2
0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-05 20:04:19 +08:00
linux-next/kernel/exit.c

1849 lines
46 KiB
C
Raw Normal View History

/*
* linux/kernel/exit.c
*
* Copyright (C) 1991, 1992 Linus Torvalds
*/
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/capability.h>
#include <linux/completion.h>
#include <linux/personality.h>
#include <linux/tty.h>
#include <linux/mnt_namespace.h>
#include <linux/iocontext.h>
#include <linux/key.h>
#include <linux/security.h>
#include <linux/cpu.h>
#include <linux/acct.h>
#include <linux/tsacct_kern.h>
#include <linux/file.h>
#include <linux/fdtable.h>
#include <linux/binfmts.h>
#include <linux/nsproxy.h>
#include <linux/pid_namespace.h>
#include <linux/ptrace.h>
#include <linux/profile.h>
#include <linux/mount.h>
#include <linux/proc_fs.h>
#include <linux/kthread.h>
#include <linux/mempolicy.h>
#include <linux/taskstats_kern.h>
#include <linux/delayacct.h>
#include <linux/freezer.h>
#include <linux/cgroup.h>
#include <linux/syscalls.h>
#include <linux/signal.h>
#include <linux/posix-timers.h>
#include <linux/cn_proc.h>
#include <linux/mutex.h>
#include <linux/futex.h>
#include <linux/pipe_fs_i.h>
#include <linux/audit.h> /* for audit_free() */
#include <linux/resource.h>
#include <linux/blkdev.h>
#include <linux/task_io_accounting_ops.h>
#include <linux/tracehook.h>
#include <trace/sched.h>
#include <asm/uaccess.h>
#include <asm/unistd.h>
#include <asm/pgtable.h>
#include <asm/mmu_context.h>
static void exit_mm(struct task_struct * tsk);
static inline int task_detached(struct task_struct *p)
{
return p->exit_signal == -1;
}
static void __unhash_process(struct task_struct *p)
{
nr_threads--;
detach_pid(p, PIDTYPE_PID);
if (thread_group_leader(p)) {
detach_pid(p, PIDTYPE_PGID);
detach_pid(p, PIDTYPE_SID);
list_del_rcu(&p->tasks);
__get_cpu_var(process_counts)--;
}
list_del_rcu(&p->thread_group);
list_del_init(&p->sibling);
}
/*
* This function expects the tasklist_lock write-locked.
*/
static void __exit_signal(struct task_struct *tsk)
{
struct signal_struct *sig = tsk->signal;
struct sighand_struct *sighand;
BUG_ON(!sig);
BUG_ON(!atomic_read(&sig->count));
sighand = rcu_dereference(tsk->sighand);
spin_lock(&sighand->siglock);
posix_cpu_timers_exit(tsk);
if (atomic_dec_and_test(&sig->count))
posix_cpu_timers_exit_group(tsk);
else {
/*
* If there is any task waiting for the group exit
* then notify it:
*/
if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count)
wake_up_process(sig->group_exit_task);
if (tsk == sig->curr_target)
sig->curr_target = next_thread(tsk);
/*
* Accumulate here the counters for all threads but the
* group leader as they die, so they can be added into
* the process-wide totals when those are taken.
* The group leader stays around as a zombie as long
* as there are other threads. When it gets reaped,
* the exit.c code will add its counts into these totals.
* We won't ever get here for the group leader, since it
* will have been the last reference on the signal_struct.
*/
sig->gtime = cputime_add(sig->gtime, task_gtime(tsk));
sig->min_flt += tsk->min_flt;
sig->maj_flt += tsk->maj_flt;
sig->nvcsw += tsk->nvcsw;
sig->nivcsw += tsk->nivcsw;
sig->inblock += task_io_get_inblock(tsk);
sig->oublock += task_io_get_oublock(tsk);
task_io_accounting_add(&sig->ioac, &tsk->ioac);
sig = NULL; /* Marker for below. */
}
__unhash_process(tsk);
/*
* Do this under ->siglock, we can race with another thread
* doing sigqueue_free() if we have SIGQUEUE_PREALLOC signals.
*/
flush_sigqueue(&tsk->pending);
tsk->signal = NULL;
tsk->sighand = NULL;
spin_unlock(&sighand->siglock);
__cleanup_sighand(sighand);
clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
if (sig) {
flush_sigqueue(&sig->shared_pending);
taskstats_tgid_free(sig);
/*
* Make sure ->signal can't go away under rq->lock,
* see account_group_exec_runtime().
*/
task_rq_unlock_wait(tsk);
__cleanup_signal(sig);
}
}
[PATCH] task: RCU protect task->usage A big problem with rcu protected data structures that are also reference counted is that you must jump through several hoops to increase the reference count. I think someone finally implemented atomic_inc_not_zero(&count) to automate the common case. Unfortunately this means you must special case the rcu access case. When data structures are only visible via rcu in a manner that is not determined by the reference count on the object (i.e. tasks are visible until their zombies are reaped) there is a much simpler technique we can employ. Simply delaying the decrement of the reference count until the rcu interval is over. What that means is that the proc code that looks up a task and later wants to sleep can now do: rcu_read_lock(); task = find_task_by_pid(some_pid); if (task) { get_task_struct(task); } rcu_read_unlock(); The effect on the rest of the kernel is that put_task_struct becomes cheaper and immediate, and in the case where the task has been reaped it frees the task immediate instead of unnecessarily waiting an until the rcu interval is over. Cleanup of task_struct does not happen when its reference count drops to zero, instead cleanup happens when release_task is called. Tasks can only be looked up via rcu before release_task is called. All rcu protected members of task_struct are freed by release_task. Therefore we can move call_rcu from put_task_struct into release_task. And we can modify release_task to not immediately release the reference count but instead have it call put_task_struct from the function it gives to call_rcu. The end result: - get_task_struct is safe in an rcu context where we have just looked up the task. - put_task_struct() simplifies into its old pre rcu self. This reorganization also makes put_task_struct uncallable from modules as it is not exported but it does not appear to be called from any modules so this should not be an issue, and is trivially fixed. Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-31 18:31:37 +08:00
static void delayed_put_task_struct(struct rcu_head *rhp)
{
struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
trace_sched_process_free(tsk);
put_task_struct(tsk);
[PATCH] task: RCU protect task->usage A big problem with rcu protected data structures that are also reference counted is that you must jump through several hoops to increase the reference count. I think someone finally implemented atomic_inc_not_zero(&count) to automate the common case. Unfortunately this means you must special case the rcu access case. When data structures are only visible via rcu in a manner that is not determined by the reference count on the object (i.e. tasks are visible until their zombies are reaped) there is a much simpler technique we can employ. Simply delaying the decrement of the reference count until the rcu interval is over. What that means is that the proc code that looks up a task and later wants to sleep can now do: rcu_read_lock(); task = find_task_by_pid(some_pid); if (task) { get_task_struct(task); } rcu_read_unlock(); The effect on the rest of the kernel is that put_task_struct becomes cheaper and immediate, and in the case where the task has been reaped it frees the task immediate instead of unnecessarily waiting an until the rcu interval is over. Cleanup of task_struct does not happen when its reference count drops to zero, instead cleanup happens when release_task is called. Tasks can only be looked up via rcu before release_task is called. All rcu protected members of task_struct are freed by release_task. Therefore we can move call_rcu from put_task_struct into release_task. And we can modify release_task to not immediately release the reference count but instead have it call put_task_struct from the function it gives to call_rcu. The end result: - get_task_struct is safe in an rcu context where we have just looked up the task. - put_task_struct() simplifies into its old pre rcu self. This reorganization also makes put_task_struct uncallable from modules as it is not exported but it does not appear to be called from any modules so this should not be an issue, and is trivially fixed. Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-31 18:31:37 +08:00
}
void release_task(struct task_struct * p)
{
struct task_struct *leader;
int zap_leader;
repeat:
tracehook_prepare_release_task(p);
atomic_dec(&p->user->processes);
proc_flush_task(p);
write_lock_irq(&tasklist_lock);
tracehook_finish_release_task(p);
__exit_signal(p);
/*
* If we are the last non-leader member of the thread
* group, and the leader is zombie, then notify the
* group leader's parent process. (if it wants notification.)
*/
zap_leader = 0;
leader = p->group_leader;
if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
BUG_ON(task_detached(leader));
do_notify_parent(leader, leader->exit_signal);
/*
* If we were the last child thread and the leader has
* exited already, and the leader's parent ignores SIGCHLD,
* then we are the one who should release the leader.
*
* do_notify_parent() will have marked it self-reaping in
* that case.
*/
zap_leader = task_detached(leader);
/*
* This maintains the invariant that release_task()
* only runs on a task in EXIT_DEAD, just for sanity.
*/
if (zap_leader)
leader->exit_state = EXIT_DEAD;
}
write_unlock_irq(&tasklist_lock);
release_thread(p);
[PATCH] task: RCU protect task->usage A big problem with rcu protected data structures that are also reference counted is that you must jump through several hoops to increase the reference count. I think someone finally implemented atomic_inc_not_zero(&count) to automate the common case. Unfortunately this means you must special case the rcu access case. When data structures are only visible via rcu in a manner that is not determined by the reference count on the object (i.e. tasks are visible until their zombies are reaped) there is a much simpler technique we can employ. Simply delaying the decrement of the reference count until the rcu interval is over. What that means is that the proc code that looks up a task and later wants to sleep can now do: rcu_read_lock(); task = find_task_by_pid(some_pid); if (task) { get_task_struct(task); } rcu_read_unlock(); The effect on the rest of the kernel is that put_task_struct becomes cheaper and immediate, and in the case where the task has been reaped it frees the task immediate instead of unnecessarily waiting an until the rcu interval is over. Cleanup of task_struct does not happen when its reference count drops to zero, instead cleanup happens when release_task is called. Tasks can only be looked up via rcu before release_task is called. All rcu protected members of task_struct are freed by release_task. Therefore we can move call_rcu from put_task_struct into release_task. And we can modify release_task to not immediately release the reference count but instead have it call put_task_struct from the function it gives to call_rcu. The end result: - get_task_struct is safe in an rcu context where we have just looked up the task. - put_task_struct() simplifies into its old pre rcu self. This reorganization also makes put_task_struct uncallable from modules as it is not exported but it does not appear to be called from any modules so this should not be an issue, and is trivially fixed. Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-31 18:31:37 +08:00
call_rcu(&p->rcu, delayed_put_task_struct);
p = leader;
if (unlikely(zap_leader))
goto repeat;
}
/*
* This checks not only the pgrp, but falls back on the pid if no
* satisfactory pgrp is found. I dunno - gdb doesn't work correctly
* without this...
*
* The caller must hold rcu lock or the tasklist lock.
*/
struct pid *session_of_pgrp(struct pid *pgrp)
{
struct task_struct *p;
struct pid *sid = NULL;
p = pid_task(pgrp, PIDTYPE_PGID);
if (p == NULL)
p = pid_task(pgrp, PIDTYPE_PID);
if (p != NULL)
sid = task_session(p);
return sid;
}
/*
* Determine if a process group is "orphaned", according to the POSIX
* definition in 2.2.2.52. Orphaned process groups are not to be affected
* by terminal-generated stop signals. Newly orphaned process groups are
* to receive a SIGHUP and a SIGCONT.
*
* "I ask you, have you ever known what it is to be an orphan?"
*/
static int will_become_orphaned_pgrp(struct pid *pgrp, struct task_struct *ignored_task)
{
struct task_struct *p;
do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
if ((p == ignored_task) ||
(p->exit_state && thread_group_empty(p)) ||
is_global_init(p->real_parent))
continue;
if (task_pgrp(p->real_parent) != pgrp &&
task_session(p->real_parent) == task_session(p))
return 0;
} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
return 1;
}
int is_current_pgrp_orphaned(void)
{
int retval;
read_lock(&tasklist_lock);
retval = will_become_orphaned_pgrp(task_pgrp(current), NULL);
read_unlock(&tasklist_lock);
return retval;
}
static int has_stopped_jobs(struct pid *pgrp)
{
int retval = 0;
struct task_struct *p;
do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
if (!task_is_stopped(p))
continue;
retval = 1;
break;
} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
return retval;
}
/*
* Check to see if any process groups have become orphaned as
* a result of our exiting, and if they have any stopped jobs,
* send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
*/
static void
kill_orphaned_pgrp(struct task_struct *tsk, struct task_struct *parent)
{
struct pid *pgrp = task_pgrp(tsk);
struct task_struct *ignored_task = tsk;
if (!parent)
/* exit: our father is in a different pgrp than
* we are and we were the only connection outside.
*/
parent = tsk->real_parent;
else
/* reparent: our child is in a different pgrp than
* we are, and it was the only connection outside.
*/
ignored_task = NULL;
if (task_pgrp(parent) != pgrp &&
task_session(parent) == task_session(tsk) &&
will_become_orphaned_pgrp(pgrp, ignored_task) &&
has_stopped_jobs(pgrp)) {
__kill_pgrp_info(SIGHUP, SEND_SIG_PRIV, pgrp);
__kill_pgrp_info(SIGCONT, SEND_SIG_PRIV, pgrp);
}
}
/**
* reparent_to_kthreadd - Reparent the calling kernel thread to kthreadd
*
* If a kernel thread is launched as a result of a system call, or if
* it ever exits, it should generally reparent itself to kthreadd so it
* isn't in the way of other processes and is correctly cleaned up on exit.
*
* The various task state such as scheduling policy and priority may have
* been inherited from a user process, so we reset them to sane values here.
*
* NOTE that reparent_to_kthreadd() gives the caller full capabilities.
*/
static void reparent_to_kthreadd(void)
{
write_lock_irq(&tasklist_lock);
ptrace_unlink(current);
/* Reparent to init */
current->real_parent = current->parent = kthreadd_task;
list_move_tail(&current->sibling, &current->real_parent->children);
/* Set the exit signal to SIGCHLD so we signal init on exit */
current->exit_signal = SIGCHLD;
if (task_nice(current) < 0)
set_user_nice(current, 0);
/* cpus_allowed? */
/* rt_priority? */
/* signals? */
security_task_reparent_to_init(current);
memcpy(current->signal->rlim, init_task.signal->rlim,
sizeof(current->signal->rlim));
atomic_inc(&(INIT_USER->__count));
write_unlock_irq(&tasklist_lock);
switch_uid(INIT_USER);
}
void __set_special_pids(struct pid *pid)
{
struct task_struct *curr = current->group_leader;
pid_t nr = pid_nr(pid);
if (task_session(curr) != pid) {
change_pid(curr, PIDTYPE_SID, pid);
set_task_session(curr, nr);
}
if (task_pgrp(curr) != pid) {
change_pid(curr, PIDTYPE_PGID, pid);
set_task_pgrp(curr, nr);
}
}
static void set_special_pids(struct pid *pid)
{
write_lock_irq(&tasklist_lock);
__set_special_pids(pid);
write_unlock_irq(&tasklist_lock);
}
/*
* Let kernel threads use this to say that they
* allow a certain signal (since daemonize() will
* have disabled all of them by default).
*/
int allow_signal(int sig)
{
if (!valid_signal(sig) || sig < 1)
return -EINVAL;
spin_lock_irq(&current->sighand->siglock);
sigdelset(&current->blocked, sig);
if (!current->mm) {
/* Kernel threads handle their own signals.
Let the signal code know it'll be handled, so
that they don't get converted to SIGKILL or
just silently dropped */
current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
}
recalc_sigpending();
spin_unlock_irq(&current->sighand->siglock);
return 0;
}
EXPORT_SYMBOL(allow_signal);
int disallow_signal(int sig)
{
if (!valid_signal(sig) || sig < 1)
return -EINVAL;
spin_lock_irq(&current->sighand->siglock);
current->sighand->action[(sig)-1].sa.sa_handler = SIG_IGN;
recalc_sigpending();
spin_unlock_irq(&current->sighand->siglock);
return 0;
}
EXPORT_SYMBOL(disallow_signal);
/*
* Put all the gunge required to become a kernel thread without
* attached user resources in one place where it belongs.
*/
void daemonize(const char *name, ...)
{
va_list args;
struct fs_struct *fs;
sigset_t blocked;
va_start(args, name);
vsnprintf(current->comm, sizeof(current->comm), name, args);
va_end(args);
/*
* If we were started as result of loading a module, close all of the
* user space pages. We don't need them, and if we didn't close them
* they would be locked into memory.
*/
exit_mm(current);
/*
* We don't want to have TIF_FREEZE set if the system-wide hibernation
* or suspend transition begins right now.
