From 72a1de39f89325a834a8c70b2a0d8f71d919f640 Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Wed, 17 Jun 2009 16:27:37 -0700 Subject: copy_process(): remove the unneeded clear_tsk_thread_flag(TIF_SIGPENDING) The forked child can have TIF_SIGPENDING if it was copied from parent's ti->flags. But this is harmless and actually almost never happens, because copy_process() can't succeed if signal_pending() == T. Signed-off-by: Oleg Nesterov Acked-by: Roland McGrath Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/fork.c | 1 - 1 file changed, 1 deletion(-) (limited to 'kernel/fork.c') diff --git a/kernel/fork.c b/kernel/fork.c index be022c200da6..467746b3f0aa 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1029,7 +1029,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, p->vfork_done = NULL; spin_lock_init(&p->alloc_lock); - clear_tsk_thread_flag(p, TIF_SIGPENDING); init_sigpending(&p->pending); p->utime = cputime_zero; -- cgit v1.2.3 From b43f3cbd21ffbd719fd4fa6642bfe6af255ded34 Mon Sep 17 00:00:00 2001 From: Alexey Dobriyan Date: Wed, 8 Jul 2009 01:54:37 +0400 Subject: headers: mnt_namespace.h redux Fix various silly problems wrt mnt_namespace.h: - exit_mnt_ns() isn't used, remove it - done that, sched.h and nsproxy.h inclusions aren't needed - mount.h inclusion was need for vfsmount_lock, but no longer - remove mnt_namespace.h inclusion from files which don't use anything from mnt_namespace.h Signed-off-by: Alexey Dobriyan Signed-off-by: Linus Torvalds --- fs/afs/mntpt.c | 1 - fs/namespace.c | 1 + fs/nfs/getroot.c | 1 - fs/reiserfs/super.c | 1 - include/linux/mnt_namespace.h | 13 ++----------- kernel/exit.c | 1 - kernel/fork.c | 1 - kernel/kmod.c | 1 - 8 files changed, 3 insertions(+), 17 deletions(-) (limited to 'kernel/fork.c') diff --git a/fs/afs/mntpt.c b/fs/afs/mntpt.c index c52be53f6946..5ffb570cd3a8 100644 --- a/fs/afs/mntpt.c +++ b/fs/afs/mntpt.c @@ -17,7 +17,6 @@ #include #include #include -#include #include "internal.h" diff --git a/fs/namespace.c b/fs/namespace.c index 3dc283fd4716..277c28a63ead 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -22,6 +22,7 @@ #include #include #include +#include #include #include #include diff --git a/fs/nfs/getroot.c b/fs/nfs/getroot.c index 46177cb87064..b35d2a616066 100644 --- a/fs/nfs/getroot.c +++ b/fs/nfs/getroot.c @@ -30,7 +30,6 @@ #include #include #include -#include #include #include diff --git a/fs/reiserfs/super.c b/fs/reiserfs/super.c index d3aeb061612b..7adea74d6a8a 100644 --- a/fs/reiserfs/super.c +++ b/fs/reiserfs/super.c @@ -24,7 +24,6 @@ #include #include #include -#include #include #include #include diff --git a/include/linux/mnt_namespace.h b/include/linux/mnt_namespace.h index 3beb2592b03f..d74785c2393a 100644 --- a/include/linux/mnt_namespace.h +++ b/include/linux/mnt_namespace.h @@ -2,10 +2,9 @@ #define _NAMESPACE_H_ #ifdef __KERNEL__ -#include -#include -#include +#include #include +#include struct mnt_namespace { atomic_t count; @@ -28,14 +27,6 @@ extern struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt); extern struct mnt_namespace *copy_mnt_ns(unsigned long, struct mnt_namespace *, struct fs_struct *); extern void put_mnt_ns(struct mnt_namespace *ns); - -static inline void exit_mnt_ns(struct task_struct *p) -{ - struct mnt_namespace *ns = p->nsproxy->mnt_ns; - if (ns) - put_mnt_ns(ns); -} - static inline void get_mnt_ns(struct mnt_namespace *ns) { atomic_inc(&ns->count); diff --git a/kernel/exit.c b/kernel/exit.c index 628d41f0dd54..869dc221733e 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -12,7 +12,6 @@ #include #include #include -#include #include #include #include diff --git a/kernel/fork.c b/kernel/fork.c index 467746b3f0aa..bd2959228871 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -17,7 +17,6 @@ #include #include #include -#include #include #include #include diff --git a/kernel/kmod.c b/kernel/kmod.c index 7e95bedb2bfc..385c31a1bdbf 100644 --- a/kernel/kmod.c +++ b/kernel/kmod.c @@ -24,7 +24,6 @@ #include #include #include -#include #include #include #include -- cgit v1.2.3 From ed900c054b541254f0ce5cedaf75206e29bd614e Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Thu, 16 Jul 2009 15:44:29 +0200 Subject: perf_counter: Log vfork as a fork event Right now we don't output vfork events. Even though we should always see an exec after a vfork, we may get perfcounter samples between the vfork and exec. These samples can lead to some confusion when parsing perfcounter data. To keep things consistent we should always log a fork event. It will result in a little more log data, but is less confusing to trace parsing tools. Signed-off-by: Anton Blanchard Signed-off-by: Peter Zijlstra LKML-Reference: <20090716104817.589309391@samba.org> Signed-off-by: Ingo Molnar --- kernel/fork.