The patch titled mm: revert "oom: move oom_adj value" has been removed from the -mm tree. Its filename was mm-revert-oom-move-oom_adj-value.patch This patch was dropped because it was merged into mainline or a subsystem tree The current -mm tree may be found at http://userweb.kernel.org/~akpm/mmotm/ ------------------------------------------------------ Subject: mm: revert "oom: move oom_adj value" From: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx> The commit 2ff05b2b (oom: move oom_adj value) moveed the oom_adj value to the mm_struct. It was a very good first step for sanitize OOM. However Paul Menage reported the commit makes regression to his job scheduler. Current OOM logic can kill OOM_DISABLED process. Why? His program has the code of similar to the following. ... set_oom_adj(OOM_DISABLE); /* The job scheduler never killed by oom */ ... if (vfork() == 0) { set_oom_adj(0); /* Invoked child can be killed */ execve("foo-bar-cmd"); } .... vfork() parent and child are shared the same mm_struct. then above set_oom_adj(0) doesn't only change oom_adj for vfork() child, it's also change oom_adj for vfork() parent. Then, vfork() parent (job scheduler) lost OOM immune and it was killed. Actually, fork-setting-exec idiom is very frequently used in userland program. We must not break this assumption. Then, this patch revert commit 2ff05b2b and related commit. Reverted commit list --------------------- - commit 2ff05b2b4e (oom: move oom_adj value from task_struct to mm_struct) - commit 4d8b9135c3 (oom: avoid unnecessary mm locking and scanning for OOM_DISABLE) - commit 8123681022 (oom: only oom kill exiting tasks with attached memory) - commit 933b787b57 (mm: copy over oom_adj value at fork time) Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx> Cc: Paul Menage <menage@xxxxxxxxxx> Cc: David Rientjes <rientjes@xxxxxxxxxx> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> Cc: Rik van Riel <riel@xxxxxxxxxx> Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> Cc: Oleg Nesterov <oleg@xxxxxxxxxx> Cc: Nick Piggin <npiggin@xxxxxxx> Cc: Mel Gorman <mel@xxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- Documentation/filesystems/proc.txt | 15 ++---- fs/proc/base.c | 19 +------- include/linux/mm_types.h | 2 include/linux/sched.h | 1 kernel/fork.c | 1 mm/oom_kill.c | 64 ++++++++++++++++----------- 6 files changed, 48 insertions(+), 54 deletions(-) diff -puN Documentation/filesystems/proc.txt~mm-revert-oom-move-oom_adj-value Documentation/filesystems/proc.txt --- a/Documentation/filesystems/proc.txt~mm-revert-oom-move-oom_adj-value +++ a/Documentation/filesystems/proc.txt @@ -1167,13 +1167,11 @@ CHAPTER 3: PER-PROCESS PARAMETERS 3.1 /proc/<pid>/oom_adj - Adjust the oom-killer score ------------------------------------------------------ -This file can be used to adjust the score used to select which processes should -be killed in an out-of-memory situation. The oom_adj value is a characteristic -of the task's mm, so all threads that share an mm with pid will have the same -oom_adj value. A high value will increase the likelihood of this process being -killed by the oom-killer. Valid values are in the range -16 to +15 as -explained below and a special value of -17, which disables oom-killing -altogether for threads sharing pid's mm. +This file can be used to adjust the score used to select which processes +should be killed in an out-of-memory situation. Giving it a high score will +increase the likelihood of this process being killed by the oom-killer. Valid +values are in the range -16 to +15, plus the special value -17, which disables +oom-killing altogether for this process. The process to be killed in an out-of-memory situation is selected among all others based on its badness score. This value equals the original memory size of the process @@ -1187,9 +1185,6 @@ the parent's score if they do not share are the prime candidates to be killed. Having only one 'hungry' child will make parent less preferable than the child. -/proc/<pid>/oom_adj cannot be changed for kthreads since they are immune from -oom-killing already. - /proc/<pid>/oom_score shows process' current