*/
current->flags |= (PF_NOFREEZE | PF_KTHREAD);
if (current->nsproxy != &init_nsproxy) {
get_nsproxy(&init_nsproxy);
switch_task_namespaces(current, &init_nsproxy);
}
set_special_pids(&init_struct_pid);
[PATCH] tty: ->signal->tty locking Fix the locking of signal->tty. Use ->sighand->siglock to protect ->signal->tty; this lock is already used by most other members of ->signal/->sighand. And unless we are 'current' or the tasklist_lock is held we need ->siglock to access ->signal anyway. (NOTE: sys_unshare() is broken wrt ->sighand locking rules) Note that tty_mutex is held over tty destruction, so while holding tty_mutex any tty pointer remains valid. Otherwise the lifetime of ttys are governed by their open file handles. This leaves some holes for tty access from signal->tty (or any other non file related tty access). It solves the tty SLAB scribbles we were seeing. (NOTE: the change from group_send_sig_info to __group_send_sig_info needs to be examined by someone familiar with the security framework, I think it is safe given the SEND_SIG_PRIV from other __group_send_sig_info invocations) [schwidefsky@de.ibm.com: 3270 fix] [akpm@osdl.org: various post-viro fixes] Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Acked-by: Alan Cox <alan@redhat.com> Cc: Oleg Nesterov <oleg@tv-sign.ru> Cc: Prarit Bhargava <prarit@redhat.com> Cc: Chris Wright <chrisw@sous-sol.org> Cc: Roland McGrath <roland@redhat.com> Cc: Stephen Smalley <sds@tycho.nsa.gov> Cc: James Morris <jmorris@namei.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Jeff Dike <jdike@addtoit.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Jan Kara <jack@ucw.cz> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 18:36:04 +08:00
proc_clear_tty(current);
/* Block and flush all signals */
sigfillset(&blocked);
sigprocmask(SIG_BLOCK, &blocked, NULL);
flush_signals(current);
/* Become as one with the init task */
exit_fs(current); /* current->fs->count--; */
fs = init_task.fs;
current->fs = fs;
atomic_inc(&fs->count);
exit_files(current);
current->files = init_task.files;
atomic_inc(&current->files->count);
reparent_to_kthreadd();
}
EXPORT_SYMBOL(daemonize);
static void close_files(struct files_struct * files)
{
int i, j;
struct fdtable *fdt;
j = 0;
/*
* It is safe to dereference the fd table without RCU or
* ->file_lock because this is the last reference to the
* files structure.
*/
fdt = files_fdtable(files);
for (;;) {
unsigned long set;
i = j * __NFDBITS;
if (i >= fdt->max_fds)
break;
set = fdt->open_fds->fds_bits[j++];
while (set) {
if (set & 1) {
struct file * file = xchg(&fdt->fd[i], NULL);
if (file) {
filp_close(file, files);
cond_resched();
}
}
i++;
set >>= 1;
}
}
}
struct files_struct *get_files_struct(struct task_struct *task)
{
struct files_struct *files;
task_lock(task);
files = task->files;
if (files)
atomic_inc(&files->count);
task_unlock(task);
return files;
}
void put_files_struct(struct files_struct *files)
{
struct fdtable *fdt;
if (atomic_dec_and_test(&files->count)) {
close_files(files);
/*
* Free the fd and fdset arrays if we expanded them.
* If the fdtable was embedded, pass files for freeing
* at the end of the RCU grace period. Otherwise,
* you can free files immediately.
*/
fdt = files_fdtable(files);
if (fdt != &files->fdtab)
kmem_cache_free(files_cachep, files);
free_fdtable(fdt);
}
}
void reset_files_struct(struct files_struct *files)
{
struct task_struct *tsk = current;
struct files_struct *old;
old = tsk->files;
task_lock(tsk);
tsk->files = files;
task_unlock(tsk);
put_files_struct(old);
}
void exit_files(struct task_struct *tsk)
{
struct files_struct * files = tsk->files;
if (files) {
task_lock(tsk);
tsk->files = NULL;
task_unlock(tsk);
put_files_struct(files);
}
}
void put_fs_struct(struct fs_struct *fs)
{
/* No need to hold fs->lock if we are killing it */
if (atomic_dec_and_test(&fs->count)) {
path_put(&fs->root);
path_put(&fs->pwd);
kmem_cache_free(fs_cachep, fs);
}
}
void exit_fs(struct task_struct *tsk)
{
struct fs_struct * fs = tsk->fs;
if (fs) {
task_lock(tsk);
tsk->fs = NULL;
task_unlock(tsk);
put_fs_struct(fs);
}
}
EXPORT_SYMBOL_GPL(exit_fs);
cgroups: add an owner to the mm_struct Remove the mem_cgroup member from mm_struct and instead adds an owner. This approach was suggested by Paul Menage. The advantage of this approach is that, once the mm->owner is known, using the subsystem id, the cgroup can be determined. It also allows several control groups that are virtually grouped by mm_struct, to exist independent of the memory controller i.e., without adding mem_cgroup's for each controller, to mm_struct. A new config option CONFIG_MM_OWNER is added and the memory resource controller selects this config option. This patch also adds cgroup callbacks to notify subsystems when mm->owner changes. The mm_cgroup_changed callback is called with the task_lock() of the new task held and is called just prior to changing the mm->owner. I am indebted to Paul Menage for the several reviews of this patchset and helping me make it lighter and simpler. This patch was tested on a powerpc box, it was compiled with both the MM_OWNER config turned on and off. After the thread group leader exits, it's moved to init_css_state by cgroup_exit(), thus all future charges from runnings threads would be redirected to the init_css_set's subsystem. Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelianov <xemul@openvz.org> Cc: Hugh Dickins <hugh@veritas.com> Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com> Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp> Cc: Hirokazu Takahashi <taka@valinux.co.jp> Cc: David Rientjes <rientjes@google.com>, Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Pekka Enberg <penberg@cs.helsinki.fi> Reviewed-by: Paul Menage <menage@google.com> Cc: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-29 16:00:16 +08:00
#ifdef CONFIG_MM_OWNER
/*
* Task p is exiting and it owned mm, lets find a new owner for it
*/
static inline int
mm_need_new_owner(struct mm_struct *mm, struct task_struct *p)
{
/*
* If there are other users of the mm and the owner (us) is exiting
* we need to find a new owner to take on the responsibility.
*/
if (atomic_read(&mm->mm_users) <= 1)
return 0;
if (mm->owner != p)
return 0;
return 1;
}
void mm_update_next_owner(struct mm_struct *mm)
{
struct task_struct *c, *g, *p = current;
retry:
if (!mm_need_new_owner(mm, p))
return;
read_lock(&tasklist_lock);
/*
* Search in the children
*/
list_for_each_entry(c, &p->children, sibling) {
if (c->mm == mm)
goto assign_new_owner;
}
/*
* Search in the siblings
*/
list_for_each_entry(c, &p->parent->children, sibling) {
if (c->mm == mm)
goto assign_new_owner;
}
/*
* Search through everything else. We should not get
* here often
*/
do_each_thread(g, c) {
if (c->mm == mm)
goto assign_new_owner;
} while_each_thread(g, c);
read_unlock(&tasklist_lock);
mm owner: fix race between swapoff and exit There's a race between mm->owner assignment and swapoff, more easily seen when task slab poisoning is turned on. The condition occurs when try_to_unuse() runs in parallel with an exiting task. A similar race can occur with callers of get_task_mm(), such as /proc/<pid>/<mmstats> or ptrace or page migration. CPU0 CPU1 try_to_unuse looks at mm = task0->mm increments mm->mm_users task 0 exits mm->owner needs to be updated, but no new owner is found (mm_users > 1, but no other task has task->mm = task0->mm) mm_update_next_owner() leaves mmput(mm) decrements mm->mm_users task0 freed dereferencing mm->owner fails The fix is to notify the subsystem via mm_owner_changed callback(), if no new owner is found, by specifying the new task as NULL. Jiri Slaby: mm->owner was set to NULL prior to calling cgroup_mm_owner_callbacks(), but must be set after that, so as not to pass NULL as old owner causing oops. Daisuke Nishimura: mm_update_next_owner() may set mm->owner to NULL, but mem_cgroup_from_task() and its callers need to take account of this situation to avoid oops. Hugh Dickins: Lockdep warning and hang below exec_mmap() when testing these patches. exit_mm() up_reads mmap_sem before calling mm_update_next_owner(), so exec_mmap() now needs to do the same. And with that repositioning, there's now no point in mm_need_new_owner() allowing for NULL mm. Reported-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Signed-off-by: Jiri Slaby <jirislaby@gmail.com> Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp> Signed-off-by: Hugh Dickins <hugh@veritas.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Paul Menage <menage@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-09-29 06:09:31 +08:00
/*
* We found no owner yet mm_users > 1: this implies that we are
* most likely racing with swapoff (try_to_unuse()) or /proc or
* ptrace or page migration (get_task_mm()). Mark owner as NULL,
* so that subsystems can understand the callback and take action.
*/
down_write(&mm->mmap_sem);
cgroup_mm_owner_callbacks(mm->owner, NULL);
mm->owner = NULL;
up_write(&mm->mmap_sem);
cgroups: add an owner to the mm_struct Remove the mem_cgroup member from mm_struct and instead adds an owner. This approach was suggested by Paul Menage. The advantage of this approach is that, once the mm->owner is known, using the subsystem id, the cgroup can be determined. It also allows several control groups that are virtually grouped by mm_struct, to exist independent of the memory controller i.e., without adding mem_cgroup's for each controller, to mm_struct. A new config option CONFIG_MM_OWNER is added and the memory resource controller selects this config option. This patch also adds cgroup callbacks to notify subsystems when mm->owner changes. The mm_cgroup_changed callback is called with the task_lock() of the new task held and is called just prior to changing the mm->owner. I am indebted to Paul Menage for the several reviews of this patchset and helping me make it lighter and simpler. This patch was tested on a powerpc box, it was compiled with both the MM_OWNER config turned on and off. After the thread group leader exits, it's moved to init_css_state by cgroup_exit(), thus all future charges from runnings threads would be redirected to the init_css_set's subsystem. Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelianov <xemul@openvz.org> Cc: Hugh Dickins <hugh@veritas.com> Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com> Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp> Cc: Hirokazu Takahashi <taka@valinux.co.jp> Cc: David Rientjes <rientjes@google.com>, Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Pekka Enberg <penberg@cs.helsinki.fi> Reviewed-by: Paul Menage <menage@google.com> Cc: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-29 16:00:16 +08:00
return;
assign_new_owner:
BUG_ON(c == p);
get_task_struct(c);
read_unlock(&tasklist_lock);
down_write(&mm->mmap_sem);
cgroups: add an owner to the mm_struct Remove the mem_cgroup member from mm_struct and instead adds an owner. This approach was suggested by Paul Menage. The advantage of this approach is that, once the mm->owner is known, using the subsystem id, the cgroup can be determined. It also allows several control groups that are virtually grouped by mm_struct, to exist independent of the memory controller i.e., without adding mem_cgroup's for each controller, to mm_struct. A new config option CONFIG_MM_OWNER is added and the memory resource controller selects this config option. This patch also adds cgroup callbacks to notify subsystems when mm->owner changes. The mm_cgroup_changed callback is called with the task_lock() of the new task held and is called just prior to changing the mm->owner. I am indebted to Paul Menage for the several reviews of this patchset and helping me make it lighter and simpler. This patch was tested on a powerpc box, it was compiled with both the MM_OWNER config turned on and off. After the thread group leader exits, it's moved to init_css_state by cgroup_exit(), thus all future charges from runnings threads would be redirected to the init_css_set's subsystem. Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelianov <xemul@openvz.org> Cc: Hugh Dickins <hugh@veritas.com> Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com> Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp> Cc: Hirokazu Takahashi <taka@valinux.co.jp> Cc: David Rientjes <rientjes@google.com>, Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Pekka Enberg <penberg@cs.helsinki.fi> Reviewed-by: Paul Menage <menage@google.com> Cc: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-29 16:00:16 +08:00
/*
* The task_lock protects c->mm from changing.
* We always want mm->owner->mm == mm
*/
task_lock(c);
if (c->mm != mm) {
task_unlock(c);
up_write(&mm->mmap_sem);
cgroups: add an owner to the mm_struct Remove the mem_cgroup member from mm_struct and instead adds an owner. This approach was suggested by Paul Menage. The advantage of this approach is that, once the mm->owner is known, using the subsystem id, the cgroup can be determined. It also allows several control groups that are virtually grouped by mm_struct, to exist independent of the memory controller i.e., without adding mem_cgroup's for each controller, to mm_struct. A new config option CONFIG_MM_OWNER is added and the memory resource controller selects this config option. This patch also adds cgroup callbacks to notify subsystems when mm->owner changes. The mm_cgroup_changed callback is called with the task_lock() of the new task held and is called just prior to changing the mm->owner. I am indebted to Paul Menage for the several reviews of this patchset and helping me make it lighter and simpler. This patch was tested on a powerpc box, it was compiled with both the MM_OWNER config turned on and off. After the thread group leader exits, it's moved to init_css_state by cgroup_exit(), thus all future charges from runnings threads would be redirected to the init_css_set's subsystem. Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelianov <xemul@openvz.org> Cc: Hugh Dickins <hugh@veritas.com> Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com> Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp> Cc: Hirokazu Takahashi <taka@valinux.co.jp> Cc: David Rientjes <rientjes@google.com>, Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Pekka Enberg <penberg@cs.helsinki.fi> Reviewed-by: Paul Menage <menage@google.com> Cc: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-29 16:00:16 +08:00
put_task_struct(c);
goto retry;
}
cgroup_mm_owner_callbacks(mm->owner, c);
mm->owner = c;
task_unlock(c);
up_write(&mm->mmap_sem);
cgroups: add an owner to the mm_struct Remove the mem_cgroup member from mm_struct and instead adds an owner. This approach was suggested by Paul Menage. The advantage of this approach is that, once the mm->owner is known, using the subsystem id, the cgroup can be determined. It also allows several control groups that are virtually grouped by mm_struct, to exist independent of the memory controller i.e., without adding mem_cgroup's for each controller, to mm_struct. A new config option CONFIG_MM_OWNER is added and the memory resource controller selects this config option. This patch also adds cgroup callbacks to notify subsystems when mm->owner changes. The mm_cgroup_changed callback is called with the task_lock() of the new task held and is called just prior to changing the mm->owner. I am indebted to Paul Menage for the several reviews of this patchset and helping me make it lighter and simpler. This patch was tested on a powerpc box, it was compiled with both the MM_OWNER config turned on and off. After the thread group leader exits, it's moved to init_css_state by cgroup_exit(), thus all future charges from runnings threads would be redirected to the init_css_set's subsystem. Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelianov <xemul@openvz.org> Cc: Hugh Dickins <hugh@veritas.com> Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com> Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp> Cc: Hirokazu Takahashi <taka@valinux.co.jp> Cc: David Rientjes <rientjes@google.com>, Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Pekka Enberg <penberg@cs.helsinki.fi> Reviewed-by: Paul Menage <menage@google.com> Cc: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-29 16:00:16 +08:00
put_task_struct(c);
}
#endif /* CONFIG_MM_OWNER */
/*
* Turn us into a lazy TLB process if we
* aren't already..
*/
static void exit_mm(struct task_struct * tsk)
{
struct mm_struct *mm = tsk->mm;
struct core_state *core_state;
mm_release(tsk, mm);
if (!mm)
return;
/*
* Serialize with any possible pending coredump.
* We must hold mmap_sem around checking core_state
* and clearing tsk->mm. The core-inducing thread
* will increment ->nr_threads for each thread in the
* group with ->mm != NULL.
*/
down_read(&mm->mmap_sem);
core_state = mm->core_state;
if (core_state) {
struct core_thread self;
up_read(&mm->mmap_sem);
self.task = tsk;
self.next = xchg(&core_state->dumper.next, &self);
/*
* Implies mb(), the result of xchg() must be visible
* to core_state->dumper.