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) (limited to 'kernel/fork.c') diff --git a/kernel/fork.c b/kernel/fork.c index 467746b3f0aa..4812d60b29f8 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1408,14 +1408,11 @@ long do_fork(unsigned long clone_flags, if (clone_flags & CLONE_VFORK) { p->vfork_done = &vfork; init_completion(&vfork); - } else if (!(clone_flags & CLONE_VM)) { - /* - * vfork will do an exec which will call - * set_task_comm() - */ - perf_counter_fork(p); } + if (!(clone_flags & CLONE_THREAD)) + perf_counter_fork(p); + audit_finish_fork(p); tracehook_report_clone(regs, clone_flags, nr, p); -- cgit v1.2.3 From 933b787b57ca8bdc0fc8fb2cbf67b5e6d21beb84 Mon Sep 17 00:00:00 2001 From: Rik van Riel Date: Wed, 29 Jul 2009 15:02:07 -0700 Subject: mm: copy over oom_adj value at fork time Fix a post-2.6.31 regression which was introduced by 2ff05b2b4eac2e63d345fc731ea151a060247f53 ("oom: move oom_adj value from task_struct to mm_struct"). After moving the oom_adj value from the task struct to the mm_struct, the oom_adj value was no longer properly inherited by child processes. Copying over the oom_adj value at fork time fixes that bug. [kosaki.motohiro@jp.fujitsu.com: test for current->mm before dereferencing it] Signed-off-by: Rik van Riel Reported-by: Paul Menage Cc: KOSAKI Motohiro Acked-by: David Rientjes Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/fork.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel/fork.c') diff --git a/kernel/fork.c b/kernel/fork.c index 9b42695f0d14..29b532e718f7 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -426,6 +426,7 @@ static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p) 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); -- cgit v1.2.3 From 9f498cc5be7e013d8d6e4c616980ed0ffc8680d2 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 23 Jul 2009 14:46:33 +0200 Subject: perf_counter: Full task tracing In order to be able to distinguish between no samples due to inactivity and no samples due to task ended, Arjan asked for PERF_EVENT_EXIT events. This is useful to the boot delay instrumentation (bootchart) app. This patch changes the PERF_EVENT_FORK to be emitted on every clone, and adds PERF_EVENT_EXIT to be emitted on task exit, after the task's counters have been closed. This task tracing is controlled through: attr.comm || attr.mmap and through the new attr.task field. Suggested-by: Arjan van de Ven Cc: Paul Mackerras Cc: Anton Blanchard Signed-off-by: Peter Zijlstra [ cleaned up perf_counter.h a bit ] Signed-off-by: Ingo Molnar --- include/linux/perf_counter.h | 13 ++++++- kernel/fork.c | 4 +- kernel/perf_counter.c | 87 +++++++++++++++++++++++++++++--------------- 3 files changed, 71 insertions(+), 33 deletions(-) (limited to 'kernel/fork.c') diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index bd15d7a5f5ce..e604e6ef72dd 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h @@ -181,8 +181,9 @@ struct perf_counter_attr { freq : 1, /* use freq, not period */ inherit_stat : 1, /* per task counts */ enable_on_exec : 1, /* next exec enables */ + task : 1, /* trace fork/exit */ - __reserved_1 : 51; + __reserved_1 : 50; __u32 wakeup_events; /* wakeup every n events */ __u32 __reserved_2; @@ -308,6 +309,15 @@ enum perf_event_type { */ PERF_EVENT_COMM = 3, + /* + * struct { + * struct perf_event_header header; + * u32 pid, ppid; + * u32 tid, ptid; + * }; + */ + PERF_EVENT_EXIT = 4, + /* * struct { * struct perf_event_header header; @@ -323,6 +333,7 @@ enum perf_event_type { * struct { * struct perf_event_header header; * u32 pid, ppid; + * u32 tid, ptid; * }; */ PERF_EVENT_FORK = 7, diff --git a/kernel/fork.c b/kernel/fork.c index 29b532e718f7..466531eb92cc 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1269,6 +1269,7 @@ static struct task_struct *copy_process(unsigned long clone_flags, write_unlock_irq(&tasklist_lock); proc_fork_connector(p); cgroup_post_fork(p); + perf_counter_fork(p); return p; bad_fork_free_pid: @@ -1410,9 +1411,6 @@ long do_fork(unsigned long clone_flags, init_completion(&vfork); } - if (!