From fd3d664fef97cf01f8e28fe0b024ad52f3bbc1bc Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 9 Dec 2008 13:14:11 -0800 Subject: percpu_counter: fix CPU unplug race in percpu_counter_destroy() We should first delete the counter from percpu_counters list before freeing memory, or a percpu_counter_hotcpu_callback() could dereference a NULL pointer. Signed-off-by: Eric Dumazet Acked-by: David S. Miller Cc: Peter Zijlstra Cc: Mingming Cao Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- lib/percpu_counter.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'lib') diff --git a/lib/percpu_counter.c b/lib/percpu_counter.c index a8663890a88c..71b265c330ce 100644 --- a/lib/percpu_counter.c +++ b/lib/percpu_counter.c @@ -104,13 +104,13 @@ void percpu_counter_destroy(struct percpu_counter *fbc) if (!fbc->counters) return; - free_percpu(fbc->counters); - fbc->counters = NULL; #ifdef CONFIG_HOTPLUG_CPU mutex_lock(&percpu_counters_lock); list_del(&fbc->list); mutex_unlock(&percpu_counters_lock); #endif + free_percpu(fbc->counters); + fbc->counters = NULL; } EXPORT_SYMBOL(percpu_counter_destroy); -- cgit v1.2.3 From 71c5576fbd809f2015f4eddf72e501e298720cf3 Mon Sep 17 00:00:00 2001 From: Andrew Morton Date: Tue, 9 Dec 2008 13:14:13 -0800 Subject: revert "percpu counter: clean up percpu_counter_sum_and_set()" Revert commit 1f7c14c62ce63805f9574664a6c6de3633d4a354 Author: Mingming Cao Date: Thu Oct 9 12:50:59 2008 -0400 percpu counter: clean up percpu_counter_sum_and_set() Before this patch we had the following: percpu_counter_sum(): return the percpu_counter's value percpu_counter_sum_and_set(): return the percpu_counter's value, copying that value into the central value and zeroing the per-cpu counters before returning. After this patch, percpu_counter_sum_and_set() has gone, and percpu_counter_sum() gets the old percpu_counter_sum_and_set() functionality. Problem is, as Eric points out, the old percpu_counter_sum_and_set() functionality was racy and wrong. It zeroes out counters on "other" cpus, without holding any locks which will prevent races agaist updates from those other CPUS. This patch reverts 1f7c14c62ce63805f9574664a6c6de3633d4a354. This means that percpu_counter_sum_and_set() still has the race, but percpu_counter_sum() does not. Note that this is not a simple revert - ext4 has since started using percpu_counter_sum() for its dirty_blocks counter as well. Note that this revert patch changes percpu_counter_sum() semantics. Before the patch, a call to percpu_counter_sum() will bring the counter's central counter mostly up-to-date, so a following percpu_counter_read() will return a close value. After this patch, a call to percpu_counter_sum() will leave the counter's central accumulator unaltered, so a subsequent call to percpu_counter_read() can now return a significantly inaccurate result. If there is any code in the tree which was introduced after e8ced39d5e8911c662d4d69a342b9d053eaaac4e was merged, and which depends upon the new percpu_counter_sum() semantics, that code will break. Reported-by: Eric Dumazet Cc: "David S. Miller" Cc: Peter Zijlstra Cc: Mingming Cao Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- lib/percpu_counter.