*/
if (atomic_dec_and_test(&core_state->nr_threads))
complete(&core_state->startup);
for (;;) {
set_task_state(tsk, TASK_UNINTERRUPTIBLE);
if (!self.task) /* see coredump_finish() */
break;
schedule();
}
__set_task_state(tsk, TASK_RUNNING);
down_read(&mm->mmap_sem);
}
atomic_inc(&mm->mm_count);
BUG_ON(mm != tsk->active_mm);
/* more a memory barrier than a real lock */
task_lock(tsk);
tsk->mm = NULL;
up_read(&mm->mmap_sem);
enter_lazy_tlb(mm, current);
Freezer: avoid freezing kernel threads prematurely Kernel threads should not have TIF_FREEZE set when user space processes are being frozen, since otherwise some of them might be frozen prematurely. To prevent this from happening we can (1) make exit_mm() unset TIF_FREEZE unconditionally just after clearing tsk->mm and (2) make try_to_freeze_tasks() check if p->mm is different from zero and PF_BORROWED_MM is unset in p->flags when user space processes are to be frozen. Namely, when user space processes are being frozen, we only should set TIF_FREEZE for tasks that have p->mm different from NULL and don't have PF_BORROWED_MM set in p->flags. For this reason task_lock() must be used to prevent try_to_freeze_tasks() from racing with use_mm()/unuse_mm(), in which p->mm and p->flags.PF_BORROWED_MM are changed under task_lock(p). Also, we need to prevent the following scenario from happening: * daemonize() is called by a task spawned from a user space code path * freezer checks if the task has p->mm set and the result is positive * task enters exit_mm() and clears its TIF_FREEZE * freezer sets TIF_FREEZE for the task * task calls try_to_freeze() and goes to the refrigerator, which is wrong at that point This requires us to acquire task_lock(p) before p->flags.PF_BORROWED_MM and p->mm are examined and release it after TIF_FREEZE is set for p (or it turns out that TIF_FREEZE should not be set). Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Cc: Gautham R Shenoy <ego@in.ibm.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: Nigel Cunningham <nigel@nigel.suspend2.net> Cc: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 16:47:33 +08:00
/* We don't want this task to be frozen prematurely */
clear_freeze_flag(tsk);
task_unlock(tsk);
cgroups: add an owner to the mm_struct Remove the mem_cgroup member from mm_struct and instead adds an owner. This approach was suggested by Paul Menage. The advantage of this approach is that, once the mm->owner is known, using the subsystem id, the cgroup can be determined. It also allows several control groups that are virtually grouped by mm_struct, to exist independent of the memory controller i.e., without adding mem_cgroup's for each controller, to mm_struct. A new config option CONFIG_MM_OWNER is added and the memory resource controller selects this config option. This patch also adds cgroup callbacks to notify subsystems when mm->owner changes. The mm_cgroup_changed callback is called with the task_lock() of the new task held and is called just prior to changing the mm->owner. I am indebted to Paul Menage for the several reviews of this patchset and helping me make it lighter and simpler. This patch was tested on a powerpc box, it was compiled with both the MM_OWNER config turned on and off. After the thread group leader exits, it's moved to init_css_state by cgroup_exit(), thus all future charges from runnings threads would be redirected to the init_css_set's subsystem. Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com> Cc: Pavel Emelianov <xemul@openvz.org> Cc: Hugh Dickins <hugh@veritas.com> Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com> Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp> Cc: Hirokazu Takahashi <taka@valinux.co.jp> Cc: David Rientjes <rientjes@google.com>, Cc: Balbir Singh <balbir@linux.vnet.ibm.com> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Acked-by: Pekka Enberg <penberg@cs.helsinki.fi> Reviewed-by: Paul Menage <menage@google.com> Cc: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-29 16:00:16 +08:00
mm_update_next_owner(mm);
mmput(mm);
}
/*
* Return nonzero if @parent's children should reap themselves.
*
* Called with write_lock_irq(&tasklist_lock) held.
*/
static int ignoring_children(struct task_struct *parent)
{
int ret;
struct sighand_struct *psig = parent->sighand;
unsigned long flags;
spin_lock_irqsave(&psig->siglock, flags);
ret = (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
(psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT));
spin_unlock_irqrestore(&psig->siglock, flags);
return ret;
}
/*
* Detach all tasks we were using ptrace on.
* Any that need to be release_task'd are put on the @dead list.
*
* Called with write_lock(&tasklist_lock) held.
*/
static void ptrace_exit(struct task_struct *parent, struct list_head *dead)
{
struct task_struct *p, *n;
int ign = -1;
list_for_each_entry_safe(p, n, &parent->ptraced, ptrace_entry) {
__ptrace_unlink(p);
if (p->exit_state != EXIT_ZOMBIE)
continue;
/*
* If it's a zombie, our attachedness prevented normal
* parent notification or self-reaping. Do notification
* now if it would have happened earlier. If it should
* reap itself, add it to the @dead list. We can't call
* release_task() here because we already hold tasklist_lock.
*
* If it's our own child, there is no notification to do.
* But if our normal children self-reap, then this child
* was prevented by ptrace and we must reap it now.
*/
if (!task_detached(p) && thread_group_empty(p)) {
if (!same_thread_group(p->real_parent, parent))
do_notify_parent(p, p->exit_signal);
else {
if (ign < 0)
ign = ignoring_children(parent);
if (ign)
p->exit_signal = -1;
}
}
if (task_detached(p)) {
/*
* Mark it as in the process of being reaped.
*/
p->exit_state = EXIT_DEAD;
list_add(&p->ptrace_entry, dead);
}
}
}
/*
* Finish up exit-time ptrace cleanup.
*
* Called without locks.
*/
static void ptrace_exit_finish(struct task_struct *parent,
struct list_head *dead)
{
struct task_struct *p, *n;
BUG_ON(!list_empty(&parent->ptraced));
list_for_each_entry_safe(p, n, dead, ptrace_entry) {
list_del_init(&p->ptrace_entry);
release_task(p);
}
}
static void reparent_thread(struct task_struct *p, struct task_struct *father)
{
if (p->pdeath_signal)
/* We already hold the tasklist_lock here. */
group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
list_move_tail(&p->sibling, &p->real_parent->children);
/* If this is a threaded reparent there is no need to
* notify anyone anything has happened.
*/
if (same_thread_group(p->real_parent, father))
return;
/* We don't want people slaying init. */
if (!task_detached(p))
p->exit_signal = SIGCHLD;
/* If we'd notified the old parent about this child's death,
* also notify the new parent.
*/
if (!ptrace_reparented(p) &&
p->exit_state == EXIT_ZOMBIE &&
!task_detached(p) && thread_group_empty(p))
do_notify_parent(p, p->exit_signal);
kill_orphaned_pgrp(p, father);
}
/*
* When we die, we re-parent all our children.
* Try to give them to another thread in our thread
* group, and if no such member exists, give it to
* the child reaper process (ie "init") in our pid
* space.
*/
static struct task_struct *find_new_reaper(struct task_struct *father)
{
struct pid_namespace *pid_ns = task_active_pid_ns(father);
struct task_struct *thread;
thread = father;
while_each_thread(father, thread) {
if (thread->flags & PF_EXITING)
continue;
if (unlikely(pid_ns->child_reaper == father))
pid_ns->child_reaper = thread;
return thread;
}
if (unlikely(pid_ns->child_reaper == father)) {
write_unlock_irq(&tasklist_lock);
if (unlikely(pid_ns == &init_pid_ns))
panic("Attempted to kill init!");
zap_pid_ns_processes(pid_ns);
write_lock_irq(&tasklist_lock);
/*
* We can not clear ->child_reaper or leave it alone.
* There may by stealth EXIT_DEAD tasks on ->children,
* forget_original_parent() must move them somewhere.
*/
pid_ns->child_reaper = init_pid_ns.child_reaper;
}
return pid_ns->child_reaper;
}
pid namespaces: rework forget_original_parent() A pid namespace is a "view" of a particular set of tasks on the system. They work in a similar way to filesystem namespaces. A file (or a process) can be accessed in multiple namespaces, but it may have a different name in each. In a filesystem, this name might be /etc/passwd in one namespace, but /chroot/etc/passwd in another. For processes, a process may have pid 1234 in one namespace, but be pid 1 in another. This allows new pid namespaces to have basically arbitrary pids, and not have to worry about what pids exist in other namespaces. This is essential for checkpoint/restart where a restarted process's pid might collide with an existing process on the system's pid. In this particular implementation, pid namespaces have a parent-child relationship, just like processes. A process in a pid namespace may see all of the processes in the same namespace, as well as all of the processes in all of the namespaces which are children of its namespace. Processes may not, however, see others which are in their parent's namespace, but not in their own. The same goes for sibling namespaces. The know issue to be solved in the nearest future is signal handling in the namespace boundary. That is, currently the namespace's init is treated like an ordinary task that can be killed from within an namespace. Ideally, the signal handling by the namespace's init should have two sides: when signaling the init from its namespace, the init should look like a real init task, i.e. receive only those signals, that is explicitly wants to; when signaling the init from one of the parent namespaces, init should look like an ordinary task, i.e. receive any signal, only taking the general permissions into account. The pid namespace was developed by Pavel Emlyanov and Sukadev Bhattiprolu and we eventually came to almost the same implementation, which differed in some details. This set is based on Pavel's patches, but it includes comments and patches that from Sukadev. Many thanks to Oleg, who reviewed the patches, pointed out many BUGs and made valuable advises on how to make this set cleaner. This patch: We have to call exit_task_namespaces() only after the exiting task has reparented all his children and is sure that no other threads will reparent theirs for it. Why this is needed is explained in appropriate patch. This one only reworks the forget_original_parent() so that after calling this a task cannot be/become parent of any other task. We check PF_EXITING instead of ->exit_state while choosing the new parent. Note that tasklits_lock acts as a barrier, everyone who takes tasklist after us (when forget_original_parent() drops it) must see PF_EXITING. The other changes are just cleanups. They just move some code from exit_notify to forget_original_parent(). It is a bit silly to declare ptrace_dead in exit_notify(), take tasklist, pass ptrace_dead to forget_original_parent(), unlock-lock-unlock tasklist, and then use ptrace_dead. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com> Cc: Paul Menage <menage@google.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-19 14:40:00 +08:00
static void forget_original_parent(struct task_struct *father)
{
struct task_struct *p, *n, *reaper;
LIST_HEAD(ptrace_dead);
pid namespaces: rework forget_original_parent() A pid namespace is a "view" of a particular set of tasks on the system. They work in a similar way to filesystem namespaces. A file (or a process) can be accessed in multiple namespaces, but it may have a different name in each. In a filesystem, this name might be /etc/passwd in one namespace, but /chroot/etc/passwd in another. For processes, a process may have pid 1234 in one namespace, but be pid 1 in another. This allows new pid namespaces to have basically arbitrary pids, and not have to worry about what pids exist in other namespaces. This is essential for checkpoint/restart where a restarted process's pid might collide with an existing process on the system's pid. In this particular implementation, pid namespaces have a parent-child relationship, just like processes. A process in a pid namespace may see all of the processes in the same namespace, as well as all of the processes in all of the namespaces which are children of its namespace. Processes may not, however, see others which are in their parent's namespace, but not in their own. The same goes for sibling namespaces. The know issue to be solved in the nearest future is signal handling in the namespace boundary. That is, currently the namespace's init is treated like an ordinary task that can be killed from within an namespace. Ideally, the signal handling by the namespace's init should have two sides: when signaling the init from its namespace, the init should look like a real init task, i.e. receive only those signals, that is explicitly wants to; when signaling the init from one of the parent namespaces, init should look like an ordinary task, i.e. receive any signal, only taking the general permissions into account. The pid namespace was developed by Pavel Emlyanov and Sukadev Bhattiprolu and we eventually came to almost the same implementation, which differed in some details. This set is based on Pavel's patches, but it includes comments and patches that from Sukadev. Many thanks to Oleg, who reviewed the patches, pointed out many BUGs and made valuable advises on how to make this set cleaner. This patch: We have to call exit_task_namespaces() only after the exiting task has reparented all his children and is sure that no other threads will reparent theirs for it. Why this is needed is explained in appropriate patch. This one only reworks the forget_original_parent() so that after calling this a task cannot be/become parent of any other task. We check PF_EXITING instead of ->exit_state while choosing the new parent. Note that tasklits_lock acts as a barrier, everyone who takes tasklist after us (when forget_original_parent() drops it) must see PF_EXITING. The other changes are just cleanups. They just move some code from exit_notify to forget_original_parent(). It is a bit silly to declare ptrace_dead in exit_notify(), take tasklist, pass ptrace_dead to forget_original_parent(), unlock-lock-unlock tasklist, and then use ptrace_dead. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com> Cc: Paul Menage <menage@google.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-19 14:40:00 +08:00
write_lock_irq(&tasklist_lock);
reaper = find_new_reaper(father);
/*
* First clean up ptrace if we were using it.
*/
ptrace_exit(father, &ptrace_dead);
list_for_each_entry_safe(p, n, &father->children, sibling) {
p->real_parent = reaper;
if (p->parent == father) {
BUG_ON(p->ptrace);
p->parent = p->real_parent;
}
reparent_thread(p, father);
}
pid namespaces: rework forget_original_parent() A pid namespace is a "view" of a particular set of tasks on the system. They work in a similar way to filesystem namespaces. A file (or a process) can be accessed in multiple namespaces, but it may have a different name in each. In a filesystem, this name might be /etc/passwd in one namespace, but /chroot/etc/passwd in another. For processes, a process may have pid 1234 in one namespace, but be pid 1 in another. This allows new pid namespaces to have basically arbitrary pids, and not have to worry about what pids exist in other namespaces. This is essential for checkpoint/restart where a restarted process's pid might collide with an existing process on the system's pid. In this particular implementation, pid namespaces have a parent-child relationship, just like processes. A process in a pid namespace may see all of the processes in the same namespace, as well as all of the processes in all of the namespaces which are children of its namespace. Processes may not, however, see others which are in their parent's namespace, but not in their own. The same goes for sibling namespaces. The know issue to be solved in the nearest future is signal handling in the namespace boundary. That is, currently the namespace's init is treated like an ordinary task that can be killed from within an namespace. Ideally, the signal handling by the namespace's init should have two sides: when signaling the init from its namespace, the init should look like a real init task, i.e. receive only those signals, that is explicitly wants to; when signaling the init from one of the parent namespaces, init should look like an ordinary task, i.e. receive any signal, only taking the general permissions into account. The pid namespace was developed by Pavel Emlyanov and Sukadev Bhattiprolu and we eventually came to almost the same implementation, which differed in some details. This set is based on Pavel's patches, but it includes comments and patches that from Sukadev. Many thanks to Oleg, who reviewed the patches, pointed out many BUGs and made valuable advises on how to make this set cleaner. This patch: We have to call exit_task_namespaces() only after the exiting task has reparented all his children and is sure that no other threads will reparent theirs for it. Why this is needed is explained in appropriate patch. This one only reworks the forget_original_parent() so that after calling this a task cannot be/become parent of any other task. We check PF_EXITING instead of ->exit_state while choosing the new parent. Note that tasklits_lock acts as a barrier, everyone who takes tasklist after us (when forget_original_parent() drops it) must see PF_EXITING. The other changes are just cleanups. They just move some code from exit_notify to forget_original_parent(). It is a bit silly to declare ptrace_dead in exit_notify(), take tasklist, pass ptrace_dead to forget_original_parent(), unlock-lock-unlock tasklist, and then use ptrace_dead. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com> Cc: Paul Menage <menage@google.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-19 14:40:00 +08:00
write_unlock_irq(&tasklist_lock);
BUG_ON(!list_empty(&father->children));
ptrace_exit_finish(father, &ptrace_dead);
}
/*
* Send signals to all our closest relatives so that they know
* to properly mourn us..
*/
static void exit_notify(struct task_struct *tsk, int group_dead)
{
int signal;
void *cookie;
/*
* This does two things:
*
* A. Make init inherit all the child processes
* B. Check to see if any process groups have become orphaned
* as a result of our exiting, and if they have any stopped
* jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
*/
pid namespaces: rework forget_original_parent() A pid namespace is a "view" of a particular set of tasks on the system. They work in a similar way to filesystem namespaces. A file (or a process) can be accessed in multiple namespaces, but it may have a different name in each. In a filesystem, this name might be /etc/passwd in one namespace, but /chroot/etc/passwd in another. For processes, a process may have pid 1234 in one namespace, but be pid 1 in another. This allows new pid namespaces to have basically arbitrary pids, and not have to worry about what pids exist in other namespaces. This is essential for checkpoint/restart where a restarted process's pid might collide with an existing process on the system's pid. In this particular implementation, pid namespaces have a parent-child relationship, just like processes. A process in a pid namespace may see all of the processes in the same namespace, as well as all of the processes in all of the namespaces which are children of its namespace. Processes may not, however, see others which are in their parent's namespace, but not in their own. The same goes for sibling namespaces. The know issue to be solved in the nearest future is signal handling in the namespace boundary. That is, currently the namespace's init is treated like an ordinary task that can be killed from within an namespace. Ideally, the signal handling by the namespace's init should have two sides: when signaling the init from its namespace, the init should look like a real init task, i.e. receive only those signals, that is explicitly wants to; when signaling the init from one of the parent namespaces, init should look like an ordinary task, i.e. receive any signal, only taking the general permissions into account. The pid namespace was developed by Pavel Emlyanov and Sukadev Bhattiprolu and we eventually came to almost the same implementation, which differed in some details. This set is based on Pavel's patches, but it includes comments and patches that from Sukadev. Many thanks to Oleg, who reviewed the patches, pointed out many BUGs and made valuable advises on how to make this set cleaner. This patch: We have to call exit_task_namespaces() only after the exiting task has reparented all his children and is sure that no other threads will reparent theirs for it. Why this is needed is explained in appropriate patch. This one only reworks the forget_original_parent() so that after calling this a task cannot be/become parent of any other task. We check PF_EXITING instead of ->exit_state while choosing the new parent. Note that tasklits_lock acts as a barrier, everyone who takes tasklist after us (when forget_original_parent() drops it) must see PF_EXITING. The other changes are just cleanups. They just move some code from exit_notify to forget_original_parent(). It is a bit silly to declare ptrace_dead in exit_notify(), take tasklist, pass ptrace_dead to forget_original_parent(), unlock-lock-unlock tasklist, and then use ptrace_dead. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com> Cc: Paul Menage <menage@google.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-19 14:40:00 +08:00
forget_original_parent(tsk);
exit_task_namespaces(tsk);
/*
* Flush inherited counters to the parent - before the parent
* gets woken up by child-exit notifications.
