattr
- aa_path_perm
- 3.53% aa_get_buffer
- 3.47% _raw_spin_lock
3.44% native_queued_spin_lock_slowpath
- 3.49% aa_put_buffer.part.0
- 3.45% _raw_spin_lock
3.43% native_queued_spin_lock_slowpath

Quoted 2 lines

Can we fix this contention?

sorry this got filtered to a wrong mailbox. Yes this is something that can be improved, and was a concern when the switch was made from per-CPU buffers to the global pool.

We can look into doing a hybrid approach where we can per cpu cache a buffer from the global pool. The trick will be coming up with when the cached buffer can be returned so we don't run into the problems that lead to df323337e507a0009d3db1ea

maintainer
#2 lore raw
In reply to Sergey Senozhatsky
Quoted 26 lines · Sergey Senozhatsky

Hi,

We've notices that apparmor has switched from using per-CPU buffer pool and per-CPU spin_lock to a global spin_lock in df323337e507a0009d3db1ea.

This seems to be causing some contention on our build machines (with quite a bit of cores). Because that global spin lock is a part of the stat() sys call (and perhaps some other)

E.g.

- 9.29% 0.00% clang++ [kernel.vmlinux]
- 9.28% entry_SYSCALL_64_after_hwframe
- 8.98% do_syscall_64
- 7.43% __do_sys_newlstat
- 7.43% vfs_statx
- 7.18% security_inode_getattr
- 7.15% apparmor_inode_getattr
- aa_path_perm
- 3.53% aa_get_buffer
- 3.47% _raw_spin_lock
3.44% native_queued_spin_lock_slowpath
- 3.49% aa_put_buffer.part.0
- 3.45% _raw_spin_lock
3.43% native_queued_spin_lock_slowpath

Quoted 1 line

Can we fix this contention?

sorry for the delay on this. Below is a proposed patch that I have been testing to deal with this issue.

From d026988196fdbda7234fb87bc3e4aea22edcbaf9 Mon Sep 17 00:00:00 2001
From: John Johansen <[email protected]>
Date: Tue, 25 Oct 2022 01:18:41 -0700
Subject: [PATCH] apparmor: cache buffers on percpu list if there is lock
contention

On a heavily loaded machine there can be lock contention on the global buffers lock. Add a percpu list to cache buffers on when lock contention is encountered.

When allocating buffers attempt to use cached buffers first, before taking the global buffers lock. When freeing buffers try to put them back to the global list but if contention is encountered, put the buffer on the percpu list.

The length of time a buffer is held on the percpu list is dynamically adjusted based on lock contention. The amount of hold time is rapidly increased and slow ramped down.

Signed-off-by John Johansen <[email protected]>
---
 security/apparmor/lsm.c | 74 ++++++++++++++++++++++++++++++++++++++---
 1 file changed, 69 insertions(+), 5 deletions(-)
security/apparmor/lsm.c +69 -5
index 25114735bc11..0ab70171bdb6 100644
@@ -49,12 +49,19 @@ union aa_buffer {
 	char buffer[1];
 };
 
+struct aa_local_cache {
+	unsigned int contention;
+	unsigned int hold;
+	struct list_head head;
+};
+
 #define RESERVE_COUNT 2
 static int reserve_count = RESERVE_COUNT;
 static int buffer_count;
 
 static LIST_HEAD(aa_global_buffers);
 static DEFINE_SPINLOCK(aa_buffers_lock);
+static DEFINE_PER_CPU(struct aa_local_cache, aa_local_buffers);
 
 /*
  * LSM hook functions

@@ -1622,14 +1629,44 @@ static int param_set_mode(const char *val, const struct kernel_param *kp)
 	return 0;
 }
 
+static void update_contention(struct aa_local_cache *cache)
+{
+	cache->contention += 3;
+	if (cache->contention > 9)
+		cache->contention = 9;
+	cache->hold += 1 << cache->contention;		/* 8, 64, 512 */
+}
+
 char *aa_get_buffer(bool in_atomic)
 {
 	union aa_buffer *aa_buf;
+	struct aa_local_cache *cache;
 	bool try_again = true;
 	gfp_t flags = (GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
 
+	/* use per cpu cached buffers first */
+	cache = get_cpu_ptr(&aa_local_buffers);
+	if (!list_empty(&cache->head)) {
+		aa_buf = list_first_entry(&cache->head, union aa_buffer, list);
+		list_del(&aa_buf->list);
+		cache->hold--;
+		put_cpu_ptr(&aa_local_buffers);
+		return &aa_buf->buffer[0];
+	}
+	put_cpu_ptr(&aa_local_buffers);
+
+	if (!spin_trylock(&aa_buffers_lock)) {
+		cache = get_cpu_ptr(&aa_local_buffers);
+		update_contention(cache);
+		put_cpu_ptr(&aa_local_buffers);
+		spin_lock(&aa_buffers_lock);
+	} else {
+		cache = get_cpu_ptr(&aa_local_buffers);
+		if (cache->contention)
+			cache->contention--;
+		put_cpu_ptr(&aa_local_buffers);
+	}
 retry:
-	spin_lock(&aa_buffers_lock);
 	if (buffer_count > reserve_count ||
 	    (in_atomic && !list_empty(&aa_global_buffers))) {
 		aa_buf = list_first_entry(&aa_global_buffers, union aa_buffer,

@@ -1655,6 +1692,7 @@ char *aa_get_buffer(bool in_atomic)
 	if (!aa_buf) {
 		if (try_again) {
 			try_again = false;
+			spin_lock(&aa_buffers_lock);
 			goto retry;
 		}
 		pr_warn_once("AppArmor: Failed to allocate a memory buffer.\n");

@@ -1666,15 +1704,32 @@ char *aa_get_buffer(bool in_atomic)
 void aa_put_buffer(char *buf)
 {
 	union aa_buffer *aa_buf;
+	struct aa_local_cache *cache;
 
 	if (!buf)
 		return;
 	aa_buf = container_of(buf, union aa_buffer, buffer[0]);
 
-	spin_lock(&aa_buffers_lock);
-	list_add(&aa_buf->list, &aa_global_buffers);
-	buffer_count++;
-	spin_unlock(&aa_buffers_lock);
+	cache = get_cpu_ptr(&aa_local_buffers);
+	if (!cache->hold) {
+		put_cpu_ptr(&aa_local_buffers);
+		if (spin_trylock(&aa_buffers_lock)) {
+			list_add(&aa_buf->list, &aa_global_buffers);
+			buffer_count++;
+			spin_unlock(&aa_buffers_lock);
+			cache = get_cpu_ptr(&aa_local_buffers);
+			if (cache->contention)
+				cache->contention--;
+			put_cpu_ptr(&aa_local_buffers);
+			return;
+		}
+		cache = get_cpu_ptr(&aa_local_buffers);
+		update_contention(cache);
+	}
+
+	/* cache in percpu list */
+	list_add(&aa_buf->list, &cache->head);
+	put_cpu_ptr(&aa_local_buffers);
 }
 