(clone_flags & CLONE_THREAD)) - perf_counter_fork(p); - audit_finish_fork(p); tracehook_report_clone(regs, clone_flags, nr, p); diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 48471d75ae01..199ed4771315 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -42,6 +42,7 @@ static int perf_overcommit __read_mostly = 1; static atomic_t nr_counters __read_mostly; static atomic_t nr_mmap_counters __read_mostly; static atomic_t nr_comm_counters __read_mostly; +static atomic_t nr_task_counters __read_mostly; /* * perf counter paranoia level: @@ -1654,6 +1655,8 @@ static void free_counter(struct perf_counter *counter) atomic_dec(&nr_mmap_counters); if (counter->attr.comm) atomic_dec(&nr_comm_counters); + if (counter->attr.task) + atomic_dec(&nr_task_counters); } if (counter->destroy) @@ -2831,10 +2834,12 @@ perf_counter_read_event(struct perf_counter *counter, } /* - * fork tracking + * task tracking -- fork/exit + * + * enabled by: attr.comm | attr.mmap | attr.task */ -struct perf_fork_event { +struct perf_task_event { struct task_struct *task; struct { @@ -2842,37 +2847,42 @@ struct perf_fork_event { u32 pid; u32 ppid; + u32 tid; + u32 ptid; } event; }; -static void perf_counter_fork_output(struct perf_counter *counter, - struct perf_fork_event *fork_event) +static void perf_counter_task_output(struct perf_counter *counter, + struct perf_task_event *task_event) { struct perf_output_handle handle; - int size = fork_event->event.header.size; - struct task_struct *task = fork_event->task; + int size = task_event->event.header.size; + struct task_struct *task = task_event->task; int ret = perf_output_begin(&handle, counter, size, 0, 0); if (ret) return; - fork_event->event.pid = perf_counter_pid(counter, task); - fork_event->event.ppid = perf_counter_pid(counter, task->real_parent); + task_event->event.pid = perf_counter_pid(counter, task); + task_event->event.ppid = perf_counter_pid(counter, task->real_parent); - perf_output_put(&handle, fork_event->event); + task_event->event.tid = perf_counter_tid(counter, task); + task_event->event.ptid = perf_counter_tid(counter, task->real_parent); + + perf_output_put(&handle, task_event->event); perf_output_end(&handle); } -static int perf_counter_fork_match(struct perf_counter *counter) +static int perf_counter_task_match(struct perf_counter *counter) { - if (counter->attr.comm || counter->attr.mmap) + if (counter->attr.comm || counter->attr.mmap || counter->attr.task) return 1; return 0; } -static void perf_counter_fork_ctx(struct perf_counter_context *ctx, - struct perf_fork_event *fork_event) +static void perf_counter_task_ctx(struct perf_counter_context *ctx, + struct perf_task_event *task_event) { struct perf_counter *counter; @@ -2881,19 +2891,19 @@ static void perf_counter_fork_ctx(struct perf_counter_context *ctx, rcu_read_lock(); list_for_each_entry_rcu(counter, &ctx->event_list, event_entry) { - if (perf_counter_fork_match(counter)) - perf_counter_fork_output(counter, fork_event); + if (perf_counter_task_match(counter)) + perf_counter_task_output(counter, task_event); } rcu_read_unlock(); } -static void perf_counter_fork_event(struct perf_fork_event *fork_event) +static void perf_counter_task_event(struct perf_task_event *task_event) { struct perf_cpu_context *cpuctx; struct perf_counter_context *ctx; cpuctx = &get_cpu_var(perf_cpu_context); - perf_counter_fork_ctx(&cpuctx->ctx, fork_event); + perf_counter_task_ctx(&cpuctx->ctx, task_event); put_cpu_var(perf_cpu_context); rcu_read_lock(); @@ -2903,32 +2913,40 @@ static void perf_counter_fork_event(struct perf_fork_event *fork_event) */ ctx = rcu_dereference(current->perf_counter_ctxp); if (ctx) - perf_counter_fork_ctx(ctx, fork_event); + perf_counter_task_ctx(ctx, task_event); rcu_read_unlock(); } -void perf_counter_fork(struct task_struct *task) +static void perf_counter_task(struct task_struct *task, int new) { - struct perf_fork_event fork_event; + struct perf_task_event task_event; if (!atomic_read(&nr_comm_counters) && - !atomic_read(&nr_mmap_counters)) + !atomic_read(&nr_mmap_counters) && + !atomic_read(&nr_task_counters)) return; - fork_event = (struct perf_fork_event){ + task_event = (struct perf_task_event){ .task = task, .event = { .header = { - .type = PERF_EVENT_FORK, + .type = new ? PERF_EVENT_FORK : PERF_EVENT_EXIT, .misc = 0, - .size = sizeof(fork_event.event), + .size = sizeof(task_event.event), }, /* .pid */ /* .ppid */ + /* .tid */ + /* .ptid */ }, }; - perf_counter_fork_event(&fork_event); + perf_counter_task_event(&task_event); +} + +void perf_counter_fork(struct task_struct *task) +{ + perf_counter_task(task, 1); } /* @@ -3887,6 +3905,8 @@ done: atomic_inc(&nr_mmap_counters); if (counter->attr.comm) atomic_inc(&nr_comm_counters); + if (counter->attr.task) + atomic_inc(&nr_task_counters); } return counter; @@ -4248,8 +4268,10 @@ void perf_counter_exit_task(struct task_struct *child) struct perf_counter_context *child_ctx; unsigned long flags; - if (likely(!child->perf_counter_ctxp)) + if (likely(!child->perf_counter_ctxp)) { + perf_counter_task(child, 0); return; + } local_irq_save(flags); /* @@ -4267,15 +4289,22 @@ void perf_counter_exit_task(struct task_struct *child) * incremented the context's refcount before we do put_ctx below. */ spin_lock(&child_ctx->lock); - child->perf_counter_ctxp = NULL; /* * If this context is a clone; unclone it so it can't get * swapped to another process while we're removing all * the counters from it. */ unclone_ctx(child_ctx); - spin_unlock(&child_ctx->lock); - local_irq_restore(flags); + spin_unlock_irqrestore(&child_ctx->lock, flags); + + /* + * Report the task dead after unscheduling the counters so that we + * won't get any samples after PERF_EVENT_EXIT. We can however still + * get a few PERF_EVENT_READ events. + */ + perf_counter_task(child, 0); + + child->perf_counter_ctxp = NULL; /* * We can recurse on the same lock type through: -- cgit v1.2.3 From 9c8a8228d0827e0d91d28527209988f672f97d28 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 6 Aug 2009 15:09:28 -0700 Subject: execve: must clear current->clear_child_tid While looking at Jens Rosenboom bug report (http://lkml.org/lkml/2009/7/27/35) about strange sys_futex call done from a dying "ps" program, we found following problem. clone() syscall has special support for TID of created threads. This support includes two features. One (CLONE_CHILD_SETTID) is to set an integer into user memory with the TID value. One (CLONE_CHILD_CLEARTID) is to clear this same integer once the created thread dies. The integer location is a user provided pointer, provided at clone() time. kernel keeps this pointer value into current->clear_child_tid. At execve() time, we should make sure kernel doesnt keep this user provided pointer, as full user memory is replaced by a new one. As glibc fork() actually uses clone() syscall with CLONE_CHILD_SETTID and CLONE_CHILD_CLEARTID set, chances are high that we might corrupt user memory in forked processes. Following sequence could happen: 1) bash (or any program) starts a new process, by a fork() call that glibc maps to a clone( ... CLONE_CHILD_SETTID | CLONE_CHILD_CLEARTID ...) syscall 2) When new process starts, its current->clear_child_tid is set to a location that has a meaning only in bash (or initial program) context (&THREAD_SELF->tid) 3) This new process does the execve() syscall to start a new program. current->clear_child_tid is left unchanged (a non NULL value) 4) If this new program creates some threads, and initial thread exits, kernel will attempt to clear the integer pointed by current->clear_child_tid from mm_release() : if (tsk->clear_child_tid && !