| Message ID | 1566294517-86418-4-git-send-email-alex.shi@linux.alibaba.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | per memcg lru_lock | expand |
On 20/08/2019 12.48, Alex Shi wrote: > Now we repeatly assign the lruvec->pgdat in memcg. Will remove the > assignment in lruvec getting function after very points are protected. > > Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com> > Cc: Johannes Weiner <hannes@cmpxchg.org> > Cc: Michal Hocko <mhocko@kernel.org> > Cc: Vladimir Davydov <vdavydov.dev@gmail.com> > Cc: Andrew Morton <akpm@linux-foundation.org> > Cc: Tejun Heo <tj@kernel.org> > Cc: cgroups@vger.kernel.org > Cc: linux-mm@kvack.org > Cc: linux-kernel@vger.kernel.org > --- > mm/memcontrol.c | 12 +++++------- > 1 file changed, 5 insertions(+), 7 deletions(-) > > diff --git a/mm/memcontrol.c b/mm/memcontrol.c > index e8a1b0d95ba8..19fd911e8098 100644 > --- a/mm/memcontrol.c > +++ b/mm/memcontrol.c > @@ -2550,12 +2550,12 @@ static void cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages) > static void lock_page_lru(struct page *page, int *isolated) > { > pg_data_t *pgdat = page_pgdat(page); > + struct lruvec *lruvec = mem_cgroup_page_lruvec(page, pgdat); > What protects lruvec from freeing at this point? After reading resolving lruvec page could be moved and cgroup deleted. In this old patches I've used RCU for that: https://lkml.org/lkml/2012/2/20/276 Pointer to lruvec should be resolved under disabled irq. Not sure this works these days. > - spin_lock_irq(&pgdat->lruvec.lru_lock); > + spin_lock_irq(&lruvec->lru_lock); > + sync_lruvec_pgdat(lruvec, pgdat); > if (PageLRU(page)) { > - struct lruvec *lruvec; > > - lruvec = mem_cgroup_page_lruvec(page, pgdat); > ClearPageLRU(page); > del_page_from_lru_list(page, lruvec, page_lru(page)); > *isolated = 1; > @@ -2566,16 +2566,14 @@ static void lock_page_lru(struct page *page, int *isolated) > static void unlock_page_lru(struct page *page, int isolated) > { > pg_data_t *pgdat = page_pgdat(page); > + struct lruvec *lruvec = mem_cgroup_page_lruvec(page, pgdat); > > if (isolated) { > - struct lruvec *lruvec; > - > - lruvec = mem_cgroup_page_lruvec(page, pgdat); > VM_BUG_ON_PAGE(PageLRU(page), page); > SetPageLRU(page); > add_page_to_lru_list(page, lruvec, page_lru(page)); > } > - spin_unlock_irq(&pgdat->lruvec.lru_lock); > + spin_unlock_irq(&lruvec->lru_lock); > } > > static void commit_charge(struct page *page, struct mem_cgroup *memcg, >
在 2019/8/26 下午4:30, Konstantin Khlebnikov 写道: >> >> > > What protects lruvec from freeing at this point? > After reading resolving lruvec page could be moved and cgroup deleted. > > In this old patches I've used RCU for that: https://lkml.org/lkml/2012/2/20/276 > Pointer to lruvec should be resolved under disabled irq. > Not sure this works these days. Thanks for reminder! I will reconsider this point and come up with changes. Thanks Alex
diff --git a/mm/memcontrol.c b/mm/memcontrol.c index e8a1b0d95ba8..19fd911e8098 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -2550,12 +2550,12 @@ static void cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages) static void lock_page_lru(struct page *page, int *isolated) { pg_data_t *pgdat = page_pgdat(page); + struct lruvec *lruvec = mem_cgroup_page_lruvec(page, pgdat); - spin_lock_irq(&pgdat->lruvec.lru_lock); + spin_lock_irq(&lruvec->lru_lock); + sync_lruvec_pgdat(lruvec, pgdat); if (PageLRU(page)) { - struct lruvec *lruvec; - lruvec = mem_cgroup_page_lruvec(page, pgdat); ClearPageLRU(page); del_page_from_lru_list(page, lruvec, page_lru(page)); *isolated = 1; @@ -2566,16 +2566,14 @@ static void lock_page_lru(struct page *page, int *isolated) static void unlock_page_lru(struct page *page, int isolated) { pg_data_t *pgdat = page_pgdat(page); + struct lruvec *lruvec = mem_cgroup_page_lruvec(page, pgdat); if (isolated) { - struct lruvec *lruvec; - - lruvec = mem_cgroup_page_lruvec(page, pgdat); VM_BUG_ON_PAGE(PageLRU(page), page); SetPageLRU(page); add_page_to_lru_list(page, lruvec, page_lru(page)); } - spin_unlock_irq(&pgdat->lruvec.lru_lock); + spin_unlock_irq(&lruvec->lru_lock); } static void commit_charge(struct page *page, struct mem_cgroup *memcg,
Now we repeatly assign the lruvec->pgdat in memcg. Will remove the assignment in lruvec getting function after very points are protected. Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Vladimir Davydov <vdavydov.dev@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Tejun Heo <tj@kernel.org> Cc: cgroups@vger.kernel.org Cc: linux-mm@kvack.org Cc: linux-kernel@vger.kernel.org --- mm/memcontrol.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-)