v2:
- Remove unnecessary rcu_read_{lock|unlock} from
cgroup_rstat_css_cpu_flush() in patch 3.
It was found that blkcg_destroy_blkgs() may be called with all blkcg
references gone. This may potentially cause user-after-free and so
should be fixed. The last 2 patches are miscellaneous cleanups of
commit 3b8cc6298724 ("blk-cgroup: Optimize blkcg_rstat_flush()").
Waiman Long (3):
bdi, blk-cgroup: Fix potential UAF of blkcg
blk-cgroup: Don't flush a blkg if destroyed
blk-cgroup: Flush stats at blkgs destruction path
block/blk-cgroup.c | 26 ++++++++++++++++++++++++++
include/linux/cgroup.h | 1 +
kernel/cgroup/rstat.c | 18 ++++++++++++++++++
mm/backing-dev.c | 8 ++++++--
4 files changed, 51 insertions(+), 2 deletions(-)
--
2.31.1
Commit 59b57717fff8 ("blkcg: delay blkg destruction until after
writeback has finished") delayed call to blkcg_destroy_blkgs() to
cgwb_release_workfn(). However, it is done after a css_put() of blkcg
which may be the final put that causes the blkcg to be freed as RCU
read lock isn't held.
Another place where blkcg_destroy_blkgs() can be called indirectly via
blkcg_unpin_online() is from the offline_css() function called from
css_killed_work_fn(). Over there, the potentially final css_put() call
is issued after offline_css().
By adding a css_tryget() into blkcg_destroy_blkgs() and warning its
failure, the following stack trace was produced in a test system on
bootup.
[ 34.254240] RIP: 0010:blkcg_destroy_blkgs+0x16a/0x1a0
:
[ 34.339943] Call Trace:
[ 34.344510] blkcg_unpin_online+0x38/0x60
[ 34.348523] cgwb_release_workfn+0x6a/0x200
[ 34.352708] process_one_work+0x1e5/0x3b0
[ 34.360758] worker_thread+0x50/0x3a0
[ 34.368447] kthread+0xd9/0x100
[ 34.376386] ret_from_fork+0x22/0x30
This confirms that a potential UAF situation can really happen in
cgwb_release_workfn().
Fix that by delaying the css_put() until after the blkcg_unpin_online()
call. Also use css_tryget() in blkcg_destroy_blkgs() and issue a warning
if css_tryget() fails.
The reproducing system can no longer produce a warning with this patch.
All the runnable block/0* tests including block/027 were run successfully
without failure.
Fixes: 59b57717fff8 ("blkcg: delay blkg destruction until after writeback has finished")
Suggested-by: Michal Koutný <[email protected]>
Reported-by: Yi Zhang <[email protected]>
Signed-off-by: Waiman Long <[email protected]>
---
block/blk-cgroup.c | 8 ++++++++
mm/backing-dev.c | 8 ++++++--
2 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index 1bb939d3b793..21cc88349f21 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -1084,6 +1084,13 @@ struct list_head *blkcg_get_cgwb_list(struct cgroup_subsys_state *css)
*/
static void blkcg_destroy_blkgs(struct blkcg *blkcg)
{
+ /*
+ * blkcg_destroy_blkgs() shouldn't be called with all the blkcg
+ * references gone.
+ */
+ if (WARN_ON_ONCE(!css_tryget(&blkcg->css)))
+ return;
+
might_sleep();
spin_lock_irq(&blkcg->lock);
@@ -1110,6 +1117,7 @@ static void blkcg_destroy_blkgs(struct blkcg *blkcg)
}
spin_unlock_irq(&blkcg->lock);
+ css_put(&blkcg->css);
}
/**
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index c30419a5e119..36f75b072325 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -390,11 +390,15 @@ static void cgwb_release_workfn(struct work_struct *work)
wb_shutdown(wb);
css_put(wb->memcg_css);
- css_put(wb->blkcg_css);
mutex_unlock(&wb->bdi->cgwb_release_mutex);
- /* triggers blkg destruction if no online users left */
+ /*
+ * Triggers blkg destruction if no online users left
+ * The final blkcg css_put() has to be done after blkcg_unpin_online()
+ * to avoid use-after-free.