 /*

@@ -1716,6 +1771,15 @@ static int __init alloc_buffers(void)
 	union aa_buffer *aa_buf;
 	int i, num;
 
+	/*
+	 * per cpu set of cached allocated buffers used to help reduce
+	 * lock contention
+	 */
+	for_each_possible_cpu(i) {
+		per_cpu(aa_local_buffers, i).contention = 0;
+		per_cpu(aa_local_buffers, i).hold = 0;
+		INIT_LIST_HEAD(&per_cpu(aa_local_buffers, i).head);
+	}
 	/*
 	 * A function may require two buffers at once. Usually the buffers are
 	 * used for a short period of time and are shared. On UP kernel buffers
Signature · 1 line
2.34.1
author
#3 lore raw
Quoted 15 lines · John Johansen

From d026988196fdbda7234fb87bc3e4aea22edcbaf9 Mon Sep 17 00:00:00 2001
From: John Johansen <[email protected]>
Date: Tue, 25 Oct 2022 01:18:41 -0700
Subject: [PATCH] apparmor: cache buffers on percpu list if there is lock
contention

On a heavily loaded machine there can be lock contention on the global buffers lock. Add a percpu list to cache buffers on when lock contention is encountered.

When allocating buffers attempt to use cached buffers first, before taking the global buffers lock. When freeing buffers try to put them back to the global list but if contention is encountered, put the buffer on the percpu list.

Quoted 5 lines

The length of time a buffer is held on the percpu list is dynamically adjusted based on lock contention. The amount of hold time is rapidly increased and slow ramped down.

Signed-off-by John Johansen <[email protected]>
Reported-by Sergey Senozhatsky <[email protected]>
maintainer
#4 lore raw
Quoted 16 lines · Sergey Senozhatsky
Quoted 15 lines · John Johansen

From d026988196fdbda7234fb87bc3e4aea22edcbaf9 Mon Sep 17 00:00:00 2001
From: John Johansen <[email protected]>
Date: Tue, 25 Oct 2022 01:18:41 -0700
Subject: [PATCH] apparmor: cache buffers on percpu list if there is lock
contention

On a heavily loaded machine there can be lock contention on the global buffers lock. Add a percpu list to cache buffers on when lock contention is encountered.

When allocating buffers attempt to use cached buffers first, before taking the global buffers lock. When freeing buffers try to put them back to the global list but if contention is encountered, put the buffer on the percpu list.

Quoted 7 lines
Quoted 5 lines

The length of time a buffer is held on the percpu list is dynamically adjusted based on lock contention. The amount of hold time is rapidly increased and slow ramped down.

Signed-off-by John Johansen <[email protected]>
Reported-by Sergey Senozhatsky <[email protected]>

yep, thanks for catching that

author
#5 lore raw
Quoted 21 lines · John Johansen
Quoted 20 lines · Sergey Senozhatsky
Quoted 19 lines · John Johansen

From d026988196fdbda7234fb87bc3e4aea22edcbaf9 Mon Sep 17 00:00:00 2001
From: John Johansen <[email protected]>
Date: Tue, 25 Oct 2022 01:18:41 -0700
Subject: [PATCH] apparmor: cache buffers on percpu list if there is lock
contention

On a heavily loaded machine there can be lock contention on the global buffers lock. Add a percpu list to cache buffers on when lock contention is encountered.

When allocating buffers attempt to use cached buffers first, before taking the global buffers lock. When freeing buffers try to put them back to the global list but if contention is encountered, put the buffer on the percpu list.

The length of time a buffer is held on the percpu list is dynamically adjusted based on lock contention. The amount of hold time is rapidly increased and slow ramped down.

Quoted 5 lines
Quoted 3 lines
Quoted 1 line
Signed-off-by John Johansen <[email protected]>
Reported-by Sergey Senozhatsky <[email protected]>

yep, thanks for catching that

Thanks for the patch! Unfortunately it'll be a bit difficult to test it right now; I'll probably have to wait until corp pushes new kernel (with the patch) to build boxes.

maintainer
#6 lore raw

I have sent up a new version of this patch that caps a situation where buffer lists could grow unbounded (at theoretically).

maintainer
#7 lore raw

PATCH v3 1/1 apparmor: global buffers spin lock may get contended

In reply to John Johansen

From f44dee132b0b55386b7ea31e68c80d367b073ee0 Mon Sep 17 00:00:00 2001
From: John Johansen <[email protected]>
Date: Tue, 25 Oct 2022 01:18:41 -0700
Subject: [PATCH] apparmor: cache buffers on percpu list if there is lock
contention

On a heavily loaded machine there can be lock contention on the global buffers lock. Add a percpu list to cache buffers on when lock contention is encountered.

When allocating buffers attempt to use cached buffers first, before taking the global buffers lock. When freeing buffers try to put them back to the global list but if contention is encountered, put the buffer on the percpu list.

The length of time a buffer is held on the percpu list is dynamically adjusted based on lock contention. The amount of hold time is rapidly increased and slow ramped down.

v3:
- limit number of buffers that can be pushed onto the percpu list. This avoids a problem on some kernels where one percpu list can inherit buffers from another cpu after a reschedule, causing more kernel memory to used than is necessary. Under normal conditions this should eventually return to normal but under pathelogical conditions the extra memory consumption may have been unbouanded
v2:
- dynamically adjust buffer hold time on percpu list based on lock contention.
v1:
- cache buffers on percpu list on lock contention

Signed-off-by John Johansen <[email protected]>
---
 security/apparmor/lsm.c | 81 ++++++++++++++++++++++++++++++++++++++---
 1 file changed, 76 insertions(+), 5 deletions(-)
security/apparmor/lsm.c +76 -5
index 25114735bc11..21f5ea20e715 100644
@@ -49,12 +49,19 @@ union aa_buffer {
 	char buffer[1];
 };
 
+struct aa_local_cache {
+	unsigned int contention;
+	unsigned int hold;
+	struct list_head head;
+};
+
 #define RESERVE_COUNT 2
 static int reserve_count = RESERVE_COUNT;
 static int buffer_count;
 
 static LIST_HEAD(aa_global_buffers);
 static DEFINE_SPINLOCK(aa_buffers_lock);
+static DEFINE_PER_CPU(struct aa_local_cache, aa_local_buffers);
 