*/
perf_counter_exit_task(tsk);
pid namespaces: rework forget_original_parent() A pid namespace is a "view" of a particular set of tasks on the system. They work in a similar way to filesystem namespaces. A file (or a process) can be accessed in multiple namespaces, but it may have a different name in each. In a filesystem, this name might be /etc/passwd in one namespace, but /chroot/etc/passwd in another. For processes, a process may have pid 1234 in one namespace, but be pid 1 in another. This allows new pid namespaces to have basically arbitrary pids, and not have to worry about what pids exist in other namespaces. This is essential for checkpoint/restart where a restarted process's pid might collide with an existing process on the system's pid. In this particular implementation, pid namespaces have a parent-child relationship, just like processes. A process in a pid namespace may see all of the processes in the same namespace, as well as all of the processes in all of the namespaces which are children of its namespace. Processes may not, however, see others which are in their parent's namespace, but not in their own. The same goes for sibling namespaces. The know issue to be solved in the nearest future is signal handling in the namespace boundary. That is, currently the namespace's init is treated like an ordinary task that can be killed from within an namespace. Ideally, the signal handling by the namespace's init should have two sides: when signaling the init from its namespace, the init should look like a real init task, i.e. receive only those signals, that is explicitly wants to; when signaling the init from one of the parent namespaces, init should look like an ordinary task, i.e. receive any signal, only taking the general permissions into account. The pid namespace was developed by Pavel Emlyanov and Sukadev Bhattiprolu and we eventually came to almost the same implementation, which differed in some details. This set is based on Pavel's patches, but it includes comments and patches that from Sukadev. Many thanks to Oleg, who reviewed the patches, pointed out many BUGs and made valuable advises on how to make this set cleaner. This patch: We have to call exit_task_namespaces() only after the exiting task has reparented all his children and is sure that no other threads will reparent theirs for it. Why this is needed is explained in appropriate patch. This one only reworks the forget_original_parent() so that after calling this a task cannot be/become parent of any other task. We check PF_EXITING instead of ->exit_state while choosing the new parent. Note that tasklits_lock acts as a barrier, everyone who takes tasklist after us (when forget_original_parent() drops it) must see PF_EXITING. The other changes are just cleanups. They just move some code from exit_notify to forget_original_parent(). It is a bit silly to declare ptrace_dead in exit_notify(), take tasklist, pass ptrace_dead to forget_original_parent(), unlock-lock-unlock tasklist, and then use ptrace_dead. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com> Cc: Paul Menage <menage@google.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-19 14:40:00 +08:00
write_lock_irq(&tasklist_lock);
if (group_dead)
kill_orphaned_pgrp(tsk->group_leader, NULL);
/* Let father know we died
*
* Thread signals are configurable, but you aren't going to use
* that to send signals to arbitary processes.
* That stops right now.
*
* If the parent exec id doesn't match the exec id we saved
* when we started then we know the parent has changed security
* domain.
*
* If our self_exec id doesn't match our parent_exec_id then
* we have changed execution domain as these two values started
* the same after a fork.
*/
if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) &&
(tsk->parent_exec_id != tsk->real_parent->self_exec_id ||
tsk->self_exec_id != tsk->parent_exec_id) &&
!capable(CAP_KILL))
tsk->exit_signal = SIGCHLD;
signal = tracehook_notify_death(tsk, &cookie, group_dead);
if (signal >= 0)
signal = do_notify_parent(tsk, signal);
tsk->exit_state = signal == DEATH_REAP ? EXIT_DEAD : EXIT_ZOMBIE;
/* mt-exec, de_thread() is waiting for us */
if (thread_group_leader(tsk) &&
tsk->signal->group_exit_task &&
tsk->signal->notify_count < 0)
wake_up_process(tsk->signal->group_exit_task);
write_unlock_irq(&tasklist_lock);
tracehook_report_death(tsk, signal, cookie, group_dead);
/* If the process is dead, release it - nobody will wait for it */
if (signal == DEATH_REAP)
release_task(tsk);
}
#ifdef CONFIG_DEBUG_STACK_USAGE
static void check_stack_usage(void)
{
static DEFINE_SPINLOCK(low_water_lock);
static int lowest_to_date = THREAD_SIZE;
unsigned long *n = end_of_stack(current);
unsigned long free;
while (*n == 0)
n++;
free = (unsigned long)n - (unsigned long)end_of_stack(current);
if (free >= lowest_to_date)
return;
spin_lock(&low_water_lock);
if (free < lowest_to_date) {
printk(KERN_WARNING "%s used greatest stack depth: %lu bytes "
"left\n",
current->comm, free);
lowest_to_date = free;
}
spin_unlock(&low_water_lock);
}
#else
static inline void check_stack_usage(void) {}
#endif
NORET_TYPE void do_exit(long code)
{
struct task_struct *tsk = current;
int group_dead;
profile_task_exit(tsk);
WARN_ON(atomic_read(&tsk->fs_excl));
if (unlikely(in_interrupt()))
panic("Aiee, killing interrupt handler!");
if (unlikely(!tsk->pid))
panic("Attempted to kill the idle task!");
tracehook_report_exit(&code);
/*
* We're taking recursive faults here in do_exit. Safest is to just
* leave this task alone and wait for reboot.
*/
if (unlikely(tsk->flags & PF_EXITING)) {
printk(KERN_ALERT
"Fixing recursive fault but reboot is needed!\n");
pi-futex: fix exit races and locking problems 1. New entries can be added to tsk->pi_state_list after task completed exit_pi_state_list(). The result is memory leakage and deadlocks. 2. handle_mm_fault() is called under spinlock. The result is obvious. 3. results in self-inflicted deadlock inside glibc. Sometimes futex_lock_pi returns -ESRCH, when it is not expected and glibc enters to for(;;) sleep() to simulate deadlock. This problem is quite obvious and I think the patch is right. Though it looks like each "if" in futex_lock_pi() got some stupid special case "else if". :-) 4. sometimes futex_lock_pi() returns -EDEADLK, when nobody has the lock. The reason is also obvious (see comment in the patch), but correct fix is far beyond my comprehension. I guess someone already saw this, the chunk: if (rt_mutex_trylock(&q.pi_state->pi_mutex)) ret = 0; is obviously from the same opera. But it does not work, because the rtmutex is really taken at this point: wake_futex_pi() of previous owner reassigned it to us. My fix works. But it looks very stupid. I would think about removal of shift of ownership in wake_futex_pi() and making all the work in context of process taking lock. From: Thomas Gleixner <tglx@linutronix.de> Fix 1) Avoid the tasklist lock variant of the exit race fix by adding an additional state transition to the exit code. This fixes also the issue, when a task with recursive segfaults is not able to release the futexes. Fix 2) Cleanup the lookup_pi_state() failure path and solve the -ESRCH problem finally. Fix 3) Solve the fixup_pi_state_owner() problem which needs to do the fixup in the lock protected section by using the in_atomic userspace access functions. This removes also the ugly lock drop / unqueue inside of fixup_pi_state() Fix 4) Fix a stale lock in the error path of futex_wake_pi() Added some error checks for verification. The -EDEADLK problem is solved by the rtmutex fixups. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Ingo Molnar <mingo@elte.hu> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Ulrich Drepper <drepper@redhat.com> Cc: Eric Dumazet <dada1@cosmosbay.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-06-09 04:47:00 +08:00
/*
* We can do this unlocked here. The futex code uses
* this flag just to verify whether the pi state
* cleanup has been done or not. In the worst case it
* loops once more. We pretend that the cleanup was
* done as there is no way to return. Either the
* OWNER_DIED bit is set by now or we push the blocked
* task into the wait for ever nirwana as well.
*/
tsk->flags |= PF_EXITPIDONE;
if (tsk->io_context)
exit_io_context();
set_current_state(TASK_UNINTERRUPTIBLE);
schedule();
}
exit_signals(tsk); /* sets PF_EXITING */
pi-futex: fix exit races and locking problems 1. New entries can be added to tsk->pi_state_list after task completed exit_pi_state_list(). The result is memory leakage and deadlocks. 2. handle_mm_fault() is called under spinlock. The result is obvious. 3. results in self-inflicted deadlock inside glibc. Sometimes futex_lock_pi returns -ESRCH, when it is not expected and glibc enters to for(;;) sleep() to simulate deadlock. This problem is quite obvious and I think the patch is right. Though it looks like each "if" in futex_lock_pi() got some stupid special case "else if". :-) 4. sometimes futex_lock_pi() returns -EDEADLK, when nobody has the lock. The reason is also obvious (see comment in the patch), but correct fix is far beyond my comprehension. I guess someone already saw this, the chunk: if (rt_mutex_trylock(&q.pi_state->pi_mutex)) ret = 0; is obviously from the same opera. But it does not work, because the rtmutex is really taken at this point: wake_futex_pi() of previous owner reassigned it to us. My fix works. But it looks very stupid. I would think about removal of shift of ownership in wake_futex_pi() and making all the work in context of process taking lock. From: Thomas Gleixner <tglx@linutronix.de> Fix 1) Avoid the tasklist lock variant of the exit race fix by adding an additional state transition to the exit code. This fixes also the issue, when a task with recursive segfaults is not able to release the futexes. Fix 2) Cleanup the lookup_pi_state() failure path and solve the -ESRCH problem finally. Fix 3) Solve the fixup_pi_state_owner() problem which needs to do the fixup in the lock protected section by using the in_atomic userspace access functions. This removes also the ugly lock drop / unqueue inside of fixup_pi_state() Fix 4) Fix a stale lock in the error path of futex_wake_pi() Added some error checks for verification. The -EDEADLK problem is solved by the rtmutex fixups. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Ingo Molnar <mingo@elte.hu> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Ulrich Drepper <drepper@redhat.com> Cc: Eric Dumazet <dada1@cosmosbay.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-06-09 04:47:00 +08:00
/*
* tsk->flags are checked in the futex code to protect against
* an exiting task cleaning up the robust pi futexes.
*/
smp_mb();
spin_unlock_wait(&tsk->pi_lock);
if (unlikely(in_atomic()))
printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
current->comm, task_pid_nr(current),
preempt_count());
acct_update_integrals(tsk);
[PATCH] mm: update_hiwaters just in time update_mem_hiwater has attracted various criticisms, in particular from those concerned with mm scalability. Originally it was called whenever rss or total_vm got raised. Then many of those callsites were replaced by a timer tick call from account_system_time. Now Frank van Maarseveen reports that to be found inadequate. How about this? Works for Frank. Replace update_mem_hiwater, a poor combination of two unrelated ops, by macros update_hiwater_rss and update_hiwater_vm. Don't attempt to keep mm->hiwater_rss up to date at timer tick, nor every time we raise rss (usually by 1): those are hot paths. Do the opposite, update only when about to lower rss (usually by many), or just before final accounting in do_exit. Handle mm->hiwater_vm in the same way, though it's much less of an issue. Demand that whoever collects these hiwater statistics do the work of taking the maximum with rss or total_vm. And there has been no collector of these hiwater statistics in the tree. The new convention needs an example, so match Frank's usage by adding a VmPeak line above VmSize to /proc/<pid>/status, and also a VmHWM line above VmRSS (High-Water-Mark or High-Water-Memory). There was a particular anomaly during mremap move, that hiwater_vm might be captured too high. A fleeting such anomaly remains, but it's quickly corrected now, whereas before it would stick. What locking? None: if the app is racy then these statistics will be racy, it's not worth any overhead to make them exact. But whenever it suits, hiwater_vm is updated under exclusive mmap_sem, and hiwater_rss under page_table_lock (for now) or with preemption disabled (later on): without going to any trouble, minimize the time between reading current values and updating, to minimize those occasions when a racing thread bumps a count up and back down in between. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 09:16:18 +08:00
if (tsk->mm) {
update_hiwater_rss(tsk->mm);
update_hiwater_vm(tsk->mm);
}
group_dead = atomic_dec_and_test(&tsk->signal->live);
if (group_dead) {
pi-futex: fix exit races and locking problems 1. New entries can be added to tsk->pi_state_list after task completed exit_pi_state_list(). The result is memory leakage and deadlocks. 2. handle_mm_fault() is called under spinlock. The result is obvious. 3. results in self-inflicted deadlock inside glibc. Sometimes futex_lock_pi returns -ESRCH, when it is not expected and glibc enters to for(;;) sleep() to simulate deadlock. This problem is quite obvious and I think the patch is right. Though it looks like each "if" in futex_lock_pi() got some stupid special case "else if". :-) 4. sometimes futex_lock_pi() returns -EDEADLK, when nobody has the lock. The reason is also obvious (see comment in the patch), but correct fix is far beyond my comprehension. I guess someone already saw this, the chunk: if (rt_mutex_trylock(&q.pi_state->pi_mutex)) ret = 0; is obviously from the same opera. But it does not work, because the rtmutex is really taken at this point: wake_futex_pi() of previous owner reassigned it to us. My fix works. But it looks very stupid. I would think about removal of shift of ownership in wake_futex_pi() and making all the work in context of process taking lock. From: Thomas Gleixner <tglx@linutronix.de> Fix 1) Avoid the tasklist lock variant of the exit race fix by adding an additional state transition to the exit code. This fixes also the issue, when a task with recursive segfaults is not able to release the futexes. Fix 2) Cleanup the lookup_pi_state() failure path and solve the -ESRCH problem finally. Fix 3) Solve the fixup_pi_state_owner() problem which needs to do the fixup in the lock protected section by using the in_atomic userspace access functions. This removes also the ugly lock drop / unqueue inside of fixup_pi_state() Fix 4) Fix a stale lock in the error path of futex_wake_pi() Added some error checks for verification. The -EDEADLK problem is solved by the rtmutex fixups. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Ingo Molnar <mingo@elte.hu> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Ulrich Drepper <drepper@redhat.com> Cc: Eric Dumazet <dada1@cosmosbay.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-06-09 04:47:00 +08:00
hrtimer_cancel(&tsk->signal->real_timer);
exit_itimers(tsk->signal);
}
acct_collect(code, group_dead);
Audit: add TTY input auditing Add TTY input auditing, used to audit system administrator's actions. This is required by various security standards such as DCID 6/3 and PCI to provide non-repudiation of administrator's actions and to allow a review of past actions if the administrator seems to overstep their duties or if the system becomes misconfigured for unknown reasons. These requirements do not make it necessary to audit TTY output as well. Compared to an user-space keylogger, this approach records TTY input using the audit subsystem, correlated with other audit events, and it is completely transparent to the user-space application (e.g. the console ioctls still work). TTY input auditing works on a higher level than auditing all system calls within the session, which would produce an overwhelming amount of mostly useless audit events. Add an "audit_tty" attribute, inherited across fork (). Data read from TTYs by process with the attribute is sent to the audit subsystem by the kernel. The audit netlink interface is extended to allow modifying the audit_tty attribute, and to allow sending explanatory audit events from user-space (for example, a shell might send an event containing the final command, after the interactive command-line editing and history expansion is performed, which might be difficult to decipher from the TTY input alone). Because the "audit_tty" attribute is inherited across fork (), it would be set e.g. for sshd restarted within an audited session. To prevent this, the audit_tty attribute is cleared when a process with no open TTY file descriptors (e.g. after daemon startup) opens a TTY. See https://www.redhat.com/archives/linux-audit/2007-June/msg00000.html for a more detailed rationale document for an older version of this patch. [akpm@linux-foundation.org: build fix] Signed-off-by: Miloslav Trmac <mitr@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk> Cc: Paul Fulghum <paulkf@microgate.com> Cc: Casey Schaufler <casey@schaufler-ca.com> Cc: Steve Grubb <sgrubb@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-16 14:40:56 +08:00
if (group_dead)
tty_audit_exit();
if (unlikely(tsk->audit_context))
audit_free(tsk);
tsk->exit_code = code;
taskstats_exit(tsk, group_dead);
exit_mm(tsk);
if (group_dead)
acct_process();
trace_sched_process_exit(tsk);
exit_sem(tsk);
exit_files(tsk);
exit_fs(tsk);
check_stack_usage();
exit_thread();
cgroup_exit(tsk, 1);
exit_keys(tsk);
if (group_dead && tsk->signal->leader)
disassociate_ctty(1);
module_put(task_thread_info(tsk)->exec_domain->module);
if (tsk->binfmt)
module_put(tsk->binfmt->module);
proc_exit_connector(tsk);
exit_notify(tsk, group_dead);
#ifdef CONFIG_NUMA
mpol_put(tsk->mempolicy);
tsk->mempolicy = NULL;
#endif
#ifdef CONFIG_FUTEX
if (unlikely(!list_empty(&tsk->pi_state_list)))
exit_pi_state_list(tsk);
if (unlikely(current->pi_state_cache))
kfree(current->pi_state_cache);
#endif
/*
* Make sure we are holding no locks:
*/
debug_check_no_locks_held(tsk);
pi-futex: fix exit races and locking problems 1. New entries can be added to tsk->pi_state_list after task completed exit_pi_state_list(). The result is memory leakage and deadlocks. 2. handle_mm_fault() is called under spinlock. The result is obvious. 3. results in self-inflicted deadlock inside glibc. Sometimes futex_lock_pi returns -ESRCH, when it is not expected and glibc enters to for(;;) sleep() to simulate deadlock. This problem is quite obvious and I think the patch is right. Though it looks like each "if" in futex_lock_pi() got some stupid special case "else if". :-) 4. sometimes futex_lock_pi() returns -EDEADLK, when nobody has the lock. The reason is also obvious (see comment in the patch), but correct fix is far beyond my comprehension. I guess someone already saw this, the chunk: if (rt_mutex_trylock(&q.pi_state->pi_mutex)) ret = 0; is obviously from the same opera. But it does not work, because the rtmutex is really taken at this point: wake_futex_pi() of previous owner reassigned it to us. My fix works. But it looks very stupid. I would think about removal of shift of ownership in wake_futex_pi() and making all the work in context of process taking lock. From: Thomas Gleixner <tglx@linutronix.de> Fix 1) Avoid the tasklist lock variant of the exit race fix by adding an additional state transition to the exit code. This fixes also the issue, when a task with recursive segfaults is not able to release the futexes. Fix 2) Cleanup the lookup_pi_state() failure path and solve the -ESRCH problem finally. Fix 3) Solve the fixup_pi_state_owner() problem which needs to do the fixup in the lock protected section by using the in_atomic userspace access functions. This removes also the ugly lock drop / unqueue inside of fixup_pi_state() Fix 4) Fix a stale lock in the error path of futex_wake_pi() Added some error checks for verification. The -EDEADLK problem is solved by the rtmutex fixups. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Ingo Molnar <mingo@elte.hu> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Ulrich Drepper <drepper@redhat.com> Cc: Eric Dumazet <dada1@cosmosbay.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-06-09 04:47:00 +08:00
/*
* We can do this unlocked here. The futex code uses this flag
* just to verify whether the pi state cleanup has been done
* or not. In the worst case it loops once more.