+ */
blkcg_unpin_online(wb->blkcg_css);
+ css_put(wb->blkcg_css);
fprop_local_destroy_percpu(&wb->memcg_completions);
--
2.31.1
As noted by Michal, the blkg_iostat_set's in the lockless list
hold reference to blkg's to protect against their removal. Those
blkg's hold reference to blkcg. When a cgroup is being destroyed,
cgroup_rstat_flush() is only called at css_release_work_fn() which is
called when the blkcg reference count reaches 0. This circular dependency
will prevent blkcg from being freed until some other events cause
cgroup_rstat_flush() to be called to flush out the pending blkcg stats.
To prevent this delayed blkcg removal, add a new cgroup_rstat_css_flush()
function to flush stats for a given css and cpu and call it at the blkgs
destruction path, blkcg_destroy_blkgs(), whenever there are still some
pending stats to be flushed. This will ensure that blkcg reference
count can reach 0 ASAP.
Signed-off-by: Waiman Long <[email protected]>
---
block/blk-cgroup.c | 12 ++++++++++++
include/linux/cgroup.h | 1 +
kernel/cgroup/rstat.c | 18 ++++++++++++++++++
3 files changed, 31 insertions(+)
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index c466aef0d467..534f3baeb84a 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -1090,6 +1090,8 @@ struct list_head *blkcg_get_cgwb_list(struct cgroup_subsys_state *css)
*/
static void blkcg_destroy_blkgs(struct blkcg *blkcg)
{
+ int cpu;
+
/*
* blkcg_destroy_blkgs() shouldn't be called with all the blkcg
* references gone.
@@ -1099,6 +1101,16 @@ static void blkcg_destroy_blkgs(struct blkcg *blkcg)
might_sleep();
+ /*
+ * Flush all the non-empty percpu lockless lists.
+ */
+ for_each_possible_cpu(cpu) {
+ struct llist_head *lhead = per_cpu_ptr(blkcg->lhead, cpu);
+
+ if (!llist_empty(lhead))
+ cgroup_rstat_css_cpu_flush(&blkcg->css, cpu);
+ }
+
spin_lock_irq(&blkcg->lock);
while (!hlist_empty(&blkcg->blkg_list)) {
diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index 528bd44b59e2..6c4e66b3fa84 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -766,6 +766,7 @@ void cgroup_rstat_flush(struct cgroup *cgrp);
void cgroup_rstat_flush_irqsafe(struct cgroup *cgrp);
void cgroup_rstat_flush_hold(struct cgroup *cgrp);
void cgroup_rstat_flush_release(void);
+void cgroup_rstat_css_cpu_flush(struct cgroup_subsys_state *css, int cpu);
/*
* Basic resource stats.
diff --git a/kernel/cgroup/rstat.c b/kernel/cgroup/rstat.c
index 793ecff29038..2e44be44351f 100644
--- a/kernel/cgroup/rstat.c
+++ b/kernel/cgroup/rstat.c
@@ -281,6 +281,24 @@ void cgroup_rstat_flush_release(void)
spin_unlock_irq(&cgroup_rstat_lock);
}
+/**
+ * cgroup_rstat_css_cpu_flush - flush stats for the given css and cpu
+ * @css: target css to be flush
+ * @cpu: the cpu that holds the stats to be flush
+ *
+ * A lightweight rstat flush operation for a given css and cpu.
+ * Only the cpu_lock is being held for mutual exclusion, the cgroup_rstat_lock
+ * isn't used.
+ */
+void cgroup_rstat_css_cpu_flush(struct cgroup_subsys_state *css, int cpu)
+{
+ raw_spinlock_t *cpu_lock = per_cpu_ptr(&cgroup_rstat_cpu_lock, cpu);
+
+ raw_spin_lock_irq(cpu_lock);
+ css->ss->css_rstat_flush(css, cpu);
+ raw_spin_unlock_irq(cpu_lock);
+}
+
int cgroup_rstat_init(struct cgroup *cgrp)
{
int cpu;
--
2.31.1
On Sun, Dec 11, 2022 at 05:20:56PM -0500, Waiman Long wrote:
> static void blkcg_destroy_blkgs(struct blkcg *blkcg)
> {
> + /*
> + * blkcg_destroy_blkgs() shouldn't be called with all the blkcg
> + * references gone.