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'lib') diff --git a/lib/percpu_counter.c b/lib/percpu_counter.c index 71b265c330ce..dba1530a5b29 100644 --- a/lib/percpu_counter.c +++ b/lib/percpu_counter.c @@ -52,7 +52,7 @@ EXPORT_SYMBOL(__percpu_counter_add); * Add up all the per-cpu counts, return the result. This is a more accurate * but much slower version of percpu_counter_read_positive() */ -s64 __percpu_counter_sum(struct percpu_counter *fbc) +s64 __percpu_counter_sum(struct percpu_counter *fbc, int set) { s64 ret; int cpu; @@ -62,9 +62,11 @@ s64 __percpu_counter_sum(struct percpu_counter *fbc) for_each_online_cpu(cpu) { s32 *pcount = per_cpu_ptr(fbc->counters, cpu); ret += *pcount; - *pcount = 0; + if (set) + *pcount = 0; } - fbc->count = ret; + if (set) + fbc->count = ret; spin_unlock(&fbc->lock); return ret; -- cgit v1.2.3 From 02d211688727ad02bb4555b1aa8ae2de16b21b39 Mon Sep 17 00:00:00 2001 From: Andrew Morton Date: Tue, 9 Dec 2008 13:14:14 -0800 Subject: revert "percpu_counter: new function percpu_counter_sum_and_set" Revert commit e8ced39d5e8911c662d4d69a342b9d053eaaac4e Author: Mingming Cao Date: Fri Jul 11 19:27:31 2008 -0400 percpu_counter: new function percpu_counter_sum_and_set As described in revert "percpu counter: clean up percpu_counter_sum_and_set()" the new percpu_counter_sum_and_set() is racy against updates to the cpu-local accumulators on other CPUs. Revert that change. This means that ext4 will be slow again. But correct. Reported-by: Eric Dumazet Cc: "David S. Miller" Cc: Peter Zijlstra Cc: Mingming Cao Cc: Cc: [2.6.27.x] Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- lib/percpu_counter.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'lib') diff --git a/lib/percpu_counter.c b/lib/percpu_counter.c index dba1530a5b29..b255b939bc1b 100644 --- a/lib/percpu_counter.c +++ b/lib/percpu_counter.c @@ -52,7 +52,7 @@ EXPORT_SYMBOL(__percpu_counter_add); * Add up all the per-cpu counts, return the result. This is a more accurate * but much slower version of percpu_counter_read_positive() */ -s64 __percpu_counter_sum(struct percpu_counter *fbc, int set) +s64 __percpu_counter_sum(struct percpu_counter *fbc) { s64 ret; int cpu; @@ -62,12 +62,7 @@ s64 __percpu_counter_sum(struct percpu_counter *fbc, int set) for_each_online_cpu(cpu) { s32 *pcount = per_cpu_ptr(fbc->counters, cpu); ret += *pcount; - if (set) - *pcount = 0; } - if (set) - fbc->count = ret; - spin_unlock(&fbc->lock); return ret; } -- cgit v1.2.3 From 711a49a07f84f914aac26a52143f6e7526571143 Mon Sep 17 00:00:00 2001 From: Manfred Spraul Date: Wed, 10 Dec 2008 18:17:06 +0100 Subject: lib/idr.c: Fix bug introduced by RCU fix The last patch to lib/idr.c caused a bug if idr_get_new_above() was called on an empty idr. Usually, nodes stay on the same layer. New layers are added to the top of the tree. The exception is idr_get_new_above() on an empty tree: In this case, the new root node is first added on layer 0, then moved upwards. p->layer was not updated. As usual: You shall never rely on the source code comments, they will only mislead you. Signed-off-by: Manfred Spraul Signed-off-by: Linus Torvalds --- lib/idr.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'lib') diff --git a/lib/idr.c b/lib/idr.c index 7a785a0c2ea0..1c4f9281f412 100644 --- a/lib/idr.c +++ b/lib/idr.c @@ -220,8 +220,14 @@ build_up: */ while ((layers < (MAX_LEVEL - 1)) && (id >= (1 << (layers*IDR_BITS)))) { layers++; - if (!p->count) + if (!p->count) { + /* special case: if the tree is currently empty, + * then we grow the tree by moving the top node + * upwards. + */ + p->layer++; continue; + } if (!(new = get_from_free_list(idp))) { /* * The allocation failed. If we built part of -- cgit v1.2.3