*/
tsk->flags |= PF_EXITPIDONE;
if (tsk->io_context)
exit_io_context();
if (tsk->splice_pipe)
__free_pipe_info(tsk->splice_pipe);
/*
* These must happen late, after the PID is not
* hashed anymore, but still at a point that may sleep:
*/
perf_counter_exit_task(tsk);
preempt_disable();
/* causes final put_task_struct in finish_task_switch(). */
tsk->state = TASK_DEAD;
schedule();
BUG();
/* Avoid "noreturn function does return". */
for (;;)
cpu_relax(); /* For when BUG is null */
}
EXPORT_SYMBOL_GPL(do_exit);
NORET_TYPE void complete_and_exit(struct completion *comp, long code)
{
if (comp)
complete(comp);
do_exit(code);
}
EXPORT_SYMBOL(complete_and_exit);
asmlinkage long sys_exit(int error_code)
{
do_exit((error_code&0xff)<<8);
}
/*
* Take down every thread in the group. This is called by fatal signals
* as well as by sys_exit_group (below).
*/
NORET_TYPE void
do_group_exit(int exit_code)
{
struct signal_struct *sig = current->signal;
BUG_ON(exit_code & 0x80); /* core dumps don't get here */
if (signal_group_exit(sig))
exit_code = sig->group_exit_code;
else if (!thread_group_empty(current)) {
struct sighand_struct *const sighand = current->sighand;
spin_lock_irq(&sighand->siglock);
if (signal_group_exit(sig))
/* Another thread got here before we took the lock. */
exit_code = sig->group_exit_code;
else {
sig->group_exit_code = exit_code;
sig->flags = SIGNAL_GROUP_EXIT;
zap_other_threads(current);
}
spin_unlock_irq(&sighand->siglock);
}
do_exit(exit_code);
/* NOTREACHED */
}
/*
* this kills every thread in the thread group. Note that any externally
* wait4()-ing process will get the correct exit code - even if this
* thread is not the thread group leader.
*/
asmlinkage void sys_exit_group(int error_code)
{
do_group_exit((error_code & 0xff) << 8);
}
static struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
{
struct pid *pid = NULL;
if (type == PIDTYPE_PID)
pid = task->pids[type].pid;
else if (type < PIDTYPE_MAX)
pid = task->group_leader->pids[type].pid;
return pid;
}
static int eligible_child(enum pid_type type, struct pid *pid, int options,
struct task_struct *p)
{
int err;
if (type < PIDTYPE_MAX) {
if (task_pid_type(p, type) != pid)
return 0;
}
/* Wait for all children (clone and not) if __WALL is set;
* otherwise, wait for clone children *only* if __WCLONE is
* set; otherwise, wait for non-clone children *only*. (Note:
* A "clone" child here is one that reports to its parent
* using a signal other than SIGCHLD.) */
if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
&& !(options & __WALL))
return 0;
err = security_task_wait(p);
if (err)
return err;
return 1;
}
static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
int why, int status,
struct siginfo __user *infop,
struct rusage __user *rusagep)
{
int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
put_task_struct(p);
if (!retval)
retval = put_user(SIGCHLD, &infop->si_signo);
if (!retval)
retval = put_user(0, &infop->si_errno);
if (!retval)
retval = put_user((short)why, &infop->si_code);
if (!retval)
retval = put_user(pid, &infop->si_pid);
if (!retval)
retval = put_user(uid, &infop->si_uid);
if (!retval)
retval = put_user(status, &infop->si_status);
if (!retval)
retval = pid;
return retval;
}
/*
* Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
* read_lock(&tasklist_lock) on entry. If we return zero, we still hold
* the lock and this task is uninteresting. If we return nonzero, we have
* released the lock and the system call should return.
*/
static int wait_task_zombie(struct task_struct *p, int options,
struct siginfo __user *infop,
int __user *stat_addr, struct rusage __user *ru)
{
unsigned long state;
wait_task_zombie: fix 2/3 races vs forget_original_parent() Two threads, T1 and T2. T2 ptraces P, and P is not a child of ptracer's thread group. P exits and goes to TASK_ZOMBIE. T1 does wait_task_zombie(P): P->exit_state = TASK_DEAD; ... read_unlock(&tasklist_lock); T2 does exit(), takes tasklist, forget_original_parent() does __ptrace_unlink(P) but doesn't call do_notify_parent(P) because p->exit_state == EXIT_DEAD. Now, P is not visible to our process: __ptrace_unlink() removed it from ->children. We should send notification to P->parent and release P if and only if SIGCHLD is ignored. And we have 3 bugs: 1. P->parent does do_wait() and gets -ECHILD (P is on ->parent->children, but its state is TASK_DEAD). 2. // wait_task_zombie() continues if (put_user(...)) { // TODO: is this safe? p->exit_state = EXIT_ZOMBIE; return; } we return without notification/release, task_struct leaked. Solution: ignore -EFAULT and proceed. It is an application's bug if we can't fill infop/stat_addr (in case of VM_FAULT_OOM we have much more problems). 3. // wait_task_zombie() continues if (p->real_parent != p->parent) { // Not taken, it was untraced'ed ... } release_task(p); we released the task which we shouldn't. Solution: check ->real_parent != ->parent before, under tasklist_lock, but use ptrace_unlink() instead of __ptrace_unlink() to check ->ptrace. This patch hopefully solves 2 and 3, the 1st bug will be fixed later, we need some cleanups in forget_original_parent/reparent_thread. However, the first race is very unlikely and not critical, so I hope it makes sense to fix 1 and 2 for now. 4. Small cleanup: don't "restore" EXIT_ZOMBIE unless we know we are not going to realease the child. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-17 14:26:58 +08:00
int retval, status, traced;
pid_t pid = task_pid_vnr(p);
if (!likely(options & WEXITED))
return 0;
if (unlikely(options & WNOWAIT)) {
uid_t uid = p->uid;
int exit_code = p->exit_code;
int why, status;
get_task_struct(p);
read_unlock(&tasklist_lock);
if ((exit_code & 0x7f) == 0) {
why = CLD_EXITED;
status = exit_code >> 8;
} else {
why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
status = exit_code & 0x7f;
}
return wait_noreap_copyout(p, pid, uid, why,
status, infop, ru);
}
/*
* Try to move the task's state to DEAD
* only one thread is allowed to do this:
*/
state = xchg(&p->exit_state, EXIT_DEAD);
if (state != EXIT_ZOMBIE) {
BUG_ON(state != EXIT_DEAD);
return 0;
}
traced = ptrace_reparented(p);
wait_task_zombie: fix 2/3 races vs forget_original_parent() Two threads, T1 and T2. T2 ptraces P, and P is not a child of ptracer's thread group. P exits and goes to TASK_ZOMBIE. T1 does wait_task_zombie(P): P->exit_state = TASK_DEAD; ... read_unlock(&tasklist_lock); T2 does exit(), takes tasklist, forget_original_parent() does __ptrace_unlink(P) but doesn't call do_notify_parent(P) because p->exit_state == EXIT_DEAD. Now, P is not visible to our process: __ptrace_unlink() removed it from ->children. We should send notification to P->parent and release P if and only if SIGCHLD is ignored. And we have 3 bugs: 1. P->parent does do_wait() and gets -ECHILD (P is on ->parent->children, but its state is TASK_DEAD). 2. // wait_task_zombie() continues if (put_user(...)) { // TODO: is this safe? p->exit_state = EXIT_ZOMBIE; return; } we return without notification/release, task_struct leaked. Solution: ignore -EFAULT and proceed. It is an application's bug if we can't fill infop/stat_addr (in case of VM_FAULT_OOM we have much more problems). 3. // wait_task_zombie() continues if (p->real_parent != p->parent) { // Not taken, it was untraced'ed ... } release_task(p); we released the task which we shouldn't. Solution: check ->real_parent != ->parent before, under tasklist_lock, but use ptrace_unlink() instead of __ptrace_unlink() to check ->ptrace. This patch hopefully solves 2 and 3, the 1st bug will be fixed later, we need some cleanups in forget_original_parent/reparent_thread. However, the first race is very unlikely and not critical, so I hope it makes sense to fix 1 and 2 for now. 4. Small cleanup: don't "restore" EXIT_ZOMBIE unless we know we are not going to realease the child. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-17 14:26:58 +08:00
if (likely(!traced)) {
struct signal_struct *psig;
struct signal_struct *sig;
timers: fix itimer/many thread hang Overview This patch reworks the handling of POSIX CPU timers, including the ITIMER_PROF, ITIMER_VIRT timers and rlimit handling. It was put together with the help of Roland McGrath, the owner and original writer of this code. The problem we ran into, and the reason for this rework, has to do with using a profiling timer in a process with a large number of threads. It appears that the performance of the old implementation of run_posix_cpu_timers() was at least O(n*3) (where "n" is the number of threads in a process) or worse. Everything is fine with an increasing number of threads until the time taken for that routine to run becomes the same as or greater than the tick time, at which point things degrade rather quickly. This patch fixes bug 9906, "Weird hang with NPTL and SIGPROF." Code Changes This rework corrects the implementation of run_posix_cpu_timers() to make it run in constant time for a particular machine. (Performance may vary between one machine and another depending upon whether the kernel is built as single- or multiprocessor and, in the latter case, depending upon the number of running processors.) To do this, at each tick we now update fields in signal_struct as well as task_struct. The run_posix_cpu_timers() function uses those fields to make its decisions. We define a new structure, "task_cputime," to contain user, system and scheduler times and use these in appropriate places: struct task_cputime { cputime_t utime; cputime_t stime; unsigned long long sum_exec_runtime; }; This is included in the structure "thread_group_cputime," which is a new substructure of signal_struct and which varies for uniprocessor versus multiprocessor kernels. For uniprocessor kernels, it uses "task_cputime" as a simple substructure, while for multiprocessor kernels it is a pointer: struct thread_group_cputime { struct task_cputime totals; }; struct thread_group_cputime { struct task_cputime *totals; }; We also add a new task_cputime substructure directly to signal_struct, to cache the earliest expiration of process-wide timers, and task_cputime also replaces the it_*_expires fields of task_struct (used for earliest expiration of thread timers). The "thread_group_cputime" structure contains process-wide timers that are updated via account_user_time() and friends. In the non-SMP case the structure is a simple aggregator; unfortunately in the SMP case that simplicity was not achievable due to cache-line contention between CPUs (in one measured case performance was actually _worse_ on a 16-cpu system than the same test on a 4-cpu system, due to this contention). For SMP, the thread_group_cputime counters are maintained as a per-cpu structure allocated using alloc_percpu(). The timer functions update only the timer field in the structure corresponding to the running CPU, obtained using per_cpu_ptr(). We define a set of inline functions in sched.h that we use to maintain the thread_group_cputime structure and hide the differences between UP and SMP implementations from the rest of the kernel. The thread_group_cputime_init() function initializes the thread_group_cputime structure for the given task. The thread_group_cputime_alloc() is a no-op for UP; for SMP it calls the out-of-line function thread_group_cputime_alloc_smp() to allocate and fill in the per-cpu structures and fields. The thread_group_cputime_free() function, also a no-op for UP, in SMP frees the per-cpu structures. The thread_group_cputime_clone_thread() function (also a UP no-op) for SMP calls thread_group_cputime_alloc() if the per-cpu structures haven't yet been allocated. The thread_group_cputime() function fills the task_cputime structure it is passed with the contents of the thread_group_cputime fields; in UP it's that simple but in SMP it must also safely check that tsk->signal is non-NULL (if it is it just uses the appropriate fields of task_struct) and, if so, sums the per-cpu values for each online CPU. Finally, the three functions account_group_user_time(), account_group_system_time() and account_group_exec_runtime() are used by timer functions to update the respective fields of the thread_group_cputime structure. Non-SMP operation is trivial and will not be mentioned further. The per-cpu structure is always allocated when a task creates its first new thread, via a call to thread_group_cputime_clone_thread() from copy_signal(). It is freed at process exit via a call to thread_group_cputime_free() from cleanup_signal(). All functions that formerly summed utime/stime/sum_sched_runtime values from from all threads in the thread group now use thread_group_cputime() to snapshot the values in the thread_group_cputime structure or the values in the task structure itself if the per-cpu structure hasn't been allocated. Finally, the code in kernel/posix-cpu-timers.c has changed quite a bit. The run_posix_cpu_timers() function has been split into a fast path and a slow path; the former safely checks whether there are any expired thread timers and, if not, just returns, while the slow path does the heavy lifting. With the dedicated thread group fields, timers are no longer "rebalanced" and the process_timer_rebalance() function and related code has gone away. All summing loops are gone and all code that used them now uses the thread_group_cputime() inline. When process-wide timers are set, the new task_cputime structure in signal_struct is used to cache the earliest expiration; this is checked in the fast path. Performance The fix appears not to add significant overhead to existing operations. It generally performs the same as the current code except in two cases, one in which it performs slightly worse (Case 5 below) and one in which it performs very significantly better (Case 2 below). Overall it's a wash except in those two cases. I've since done somewhat more involved testing on a dual-core Opteron system. Case 1: With no itimer running, for a test with 100,000 threads, the fixed kernel took 1428.5 seconds, 513 seconds more than the unfixed system, all of which was spent in the system. There were twice as many voluntary context switches with the fix as without it. Case 2: With an itimer running at .01 second ticks and 4000 threads (the most an unmodified kernel can handle), the fixed kernel ran the test in eight percent of the time (5.8 seconds as opposed to 70 seconds) and had better tick accuracy (.012 seconds per tick as opposed to .023 seconds per tick). Case 3: A 4000-thread test with an initial timer tick of .01 second and an interval of 10,000 seconds (i.e. a timer that ticks only once) had very nearly the same performance in both cases: 6.3 seconds elapsed for the fixed kernel versus 5.5 seconds for the unfixed kernel. With fewer threads (eight in these tests), the Case 1 test ran in essentially the same time on both the modified and unmodified kernels (5.2 seconds versus 5.8 seconds). The Case 2 test ran in about the same time as well, 5.9 seconds versus 5.4 seconds but again with much better tick accuracy, .013 seconds per tick versus .025 seconds per tick for the unmodified kernel. Since the fix affected the rlimit code, I also tested soft and hard CPU limits. Case 4: With a hard CPU limit of 20 seconds and eight threads (and an itimer running), the modified kernel was very slightly favored in that while it killed the process in 19.997 seconds of CPU time (5.002 seconds of wall time), only .003 seconds of that was system time, the rest was user time. The unmodified kernel killed the process in 20.001 seconds of CPU (5.014 seconds of wall time) of which .016 seconds was system time. Really, though, the results were too close to call. The results were essentially the same with no itimer running. Case 5: With a soft limit of 20 seconds and a hard limit of 2000 seconds (where the hard limit would never be reached) and an itimer running, the modified kernel exhibited worse tick accuracy than the unmodified kernel: .050 seconds/tick versus .028 seconds/tick. Otherwise, performance was almost indistinguishable. With no itimer running this test exhibited virtually identical behavior and times in both cases. In times past I did some limited performance testing. those results are below. On a four-cpu Opteron system without this fix, a sixteen-thread test executed in 3569.991 seconds, of which user was 3568.435s and system was 1.556s. On the same system with the fix, user and elapsed time were about the same, but system time dropped to 0.007 seconds. Performance with eight, four and one thread were comparable. Interestingly, the timer ticks with the fix seemed more accurate: The sixteen-thread test with the fix received 149543 ticks for 0.024 seconds per tick, while the same test without the fix received 58720 for 0.061 seconds per tick. Both cases were configured for an interval of 0.01 seconds. Again, the other tests were comparable. Each thread in this test computed the primes up to 25,000,000. I also did a test with a large number of threads, 100,000 threads, which is impossible without the fix. In this case each thread computed the primes only up to 10,000 (to make the runtime manageable). System time dominated, at 1546.968 seconds out of a total 2176.906 seconds (giving a user time of 629.938s). It received 147651 ticks for 0.015 seconds per tick, still quite accurate. There is obviously no comparable test without the fix. Signed-off-by: Frank Mayhar <fmayhar@google.com> Cc: Roland McGrath <roland@redhat.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-09-13 00:54:39 +08:00
struct task_cputime cputime;
/*
* The resource counters for the group leader are in its
* own task_struct. Those for dead threads in the group
* are in its signal_struct, as are those for the child
* processes it has previously reaped. All these
* accumulate in the parent's signal_struct c* fields.