> + */
> + if (WARN_ON_ONCE(!css_tryget(&blkcg->css)))
> + return;
Wouldn't it make more sense to use percpu_ref_is_zero()? It's not like the
obtained extra reference does anything, right?
Thanks.
--
tejun
On Sun, Dec 11, 2022 at 05:20:58PM -0500, Waiman Long wrote:
> As noted by Michal, the blkg_iostat_set's in the lockless list
> hold reference to blkg's to protect against their removal. Those
> blkg's hold reference to blkcg. When a cgroup is being destroyed,
> cgroup_rstat_flush() is only called at css_release_work_fn() which is
> called when the blkcg reference count reaches 0. This circular dependency
> will prevent blkcg from being freed until some other events cause
> cgroup_rstat_flush() to be called to flush out the pending blkcg stats.
>
> To prevent this delayed blkcg removal, add a new cgroup_rstat_css_flush()
> function to flush stats for a given css and cpu and call it at the blkgs
> destruction path, blkcg_destroy_blkgs(), whenever there are still some
> pending stats to be flushed. This will ensure that blkcg reference
> count can reach 0 ASAP.
>
> Signed-off-by: Waiman Long <[email protected]>
Acked-by: Tejun Heo <[email protected]>
But a nit below
> + /*
> + * Flush all the non-empty percpu lockless lists.
> + */
Can you please explain the deadlock that's being avoided in the above
comment? ie. it should say why this flush is necessary.
> + for_each_possible_cpu(cpu) {
> + struct llist_head *lhead = per_cpu_ptr(blkcg->lhead, cpu);
> +
> + if (!llist_empty(lhead))
> + cgroup_rstat_css_cpu_flush(&blkcg->css, cpu);
> + }
Thanks.
--
tejun
On 12/12/22 17:13, Tejun Heo wrote:
> On Sun, Dec 11, 2022 at 05:20:56PM -0500, Waiman Long wrote:
>> static void blkcg_destroy_blkgs(struct blkcg *blkcg)
>> {
>> + /*
>> + * blkcg_destroy_blkgs() shouldn't be called with all the blkcg
>> + * references gone.
>> + */
>> + if (WARN_ON_ONCE(!css_tryget(&blkcg->css)))
>> + return;
> Wouldn't it make more sense to use percpu_ref_is_zero()? It's not like the
> obtained extra reference does anything, right?
Yes, that makes sense. Will incorporate the change in the next version.
Thanks,
Longman
On 12/12/22 17:24, Tejun Heo wrote:
> On Sun, Dec 11, 2022 at 05:20:58PM -0500, Waiman Long wrote:
>> As noted by Michal, the blkg_iostat_set's in the lockless list
>> hold reference to blkg's to protect against their removal. Those
>> blkg's hold reference to blkcg. When a cgroup is being destroyed,
>> cgroup_rstat_flush() is only called at css_release_work_fn() which is
>> called when the blkcg reference count reaches 0. This circular dependency
>> will prevent blkcg from being freed until some other events cause
>> cgroup_rstat_flush() to be called to flush out the pending blkcg stats.
>>
>> To prevent this delayed blkcg removal, add a new cgroup_rstat_css_flush()
>> function to flush stats for a given css and cpu and call it at the blkgs
>> destruction path, blkcg_destroy_blkgs(), whenever there are still some
>> pending stats to be flushed. This will ensure that blkcg reference
>> count can reach 0 ASAP.
>>
>> Signed-off-by: Waiman Long <[email protected]>
> Acked-by: Tejun Heo <[email protected]>
>
> But a nit below
>
>> + /*
>> + * Flush all the non-empty percpu lockless lists.
>> + */
> Can you please explain the deadlock that's being avoided in the above
> comment? ie. it should say why this flush is necessary.
Sure. I will expand the comment to elaborate a bit more.
Cheers,
Longman
>
>> + for_each_possible_cpu(cpu) {
>> + struct llist_head *lhead = per_cpu_ptr(blkcg->lhead, cpu);
>> +
>> + if (!llist_empty(lhead))
>> + cgroup_rstat_css_cpu_flush(&blkcg->css, cpu);
>> + }