adulau/aha
1
0
Fork 0
mirror of https://github.com/adulau/aha.git synced 2024-12-28 03:36:19 +00:00
Code Issues Projects Releases Packages Wiki Activity

cgroup avoid permanent sleep at rmdir

Browse source 
After commit ec64f51545 ("cgroup: fix
frequent -EBUSY at rmdir"), cgroup's rmdir (especially against memcg)
doesn't return -EBUSY by temporary ref counts.  That commit expects all
refs after pre_destroy() is temporary but...it wasn't.  Then, rmdir can
wait permanently.  This patch tries to fix that and change followings.

 - set CGRP_WAIT_ON_RMDIR flag before pre_destroy().
 - clear CGRP_WAIT_ON_RMDIR flag when the subsys finds racy case.
   if there are sleeping ones, wakes them up.
 - rmdir() sleeps only when CGRP_WAIT_ON_RMDIR flag is set.

Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Reported-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Reviewed-by: Paul Menage <menage@google.com>
Acked-by: Balbir Sigh <balbir@linux.vnet.ibm.com>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This commit is contained in:
KAMEZAWA Hiroyuki 2009-07-29 15:04:06 -07:00 committed by Linus Torvalds
parent f0d83679a8
commit 887032670d
3 changed files with 75 additions and 20 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

17
include/linux/cgroup.h
View file

@ -362,6 +362,23 @@ int cgroup_task_count(const struct cgroup *cgrp);
/* Return true if cgrp is a descendant of the task's cgroup */
int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task);
/*
* When the subsys has to access css and may add permanent refcnt to css,
* it should take care of racy conditions with rmdir(). Following set of
* functions, is for stop/restart rmdir if necessary.
* Because these will call css_get/put, "css" should be alive css.
*
* cgroup_exclude_rmdir();
* ...do some jobs which may access arbitrary empty cgroup
* cgroup_release_and_wakeup_rmdir();
*
* When someone removes a cgroup while cgroup_exclude_rmdir() holds it,
* it sleeps and cgroup_release_and_wakeup_rmdir() will wake him up.
*/
void cgroup_exclude_rmdir(struct cgroup_subsys_state *css);
void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css);
/*
* Control Group subsystem type.
* See Documentation/cgroups/cgroups.txt for details