badness score. The following heuristics are then applied: diff -puN fs/proc/base.c~mm-revert-oom-move-oom_adj-value fs/proc/base.c --- a/fs/proc/base.c~mm-revert-oom-move-oom_adj-value +++ a/fs/proc/base.c @@ -1003,12 +1003,7 @@ static ssize_t oom_adjust_read(struct fi if (!task) return -ESRCH; - task_lock(task); - if (task->mm) - oom_adjust = task->mm->oom_adj; - else - oom_adjust = OOM_DISABLE; - task_unlock(task); + oom_adjust = task->oomkilladj; put_task_struct(task); len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust); @@ -1037,19 +1032,11 @@ static ssize_t oom_adjust_write(struct f task = get_proc_task(file->f_path.dentry->d_inode); if (!task) return -ESRCH; - task_lock(task); - if (!task->mm) { - task_unlock(task); - put_task_struct(task); - return -EINVAL; - } - if (oom_adjust < task->mm->oom_adj && !capable(CAP_SYS_RESOURCE)) { - task_unlock(task); + if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) { put_task_struct(task); return -EACCES; } - task->mm->oom_adj = oom_adjust; - task_unlock(task); + task->oomkilladj = oom_adjust; put_task_struct(task); if (end - buffer == 0) return -EIO; diff -puN include/linux/mm_types.h~mm-revert-oom-move-oom_adj-value include/linux/mm_types.h --- a/include/linux/mm_types.h~mm-revert-oom-move-oom_adj-value +++ a/include/linux/mm_types.h @@ -240,8 +240,6 @@ struct mm_struct { unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ - s8 oom_adj; /* OOM kill score adjustment (bit shift) */ - cpumask_t cpu_vm_mask; /* Architecture-specific MM context */ diff -puN include/linux/sched.h~mm-revert-oom-move-oom_adj-value include/linux/sched.h --- a/include/linux/sched.h~mm-revert-oom-move-oom_adj-value +++ a/include/linux/sched.h @@ -1198,6 +1198,7 @@ struct task_struct { * a short time */ unsigned char fpu_counter; + s8 oomkilladj; /* OOM kill score adjustment (bit shift). */ #ifdef CONFIG_BLK_DEV_IO_TRACE unsigned int btrace_seq; #endif diff -puN kernel/fork.c~mm-revert-oom-move-oom_adj-value kernel/fork.c --- a/kernel/fork.c~mm-revert-oom-move-oom_adj-value +++ a/kernel/fork.c @@ -426,7 +426,6 @@ static struct mm_struct * mm_init(struct init_rwsem(&mm->mmap_sem); INIT_LIST_HEAD(&mm->mmlist); mm->flags = (current->mm) ? current->mm->flags : default_dump_filter; - mm->oom_adj = (current->mm) ? current->mm->oom_adj : 0; mm->core_state = NULL; mm->nr_ptes = 0; set_mm_counter(mm, file_rss, 0); diff -puN mm/oom_kill.c~mm-revert-oom-move-oom_adj-value mm/oom_kill.c --- a/mm/oom_kill.c~mm-revert-oom-move-oom_adj-value +++ a/mm/oom_kill.c @@ -58,7 +58,6 @@ unsigned long badness(struct task_struct unsigned long points, cpu_time, run_time; struct mm_struct *mm; struct task_struct *child; - int oom_adj; task_lock(p); mm = p->mm; @@ -66,11 +65,6 @@ unsigned long badness(struct task_struct task_unlock(p); return 0; } - oom_adj = mm->oom_adj; - if (oom_adj == OOM_DISABLE) { - task_unlock(p); - return 0; - } /* * The memory size of the process is the basis for the badness. @@ -154,15 +148,15 @@ unsigned long badness(struct task_struct points /= 8; /* - * Adjust the score by oom_adj. + * Adjust the score by oomkilladj. */ - if (oom_adj) { - if (oom_adj > 0) { + if (p->oomkilladj) { + if (p->oomkilladj > 0) { if (!points) points = 1; - points <<= oom_adj; + points <<= p->oomkilladj; } else - points >>= -(oom_adj); + points >>= -(p->oomkilladj); } #ifdef DEBUG @@ -257,8 +251,11 @@ static struct task_struct *select_bad_pr *ppoints = ULONG_MAX; } + if (p->oomkilladj == OOM_DISABLE) + continue; + points = badness(p, uptime.tv_sec); - if (points > *ppoints) { + if (points > *ppoints || !chosen) { chosen = p; *ppoints = points; } @@ -307,7 +304,8 @@ static void dump_tasks(const struct mem_ } printk(KERN_INFO "[%5d] %5d %5d %8lu %8lu %3d %3d %s\n", p->pid, __task_cred(p)->uid, p->tgid, mm->total_vm, - get_mm_rss(mm), (int)task_cpu(p), mm->oom_adj, p->comm); + get_mm_rss(mm), (int)task_cpu(p), p->oomkilladj, + p->comm); task_unlock(p); } while_each_thread(g, p); } @@ -325,8 +323,11 @@ static void __oom_kill_task(struct task_ return; } - if (!p->mm) + if (!p->mm) { + WARN_ON(1); + printk(KERN_WARNING "tried to kill an mm-less task!