*
* We don't bother to take a lock here to protect these
* p->signal fields, because they are only touched by
* __exit_signal, which runs with tasklist_lock
* write-locked anyway, and so is excluded here. We do
* need to protect the access to p->parent->signal fields,
* as other threads in the parent group can be right
* here reaping other children at the same time.
timers: fix itimer/many thread hang Overview This patch reworks the handling of POSIX CPU timers, including the ITIMER_PROF, ITIMER_VIRT timers and rlimit handling. It was put together with the help of Roland McGrath, the owner and original writer of this code. The problem we ran into, and the reason for this rework, has to do with using a profiling timer in a process with a large number of threads. It appears that the performance of the old implementation of run_posix_cpu_timers() was at least O(n*3) (where "n" is the number of threads in a process) or worse. Everything is fine with an increasing number of threads until the time taken for that routine to run becomes the same as or greater than the tick time, at which point things degrade rather quickly. This patch fixes bug 9906, "Weird hang with NPTL and SIGPROF." Code Changes This rework corrects the implementation of run_posix_cpu_timers() to make it run in constant time for a particular machine. (Performance may vary between one machine and another depending upon whether the kernel is built as single- or multiprocessor and, in the latter case, depending upon the number of running processors.) To do this, at each tick we now update fields in signal_struct as well as task_struct. The run_posix_cpu_timers() function uses those fields to make its decisions. We define a new structure, "task_cputime," to contain user, system and scheduler times and use these in appropriate places: struct task_cputime { cputime_t utime; cputime_t stime; unsigned long long sum_exec_runtime; }; This is included in the structure "thread_group_cputime," which is a new substructure of signal_struct and which varies for uniprocessor versus multiprocessor kernels. For uniprocessor kernels, it uses "task_cputime" as a simple substructure, while for multiprocessor kernels it is a pointer: struct thread_group_cputime { struct task_cputime totals; }; struct thread_group_cputime { struct task_cputime *totals; }; We also add a new task_cputime substructure directly to signal_struct, to cache the earliest expiration of process-wide timers, and task_cputime also replaces the it_*_expires fields of task_struct (used for earliest expiration of thread timers). The "thread_group_cputime" structure contains process-wide timers that are updated via account_user_time() and friends. In the non-SMP case the structure is a simple aggregator; unfortunately in the SMP case that simplicity was not achievable due to cache-line contention between CPUs (in one measured case performance was actually _worse_ on a 16-cpu system than the same test on a 4-cpu system, due to this contention). For SMP, the thread_group_cputime counters are maintained as a per-cpu structure allocated using alloc_percpu(). The timer functions update only the timer field in the structure corresponding to the running CPU, obtained using per_cpu_ptr(). We define a set of inline functions in sched.h that we use to maintain the thread_group_cputime structure and hide the differences between UP and SMP implementations from the rest of the kernel. The thread_group_cputime_init() function initializes the thread_group_cputime structure for the given task. The thread_group_cputime_alloc() is a no-op for UP; for SMP it calls the out-of-line function thread_group_cputime_alloc_smp() to allocate and fill in the per-cpu structures and fields. The thread_group_cputime_free() function, also a no-op for UP, in SMP frees the per-cpu structures. The thread_group_cputime_clone_thread() function (also a UP no-op) for SMP calls thread_group_cputime_alloc() if the per-cpu structures haven't yet been allocated. The thread_group_cputime() function fills the task_cputime structure it is passed with the contents of the thread_group_cputime fields; in UP it's that simple but in SMP it must also safely check that tsk->signal is non-NULL (if it is it just uses the appropriate fields of task_struct) and, if so, sums the per-cpu values for each online CPU. Finally, the three functions account_group_user_time(), account_group_system_time() and account_group_exec_runtime() are used by timer functions to update the respective fields of the thread_group_cputime structure. Non-SMP operation is trivial and will not be mentioned further. The per-cpu structure is always allocated when a task creates its first new thread, via a call to thread_group_cputime_clone_thread() from copy_signal(). It is freed at process exit via a call to thread_group_cputime_free() from cleanup_signal(). All functions that formerly summed utime/stime/sum_sched_runtime values from from all threads in the thread group now use thread_group_cputime() to snapshot the values in the thread_group_cputime structure or the values in the task structure itself if the per-cpu structure hasn't been allocated. Finally, the code in kernel/posix-cpu-timers.c has changed quite a bit. The run_posix_cpu_timers() function has been split into a fast path and a slow path; the former safely checks whether there are any expired thread timers and, if not, just returns, while the slow path does the heavy lifting. With the dedicated thread group fields, timers are no longer "rebalanced" and the process_timer_rebalance() function and related code has gone away. All summing loops are gone and all code that used them now uses the thread_group_cputime() inline. When process-wide timers are set, the new task_cputime structure in signal_struct is used to cache the earliest expiration; this is checked in the fast path. Performance The fix appears not to add significant overhead to existing operations. It generally performs the same as the current code except in two cases, one in which it performs slightly worse (Case 5 below) and one in which it performs very significantly better (Case 2 below). Overall it's a wash except in those two cases. I've since done somewhat more involved testing on a dual-core Opteron system. Case 1: With no itimer running, for a test with 100,000 threads, the fixed kernel took 1428.5 seconds, 513 seconds more than the unfixed system, all of which was spent in the system. There were twice as many voluntary context switches with the fix as without it. Case 2: With an itimer running at .01 second ticks and 4000 threads (the most an unmodified kernel can handle), the fixed kernel ran the test in eight percent of the time (5.8 seconds as opposed to 70 seconds) and had better tick accuracy (.012 seconds per tick as opposed to .023 seconds per tick). Case 3: A 4000-thread test with an initial timer tick of .01 second and an interval of 10,000 seconds (i.e. a timer that ticks only once) had very nearly the same performance in both cases: 6.3 seconds elapsed for the fixed kernel versus 5.5 seconds for the unfixed kernel. With fewer threads (eight in these tests), the Case 1 test ran in essentially the same time on both the modified and unmodified kernels (5.2 seconds versus 5.8 seconds). The Case 2 test ran in about the same time as well, 5.9 seconds versus 5.4 seconds but again with much better tick accuracy, .013 seconds per tick versus .025 seconds per tick for the unmodified kernel. Since the fix affected the rlimit code, I also tested soft and hard CPU limits. Case 4: With a hard CPU limit of 20 seconds and eight threads (and an itimer running), the modified kernel was very slightly favored in that while it killed the process in 19.997 seconds of CPU time (5.002 seconds of wall time), only .003 seconds of that was system time, the rest was user time. The unmodified kernel killed the process in 20.001 seconds of CPU (5.014 seconds of wall time) of which .016 seconds was system time. Really, though, the results were too close to call. The results were essentially the same with no itimer running. Case 5: With a soft limit of 20 seconds and a hard limit of 2000 seconds (where the hard limit would never be reached) and an itimer running, the modified kernel exhibited worse tick accuracy than the unmodified kernel: .050 seconds/tick versus .028 seconds/tick. Otherwise, performance was almost indistinguishable. With no itimer running this test exhibited virtually identical behavior and times in both cases. In times past I did some limited performance testing. those results are below. On a four-cpu Opteron system without this fix, a sixteen-thread test executed in 3569.991 seconds, of which user was 3568.435s and system was 1.556s. On the same system with the fix, user and elapsed time were about the same, but system time dropped to 0.007 seconds. Performance with eight, four and one thread were comparable. Interestingly, the timer ticks with the fix seemed more accurate: The sixteen-thread test with the fix received 149543 ticks for 0.024 seconds per tick, while the same test without the fix received 58720 for 0.061 seconds per tick. Both cases were configured for an interval of 0.01 seconds. Again, the other tests were comparable. Each thread in this test computed the primes up to 25,000,000. I also did a test with a large number of threads, 100,000 threads, which is impossible without the fix. In this case each thread computed the primes only up to 10,000 (to make the runtime manageable). System time dominated, at 1546.968 seconds out of a total 2176.906 seconds (giving a user time of 629.938s). It received 147651 ticks for 0.015 seconds per tick, still quite accurate. There is obviously no comparable test without the fix. Signed-off-by: Frank Mayhar <fmayhar@google.com> Cc: Roland McGrath <roland@redhat.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-09-13 00:54:39 +08:00
*
* We use thread_group_cputime() to get times for the thread
* group, which consolidates times for all threads in the
* group including the group leader.
*/
spin_lock_irq(&p->parent->sighand->siglock);
psig = p->parent->signal;
sig = p->signal;
timers: fix itimer/many thread hang Overview This patch reworks the handling of POSIX CPU timers, including the ITIMER_PROF, ITIMER_VIRT timers and rlimit handling. It was put together with the help of Roland McGrath, the owner and original writer of this code. The problem we ran into, and the reason for this rework, has to do with using a profiling timer in a process with a large number of threads. It appears that the performance of the old implementation of run_posix_cpu_timers() was at least O(n*3) (where "n" is the number of threads in a process) or worse. Everything is fine with an increasing number of threads until the time taken for that routine to run becomes the same as or greater than the tick time, at which point things degrade rather quickly. This patch fixes bug 9906, "Weird hang with NPTL and SIGPROF." Code Changes This rework corrects the implementation of run_posix_cpu_timers() to make it run in constant time for a particular machine. (Performance may vary between one machine and another depending upon whether the kernel is built as single- or multiprocessor and, in the latter case, depending upon the number of running processors.) To do this, at each tick we now update fields in signal_struct as well as task_struct. The run_posix_cpu_timers() function uses those fields to make its decisions. We define a new structure, "task_cputime," to contain user, system and scheduler times and use these in appropriate places: struct task_cputime { cputime_t utime; cputime_t stime; unsigned long long sum_exec_runtime; }; This is included in the structure "thread_group_cputime," which is a new substructure of signal_struct and which varies for uniprocessor versus multiprocessor kernels. For uniprocessor kernels, it uses "task_cputime" as a simple substructure, while for multiprocessor kernels it is a pointer: struct thread_group_cputime { struct task_cputime totals; }; struct thread_group_cputime { struct task_cputime *totals; }; We also add a new task_cputime substructure directly to signal_struct, to cache the earliest expiration of process-wide timers, and task_cputime also replaces the it_*_expires fields of task_struct (used for earliest expiration of thread timers). The "thread_group_cputime" structure contains process-wide timers that are updated via account_user_time() and friends. In the non-SMP case the structure is a simple aggregator; unfortunately in the SMP case that simplicity was not achievable due to cache-line contention between CPUs (in one measured case performance was actually _worse_ on a 16-cpu system than the same test on a 4-cpu system, due to this contention). For SMP, the thread_group_cputime counters are maintained as a per-cpu structure allocated using alloc_percpu(). The timer functions update only the timer field in the structure corresponding to the running CPU, obtained using per_cpu_ptr(). We define a set of inline functions in sched.h that we use to maintain the thread_group_cputime structure and hide the differences between UP and SMP implementations from the rest of the kernel. The thread_group_cputime_init() function initializes the thread_group_cputime structure for the given task. The thread_group_cputime_alloc() is a no-op for UP; for SMP it calls the out-of-line function thread_group_cputime_alloc_smp() to allocate and fill in the per-cpu structures and fields. The thread_group_cputime_free() function, also a no-op for UP, in SMP frees the per-cpu structures. The thread_group_cputime_clone_thread() function (also a UP no-op) for SMP calls thread_group_cputime_alloc() if the per-cpu structures haven't yet been allocated. The thread_group_cputime() function fills the task_cputime structure it is passed with the contents of the thread_group_cputime fields; in UP it's that simple but in SMP it must also safely check that tsk->signal is non-NULL (if it is it just uses the appropriate fields of task_struct) and, if so, sums the per-cpu values for each online CPU. Finally, the three functions account_group_user_time(), account_group_system_time() and account_group_exec_runtime() are used by timer functions to update the respective fields of the thread_group_cputime structure. Non-SMP operation is trivial and will not be mentioned further. The per-cpu structure is always allocated when a task creates its first new thread, via a call to thread_group_cputime_clone_thread() from copy_signal(). It is freed at process exit via a call to thread_group_cputime_free() from cleanup_signal(). All functions that formerly summed utime/stime/sum_sched_runtime values from from all threads in the thread group now use thread_group_cputime() to snapshot the values in the thread_group_cputime structure or the values in the task structure itself if the per-cpu structure hasn't been allocated. Finally, the code in kernel/posix-cpu-timers.c has changed quite a bit. The run_posix_cpu_timers() function has been split into a fast path and a slow path; the former safely checks whether there are any expired thread timers and, if not, just returns, while the slow path does the heavy lifting. With the dedicated thread group fields, timers are no longer "rebalanced" and the process_timer_rebalance() function and related code has gone away. All summing loops are gone and all code that used them now uses the thread_group_cputime() inline. When process-wide timers are set, the new task_cputime structure in signal_struct is used to cache the earliest expiration; this is checked in the fast path. Performance The fix appears not to add significant overhead to existing operations. It generally performs the same as the current code except in two cases, one in which it performs slightly worse (Case 5 below) and one in which it performs very significantly better (Case 2 below). Overall it's a wash except in those two cases. I've since done somewhat more involved testing on a dual-core Opteron system. Case 1: With no itimer running, for a test with 100,000 threads, the fixed kernel took 1428.5 seconds, 513 seconds more than the unfixed system, all of which was spent in the system. There were twice as many voluntary context switches with the fix as without it. Case 2: With an itimer running at .01 second ticks and 4000 threads (the most an unmodified kernel can handle), the fixed kernel ran the test in eight percent of the time (5.8 seconds as opposed to 70 seconds) and had better tick accuracy (.012 seconds per tick as opposed to .023 seconds per tick). Case 3: A 4000-thread test with an initial timer tick of .01 second and an interval of 10,000 seconds (i.e. a timer that ticks only once) had very nearly the same performance in both cases: 6.3 seconds elapsed for the fixed kernel versus 5.5 seconds for the unfixed kernel. With fewer threads (eight in these tests), the Case 1 test ran in essentially the same time on both the modified and unmodified kernels (5.2 seconds versus 5.8 seconds). The Case 2 test ran in about the same time as well, 5.9 seconds versus 5.4 seconds but again with much better tick accuracy, .013 seconds per tick versus .025 seconds per tick for the unmodified kernel. Since the fix affected the rlimit code, I also tested soft and hard CPU limits. Case 4: With a hard CPU limit of 20 seconds and eight threads (and an itimer running), the modified kernel was very slightly favored in that while it killed the process in 19.997 seconds of CPU time (5.002 seconds of wall time), only .003 seconds of that was system time, the rest was user time. The unmodified kernel killed the process in 20.001 seconds of CPU (5.014 seconds of wall time) of which .016 seconds was system time. Really, though, the results were too close to call. The results were essentially the same with no itimer running. Case 5: With a soft limit of 20 seconds and a hard limit of 2000 seconds (where the hard limit would never be reached) and an itimer running, the modified kernel exhibited worse tick accuracy than the unmodified kernel: .050 seconds/tick versus .028 seconds/tick. Otherwise, performance was almost indistinguishable. With no itimer running this test exhibited virtually identical behavior and times in both cases. In times past I did some limited performance testing. those results are below. On a four-cpu Opteron system without this fix, a sixteen-thread test executed in 3569.991 seconds, of which user was 3568.435s and system was 1.556s. On the same system with the fix, user and elapsed time were about the same, but system time dropped to 0.007 seconds. Performance with eight, four and one thread were comparable. Interestingly, the timer ticks with the fix seemed more accurate: The sixteen-thread test with the fix received 149543 ticks for 0.024 seconds per tick, while the same test without the fix received 58720 for 0.061 seconds per tick. Both cases were configured for an interval of 0.01 seconds. Again, the other tests were comparable. Each thread in this test computed the primes up to 25,000,000. I also did a test with a large number of threads, 100,000 threads, which is impossible without the fix. In this case each thread computed the primes only up to 10,000 (to make the runtime manageable). System time dominated, at 1546.968 seconds out of a total 2176.906 seconds (giving a user time of 629.938s). It received 147651 ticks for 0.015 seconds per tick, still quite accurate. There is obviously no comparable test without the fix. Signed-off-by: Frank Mayhar <fmayhar@google.com> Cc: Roland McGrath <roland@redhat.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-09-13 00:54:39 +08:00
thread_group_cputime(p, &cputime);
psig->cutime =
cputime_add(psig->cutime,