 /*
  * LSM hook functions

@@ -1622,14 +1629,44 @@ static int param_set_mode(const char *val, const struct kernel_param *kp)
 	return 0;
 }
 
+static void update_contention(struct aa_local_cache *cache)
+{
+	cache->contention += 3;
+	if (cache->contention > 9)
+		cache->contention = 9;
+	cache->hold += 1 << cache->contention;		/* 8, 64, 512 */
+}
+
 char *aa_get_buffer(bool in_atomic)
 {
 	union aa_buffer *aa_buf;
+	struct aa_local_cache *cache;
 	bool try_again = true;
 	gfp_t flags = (GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
 
+	/* use per cpu cached buffers first */
+	cache = get_cpu_ptr(&aa_local_buffers);
+	if (!list_empty(&cache->head)) {
+		aa_buf = list_first_entry(&cache->head, union aa_buffer, list);
+		list_del(&aa_buf->list);
+		cache->hold--;
+		put_cpu_ptr(&aa_local_buffers);
+		return &aa_buf->buffer[0];
+	}
+	put_cpu_ptr(&aa_local_buffers);
+
+	if (!spin_trylock(&aa_buffers_lock)) {
+		cache = get_cpu_ptr(&aa_local_buffers);
+		update_contention(cache);
+		put_cpu_ptr(&aa_local_buffers);
+		spin_lock(&aa_buffers_lock);
+	} else {
+		cache = get_cpu_ptr(&aa_local_buffers);
+		if (cache->contention)
+			cache->contention--;
+		put_cpu_ptr(&aa_local_buffers);
+	}
 retry:
-	spin_lock(&aa_buffers_lock);
 	if (buffer_count > reserve_count ||
 	    (in_atomic && !list_empty(&aa_global_buffers))) {
 		aa_buf = list_first_entry(&aa_global_buffers, union aa_buffer,

@@ -1655,6 +1692,7 @@ char *aa_get_buffer(bool in_atomic)
 	if (!aa_buf) {
 		if (try_again) {
 			try_again = false;
+			spin_lock(&aa_buffers_lock);
 			goto retry;
 		}
 		pr_warn_once("AppArmor: Failed to allocate a memory buffer.\n");

@@ -1666,15 +1704,39 @@ char *aa_get_buffer(bool in_atomic)
 void aa_put_buffer(char *buf)
 {
 	union aa_buffer *aa_buf;
+	struct aa_local_cache *cache;
 
 	if (!buf)
 		return;
 	aa_buf = container_of(buf, union aa_buffer, buffer[0]);
 
-	spin_lock(&aa_buffers_lock);
-	list_add(&aa_buf->list, &aa_global_buffers);
-	buffer_count++;
-	spin_unlock(&aa_buffers_lock);
+	cache = get_cpu_ptr(&aa_local_buffers);
+	if (!cache->hold || cache->count >= 2) {
+		put_cpu_ptr(&aa_local_buffers);
+		if (spin_trylock(&aa_buffers_lock)) {
+		locked:
+			list_add(&aa_buf->list, &aa_global_buffers);
+			buffer_count++;
+			spin_unlock(&aa_buffers_lock);
+			cache = get_cpu_ptr(&aa_local_buffers);
+			if (cache->contention)
+				cache->contention--;
+			put_cpu_ptr(&aa_local_buffers);
+			return;
+		}
+		cache = get_cpu_ptr(&aa_local_buffers);
+		update_contention(cache);
+		if (cache->count >= 2) {
+			put_cpu_ptr(&aa_local_buffers);
+			spin_lock(&aa_buffers_lock);
+			/* force putting the buffer to global */
+			goto locked;
+		}
+	}
+
+	/* cache in percpu list */
+	list_add(&aa_buf->list, &cache->head);
+	put_cpu_ptr(&aa_local_buffers);
 }
 
 /*

@@ -1716,6 +1778,15 @@ static int __init alloc_buffers(void)
 	union aa_buffer *aa_buf;
 	int i, num;
 
+	/*
+	 * per cpu set of cached allocated buffers used to help reduce
+	 * lock contention
+	 */
+	for_each_possible_cpu(i) {
+		per_cpu(aa_local_buffers, i).contention = 0;
+		per_cpu(aa_local_buffers, i).hold = 0;
+		INIT_LIST_HEAD(&per_cpu(aa_local_buffers, i).head);
+	}
 	/*
 	 * A function may require two buffers at once. Usually the buffers are
 	 * used for a short period of time and are shared. On UP kernel buffers
Signature · 1 line
2.34.1
#8 lore raw

security/apparmor/lsm.c union aa_buffer { · line 55

Quoted 10 lines · John Johansen
security/apparmor/lsm.c +5 -0
@@ -49,12 +49,19 @@ union aa_buffer {
 	char buffer[1];
 };
+struct aa_local_cache {
+	unsigned int contention;
+	unsigned int hold;
+	struct list_head head;
+};

if you stick a local_lock_t into that struct, then you could replace

	cache = get_cpu_ptr(&aa_local_buffers);
with
	local_lock(&aa_local_buffers.lock);
	cache = this_cpu_ptr(&aa_local_buffers);

You would get the preempt_disable() based locking for the per-CPU variable (as with get_cpu_ptr()) and additionally some lockdep validation which would warn if it is used outside of task context (IRQ).

I didn't parse completely the hold/contention logic but it seems to work ;)
You check "cache->count >= 2" twice but I don't see an inc/ dec of it nor is it part of aa_local_cache.

I can't parse how many items can end up on the local list if the global list is locked. My guess would be more than 2 due the ->hold parameter.

Do you have any numbers on the machine and performance it improved? It sure will be a good selling point.

Sebastian

maintainer
#9 lore raw
Quoted 13 lines · Sebastian Andrzej Siewior

security/apparmor/lsm.c union aa_buffer { · line 55

Quoted 10 lines · John Johansen
security/apparmor/lsm.c +5 -0
@@ -49,12 +49,19 @@ union aa_buffer {
  	char buffer[1];
  };
+struct aa_local_cache {
+	unsigned int contention;
+	unsigned int hold;
+	struct list_head head;
+};

if you stick a local_lock_t into that struct, then you could replace

Quoted 9 lines
	cache = get_cpu_ptr(&aa_local_buffers);
with
	local_lock(&aa_local_buffers.lock);
	cache = this_cpu_ptr(&aa_local_buffers);

You would get the preempt_disable() based locking for the per-CPU variable (as with get_cpu_ptr()) and additionally some lockdep validation which would warn if it is used outside of task context (IRQ).

