diff options
author | 2018-12-05 12:10:26 -0500 | |
---|---|---|
committer | 2018-12-07 22:26:36 -0700 | |
commit | 0fe061b9f03c27d0370888efc22d4b3ac7af90cf (patch) | |
tree | 5cc436fd2fc45eaee9d8796c3ee74635712ea18c /include/linux/blk-cgroup.h | |
parent | blk-mq: remove QUEUE_FLAG_POLL from default MQ flags (diff) | |
download | linux-0fe061b9f03c27d0370888efc22d4b3ac7af90cf.tar.gz linux-0fe061b9f03c27d0370888efc22d4b3ac7af90cf.tar.bz2 linux-0fe061b9f03c27d0370888efc22d4b3ac7af90cf.zip |
blkcg: fix ref count issue with bio_blkcg() using task_css
The bio_blkcg() function turns out to be inconsistent and consequently
dangerous to use. The first part returns a blkcg where a reference is
owned by the bio meaning it does not need to be rcu protected. However,
the third case, the last line, is problematic:
return css_to_blkcg(task_css(current, io_cgrp_id));
This can race against task migration and the cgroup dying. It is also
semantically different as it must be called rcu protected and is
susceptible to failure when trying to get a reference to it.
This patch adds association ahead of calling bio_blkcg() rather than
after. This makes association a required and explicit step along the
code paths for calling bio_blkcg(). In blk-iolatency, association is
moved above the bio_blkcg() call to ensure it will not return %NULL.
BFQ uses the old bio_blkcg() function, but I do not want to address it
in this series due to the complexity. I have created a private version
documenting the inconsistency and noting not to use it.
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Diffstat (limited to 'include/linux/blk-cgroup.h')
-rw-r--r-- | include/linux/blk-cgroup.h | 98 |
1 files changed, 90 insertions, 8 deletions
diff --git a/include/linux/blk-cgroup.h b/include/linux/blk-cgroup.h index a9e2e2037129..f619307171a6 100644 --- a/include/linux/blk-cgroup.h +++ b/include/linux/blk-cgroup.h @@ -227,22 +227,103 @@ int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol, char *input, struct blkg_conf_ctx *ctx); void blkg_conf_finish(struct blkg_conf_ctx *ctx); +/** + * blkcg_css - find the current css + * + * Find the css associated with either the kthread or the current task. + * This may return a dying css, so it is up to the caller to use tryget logic + * to confirm it is alive and well. + */ +static inline struct cgroup_subsys_state *blkcg_css(void) +{ + struct cgroup_subsys_state *css; + + css = kthread_blkcg(); + if (css) + return css; + return task_css(current, io_cgrp_id); +} + +/** + * blkcg_get_css - find and get a reference to the css + * + * Find the css associated with either the kthread or the current task. + * This takes a reference on the blkcg which will need to be managed by the + * caller. + */ +static inline struct cgroup_subsys_state *blkcg_get_css(void) +{ + struct cgroup_subsys_state *css; + + rcu_read_lock(); + + css = kthread_blkcg(); + if (css) { + css_get(css); + } else { + /* + * This is a bit complicated. It is possible task_css() is + * seeing an old css pointer here. This is caused by the + * current thread migrating away from this cgroup and this + * cgroup dying. css_tryget() will fail when trying to take a + * ref on a cgroup that's ref count has hit 0. + * + * Therefore, if it does fail, this means current must have + * been swapped away already and this is waiting for it to + * propagate on the polling cpu. Hence the use of cpu_relax(). + */ + while (true) { + css = task_css(current, io_cgrp_id); + if (likely(css_tryget(css))) + break; + cpu_relax(); + } + } + + rcu_read_unlock(); + + return css; +} static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css) { return css ? container_of(css, struct blkcg, css) : NULL; } -static inline struct blkcg *bio_blkcg(struct bio *bio) +/** + * __bio_blkcg - internal, inconsistent version to get blkcg + * + * DO NOT USE. + * This function is inconsistent and consequently is dangerous to use. The + * first part of the function returns a blkcg where a reference is owned by the + * bio. This means it does not need to be rcu protected as it cannot go away + * with the bio owning a reference to it. However, the latter potentially gets + * it from task_css(). This can race against task migration and the cgroup + * dying. It is also semantically different as it must be called rcu protected + * and is susceptible to failure when trying to get a reference to it. + * Therefore, it is not ok to assume that *_get() will always succeed on the + * blkcg returned here. + */ +static inline struct blkcg *__bio_blkcg(struct bio *bio) { - struct cgroup_subsys_state *css; + if (bio && bio->bi_css) + return css_to_blkcg(bio->bi_css); + return css_to_blkcg(blkcg_css()); +} +/** + * bio_blkcg - grab the blkcg associated with a bio + * @bio: target bio + * + * This returns the blkcg associated with a bio, %NULL if not associated. + * Callers are expected to either handle %NULL or know association has been + * done prior to calling this. + */ +static inline struct blkcg *bio_blkcg(struct bio *bio) +{ if (bio && bio->bi_css) return css_to_blkcg(bio->bi_css); - css = kthread_blkcg(); - if (css) - return css_to_blkcg(css); - return css_to_blkcg(task_css(current, io_cgrp_id)); + return NULL; } static inline bool blk_cgroup_congested(void) @@ -710,10 +791,10 @@ static inline bool blkcg_bio_issue_check(struct request_queue *q, bool throtl = false; rcu_read_lock(); - blkcg = bio_blkcg(bio); /* associate blkcg if bio hasn't attached one */ - bio_associate_blkcg(bio, &blkcg->css); + bio_associate_blkcg(bio, NULL); + blkcg = bio_blkcg(bio); blkg = blkg_lookup(blkcg, q); if (unlikely(!blkg)) { @@ -835,6 +916,7 @@ static inline int blkcg_activate_policy(struct request_queue *q, static inline void blkcg_deactivate_policy(struct request_queue *q, const struct blkcg_policy *pol) { } +static inline struct blkcg *__bio_blkcg(struct bio *bio) { return NULL; } static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; } static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |