The patch titled memcg: make resize limit hold mutex has been added to the -mm tree. Its filename is memcg-memswap-controller-core-make-resize-limit-hold-mutex.patch Before you just go and hit "reply", please: a) Consider who else should be cc'ed b) Prefer to cc a suitable mailing list as well c) Ideally: find the original patch on the mailing list and do a reply-to-all to that, adding suitable additional cc's *** Remember to use Documentation/SubmitChecklist when testing your code *** See http://userweb.kernel.org/~akpm/stuff/added-to-mm.txt to find out what to do about this The current -mm tree may be found at http://userweb.kernel.org/~akpm/mmotm/ ------------------------------------------------------ Subject: memcg: make resize limit hold mutex From: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx> mem_cgroup_resize_memsw_limit() try to hold memsw.lock while holding res.lock, so below message is showed when trying to write memory.memsw.limit_in_bytes file. [ INFO: possible recursive locking detected ] 2.6.28-rc4-mm1-mmotm-2008-11-14-20-50-ef4e17ef #1 bash/4406 is trying to acquire lock: (&counter->lock){....}, at: [<c0498408>] mem_cgroup_resize_memsw_limit+0x8d/0x113 but task is already holding lock: (&counter->lock){....}, at: [<c04983d6>] mem_cgroup_resize_memsw_limit+0x5b/0x113 other info that might help us debug this: 1 lock held by bash/4406: #0: (&counter->lock){....}, at: [<c04983d6>] mem_cgroup_resize_memsw_limit+0x5b/0x113 stack backtrace: Pid: 4406, comm: bash Not tainted 2.6.28-rc4-mm1-mmotm-2008-11-14-20-50-ef4e17ef #1 Call Trace: [<c066e60f>] ? printk+0xf/0x18 [<c044d0c0>] __lock_acquire+0xc67/0x1353 [<c044d793>] ? __lock_acquire+0x133a/0x1353 [<c044d81c>] lock_acquire+0x70/0x97 [<c0498408>] ? mem_cgroup_resize_memsw_limit+0x8d/0x113 [<c0671519>] _spin_lock_irqsave+0x3a/0x6d [<c0498408>] ? mem_cgroup_resize_memsw_limit+0x8d/0x113 [<c0498408>] mem_cgroup_resize_memsw_limit+0x8d/0x113 [<c0518a6c>] ? memparse+0x14/0x66 [<c0498594>] mem_cgroup_write+0x4a/0x50 [<c045e063>] cgroup_file_write+0x181/0x1c6 [<c0449e43>] ? lock_release_holdtime+0x1a/0x168 [<c04ec725>] ? security_file_permission+0xf/0x11 [<c049b5f0>] ? rw_verify_area+0x76/0x97 [<c045dee2>] ? cgroup_file_write+0x0/0x1c6 [<c049bce6>] vfs_write+0x8a/0x12e [<c049be23>] sys_write+0x3b/0x60 [<c0403867>] sysenter_do_call+0x12/0x3f This patch define a new mutex and make both mem_cgroup_resize_limit and mem_cgroup_memsw_resize_limit hold it to remove spin_lock_irqsave. Signed-off-by: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> Cc: Hugh Dickins <hugh@xxxxxxxxxxx> Cc: Li Zefan <lizf@xxxxxxxxxxxxxx> Cc: Balbir Singh <balbir@xxxxxxxxxx> Cc: Pavel Emelyanov <xemul@xxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- mm/memcontrol.c | 40 ++++++++++++++++++++++++++-------------- 1 file changed, 26 insertions(+), 14 deletions(-) diff -puN mm/memcontrol.c~memcg-memswap-controller-core-make-resize-limit-hold-mutex mm/memcontrol.c --- a/mm/memcontrol.c~memcg-memswap-controller-core-make-resize-limit-hold-mutex +++ a/mm/memcontrol.c @@ -27,6 +27,7 @@ #include <linux/backing-dev.h> #include <linux/bit_spinlock.h> #include <linux/rcupdate.h> +#include <linux/mutex.h> #include <linux/slab.h> #include <linux/swap.h> #include <linux/spinlock.h> @@ -1189,27 +1190,38 @@ int mem_cgroup_shrink_usage(struct mm_st return 0; } +static DEFINE_MUTEX(set_limit_mutex); int mem_cgroup_resize_limit(struct mem_cgroup *memcg, unsigned long long val) { int retry_count = MEM_CGROUP_RECLAIM_RETRIES; int progress; + u64 memswlimit; int ret = 0; - if (do_swap_account) { - if (val > memcg->memsw.limit) - return -EINVAL; - } - - while (res_counter_set_limit(&memcg->res, val)) { + while (retry_count) { if (signal_pending(current)) { ret = -EINTR; break; } - if (!retry_count) { - ret = -EBUSY; + /* + * Rather than hide all in some function, I do this in + * open coded manner. You see what this really does. + * We have to guarantee mem->res.limit < mem->memsw.limit. + */ + mutex_lock(&set_limit_mutex); + memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT); + if (memswlimit < val) { + ret = -EINVAL; + mutex_unlock(&set_limit_mutex); break; } + ret = res_counter_set_limit(&memcg->res, val); + mutex_unlock(&set_limit_mutex); + + if (!ret) + break; + progress = try_to_free_mem_cgroup_pages(memcg, GFP_HIGHUSER_MOVABLE, false); if (!progress) @@ -1222,7 +1234,6 @@ int mem_cgroup_resize_memsw_limit(struct unsigned long long val) { int retry_count = MEM_CGROUP_RECLAIM_RETRIES; - unsigned long flags; u64 memlimit, oldusage, curusage; int ret; @@ -1239,19 +1250,20 @@ int mem_cgroup_resize_memsw_limit(struct * open coded manner. You see what this really does. * We have to guarantee mem->res.limit < mem->memsw.limit. */ - spin_lock_irqsave(&memcg->res.lock, flags); - memlimit = memcg->res.limit; + mutex_lock(&set_limit_mutex); + memlimit = res_counter_read_u64(&memcg->res, RES_LIMIT); if (memlimit > val) { - spin_unlock_irqrestore(&memcg->res.lock, flags); ret = -EINVAL; + mutex_unlock(&set_limit_mutex); break; } ret = res_counter_set_limit(&memcg->memsw, val); - oldusage = memcg->memsw.usage; - spin_unlock_irqrestore(&memcg->res.lock, flags); + mutex_unlock(&set_limit_mutex); if (!ret) break; + + oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE); try_to_free_mem_cgroup_pages(memcg, GFP_HIGHUSER_MOVABLE, true); curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE); if (curusage >= oldusage) _ Patches currently in -mm which might be from nishimura@xxxxxxxxxxxxxxxxx are cgroup-fix-potential-deadlock-in-pre_destroy-v2.patch cgroups-make-cgroup-config-a-submenu.patch memcg-introduce-charge-commit-cancel-style-of-functions.patch memcg-introduce-charge-commit-cancel-style-of-functions-fix.patch memcg-fix-gfp_mask-of-callers-of-charge.patch memcg-simple-migration-handling.patch memcg-do-not-recalculate-section-unnecessarily-in-init_section_page_cgroup.patch memcg-move-all-acccounts-to-parent-at-rmdir.patch memcg-handle-swap-caches.patch memcg-memswap-controller-kconfig.patch memcg-memswap-controller-core-make-resize-limit-hold-mutex.patch -- To unsubscribe from this list: send the line "unsubscribe mm-commits" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html