(tsk->flags & PF_SIGNALED) && atomic_read(&mm->mm_users) > 1) { u32 __user * tidptr = tsk->clear_child_tid; tsk->clear_child_tid = NULL; /* * We don't check the error code - if userspace has * not set up a proper pointer then tough luck. */ << here >> put_user(0, tidptr); sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0); } 5) OR : if new program is not multi-threaded, but spied by /proc/pid users (ps command for example), mm_users > 1, and the exiting program could corrupt 4 bytes in a persistent memory area (shm or memory mapped file) If current->clear_child_tid points to a writeable portion of memory of the new program, kernel happily and silently corrupts 4 bytes of memory, with unexpected effects. Fix is straightforward and should not break any sane program. Reported-by: Jens Rosenboom Acked-by: Linus Torvalds Signed-off-by: Eric Dumazet Signed-off-by: Oleg Nesterov Cc: Peter Zijlstra Cc: Sonny Rao Cc: Ingo Molnar Cc: Thomas Gleixner Cc: Ulrich Drepper Cc: Oleg Nesterov Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/fork.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'kernel/fork.c') diff --git a/kernel/fork.c b/kernel/fork.c index 466531eb92cc..021e1138556e 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -568,18 +568,18 @@ void mm_release(struct task_struct *tsk, struct mm_struct *mm) * the value intact in a core dump, and to save the unnecessary * trouble otherwise. Userland only wants this done for a sys_exit. */ - if (tsk->clear_child_tid - && !(tsk->flags & PF_SIGNALED) - && atomic_read(&mm->mm_users) > 1) { - u32 __user * tidptr = tsk->clear_child_tid; + if (tsk->clear_child_tid) { + if (!(tsk->flags & PF_SIGNALED) && + atomic_read(&mm->mm_users) > 1) { + /* + * We don't check the error code - if userspace has + * not set up a proper pointer then tough luck. + */ + put_user(0, tsk->clear_child_tid); + sys_futex(tsk->clear_child_tid, FUTEX_WAKE, + 1, NULL, NULL, 0); + } tsk->clear_child_tid = NULL; - - /* - * We don't check the error code - if userspace has - * not set up a proper pointer then tough luck. - */ - put_user(0, tidptr); - sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0); } } -- cgit v1.2.3 From 0753ba01e126020bf0f8150934903b48935b697d Mon Sep 17 00:00:00 2001 From: KOSAKI Motohiro Date: Tue, 18 Aug 2009 14:11:10 -0700 Subject: mm: revert "oom: move oom_adj value" 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 Cc: Paul Menage Cc: David Rientjes Cc: KAMEZAWA Hiroyuki Cc: Rik van Riel Cc: Linus Torvalds Cc: Oleg Nesterov Cc: Nick Piggin Cc: Mel Gorman Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- 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(-) (limited to 'kernel/fork.c') diff --git a/Documentation/filesystems/proc.txt b/Documentation/filesystems/proc.txt index fad18f9456e4..ffead13f9443 100644 --- a/Documentation/filesystems/proc.txt +++ b/Documentation/filesystems/proc.txt @@ -1167,13 +1167,11 @@ CHAPTER 3: PER-PROCESS PARAMETERS 3.1 /proc//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 the same memory. Thus forking servers are the prime candidates to be killed. Having only one 'hungry' child will make parent less preferable than the child. -/proc//oom_adj cannot be changed for kthreads since they are immune from -oom-killing already. - /proc//oom_score shows process' current badness score. The following heuristics are then applied: diff --git a/fs/proc/base.c b/fs/proc/base.c index 175db258942f..6f742f6658a9 100644 --- a/fs/proc/base.c +++ b/fs/proc/base.c @@ -1003,12 +1003,7 @@ static ssize_t oom_adjust_read(struct file *file, char __user *buf, 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 file *file, const char __user *buf, 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 --git a/include/linux/mm_types.h b/include/linux/mm_types.h index 7acc8439d9b3..0042090a4d70 100644 --- a/include/linux/mm_types.h +++ b/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 --git a/include/linux/sched.h b/include/linux/sched.h index 3ab08e4bb6b8..0f1ea4a66957 100644 --- a/include/linux/sched.h +++ b/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 --git a/kernel/fork.c b/kernel/fork.c index 021e1138556e..144326b7af50 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -426,7 +426,6 @@ static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p) 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 --git a/mm/oom_kill.c b/mm/oom_kill.c index 175a67a78a99..a7b2460e922b 100644 --- a/mm/oom_kill.c +++ b/mm/oom_kill.c @@ -58,7 +58,6 @@ unsigned long badness(struct