timers: fix itimer/many thread hang Overview This patch reworks the handling of POSIX CPU timers, including the ITIMER_PROF, ITIMER_VIRT timers and rlimit handling. It was put together with the help of Roland McGrath, the owner and original writer of this code. The problem we ran into, and the reason for this rework, has to do with using a profiling timer in a process with a large number of threads. It appears that the performance of the old implementation of run_posix_cpu_timers() was at least O(n*3) (where "n" is the number of threads in a process) or worse. Everything is fine with an increasing number of threads until the time taken for that routine to run becomes the same as or greater than the tick time, at which point things degrade rather quickly. This patch fixes bug 9906, "Weird hang with NPTL and SIGPROF." Code Changes This rework corrects the implementation of run_posix_cpu_timers() to make it run in constant time for a particular machine. (Performance may vary between one machine and another depending upon whether the kernel is built as single- or multiprocessor and, in the latter case, depending upon the number of running processors.) To do this, at each tick we now update fields in signal_struct as well as task_struct. The run_posix_cpu_timers() function uses those fields to make its decisions. We define a new structure, "task_cputime," to contain user, system and scheduler times and use these in appropriate places: struct task_cputime { cputime_t utime; cputime_t stime; unsigned long long sum_exec_runtime; }; This is included in the structure "thread_group_cputime," which is a new substructure of signal_struct and which varies for uniprocessor versus multiprocessor kernels. For uniprocessor kernels, it uses "task_cputime" as a simple substructure, while for multiprocessor kernels it is a pointer: struct thread_group_cputime { struct task_cputime totals; }; struct thread_group_cputime { struct task_cputime *totals; }; We also add a new task_cputime substructure directly to signal_struct, to cache the earliest expiration of process-wide timers, and task_cputime also replaces the it_*_expires fields of task_struct (used for earliest expiration of thread timers). The "thread_group_cputime" structure contains process-wide timers that are updated via account_user_time() and friends. In the non-SMP case the structure is a simple aggregator; unfortunately in the SMP case that simplicity was not achievable due to cache-line contention between CPUs (in one measured case performance was actually _worse_ on a 16-cpu system than the same test on a 4-cpu system, due to this contention). For SMP, the thread_group_cputime counters are maintained as a per-cpu structure allocated using alloc_percpu(). The timer functions update only the timer field in the structure corresponding to the running CPU, obtained using per_cpu_ptr(). We define a set of inline functions in sched.h that we use to maintain the thread_group_cputime structure and hide the differences between UP and SMP implementations from the rest of the kernel. The thread_group_cputime_init() function initializes the thread_group_cputime structure for the given task. The thread_group_cputime_alloc() is a no-op for UP; for SMP it calls the out-of-line function thread_group_cputime_alloc_smp() to allocate and fill in the per-cpu structures and fields. The thread_group_cputime_free() function, also a no-op for UP, in SMP frees the per-cpu structures. The thread_group_cputime_clone_thread() function (also a UP no-op) for SMP calls thread_group_cputime_alloc() if the per-cpu structures haven't yet been allocated. The thread_group_cputime() function fills the task_cputime structure it is passed with the contents of the thread_group_cputime fields; in UP it's that simple but in SMP it must also safely check that tsk->signal is non-NULL (if it is it just uses the appropriate fields of task_struct) and, if so, sums the per-cpu values for each online CPU. Finally, the three functions account_group_user_time(), account_group_system_time() and account_group_exec_runtime() are used by timer functions to update the respective fields of the thread_group_cputime structure. Non-SMP operation is trivial and will not be mentioned further. The per-cpu structure is always allocated when a task creates its first new thread, via a call to thread_group_cputime_clone_thread() from copy_signal(). It is freed at process exit via a call to thread_group_cputime_free() from cleanup_signal(). All functions that formerly summed utime/stime/sum_sched_runtime values from from all threads in the thread group now use thread_group_cputime() to snapshot the values in the thread_group_cputime structure or the values in the task structure itself if the per-cpu structure hasn't been allocated. Finally, the code in kernel/posix-cpu-timers.c has changed quite a bit. The run_posix_cpu_timers() function has been split into a fast path and a slow path; the former safely checks whether there are any expired thread timers and, if not, just returns, while the slow path does the heavy lifting. With the dedicated thread group fields, timers are no longer "rebalanced" and the process_timer_rebalance() function and related code has gone away. All summing loops are gone and all code that used them now uses the thread_group_cputime() inline. When process-wide timers are set, the new task_cputime structure in signal_struct is used to cache the earliest expiration; this is checked in the fast path. Performance The fix appears not to add significant overhead to existing operations. It generally performs the same as the current code except in two cases, one in which it performs slightly worse (Case 5 below) and one in which it performs very significantly better (Case 2 below). Overall it's a wash except in those two cases. I've since done somewhat more involved testing on a dual-core Opteron system. Case 1: With no itimer running, for a test with 100,000 threads, the fixed kernel took 1428.5 seconds, 513 seconds more than the unfixed system, all of which was spent in the system. There were twice as many voluntary context switches with the fix as without it. Case 2: With an itimer running at .01 second ticks and 4000 threads (the most an unmodified kernel can handle), the fixed kernel ran the test in eight percent of the time (5.8 seconds as opposed to 70 seconds) and had better tick accuracy (.012 seconds per tick as opposed to .023 seconds per tick). Case 3: A 4000-thread test with an initial timer tick of .01 second and an interval of 10,000 seconds (i.e. a timer that ticks only once) had very nearly the same performance in both cases: 6.3 seconds elapsed for the fixed kernel versus 5.5 seconds for the unfixed kernel. With fewer threads (eight in these tests), the Case 1 test ran in essentially the same time on both the modified and unmodified kernels (5.2 seconds versus 5.8 seconds). The Case 2 test ran in about the same time as well, 5.9 seconds versus 5.4 seconds but again with much better tick accuracy, .013 seconds per tick versus .025 seconds per tick for the unmodified kernel. Since the fix affected the rlimit code, I also tested soft and hard CPU limits. Case 4: With a hard CPU limit of 20 seconds and eight threads (and an itimer running), the modified kernel was very slightly favored in that while it killed the process in 19.997 seconds of CPU time (5.002 seconds of wall time), only .003 seconds of that was system time, the rest was user time. The unmodified kernel killed the process in 20.001 seconds of CPU (5.014 seconds of wall time) of which .016 seconds was system time. Really, though, the results were too close to call. The results were essentially the same with no itimer running. Case 5: With a soft limit of 20 seconds and a hard limit of 2000 seconds (where the hard limit would never be reached) and an itimer running, the modified kernel exhibited worse tick accuracy than the unmodified kernel: .050 seconds/tick versus .028 seconds/tick. Otherwise, performance was almost indistinguishable. With no itimer running this test exhibited virtually identical behavior and times in both cases. In times past I did some limited performance testing. those results are below. On a four-cpu Opteron system without this fix, a sixteen-thread test executed in 3569.991 seconds, of which user was 3568.435s and system was 1.556s. On the same system with the fix, user and elapsed time were about the same, but system time dropped to 0.007 seconds. Performance with eight, four and one thread were comparable. Interestingly, the timer ticks with the fix seemed more accurate: The sixteen-thread test with the fix received 149543 ticks for 0.024 seconds per tick, while the same test without the fix received 58720 for 0.061 seconds per tick. Both cases were configured for an interval of 0.01 seconds. Again, the other tests were comparable. Each thread in this test computed the primes up to 25,000,000. I also did a test with a large number of threads, 100,000 threads, which is impossible without the fix. In this case each thread computed the primes only up to 10,000 (to make the runtime manageable). System time dominated, at 1546.968 seconds out of a total 2176.906 seconds (giving a user time of 629.938s). It received 147651 ticks for 0.015 seconds per tick, still quite accurate. There is obviously no comparable test without the fix. Signed-off-by: Frank Mayhar <fmayhar@google.com> Cc: Roland McGrath <roland@redhat.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-09-13 00:54:39 +08:00
cputime_add(cputime.utime,
sig->cutime));
psig->cstime =
cputime_add(psig->cstime,
timers: fix itimer/many thread hang Overview This patch reworks the handling of POSIX CPU timers, including the ITIMER_PROF, ITIMER_VIRT timers and rlimit handling. It was put together with the help of Roland McGrath, the owner and original writer of this code. The problem we ran into, and the reason for this rework, has to do with using a profiling timer in a process with a large number of threads. It appears that the performance of the old implementation of run_posix_cpu_timers() was at least O(n*3) (where "n" is the number of threads in a process) or worse. Everything is fine with an increasing number of threads until the time taken for that routine to run becomes the same as or greater than the tick time, at which point things degrade rather quickly. This patch fixes bug 9906, "Weird hang with NPTL and SIGPROF." Code Changes This rework corrects the implementation of run_posix_cpu_timers() to make it run in constant time for a particular machine. (Performance may vary between one machine and another depending upon whether the kernel is built as single- or multiprocessor and, in the latter case, depending upon the number of running processors.) To do this, at each tick we now update fields in signal_struct as well as task_struct. The run_posix_cpu_timers() function uses those fields to make its decisions. We define a new structure, "task_cputime," to contain user, system and scheduler times and use these in appropriate places: struct task_cputime { cputime_t utime; cputime_t stime; unsigned long long sum_exec_runtime; }; This is included in the structure "thread_group_cputime," which is a new substructure of signal_struct and which varies for uniprocessor versus multiprocessor kernels. For uniprocessor kernels, it uses "task_cputime" as a simple substructure, while for multiprocessor kernels it is a pointer: struct thread_group_cputime { struct task_cputime totals; }; struct thread_group_cputime { struct task_cputime *totals; }; We also add a new task_cputime substructure directly to signal_struct, to cache the earliest expiration of process-wide timers, and task_cputime also replaces the it_*_expires fields of task_struct (used for earliest expiration of thread timers). The "thread_group_cputime" structure contains process-wide timers that are updated via account_user_time() and friends. In the non-SMP case the structure is a simple aggregator; unfortunately in the SMP case that simplicity was not achievable due to cache-line contention between CPUs (in one measured case performance was actually _worse_ on a 16-cpu system than the same test on a 4-cpu system, due to this contention). For SMP, the thread_group_cputime counters are maintained as a per-cpu structure allocated using alloc_percpu(). The timer functions update only the timer field in the structure corresponding to the running CPU, obtained using per_cpu_ptr(). We define a set of inline functions in sched.h that we use to maintain the thread_group_cputime structure and hide the differences between UP and SMP implementations from the rest of the kernel. The thread_group_cputime_init() function initializes the thread_group_cputime structure for the given task. The thread_group_cputime_alloc() is a no-op for UP; for SMP it calls the out-of-line function thread_group_cputime_alloc_smp() to allocate and fill in the per-cpu structures and fields. The thread_group_cputime_free() function, also a no-op for UP, in SMP frees the per-cpu structures. The thread_group_cputime_clone_thread() function (also a UP no-op) for SMP calls thread_group_cputime_alloc() if the per-cpu structures haven't yet been allocated. The thread_group_cputime() function fills the task_cputime structure it is passed with the contents of the thread_group_cputime fields; in UP it's that simple but in SMP it must also safely check that tsk->signal is non-NULL (if it is it just uses the appropriate fields of task_struct) and, if so, sums the per-cpu values for each online CPU. Finally, the three functions account_group_user_time(), account_group_system_time() and account_group_exec_runtime() are used by timer functions to update the respective fields of the thread_group_cputime structure. Non-SMP operation is trivial and will not be mentioned further. The per-cpu structure is always allocated when a task creates its first new thread, via a call to thread_group_cputime_clone_thread() from copy_signal(). It is freed at process exit via a call to thread_group_cputime_free() from cleanup_signal(). All functions that formerly summed utime/stime/sum_sched_runtime values from from all threads in the thread group now use thread_group_cputime() to snapshot the values in the thread_group_cputime structure or the values in the task structure itself if the per-cpu structure hasn't been allocated. Finally, the code in kernel/posix-cpu-timers.c has changed quite a bit. The run_posix_cpu_timers() function has been split into a fast path and a slow path; the former safely checks whether there are any expired thread timers and, if not, just returns, while the slow path does the heavy lifting. With the dedicated thread group fields, timers are no longer "rebalanced" and the process_timer_rebalance() function and related code has gone away. All summing loops are gone and all code that used them now uses the thread_group_cputime() inline. When process-wide timers are set, the new task_cputime structure in signal_struct is used to cache the earliest expiration; this is checked in the fast path. Performance The fix appears not to add significant overhead to existing operations. It generally performs the same as the current code except in two cases, one in which it performs slightly worse (Case 5 below) and one in which it performs very significantly better (Case 2 below). Overall it's a wash except in those two cases. I've since done somewhat more involved testing on a dual-core Opteron system. Case 1: With no itimer running, for a test with 100,000 threads, the fixed kernel took 1428.5 seconds, 513 seconds more than the unfixed system, all of which was spent in the system. There were twice as many voluntary context switches with the fix as without it. Case 2: With an itimer running at .01 second ticks and 4000 threads (the most an unmodified kernel can handle), the fixed kernel ran the test in eight percent of the time (5.8 seconds as opposed to 70 seconds) and had better tick accuracy (.012 seconds per tick as opposed to .023 seconds per tick). Case 3: A 4000-thread test with an initial timer tick of .01 second and an interval of 10,000 seconds (i.e. a timer that ticks only once) had very nearly the same performance in both cases: 6.3 seconds elapsed for the fixed kernel versus 5.5 seconds for the unfixed kernel. With fewer threads (eight in these tests), the Case 1 test ran in essentially the same time on both the modified and unmodified kernels (5.2 seconds versus 5.8 seconds). The Case 2 test ran in about the same time as well, 5.9 seconds versus 5.4 seconds but again with much better tick accuracy, .013 seconds per tick versus .025 seconds per tick for the unmodified kernel. Since the fix affected the rlimit code, I also tested soft and hard CPU limits. Case 4: With a hard CPU limit of 20 seconds and eight threads (and an itimer running), the modified kernel was very slightly favored in that while it killed the process in 19.997 seconds of CPU time (5.002 seconds of wall time), only .003 seconds of that was system time, the rest was user time. The unmodified kernel killed the process in 20.001 seconds of CPU (5.014 seconds of wall time) of which .016 seconds was system time. Really, though, the results were too close to call. The results were essentially the same with no itimer running. Case 5: With a soft limit of 20 seconds and a hard limit of 2000 seconds (where the hard limit would never be reached) and an itimer running, the modified kernel exhibited worse tick accuracy than the unmodified kernel: .050 seconds/tick versus .028 seconds/tick. Otherwise, performance was almost indistinguishable. With no itimer running this test exhibited virtually identical behavior and times in both cases. In times past I did some limited performance testing. those results are below. On a four-cpu Opteron system without this fix, a sixteen-thread test executed in 3569.991 seconds, of which user was 3568.435s and system was 1.556s. On the same system with the fix, user and elapsed time were about the same, but system time dropped to 0.007 seconds. Performance with eight, four and one thread were comparable. Interestingly, the timer ticks with the fix seemed more accurate: The sixteen-thread test with the fix received 149543 ticks for 0.024 seconds per tick, while the same test without the fix received 58720 for 0.061 seconds per tick. Both cases were configured for an interval of 0.01 seconds. Again, the other tests were comparable. Each thread in this test computed the primes up to 25,000,000. I also did a test with a large number of threads, 100,000 threads, which is impossible without the fix. In this case each thread computed the primes only up to 10,000 (to make the runtime manageable). System time dominated, at 1546.968 seconds out of a total 2176.906 seconds (giving a user time of 629.938s). It received 147651 ticks for 0.015 seconds per tick, still quite accurate. There is obviously no comparable test without the fix. Signed-off-by: Frank Mayhar <fmayhar@google.com> Cc: Roland McGrath <roland@redhat.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-09-13 00:54:39 +08:00
cputime_add(cputime.stime,
sig->cstime));
psig->cgtime =
cputime_add(psig->cgtime,
cputime_add(p->gtime,
cputime_add(sig->gtime,
sig->cgtime)));
psig->cmin_flt +=
p->min_flt + sig->min_flt + sig->cmin_flt;
psig->cmaj_flt +=
p->maj_flt + sig->maj_flt + sig->cmaj_flt;
psig->cnvcsw +=
p->nvcsw + sig->nvcsw + sig->cnvcsw;
psig->cnivcsw +=
p->nivcsw + sig->nivcsw + sig->cnivcsw;
psig->cinblock +=
task_io_get_inblock(p) +
sig->inblock + sig->cinblock;
psig->coublock +=
task_io_get_oublock(p) +
sig->oublock + sig->coublock;
task_io_accounting_add(&psig->ioac, &p->ioac);
task_io_accounting_add(&psig->ioac, &sig->ioac);
spin_unlock_irq(&p->parent->sighand->siglock);
}
/*
* Now we are sure this task is interesting, and no other
* thread can reap it because we set its state to EXIT_DEAD.