I did look at local_locks and there was a reason I didn't use them. I can't recall as the original iteration of this is over a year old now. I will have to dig into it again.

Quoted 5 lines

I didn't parse completely the hold/contention logic but it seems to work ;)
You check "cache->count >= 2" twice but I don't see an inc/ dec of it nor is it part of aa_local_cache.

sadly I messed up the reordering of this and the debug patch. This will be fixed in v4.

Quoted 3 lines

I can't parse how many items can end up on the local list if the global list is locked. My guess would be more than 2 due the ->hold parameter.

So this iteration, forces pushing back to global list if there are already two on the local list. The hold parameter just affects how long the buffers remain on the local list, before trying to place them back on the global list.

Originally before the count was added more than 2 buffers could end up on the local list, and having too many local buffers is a waste of memory. The count got added to address this. The value of 2 (which should be switched to a define) was chosen because no mediation routine currently uses more than 2 buffers.

Note that this doesn't mean that more than two buffers can be allocated to a tasks on a cpu. Its possible in some cases to have a task have allocated buffers and to still have buffers on the local cache list.

Quoted 3 lines

Do you have any numbers on the machine and performance it improved? It sure will be a good selling point.

I can include some supporting info, for a 16 core machine. But it will take some time to for me to get access to a bigger machine, where this is much more important. Hence the call for some of the other people on this thread to test.

thanks for the feedback

#10 lore raw

I can test the patch with 5.10 and 5.15 kernels in different machines. Just let me know which machine types you would like me to test.

Quoted 65 lines · John Johansen
Quoted 13 lines · Sebastian Andrzej Siewior

security/apparmor/lsm.c union aa_buffer { · line 55

Quoted 10 lines · John Johansen
security/apparmor/lsm.c +5 -0
@@ -49,12 +49,19 @@ union aa_buffer {
     char buffer[1];
  };
+struct aa_local_cache {
+    unsigned int contention;
+    unsigned int hold;
+    struct list_head head;
+};

if you stick a local_lock_t into that struct, then you could replace

Quoted 9 lines
      cache = get_cpu_ptr(&aa_local_buffers);
with
      local_lock(&aa_local_buffers.lock);
      cache = this_cpu_ptr(&aa_local_buffers);

You would get the preempt_disable() based locking for the per-CPU variable (as with get_cpu_ptr()) and additionally some lockdep validation which would warn if it is used outside of task context (IRQ).

I did look at local_locks and there was a reason I didn't use them. I can't recall as the original iteration of this is over a year old now. I will have to dig into it again.

Quoted 5 lines

I didn't parse completely the hold/contention logic but it seems to work ;)
You check "cache->count >= 2" twice but I don't see an inc/ dec of it nor is it part of aa_local_cache.

sadly I messed up the reordering of this and the debug patch. This will be fixed in v4.

Quoted 3 lines

I can't parse how many items can end up on the local list if the global list is locked. My guess would be more than 2 due the ->hold parameter.

So this iteration, forces pushing back to global list if there are already two on the local list. The hold parameter just affects how long the buffers remain on the local list, before trying to place them back on the global list.

Originally before the count was added more than 2 buffers could end up on the local list, and having too many local buffers is a waste of memory. The count got added to address this. The value of 2 (which should be switched to a define) was chosen because no mediation routine currently uses more than 2 buffers.

Note that this doesn't mean that more than two buffers can be allocated to a tasks on a cpu. Its possible in some cases to have a task have allocated buffers and to still have buffers on the local cache list.

Quoted 3 lines

Do you have any numbers on the machine and performance it improved? It sure will be a good selling point.

I can include some supporting info, for a 16 core machine. But it will take some time to for me to get access to a bigger machine, where this is much more important. Hence the call for some of the other people on this thread to test.

thanks for the feedback

#11 lore raw

Hi John,

I was wondering if you get a chance to work on patch v4. Please let me know if you need help with testing.

Best,
Anil

Quoted 71 lines · Anil Altinay

I can test the patch with 5.10 and 5.15 kernels in different machines. Just let me know which machine types you would like me to test.

Quoted 65 lines · John Johansen
Quoted 13 lines · Sebastian Andrzej Siewior

security/apparmor/lsm.c union aa_buffer { · line 55

Quoted 10 lines · John Johansen
security/apparmor/lsm.c +5 -0
@@ -49,12 +49,19 @@ union aa_buffer {
     char buffer[1];
  };
+struct aa_local_cache {
+    unsigned int contention;
+    unsigned int hold;
+    struct list_head head;
+};

if you stick a local_lock_t into that struct, then you could replace

Quoted 9 lines
      cache = get_cpu_ptr(&aa_local_buffers);
with
      local_lock(&aa_local_buffers.lock);
      cache = this_cpu_ptr(&aa_local_buffers);

You would get the preempt_disable() based locking for the per-CPU variable (as with get_cpu_ptr()) and additionally some lockdep validation which would warn if it is used outside of task context (IRQ).

I did look at local_locks and there was a reason I didn't use them. I can't recall as the original iteration of this is over a year old now. I will have to dig into it again.

Quoted 5 lines

I didn't parse completely the hold/contention logic but it seems to work ;)
You check "cache->count >= 2" twice but I don't see an inc/ dec of it nor is it part of aa_local_cache.

sadly I messed up the reordering of this and the debug patch. This will be fixed in v4.

Quoted 3 lines

I can't parse how many items can end up on the local list if the global list is locked. My guess would be more than 2 due the ->hold parameter.

So this iteration, forces pushing back to global list if there are already two on the local list. The hold parameter just affects how long the buffers remain on the local list, before trying to place them back on the global list.

Originally before the count was added more than 2 buffers could end up on the local list, and having too many local buffers is a waste of memory. The count got added to address this. The value of 2 (which should be switched to a define) was chosen because no mediation routine currently uses more than 2 buffers.

Note that this doesn't mean that more than two buffers can be allocated to a tasks on a cpu. Its possible in some cases to have a task have allocated buffers and to still have buffers on the local cache list.

Quoted 3 lines

Do you have any numbers on the machine and performance it improved? It sure will be a good selling point.