task_struct *p, unsigned long uptime) 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 *p, unsigned long uptime) 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 *p, unsigned long uptime) 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_process(unsigned long *ppoints, *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_cgroup *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_struct *p, int verbose) 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_struct *p) 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 *p, gfp_t gfp_mask, int order, 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_struct *p, gfp_t gfp_mask, int order, /* * 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; } -- cgit v1.2.3 From 4ab6c08336535f8c8e42cf45d7adeda882eff06e Mon Sep 17 00:00:00 2001 From: Oleg Nesterov Date: Wed, 26 Aug 2009 14:29:24 -0700 Subject: clone(): fix race between copy_process() and de_thread() Spotted by Hiroshi Shimamoto who also provided the test-case below. copy_process() uses signal->count as a reference counter, but it is not. This test case #include #include #include #include #include #include void *null_thread(void *p) { for (;;) sleep(1); return NULL; } void *exec_thread(void *p) { execl("/bin/true", "/bin/true", NULL); return null_thread(p); } int main(int argc, char **argv) { for (;;) { pid_t pid; int ret, status; pid = fork(); if (pid < 0) break; if (!pid) { pthread_t tid; pthread_create(&tid, NULL, exec_thread, NULL); for (;;) pthread_create(&tid, NULL, null_thread, NULL); } do { ret = waitpid(pid, &status, 0); } while (ret == -1 && errno == EINTR); } return 0; } quickly creates an unkillable task. If copy_process(CLONE_THREAD) races with de_thread() copy_signal()->atomic(signal->count) breaks the signal->notify_count logic, and the execing thread can hang forever in kernel space. Change copy_process() to increment count/live only when we know for sure we can't fail. In this case the forked thread will take care of its reference to signal correctly. If copy_process() fails, check CLONE_THREAD flag. If it it set - do nothing, the counters were not changed and current belongs to the same thread group. If it is not set, ->signal must be released in any case (and ->count must be == 1), the forked child is the only thread in the thread group. We need more cleanups here, in particular signal->count should not be used by de_thread/__exit_signal at all. This patch only fixes the bug. Reported-by: Hiroshi Shimamoto Tested-by: Hiroshi Shimamoto Signed-off-by: Oleg Nesterov Acked-by: Roland McGrath Cc: KAMEZAWA Hiroyuki Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/fork.c | 20 +++++--------------- 1 file changed, 5 insertions(+), 15 deletions(-) (limited to 'kernel/fork.c') diff --git a/kernel/fork.c b/kernel/fork.c index 144326b7af50..e6c04d462ab2 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -815,11 +815,8 @@ static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) { struct signal_struct *sig; - if (clone_flags & CLONE_THREAD) { - atomic_inc(¤t->signal->count); - atomic_inc(¤t->signal->live); + if (clone_flags & CLONE_THREAD) return 0; - } sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL); tsk->signal = sig; @@ -877,16 +874,6 @@ void __cleanup_signal(struct signal_struct *sig) kmem_cache_free(signal_cachep, sig); } -static void cleanup_signal(struct task_struct *tsk) -{ - struct signal_struct *sig = tsk->signal; - - atomic_dec(&sig->live); - - if (atomic_dec_and_test(&sig->count)) - __cleanup_signal(sig); -} - static void copy_flags(unsigned long clone_flags, struct task_struct *p) { unsigned long new_flags = p->flags; @@ -1239,6 +1226,8 @@ static struct task_struct *copy_process(unsigned long clone_flags, } if (clone_flags & CLONE_THREAD) { + atomic_inc(¤t->signal->count); + atomic_inc(¤t->signal->live); p->group_leader = current->group_leader; list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); } @@ -1282,7 +1271,8 @@ bad_fork_cleanup_mm: if (p->mm) mmput(p->mm); bad_fork_cleanup_signal: - cleanup_signal(p); + if (!(clone_flags & CLONE_THREAD)) + __cleanup_signal(p->signal); bad_fork_cleanup_sighand: __cleanup_sighand(p->sighand); bad_fork_cleanup_fs: -- cgit v1.2.3