Re: [PATCH 2/2] cgroup: exlclude release rmdir

[KAMEZAWA Hiroyuki]

On Tue, 30 Jun 2009 02:15:03 -0700
Paul Menage <menage@xxxxxxxxxx> wrote:

> On Tue, Jun 30, 2009 at 2:03 AM, KAMEZAWA
> Hiroyuki<kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
> > From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> >
> > Paul Menage pointed out that css_get()/put() only for avoiding race with
> > rmdir() is complicated and these should be treated as it is for.
> >
> > This adds
> > Â - cgroup_exclude_rmdir() ....prevent rmdir() for a while.
> > Â - cgroup_release_rmdir() ....rerun rmdir() if necessary.
> > And hides cgroup_wakeup_rmdir_waiter() into kernel/cgroup.c, again.
> 
> Wouldn't it be better to merge these into a single patch? Having one
> patch that exposes complexity only to take it away in the following
> patch seems unnecessary; the combined patch would be simpler than the
> constituents.
> 
This patch is _not_ tested by Nishimura.
What I want is patch 1/2, it's BUGFIX and passed tests by him.
I trust his test very much.
I want the patch 1/2 should be on fast-path as BUGFIX.

But, I think this patch 2/2 is not for fast-path.
This is something new but just a refactoring.

Anyway, I can postpone this until things are settled. Only merging patch 1/2
is okay for me now.

Thanks,
-Kame


> Paul
> 
> >
> > Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> >
> > ---
> > Âinclude/linux/cgroup.h | Â 21 +++++++++++----------
> > Âkernel/cgroup.c    Â|  17 +++++++++++++++--
> > Âmm/memcontrol.c    Â|  12 ++++--------
> > Â3 files changed, 30 insertions(+), 20 deletions(-)
> >
> > Index: mmotm-2.6.31-Jun25/include/linux/cgroup.h
> > ===================================================================
> > --- mmotm-2.6.31-Jun25.orig/include/linux/cgroup.h
> > +++ mmotm-2.6.31-Jun25/include/linux/cgroup.h
> > @@ -366,17 +366,18 @@ int cgroup_task_count(const struct cgrou
> > Âint cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task);
> >
> > Â/*
> > - * Allow to use CGRP_WAIT_ON_RMDIR flag to check race with rmdir() for subsys.
> > - * Subsys can call this function if it's necessary to call pre_destroy() again
> > - * because it adds not-temporary refs to css after or while pre_destroy().
> > - * The caller of this function should use css_tryget(), too.
> > + * 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.
> > + *
> > + * Âcgroup_exclude_rmdir();
> > + * Â...do some jobs which may access arbitrary empty cgroup
> > + * Âcgroup_release_rmdir();
> > Â*/
> > -void __cgroup_wakeup_rmdir_waiters(void);
> > -static inline void cgroup_wakeup_rmdir_waiter(struct cgroup *cgrp)
> > -{
> > - Â Â Â if (unlikely(test_and_clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags)))
> > - Â Â Â Â Â Â Â __cgroup_wakeup_rmdir_waiters();
> > -}
> > +
> > +void cgroup_exclude_rmdir(struct cgroup_subsys_state *css);
> > +void cgroup_release_rmdir(struct cgroup_subsys_state *css);
> >
> > Â/*
> > Â* Control Group subsystem type.
> > Index: mmotm-2.6.31-Jun25/kernel/cgroup.c
> > ===================================================================
> > --- mmotm-2.6.31-Jun25.orig/kernel/cgroup.c
> > +++ mmotm-2.6.31-Jun25/kernel/cgroup.c
> > @@ -738,11 +738,24 @@ static void cgroup_d_remove_dir(struct d
> > Â*/
> > ÂDECLARE_WAIT_QUEUE_HEAD(cgroup_rmdir_waitq);
> >
> > -void __cgroup_wakeup_rmdir_waiters(void)
> > +static void cgroup_wakeup_rmdir_waiter(struct cgroup *cgrp)
> > Â{
> > - Â Â Â wake_up_all(&cgroup_rmdir_waitq);
> > + Â Â Â if (unlikely(test_and_clear_bit(CGRP_WAIT_ON_RMDIR, &cgrp->flags)))
> > + Â Â Â Â Â Â Â wake_up_all(&cgroup_rmdir_waitq);
> > Â}
> >
> > +void cgroup_exclude_rmdir(struct cgroup_subsys_state *css)
> > +{
> > + Â Â Â css_get(css);
> > +}
> > +
> > +void cgroup_release_rmdir(struct cgroup_subsys_state *css)
> > +{
> > + Â Â Â cgroup_wakeup_rmdir_waiter(css->cgroup);
> > + Â Â Â css_put(css);
> > +}
> > +
> > +
> > Âstatic int rebind_subsystems(struct cgroupfs_root *root,
> > Â Â Â Â Â Â Â Â Â Â Â Â Â Â Âunsigned long final_bits)
> > Â{
> > Index: mmotm-2.6.31-Jun25/mm/memcontrol.c
> > ===================================================================
> > --- mmotm-2.6.31-Jun25.orig/mm/memcontrol.c
> > +++ mmotm-2.6.31-Jun25/mm/memcontrol.c
> > @@ -1461,7 +1461,7 @@ __mem_cgroup_commit_charge_swapin(struct
> > Â Â Â Â Â Â Â Âreturn;
> > Â Â Â Âif (!ptr)
> > Â Â Â Â Â Â Â Âreturn;
> > - Â Â Â css_get(&ptr->css);
> > + Â Â Â cgroup_exclude_rmdir(&ptr->css);
> > Â Â Â Âpc = lookup_page_cgroup(page);
> > Â Â Â Âmem_cgroup_lru_del_before_commit_swapcache(page);
> > Â Â Â Â__mem_cgroup_commit_charge(ptr, pc, ctype);
> > @@ -1496,9 +1496,7 @@ __mem_cgroup_commit_charge_swapin(struct
> > Â Â Â Â * So, rmdir()->pre_destroy() can be called while we do this charge.
> > Â Â Â Â * In that case, we need to call pre_destroy() again. check it here.
> > Â Â Â Â */
> > - Â Â Â cgroup_wakeup_rmdir_waiter(ptr->css.cgroup);
> > - Â Â Â css_put(&ptr->css);
> > -
> > + Â Â Â cgroup_release_rmdir(&ptr->css);
> > Â}
> >
> > Âvoid mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
> > @@ -1704,7 +1702,7 @@ void mem_cgroup_end_migration(struct mem
> >
> > Â Â Â Âif (!mem)
> > Â Â Â Â Â Â Â Âreturn;
> > - Â Â Â css_get(&mem->css);
> > + Â Â Â cgroup_exclude_rmdir(&mem->css);
> > Â Â Â Â/* at migration success, oldpage->mapping is NULL. */
> > Â Â Â Âif (oldpage->mapping) {
> > Â Â Â Â Â Â Â Âtarget = oldpage;
> > @@ -1749,9 +1747,7 @@ void mem_cgroup_end_migration(struct mem
> > Â Â Â Â * So, rmdir()->pre_destroy() can be called while we do this charge.
> > Â Â Â Â * In that case, we need to call pre_destroy() again. check it here.
> > Â Â Â Â */
> > - Â Â Â cgroup_wakeup_rmdir_waiter(mem->css.cgroup);
> > - Â Â Â css_put(&mem->css);
> > -
> > + Â Â Â cgroup_release_rmdir(&mem->css);
> > Â}
> >
> > Â/*
> >
> >
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/