I can include some supporting info, for a 16 core machine. But it will take some time to for me to get access to a bigger machine, where this is much more important. Hence the call for some of the other people on this thread to test.

thanks for the feedback

A message that is not in the archive (sent off-list or not yet received): CACCxZWO-+M-J_enENr7q1WDcu1U8vYFoytqJxAh=x-nuP268zA@mail.gmail.com

maintainer
#13 lore raw
Quoted 4 lines · Anil Altinay

Hi John,

I was wondering if you get a chance to work on patch v4. Please let me know if you need help with testing.

yeah, testing help is always much appreciated. I have a v4, and I am working on 3 alternate version to compare against, to help give a better sense if we can get away with simplifying or tweak the scaling. I should be able to post them out some time tonight.

Quoted 76 lines

Best,
Anil

On Tue, Feb 21, 2023 at 1:27 PM Anil Altinay <[email protected] <mailto:[email protected]>> wrote:

I can test the patch with 5.10 and 5.15 kernels in different machines. Just let me know which machine types you would like me to test.

On Mon, Feb 20, 2023 at 12:42 AM John Johansen <[email protected] <mailto:[email protected]>> wrote:
>
> On 2/17/23 02:44, Sebastian Andrzej Siewior wrote:
> > On 2023-02-16 16:08:10 [-0800], John Johansen wrote:
> >> --- a/security/apparmor/lsm.c
> >> +++ b/security/apparmor/lsm.c
> >> @@ -49,12 +49,19 @@ union aa_buffer {
> >>      char buffer[1];
> >>   };
> >> +struct aa_local_cache {
> >> +    unsigned int contention;
> >> +    unsigned int hold;
> >> +    struct list_head head;
> >> +};
> >
> > if you stick a local_lock_t into that struct, then you could replace > >       cache = get_cpu_ptr(&aa_local_buffers);
> > with
> >       local_lock(&aa_local_buffers.lock);
> >       cache = this_cpu_ptr(&aa_local_buffers);
> >
> > You would get the preempt_disable() based locking for the per-CPU > > variable (as with get_cpu_ptr()) and additionally some lockdep > > validation which would warn if it is used outside of task context (IRQ). > >
> I did look at local_locks and there was a reason I didn't use them. I > can't recall as the original iteration of this is over a year old now. > I will have to dig into it again.
>
> > I didn't parse completely the hold/contention logic but it seems to work > > ;)
> > You check "cache->count >=  2" twice but I don't see an inc/ dec of it > > nor is it part of aa_local_cache.
> >
> sadly I messed up the reordering of this and the debug patch. This will be > fixed in v4.
>
> > I can't parse how many items can end up on the local list if the global > > list is locked. My guess would be more than 2 due the ->hold parameter. > >
> So this iteration, forces pushing back to global list if there are already > two on the local list. The hold parameter just affects how long the > buffers remain on the local list, before trying to place them back on > the global list.
>
> Originally before the count was added more than 2 buffers could end up > on the local list, and having too many local buffers is a waste of > memory. The count got added to address this. The value of 2 (which should > be switched to a define) was chosen because no mediation routine currently > uses more than 2 buffers.
>
> Note that this doesn't mean that more than two buffers can be allocated > to a tasks on a cpu. Its possible in some cases to have a task have > allocated buffers and to still have buffers on the local cache list. >
> > Do you have any numbers on the machine and performance it improved? It > > sure will be a good selling point.
> >
>
> I can include some supporting info, for a 16 core machine. But it will > take some time to for me to get access to a bigger machine, where this > is much more important. Hence the call for some of the other people > on this thread to test.
>
> thanks for the feedback
>

author
#14 lore raw
Quoted 11 lines · John Johansen
Quoted 4 lines · Anil Altinay

Hi John,

I was wondering if you get a chance to work on patch v4. Please let me know if you need help with testing.

yeah, testing help is always much appreciated. I have a v4, and I am working on 3 alternate version to compare against, to help give a better sense if we can get away with simplifying or tweak the scaling.

I should be able to post them out some time tonight.

Hi John,

Did you get a chance to post v4? I may be able to give it some testing on our real-life case.

maintainer
#15 lore raw

v5 cover apparmor: cache buffers on percpu list if there is lock, contention

Quoted 11 lines · Sergey Senozhatsky
Quoted 10 lines · John Johansen
Quoted 4 lines · Anil Altinay

Hi John,

I was wondering if you get a chance to work on patch v4. Please let me know if you need help with testing.

yeah, testing help is always much appreciated. I have a v4, and I am working on 3 alternate version to compare against, to help give a better sense if we can get away with simplifying or tweak the scaling.

Quoted 6 lines
Quoted 1 line

I should be able to post them out some time tonight.

Hi John,

Did you get a chance to post v4? I may be able to give it some testing on our real-life case.

sorry yes, how about a v5. That is simplified with 3 follow on patches that aren't strictly necessary, but some combination of them might be better than just the base patch, but splitting them out makes the individual changes easier to review.

---

df323337e507 ("apparmor: Use a memory pool instead per-CPU caches") changed buffer allocation to use a memory pool, however on a heavily loaded machine there can be lock contention on the global buffers lock. Add a percpu list to cache buffers on when lock contention is encountered.

When allocating buffers attempt to use cached buffers first, before taking the global buffers lock. When freeing buffers try to put them back to the global list but if contention is encountered, put the buffer on the percpu list.

The length of time a buffer is held on the percpu list is dynamically adjusted based on lock contention.

v5:
- simplify base patch by removing: improvements can be added later
- MAX_LOCAL and must lock
- contention scaling.
v4:
- fix percpu ->count buffer count which had been spliced across a debug patch.
- introduce define for MAX_LOCAL_COUNT
- rework count check and locking around it.
- update commit message to reference commit that introduced the memory.
v3:
- limit number of buffers that can be pushed onto the percpu list. This avoids a problem on some kernels where one percpu list can inherit buffers from another cpu after a reschedule, causing more kernel memory to used than is necessary. Under normal conditions this should eventually return to normal but under pathelogical conditions the extra memory consumption may have been unbouanded
v2:
- dynamically adjust buffer hold time on percpu list based on lock contention.
v1:
- cache buffers on percpu list on lock contention

Reported-by Sergey Senozhatsky <[email protected]>
Signed-off-by John Johansen <[email protected]>
maintainer
#16 lore raw

Re: [PATCH v5 1/4] apparmor: cache buffers on percpu list if there is lock, contention

df323337e507 ("apparmor: Use a memory pool instead per-CPU caches") changed buffer allocation to use a memory pool, however on a heavily loaded machine there can be lock contention on the global buffers lock. Add a percpu list to cache buffers on when lock contention is encountered.

When allocating buffers attempt to use cached buffers first, before taking the global buffers lock. When freeing buffers try to put them back to the global list but if contention is encountered, put the buffer on the percpu list.

The length of time a buffer is held on the percpu list is dynamically adjusted based on lock contention. The amount of hold time is increased and decreased linearly.