\n"); return; + } if (verbose) printk(KERN_ERR "Killed process %d (%s)\n", @@ -348,13 +349,28 @@ static int oom_kill_task(struct task_str struct mm_struct *mm; struct task_struct *g, *q; - task_lock(p); mm = p->mm; - if (!mm || mm->oom_adj == OOM_DISABLE) { - task_unlock(p); + + /* WARNING: mm may not be dereferenced since we did not obtain its + * value from get_task_mm(p). This is OK since all we need to do is + * compare mm to q->mm below. + * + * Furthermore, even if mm contains a non-NULL value, p->mm may + * change to NULL at any time since we do not hold task_lock(p). + * However, this is of no concern to us. + */ + + if (mm == NULL) return 1; - } - task_unlock(p); + + /* + * Don't kill the process if any threads are set to OOM_DISABLE + */ + do_each_thread(g, q) { + if (q->mm == mm && q->oomkilladj == OOM_DISABLE) + return 1; + } while_each_thread(g, q); + __oom_kill_task(p, 1); /* @@ -377,11 +393,10 @@ static int oom_kill_process(struct task_ struct task_struct *c; if (printk_ratelimit()) { - task_lock(current); printk(KERN_WARNING "%s invoked oom-killer: " - "gfp_mask=0x%x, order=%d, oom_adj=%d\n", - current->comm, gfp_mask, order, - current->mm ? current->mm->oom_adj : OOM_DISABLE); + "gfp_mask=0x%x, order=%d, oomkilladj=%d\n", + current->comm, gfp_mask, order, current->oomkilladj); + task_lock(current); cpuset_print_task_mems_allowed(current); task_unlock(current); dump_stack(); @@ -394,9 +409,8 @@ static int oom_kill_process(struct task_ /* * If the task is already exiting, don't alarm the sysadmin or kill * its children or threads, just set TIF_MEMDIE so it can die quickly - * if its mm is still attached. */ - if (p->mm && (p->flags & PF_EXITING)) { + if (p->flags & PF_EXITING) { __oom_kill_task(p, 0); return 0; } _ Patches currently in -mm which might be from kosaki.motohiro@xxxxxxxxxxxxxx are origin.patch linux-next.patch mm-fix-for-infinite-churning-of-mlocked-pages.patch readahead-add-blk_run_backing_dev.patch readahead-add-blk_run_backing_dev-fix.patch readahead-add-blk_run_backing_dev-fix-fix-2.patch mm-clean-up-page_remove_rmap.patch mm-show_free_areas-display-slab-pages-in-two-separate-fields.patch mm-oom-analysis-add-per-zone-statistics-to-show_free_areas.patch mm-oom-analysis-add-buffer-cache-information-to-show_free_areas.patch mm-oom-analysis-show-kernel-stack-usage-in-proc-meminfo-and-oom-log-output.patch mm-oom-analysis-add-shmem-vmstat.patch mm-rename-pgmoved-variable-in-shrink_active_list.patch mm-shrink_inactive_list-nr_scan-accounting-fix-fix.patch mm-vmstat-add-isolate-pages.patch mm-vmstat-add-isolate-pages-fix.patch vmscan-throttle-direct-reclaim-when-too-many-pages-are-isolated-already.patch mm-remove-__addsub_zone_page_state.patch mm-count-only-reclaimable-lru-pages-v2.patch vmscan-dont-attempt-to-reclaim-anon-page-in-lumpy-reclaim-when-no-swap-space-is-avilable.patch vmscan-move-clearpageactive-from-move_active_pages-to-shrink_active_list.patch vmscan-kill-unnecessary-page-flag-test.patch vmscan-kill-unnecessary-prefetch.patch mm-perform-non-atomic-test-clear-of-pg_mlocked-on-free.patch tracing-page-allocator-add-trace-events-for-page-allocation-and-page-freeing.patch tracing-page-allocator-add-trace-event-for-page-traffic-related-to-the-buddy-lists.patch mm-drop-unneeded-double-negations.patch mm-introduce-page_lru_base_type.patch mm-return-boolean-from-page_is_file_cache.patch mm-return-boolean-from-page_has_private.patch mm-document-is_page_cache_freeable.patch getrusage-fill-ru_maxrss-value.patch getrusage-fill-ru_maxrss-value-update.patch memory-controller-soft-limit-documentation-v9.patch memory-controller-soft-limit-interface-v9.patch memory-controller-soft-limit-organize-cgroups-v9.patch memory-controller-soft-limit-organize-cgroups-v9-fix.patch memory-controller-soft-limit-refactor-reclaim-flags-v9.patch memory-controller-soft-limit-reclaim-on-contention-v9.patch memory-controller-soft-limit-reclaim-on-contention-v9-fix.patch memcg-improve-resource-counter-scalability.patch memcg-improve-resource-counter-scalability-v5.patch fs-symlink-write_begin-allocation-context-fix-reiser4-fix.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