53
kernel/cgroup.c
View file

@ -735,16 +735,28 @@ static void cgroup_d_remove_dir(struct dentry *dentry)
* reference to css->refcnt. In general, this refcnt is expected to goes down
* to zero, soon.
*
* CGRP_WAIT_ON_RMDIR flag is modified under cgroup's inode->i_mutex;
* CGRP_WAIT_ON_RMDIR flag is set under cgroup's inode->i_mutex;
*/
DECLARE_WAIT_QUEUE_HEAD(cgroup_rmdir_waitq);
static void cgroup_wakeup_rmdir_waiters(const struct cgroup *cgrp)
static void cgroup_wakeup_rmdir_waiter(struct cgroup *cgrp)
{
if (unlikely(test_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags)))
if (unlikely(test_and_clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags)))
wake_up_all(&cgroup_rmdir_waitq);
}
void cgroup_exclude_rmdir(struct cgroup_subsys_state *css)
{
css_get(css);
}
void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css)
{
cgroup_wakeup_rmdir_waiter(css->cgroup);
css_put(css);
}
static int rebind_subsystems(struct cgroupfs_root *root,
unsigned long final_bits)
{
@ -1359,7 +1371,7 @@ int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
* wake up rmdir() waiter. the rmdir should fail since the cgroup
* is no longer empty.
*/
cgroup_wakeup_rmdir_waiters(cgrp);
cgroup_wakeup_rmdir_waiter(cgrp);
return 0;
}
@ -2743,33 +2755,42 @@ again:
}
mutex_unlock(&cgroup_mutex);
/*
* In general, subsystem has no css->refcnt after pre_destroy(). But
* in racy cases, subsystem may have to get css->refcnt after
* pre_destroy() and it makes rmdir return with -EBUSY. This sometimes
* make rmdir return -EBUSY too often. To avoid that, we use waitqueue
* for cgroup's rmdir. CGRP_WAIT_ON_RMDIR is for synchronizing rmdir
* and subsystem's reference count handling. Please see css_get/put
* and css_tryget() and cgroup_wakeup_rmdir_waiter() implementation.
*/
set_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
/*
* Call pre_destroy handlers of subsys. Notify subsystems
* that rmdir() request comes.
*/
ret = cgroup_call_pre_destroy(cgrp);
if (ret)
if (ret) {
clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
return ret;
}
mutex_lock(&cgroup_mutex);
parent = cgrp->parent;
if (atomic_read(&cgrp->count) || !list_empty(&cgrp->children)) {
clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
mutex_unlock(&cgroup_mutex);
return -EBUSY;
}
/*
* css_put/get is provided for subsys to grab refcnt to css. In typical
* case, subsystem has no reference after pre_destroy(). But, under
* hierarchy management, some *temporal* refcnt can be hold.
* To avoid returning -EBUSY to a user, waitqueue is used. If subsys
* is really busy, it should return -EBUSY at pre_destroy(). wake_up
* is called when css_put() is called and refcnt goes down to 0.
*/
set_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
prepare_to_wait(&cgroup_rmdir_waitq, &wait, TASK_INTERRUPTIBLE);
if (!cgroup_clear_css_refs(cgrp)) {
mutex_unlock(&cgroup_mutex);
/*
* Because someone may call cgroup_wakeup_rmdir_waiter() before
* prepare_to_wait(), we need to check this flag.
*/
if (test_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags))
schedule();
finish_wait(&cgroup_rmdir_waitq, &wait);
clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags);
@ -3342,7 +3363,7 @@ void __css_put(struct cgroup_subsys_state *css)
set_bit(CGRP_RELEASABLE, &cgrp->flags);
check_for_release(cgrp);
}
cgroup_wakeup_rmdir_waiters(cgrp);
cgroup_wakeup_rmdir_waiter(cgrp);
}
rcu_read_unlock();
}

23
mm/memcontrol.c
View file

@ -1207,6 +1207,12 @@ static int mem_cgroup_move_account(struct page_cgroup *pc,
ret = 0;
out:
unlock_page_cgroup(pc);
/*
* We charges against "to" which may not have any tasks. Then, "to"
* can be under rmdir(). But in current implementation, caller of
* this function is just force_empty() and it's garanteed that
* "to" is never removed. So, we don't check rmdir status here.
*/
return ret;
}
@ -1428,6 +1434,7 @@ __mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
return;
if (!ptr)
return;
cgroup_exclude_rmdir(&ptr->css);
pc = lookup_page_cgroup(page);
mem_cgroup_lru_del_before_commit_swapcache(page);
__mem_cgroup_commit_charge(ptr, pc, ctype);
@ -1457,8 +1464,12 @@ __mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
}
rcu_read_unlock();
}
/* add this page(page_cgroup) to the LRU we want. */
/*
* At swapin, we may charge account against cgroup which has no tasks.
* So, rmdir()->pre_destroy() can be called while we do this charge.
* In that case, we need to call pre_destroy() again. check it here.
*/
cgroup_release_and_wakeup_rmdir(&ptr->css);
}
void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
@ -1664,7 +1675,7 @@ void mem_cgroup_end_migration(struct mem_cgroup *mem,
if (!mem)
return;
cgroup_exclude_rmdir(&mem->css);
/* at migration success, oldpage->mapping is NULL. */
if (oldpage->mapping) {
target = oldpage;
@ -1704,6 +1715,12 @@ void mem_cgroup_end_migration(struct mem_cgroup *mem,
*/
if (ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED)
mem_cgroup_uncharge_page(target);
/*
* At migration, we may charge account against cgroup which has no tasks
* So, rmdir()->pre_destroy() can be called while we do this charge.
* In that case, we need to call pre_destroy() again. check it here.
*/
cgroup_release_and_wakeup_rmdir(&mem->css);
}
/*