Reported-by Sergey Senozhatsky <[email protected]>
Signed-off-by John Johansen <[email protected]>
---
 security/apparmor/lsm.c | 67 ++++++++++++++++++++++++++++++++++++++---
 1 file changed, 62 insertions(+), 5 deletions(-)
security/apparmor/lsm.c +62 -5
index c80c1bd3024a..ce4f3e7a784d 100644
@@ -49,12 +49,19 @@ union aa_buffer {
 	DECLARE_FLEX_ARRAY(char, buffer);
 };
 
+struct aa_local_cache {
+	unsigned int hold;
+	unsigned int count;
+	struct list_head head;
+};
+
 #define RESERVE_COUNT 2
 static int reserve_count = RESERVE_COUNT;
 static int buffer_count;
 
 static LIST_HEAD(aa_global_buffers);
 static DEFINE_SPINLOCK(aa_buffers_lock);
+static DEFINE_PER_CPU(struct aa_local_cache, aa_local_buffers);
 
 /*
  * LSM hook functions

@@ -1789,11 +1796,32 @@ static int param_set_mode(const char *val, const struct kernel_param *kp)
 char *aa_get_buffer(bool in_atomic)
 {
 	union aa_buffer *aa_buf;
+	struct aa_local_cache *cache;
 	bool try_again = true;
 	gfp_t flags = (GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
 
+	/* use per cpu cached buffers first */
+	cache = get_cpu_ptr(&aa_local_buffers);
+	if (!list_empty(&cache->head)) {
+		aa_buf = list_first_entry(&cache->head, union aa_buffer, list);
+		list_del(&aa_buf->list);
+		cache->hold--;
+		cache->count--;
+		put_cpu_ptr(&aa_local_buffers);
+		return &aa_buf->buffer[0];
+	}
+	put_cpu_ptr(&aa_local_buffers);
+
+	if (!spin_trylock(&aa_buffers_lock)) {
+		cache = get_cpu_ptr(&aa_local_buffers);
+		cache->hold += 1;
+		put_cpu_ptr(&aa_local_buffers);
+		spin_lock(&aa_buffers_lock);
+	} else {
+		cache = get_cpu_ptr(&aa_local_buffers);
+		put_cpu_ptr(&aa_local_buffers);
+	}
 retry:
-	spin_lock(&aa_buffers_lock);
 	if (buffer_count > reserve_count ||
 	    (in_atomic && !list_empty(&aa_global_buffers))) {
 		aa_buf = list_first_entry(&aa_global_buffers, union aa_buffer,

@@ -1819,6 +1847,7 @@ char *aa_get_buffer(bool in_atomic)
 	if (!aa_buf) {
 		if (try_again) {
 			try_again = false;
+			spin_lock(&aa_buffers_lock);
 			goto retry;
 		}
 		pr_warn_once("AppArmor: Failed to allocate a memory buffer.\n");

@@ -1830,15 +1859,34 @@ char *aa_get_buffer(bool in_atomic)
 void aa_put_buffer(char *buf)
 {
 	union aa_buffer *aa_buf;
+	struct aa_local_cache *cache;
 
 	if (!buf)
 		return;
 	aa_buf = container_of(buf, union aa_buffer, buffer[0]);
 
-	spin_lock(&aa_buffers_lock);
-	list_add(&aa_buf->list, &aa_global_buffers);
-	buffer_count++;
-	spin_unlock(&aa_buffers_lock);
+	cache = get_cpu_ptr(&aa_local_buffers);
+	if (!cache->hold) {
+		put_cpu_ptr(&aa_local_buffers);
+
+		if (spin_trylock(&aa_buffers_lock)) {
+			/* put back on global list */
+			list_add(&aa_buf->list, &aa_global_buffers);
+			buffer_count++;
+			spin_unlock(&aa_buffers_lock);
+			cache = get_cpu_ptr(&aa_local_buffers);
+			put_cpu_ptr(&aa_local_buffers);
+			return;
+		}
+		/* contention on global list, fallback to percpu */
+		cache = get_cpu_ptr(&aa_local_buffers);
+		cache->hold += 1;
+	}
+
+	/* cache in percpu list */
+	list_add(&aa_buf->list, &cache->head);
+	cache->count++;
+	put_cpu_ptr(&aa_local_buffers);
 }
 
 /*

@@ -1880,6 +1928,15 @@ static int __init alloc_buffers(void)
 	union aa_buffer *aa_buf;
 	int i, num;
 
+	/*
+	 * per cpu set of cached allocated buffers used to help reduce
+	 * lock contention
+	 */
+	for_each_possible_cpu(i) {
+		per_cpu(aa_local_buffers, i).hold = 0;
+		per_cpu(aa_local_buffers, i).count = 0;
+		INIT_LIST_HEAD(&per_cpu(aa_local_buffers, i).head);
+	}
 	/*
 	 * A function may require two buffers at once. Usually the buffers are
 	 * used for a short period of time and are shared. On UP kernel buffers
Signature · 1 line
2.34.1
maintainer
#17 lore raw

Re: [PATCH v5 2/4] apparmor: exponential backoff on cache buffer contention

In reply to John Johansen

Reduce contention on the global buffers lock by using an exponential back off strategy where the amount tries to hold is doubled when contention is encoutered and backed off linearly when there isn't contention.

Signed-off-by John Johansen <[email protected]>
---
 security/apparmor/lsm.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)
security/apparmor/lsm.c +16 -2
index ce4f3e7a784d..fd6779ff0da4 100644
@@ -50,6 +50,7 @@ union aa_buffer {
 };
 
 struct aa_local_cache {
+	unsigned int contention;
 	unsigned int hold;
 	unsigned int count;
 	struct list_head head;

@@ -1793,6 +1794,14 @@ static int param_set_mode(const char *val, const struct kernel_param *kp)
 	return 0;
 }
 
+static void update_contention(struct aa_local_cache *cache)
+{
+	cache->contention += 1;
+	if (cache->contention > 9)
+		cache->contention = 9;
+	cache->hold += 1 << cache->contention;		/* 2, 4, 8, ... */
+}
+
 char *aa_get_buffer(bool in_atomic)
 {
 	union aa_buffer *aa_buf;

@@ -1814,11 +1823,13 @@ char *aa_get_buffer(bool in_atomic)
 
 	if (!spin_trylock(&aa_buffers_lock)) {
 		cache = get_cpu_ptr(&aa_local_buffers);
-		cache->hold += 1;
+		update_contention(cache);
 		put_cpu_ptr(&aa_local_buffers);
 		spin_lock(&aa_buffers_lock);
 	} else {
 		cache = get_cpu_ptr(&aa_local_buffers);
+		if (cache->contention)
+			cache->contention--;
 		put_cpu_ptr(&aa_local_buffers);
 	}
 retry:

@@ -1875,12 +1886,14 @@ void aa_put_buffer(char *buf)
 			buffer_count++;
 			spin_unlock(&aa_buffers_lock);
 			cache = get_cpu_ptr(&aa_local_buffers);
+			if (cache->contention)
+				cache->contention--;
 			put_cpu_ptr(&aa_local_buffers);
 			return;
 		}
 		/* contention on global list, fallback to percpu */
 		cache = get_cpu_ptr(&aa_local_buffers);
-		cache->hold += 1;
+		update_contention(cache);
 	}
 
 	/* cache in percpu list */

@@ -1933,6 +1946,7 @@ static int __init alloc_buffers(void)
 	 * lock contention
 	 */
 	for_each_possible_cpu(i) {
+		per_cpu(aa_local_buffers, i).contention = 0;
 		per_cpu(aa_local_buffers, i).hold = 0;
 		per_cpu(aa_local_buffers, i).count = 0;
 		INIT_LIST_HEAD(&per_cpu(aa_local_buffers, i).head);
Signature · 1 line
2.34.1
maintainer
#18 lore raw

Re: [PATCH v5 3/4] apparmor: experiment with faster backoff on global buffer

In reply to John Johansen

Instead of doubling hold count when contention is encounter increase it by 8x. This makes for a faster back off, but results in buffers being held longer.

Signed-off-by John Johansen <[email protected]>
---
 security/apparmor/lsm.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
security/apparmor/lsm.c +2 -2
index fd6779ff0da4..52423d88854a 100644
@@ -1796,10 +1796,10 @@ static int param_set_mode(const char *val, const struct kernel_param *kp)
 
 static void update_contention(struct aa_local_cache *cache)
 {
-	cache->contention += 1;
+	cache->contention += 3;
 	if (cache->contention > 9)
 		cache->contention = 9;
-	cache->hold += 1 << cache->contention;		/* 2, 4, 8, ... */
+	cache->hold += 1 << cache->contention;		/* 8, 64, 512, ... */
 }
 
 char *aa_get_buffer(bool in_atomic)
Signature · 1 line
2.34.1
maintainer
#19 lore raw

Re: [PATCH v5 4/4] apparmor: limit the number of buffers in percpu cache

In reply to John Johansen

Force buffers to be returned to the global pool, regardless of contention when the percpu cache is full. This ensures that the percpu buffer list never grows longer than needed.

Signed-off-by John Johansen <[email protected]>
---
 security/apparmor/lsm.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
security/apparmor/lsm.c +8 -1
index 52423d88854a..e6765f64f6bf 100644
@@ -56,6 +56,7 @@ struct aa_local_cache {
 	struct list_head head;
 };
 
+#define MAX_LOCAL_COUNT 2
 #define RESERVE_COUNT 2
 static int reserve_count = RESERVE_COUNT;
 static int buffer_count;

@@ -1878,9 +1879,15 @@ void aa_put_buffer(char *buf)
 
 	cache = get_cpu_ptr(&aa_local_buffers);
 	if (!cache->hold) {
+		bool must_lock = cache->count >= MAX_LOCAL_COUNT;
+
 		put_cpu_ptr(&aa_local_buffers);
 
-		if (spin_trylock(&aa_buffers_lock)) {
+		if (must_lock) {
+			spin_lock(&aa_buffers_lock);
+			goto locked;
+		} else if (spin_trylock(&aa_buffers_lock)) {
+		locked:
 			/* put back on global list */
 			list_add(&aa_buf->list, &aa_global_buffers);
 			buffer_count++;
Signature · 1 line
2.34.1
author
#20 lore raw

Re: [PATCH v5 0/4] apparmor: cache buffers on percpu list if there is lock, contention

In reply to John Johansen
Quoted 8 lines · John Johansen
Quoted 8 lines
Quoted 5 lines

yeah, testing help is always much appreciated. I have a v4, and I am working on 3 alternate version to compare against, to help give a better sense if we can get away with simplifying or tweak the scaling.

I should be able to post them out some time tonight.

Hi John,

Quoted 7 lines
Quoted 2 lines

Did you get a chance to post v4? I may be able to give it some testing on our real-life case.

sorry yes, how about a v5. That is simplified with 3 follow on patches that aren't strictly necessary, but some combination of them might be better than just the base patch, but splitting them out makes the individual changes easier to review.

Sorry for late reply. So I gave it a try but, apparently, our build environment has changed quite significantly since the last time I looked into it.

I don't see that many aa_get/put_buffer() anymore. apparmor buffer functions are mostly called form the exec path:

	security_bprm_creds_for_exec()
	 apparmor_bprm_creds_for_exec()
	  make_vfsuid()
	   aa_get_buffer()

As for vfs_statx()->...->apparmor_inode_getattr()->aa_path_perm(), that path is bpf_lsm_inode_getsecid() now.

A message that is not in the archive (sent off-list or not yet received): [email protected]

maintainer
#22 lore raw
Quoted 145 lines · Hillf Danton

On 28 Oct 2022 02:34:07 -0700 John Johansen <[email protected]>

Quoted 144 lines · Sergey Senozhatsky
Quoted 26 lines

Hi,

We've notices that apparmor has switched from using per-CPU buffer pool and per-CPU spin_lock to a global spin_lock in df323337e507a0009d3db1ea.

This seems to be causing some contention on our build machines (with quite a bit of cores). Because that global spin lock is a part of the stat() sys call (and perhaps some other)

E.g.

- 9.29% 0.00% clang++ [kernel.vmlinux]
- 9.28% entry_SYSCALL_64_after_hwframe
- 8.98% do_syscall_64
- 7.43% __do_sys_newlstat
- 7.43% vfs_statx
- 7.18% security_inode_getattr
- 7.15% apparmor_inode_getattr
- aa_path_perm
- 3.53% aa_get_buffer
- 3.47% _raw_spin_lock
3.44% native_queued_spin_lock_slowpath
- 3.49% aa_put_buffer.part.0
- 3.45% _raw_spin_lock
3.43% native_queued_spin_lock_slowpath

Quoted 1 line

Can we fix this contention?

sorry for the delay on this. Below is a proposed patch that I have been testing to deal with this issue.

From d026988196fdbda7234fb87bc3e4aea22edcbaf9 Mon Sep 17 00:00:00 2001
From: John Johansen <[email protected]>
Date: Tue, 25 Oct 2022 01:18:41 -0700
Subject: [PATCH] apparmor: cache buffers on percpu list if there is lock contention

On a heavily loaded machine there can be lock contention on the global buffers lock. Add a percpu list to cache buffers on when lock contention is encountered.

When allocating buffers attempt to use cached buffers first, before taking the global buffers lock. When freeing buffers try to put them back to the global list but if contention is encountered, put the buffer on the percpu list.

The length of time a buffer is held on the percpu list is dynamically adjusted based on lock contention. The amount of hold time is rapidly increased and slow ramped down.

Signed-off-by John Johansen <[email protected]>
---
   security/apparmor/lsm.c | 74 ++++++++++++++++++++++++++++++++++++++---
   1 file changed, 69 insertions(+), 5 deletions(-)
security/apparmor/lsm.c +40 -1
index 25114735bc11..0ab70171bdb6 100644
@@ -49,12 +49,19 @@ union aa_buffer {
   	char buffer[1];
   };
 
+struct aa_local_cache {
+	unsigned int contention;
+	unsigned int hold;
+	struct list_head head;
+};
+
   #define RESERVE_COUNT 2
   static int reserve_count = RESERVE_COUNT;
   static int buffer_count;
 
   static LIST_HEAD(aa_global_buffers);
   static DEFINE_SPINLOCK(aa_buffers_lock);
+static DEFINE_PER_CPU(struct aa_local_cache, aa_local_buffers);
 
   /*
    * LSM hook functions

@@ -1622,14 +1629,44 @@ static int param_set_mode(const char *val, const struct kernel_param *kp)
   	return 0;
   }
 
+static void update_contention(struct aa_local_cache *cache)
+{
+	cache->contention += 3;
+	if (cache->contention > 9)
+		cache->contention = 9;
+	cache->hold += 1 << cache->contention;		/* 8, 64, 512 */
+}
+
   char *aa_get_buffer(bool in_atomic)
   {
   	union aa_buffer *aa_buf;
+	struct aa_local_cache *cache;
   	bool try_again = true;
   	gfp_t flags = (GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
 
+	/* use per cpu cached buffers first */
+	cache = get_cpu_ptr(&aa_local_buffers);
+	if (!list_empty(&cache->head)) {
+		aa_buf = list_first_entry(&cache->head, union aa_buffer, list);
+		list_del(&aa_buf->list);
+		cache->hold--;
+		put_cpu_ptr(&aa_local_buffers);
+		return &aa_buf->buffer[0];
+	}
+	put_cpu_ptr(&aa_local_buffers);
+
+	if (!spin_trylock(&aa_buffers_lock)) {
+		cache = get_cpu_ptr(&aa_local_buffers);
+		update_contention(cache);
+		put_cpu_ptr(&aa_local_buffers);
+		spin_lock(&aa_buffers_lock);
+	} else {
+		cache = get_cpu_ptr(&aa_local_buffers);
+		if (cache->contention)
+			cache->contention--;
+		put_cpu_ptr(&aa_local_buffers);
+	}
   retry:
-	spin_lock(&aa_buffers_lock);
   	if (buffer_count > reserve_count ||
   	    (in_atomic && !list_empty(&aa_global_buffers))) {
   		aa_buf = list_first_entry(&aa_global_buffers, union aa_buffer,

@@ -1655,6 +1692,7 @@ char *aa_get_buffer(bool in_atomic)
   	if (!aa_buf) {
   		if (try_again) {
   			try_again = false;
+			spin_lock(&aa_buffers_lock);
   			goto retry;
   		}
   		pr_warn_once("AppArmor: Failed to allocate a memory buffer.\n");

@@ -1666,15 +1704,32 @@ char *aa_get_buffer(bool in_atomic)
   void aa_put_buffer(char *buf)
   {
   	union aa_buffer *aa_buf;
+	struct aa_local_cache *cache;
 
   	if (!buf)
   		return;
   	aa_buf = container_of(buf, union aa_buffer, buffer[0]);
Quoted 13 lines

security/apparmor/lsm.c char *aa_get_buffer(bool in_atomic) · line 1718

Quoted 10 lines
security/apparmor/lsm.c +6 -4
-	spin_lock(&aa_buffers_lock);
-	list_add(&aa_buf->list, &aa_global_buffers);
-	buffer_count++;
-	spin_unlock(&aa_buffers_lock);
+	cache = get_cpu_ptr(&aa_local_buffers);
+	if (!cache->hold) {
+		put_cpu_ptr(&aa_local_buffers);
+		if (spin_trylock(&aa_buffers_lock)) {
+			list_add(&aa_buf->list, &aa_global_buffers);
+			buffer_count++;

Given !hold and trylock, right time to drain the perpcu cache?

yes hold is a count of how long (or in this case a count of how many times) to allocate from the local from the percpu cache before trying to return to the global buffer pool. When the time/count hits zero its time to try and return it.

If we succeed the try lock then we succeeded taking the global buffer pool lock without contention and we can add the buffer back in.

As for the other cases

hold == 0 and fail to grab the lock
- contention is recorded and we add the buffer back to the percpu cache

hold > 0
- decrease hold and add back to the percpu cache

Since we never try and grab the spinlock if hold > 0, the lock variations do not need to be considered.

Quoted 35 lines
Quoted 35 lines
security/apparmor/lsm.c +23 -0
+			spin_unlock(&aa_buffers_lock);
+			cache = get_cpu_ptr(&aa_local_buffers);
+			if (cache->contention)
+				cache->contention--;
+			put_cpu_ptr(&aa_local_buffers);
+			return;
+		}
+		cache = get_cpu_ptr(&aa_local_buffers);
+		update_contention(cache);
+	}
+
+	/* cache in percpu list */
+	list_add(&aa_buf->list, &cache->head);
+	put_cpu_ptr(&aa_local_buffers);
   }
 
   /*

@@ -1716,6 +1771,15 @@ static int __init alloc_buffers(void)
   	union aa_buffer *aa_buf;
   	int i, num;
 
+	/*
+	 * per cpu set of cached allocated buffers used to help reduce
+	 * lock contention
+	 */
+	for_each_possible_cpu(i) {
+		per_cpu(aa_local_buffers, i).contention = 0;
+		per_cpu(aa_local_buffers, i).hold = 0;
+		INIT_LIST_HEAD(&per_cpu(aa_local_buffers, i).head);
+	}
   	/*
   	 * A function may require two buffers at once. Usually the buffers are
   	 * used for a short period of time and are shared. On UP kernel buffers
Signature · 1 line
2.34.1