*/
read_unlock(&tasklist_lock);
retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
status = (p->signal->flags & SIGNAL_GROUP_EXIT)
? p->signal->group_exit_code : p->exit_code;
if (!retval && stat_addr)
retval = put_user(status, stat_addr);
if (!retval && infop)
retval = put_user(SIGCHLD, &infop->si_signo);
if (!retval && infop)
retval = put_user(0, &infop->si_errno);
if (!retval && infop) {
int why;
if ((status & 0x7f) == 0) {
why = CLD_EXITED;
status >>= 8;
} else {
why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
status &= 0x7f;
}
retval = put_user((short)why, &infop->si_code);
if (!retval)
retval = put_user(status, &infop->si_status);
}
if (!retval && infop)
retval = put_user(pid, &infop->si_pid);
if (!retval && infop)
retval = put_user(p->uid, &infop->si_uid);
wait_task_zombie: fix 2/3 races vs forget_original_parent() Two threads, T1 and T2. T2 ptraces P, and P is not a child of ptracer's thread group. P exits and goes to TASK_ZOMBIE. T1 does wait_task_zombie(P): P->exit_state = TASK_DEAD; ... read_unlock(&tasklist_lock); T2 does exit(), takes tasklist, forget_original_parent() does __ptrace_unlink(P) but doesn't call do_notify_parent(P) because p->exit_state == EXIT_DEAD. Now, P is not visible to our process: __ptrace_unlink() removed it from ->children. We should send notification to P->parent and release P if and only if SIGCHLD is ignored. And we have 3 bugs: 1. P->parent does do_wait() and gets -ECHILD (P is on ->parent->children, but its state is TASK_DEAD). 2. // wait_task_zombie() continues if (put_user(...)) { // TODO: is this safe? p->exit_state = EXIT_ZOMBIE; return; } we return without notification/release, task_struct leaked. Solution: ignore -EFAULT and proceed. It is an application's bug if we can't fill infop/stat_addr (in case of VM_FAULT_OOM we have much more problems). 3. // wait_task_zombie() continues if (p->real_parent != p->parent) { // Not taken, it was untraced'ed ... } release_task(p); we released the task which we shouldn't. Solution: check ->real_parent != ->parent before, under tasklist_lock, but use ptrace_unlink() instead of __ptrace_unlink() to check ->ptrace. This patch hopefully solves 2 and 3, the 1st bug will be fixed later, we need some cleanups in forget_original_parent/reparent_thread. However, the first race is very unlikely and not critical, so I hope it makes sense to fix 1 and 2 for now. 4. Small cleanup: don't "restore" EXIT_ZOMBIE unless we know we are not going to realease the child. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-17 14:26:58 +08:00
if (!retval)
retval = pid;
wait_task_zombie: fix 2/3 races vs forget_original_parent() Two threads, T1 and T2. T2 ptraces P, and P is not a child of ptracer's thread group. P exits and goes to TASK_ZOMBIE. T1 does wait_task_zombie(P): P->exit_state = TASK_DEAD; ... read_unlock(&tasklist_lock); T2 does exit(), takes tasklist, forget_original_parent() does __ptrace_unlink(P) but doesn't call do_notify_parent(P) because p->exit_state == EXIT_DEAD. Now, P is not visible to our process: __ptrace_unlink() removed it from ->children. We should send notification to P->parent and release P if and only if SIGCHLD is ignored. And we have 3 bugs: 1. P->parent does do_wait() and gets -ECHILD (P is on ->parent->children, but its state is TASK_DEAD). 2. // wait_task_zombie() continues if (put_user(...)) { // TODO: is this safe? p->exit_state = EXIT_ZOMBIE; return; } we return without notification/release, task_struct leaked. Solution: ignore -EFAULT and proceed. It is an application's bug if we can't fill infop/stat_addr (in case of VM_FAULT_OOM we have much more problems). 3. // wait_task_zombie() continues if (p->real_parent != p->parent) { // Not taken, it was untraced'ed ... } release_task(p); we released the task which we shouldn't. Solution: check ->real_parent != ->parent before, under tasklist_lock, but use ptrace_unlink() instead of __ptrace_unlink() to check ->ptrace. This patch hopefully solves 2 and 3, the 1st bug will be fixed later, we need some cleanups in forget_original_parent/reparent_thread. However, the first race is very unlikely and not critical, so I hope it makes sense to fix 1 and 2 for now. 4. Small cleanup: don't "restore" EXIT_ZOMBIE unless we know we are not going to realease the child. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-17 14:26:58 +08:00
if (traced) {
write_lock_irq(&tasklist_lock);
wait_task_zombie: fix 2/3 races vs forget_original_parent() Two threads, T1 and T2. T2 ptraces P, and P is not a child of ptracer's thread group. P exits and goes to TASK_ZOMBIE. T1 does wait_task_zombie(P): P->exit_state = TASK_DEAD; ... read_unlock(&tasklist_lock); T2 does exit(), takes tasklist, forget_original_parent() does __ptrace_unlink(P) but doesn't call do_notify_parent(P) because p->exit_state == EXIT_DEAD. Now, P is not visible to our process: __ptrace_unlink() removed it from ->children. We should send notification to P->parent and release P if and only if SIGCHLD is ignored. And we have 3 bugs: 1. P->parent does do_wait() and gets -ECHILD (P is on ->parent->children, but its state is TASK_DEAD). 2. // wait_task_zombie() continues if (put_user(...)) { // TODO: is this safe? p->exit_state = EXIT_ZOMBIE; return; } we return without notification/release, task_struct leaked. Solution: ignore -EFAULT and proceed. It is an application's bug if we can't fill infop/stat_addr (in case of VM_FAULT_OOM we have much more problems). 3. // wait_task_zombie() continues if (p->real_parent != p->parent) { // Not taken, it was untraced'ed ... } release_task(p); we released the task which we shouldn't. Solution: check ->real_parent != ->parent before, under tasklist_lock, but use ptrace_unlink() instead of __ptrace_unlink() to check ->ptrace. This patch hopefully solves 2 and 3, the 1st bug will be fixed later, we need some cleanups in forget_original_parent/reparent_thread. However, the first race is very unlikely and not critical, so I hope it makes sense to fix 1 and 2 for now. 4. Small cleanup: don't "restore" EXIT_ZOMBIE unless we know we are not going to realease the child. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-17 14:26:58 +08:00
/* We dropped tasklist, ptracer could die and untrace */
ptrace_unlink(p);
/*
* If this is not a detached task, notify the parent.
* If it's still not detached after that, don't release
* it now.
*/
if (!task_detached(p)) {
wait_task_zombie: fix 2/3 races vs forget_original_parent() Two threads, T1 and T2. T2 ptraces P, and P is not a child of ptracer's thread group. P exits and goes to TASK_ZOMBIE. T1 does wait_task_zombie(P): P->exit_state = TASK_DEAD; ... read_unlock(&tasklist_lock); T2 does exit(), takes tasklist, forget_original_parent() does __ptrace_unlink(P) but doesn't call do_notify_parent(P) because p->exit_state == EXIT_DEAD. Now, P is not visible to our process: __ptrace_unlink() removed it from ->children. We should send notification to P->parent and release P if and only if SIGCHLD is ignored. And we have 3 bugs: 1. P->parent does do_wait() and gets -ECHILD (P is on ->parent->children, but its state is TASK_DEAD). 2. // wait_task_zombie() continues if (put_user(...)) { // TODO: is this safe? p->exit_state = EXIT_ZOMBIE; return; } we return without notification/release, task_struct leaked. Solution: ignore -EFAULT and proceed. It is an application's bug if we can't fill infop/stat_addr (in case of VM_FAULT_OOM we have much more problems). 3. // wait_task_zombie() continues if (p->real_parent != p->parent) { // Not taken, it was untraced'ed ... } release_task(p); we released the task which we shouldn't. Solution: check ->real_parent != ->parent before, under tasklist_lock, but use ptrace_unlink() instead of __ptrace_unlink() to check ->ptrace. This patch hopefully solves 2 and 3, the 1st bug will be fixed later, we need some cleanups in forget_original_parent/reparent_thread. However, the first race is very unlikely and not critical, so I hope it makes sense to fix 1 and 2 for now. 4. Small cleanup: don't "restore" EXIT_ZOMBIE unless we know we are not going to realease the child. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-17 14:26:58 +08:00
do_notify_parent(p, p->exit_signal);
if (!task_detached(p)) {
wait_task_zombie: fix 2/3 races vs forget_original_parent() Two threads, T1 and T2. T2 ptraces P, and P is not a child of ptracer's thread group. P exits and goes to TASK_ZOMBIE. T1 does wait_task_zombie(P): P->exit_state = TASK_DEAD; ... read_unlock(&tasklist_lock); T2 does exit(), takes tasklist, forget_original_parent() does __ptrace_unlink(P) but doesn't call do_notify_parent(P) because p->exit_state == EXIT_DEAD. Now, P is not visible to our process: __ptrace_unlink() removed it from ->children. We should send notification to P->parent and release P if and only if SIGCHLD is ignored. And we have 3 bugs: 1. P->parent does do_wait() and gets -ECHILD (P is on ->parent->children, but its state is TASK_DEAD). 2. // wait_task_zombie() continues if (put_user(...)) { // TODO: is this safe? p->exit_state = EXIT_ZOMBIE; return; } we return without notification/release, task_struct leaked. Solution: ignore -EFAULT and proceed. It is an application's bug if we can't fill infop/stat_addr (in case of VM_FAULT_OOM we have much more problems). 3. // wait_task_zombie() continues if (p->real_parent != p->parent) { // Not taken, it was untraced'ed ... } release_task(p); we released the task which we shouldn't. Solution: check ->real_parent != ->parent before, under tasklist_lock, but use ptrace_unlink() instead of __ptrace_unlink() to check ->ptrace. This patch hopefully solves 2 and 3, the 1st bug will be fixed later, we need some cleanups in forget_original_parent/reparent_thread. However, the first race is very unlikely and not critical, so I hope it makes sense to fix 1 and 2 for now. 4. Small cleanup: don't "restore" EXIT_ZOMBIE unless we know we are not going to realease the child. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-17 14:26:58 +08:00
p->exit_state = EXIT_ZOMBIE;
p = NULL;
}
}
write_unlock_irq(&tasklist_lock);
}
if (p != NULL)
release_task(p);
wait_task_zombie: fix 2/3 races vs forget_original_parent() Two threads, T1 and T2. T2 ptraces P, and P is not a child of ptracer's thread group. P exits and goes to TASK_ZOMBIE. T1 does wait_task_zombie(P): P->exit_state = TASK_DEAD; ... read_unlock(&tasklist_lock); T2 does exit(), takes tasklist, forget_original_parent() does __ptrace_unlink(P) but doesn't call do_notify_parent(P) because p->exit_state == EXIT_DEAD. Now, P is not visible to our process: __ptrace_unlink() removed it from ->children. We should send notification to P->parent and release P if and only if SIGCHLD is ignored. And we have 3 bugs: 1. P->parent does do_wait() and gets -ECHILD (P is on ->parent->children, but its state is TASK_DEAD). 2. // wait_task_zombie() continues if (put_user(...)) { // TODO: is this safe? p->exit_state = EXIT_ZOMBIE; return; } we return without notification/release, task_struct leaked. Solution: ignore -EFAULT and proceed. It is an application's bug if we can't fill infop/stat_addr (in case of VM_FAULT_OOM we have much more problems). 3. // wait_task_zombie() continues if (p->real_parent != p->parent) { // Not taken, it was untraced'ed ... } release_task(p); we released the task which we shouldn't. Solution: check ->real_parent != ->parent before, under tasklist_lock, but use ptrace_unlink() instead of __ptrace_unlink() to check ->ptrace. This patch hopefully solves 2 and 3, the 1st bug will be fixed later, we need some cleanups in forget_original_parent/reparent_thread. However, the first race is very unlikely and not critical, so I hope it makes sense to fix 1 and 2 for now. 4. Small cleanup: don't "restore" EXIT_ZOMBIE unless we know we are not going to realease the child. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-17 14:26:58 +08:00
return retval;
}
/*
* Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
* read_lock(&tasklist_lock) on entry. If we return zero, we still hold
* the lock and this task is uninteresting. If we return nonzero, we have
* released the lock and the system call should return.
*/
static int wait_task_stopped(int ptrace, struct task_struct *p,
int options, struct siginfo __user *infop,
int __user *stat_addr, struct rusage __user *ru)
{
int retval, exit_code, why;
uid_t uid = 0; /* unneeded, required by compiler */
pid_t pid;
if (!(options & WUNTRACED))
return 0;
exit_code = 0;
spin_lock_irq(&p->sighand->siglock);
if (unlikely(!task_is_stopped_or_traced(p)))
goto unlock_sig;
if (!ptrace && p->signal->group_stop_count > 0)
/*
* A group stop is in progress and this is the group leader.
* We won't report until all threads have stopped.
*/
goto unlock_sig;
exit_code = p->exit_code;
if (!exit_code)
goto unlock_sig;
if (!unlikely(options & WNOWAIT))
p->exit_code = 0;
uid = p->uid;
unlock_sig:
spin_unlock_irq(&p->sighand->siglock);
if (!exit_code)
return 0;
/*
* Now we are pretty sure this task is interesting.
* Make sure it doesn't get reaped out from under us while we
* give up the lock and then examine it below. We don't want to
* keep holding onto the tasklist_lock while we call getrusage and
* possibly take page faults for user memory.
*/
get_task_struct(p);
pid = task_pid_vnr(p);
why = ptrace ? CLD_TRAPPED : CLD_STOPPED;
read_unlock(&tasklist_lock);
if (unlikely(options & WNOWAIT))
return wait_noreap_copyout(p, pid, uid,
why, exit_code,
infop, ru);
retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
if (!retval && stat_addr)
retval = put_user((exit_code << 8) | 0x7f, stat_addr);
if (!retval && infop)
retval = put_user(SIGCHLD, &infop->si_signo);
if (!retval && infop)
retval = put_user(0, &infop->si_errno);
if (!retval && infop)
retval = put_user((short)why, &infop->si_code);
if (!retval && infop)
retval = put_user(exit_code, &infop->si_status);
if (!retval && infop)
retval = put_user(pid, &infop->si_pid);
if (!retval && infop)
retval = put_user(uid, &infop->si_uid);
if (!retval)
retval = pid;
put_task_struct(p);
BUG_ON(!retval);
return retval;
}
/*
* Handle do_wait work for one task in a live, non-stopped state.
* read_lock(&tasklist_lock) on entry. If we return zero, we still hold
* the lock and this task is uninteresting. If we return nonzero, we have
* released the lock and the system call should return.
*/
static int wait_task_continued(struct task_struct *p, int options,
struct siginfo __user *infop,
int __user *stat_addr, struct rusage __user *ru)
{
int retval;
pid_t pid;
uid_t uid;
if (!unlikely(options & WCONTINUED))
return 0;
if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
return 0;
spin_lock_irq(&p->sighand->siglock);
/* Re-check with the lock held. */
if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
spin_unlock_irq(&p->sighand->siglock);
return 0;
}
if (!unlikely(options & WNOWAIT))
p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
spin_unlock_irq(&p->sighand->siglock);
pid = task_pid_vnr(p);
uid = p->uid;
get_task_struct(p);
read_unlock(&tasklist_lock);
if (!infop) {
retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
put_task_struct(p);
if (!retval && stat_addr)
retval = put_user(0xffff, stat_addr);
if (!retval)
retval = pid;
} else {
retval = wait_noreap_copyout(p, pid, uid,
CLD_CONTINUED, SIGCONT,
infop, ru);
BUG_ON(retval == 0);
}
return retval;
}
/*
* Consider @p for a wait by @parent.
*
* -ECHILD should be in *@notask_error before the first call.
* Returns nonzero for a final return, when we have unlocked tasklist_lock.
* Returns zero if the search for a child should continue;
* then *@notask_error is 0 if @p is an eligible child,
* or another error from security_task_wait(), or still -ECHILD.
*/
static int wait_consider_task(struct task_struct *parent, int ptrace,
struct task_struct *p, int *notask_error,
enum pid_type type, struct pid *pid, int options,
struct siginfo __user *infop,
int __user *stat_addr, struct rusage __user *ru)
{
int ret = eligible_child(type, pid, options, p);
if (!ret)
return ret;
if (unlikely(ret < 0)) {
/*
* If we have not yet seen any eligible child,
* then let this error code replace -ECHILD.
* A permission error will give the user a clue
* to look for security policy problems, rather
* than for mysterious wait bugs.
*/
if (*notask_error)
*notask_error = ret;
}
if (likely(!ptrace) && unlikely(p->ptrace)) {
/*
* This child is hidden by ptrace.
* We aren't allowed to see it now, but eventually we will.
*/
*notask_error = 0;
return 0;
}
if (p->exit_state == EXIT_DEAD)
return 0;
/*
* We don't reap group leaders with subthreads.
*/
if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
return wait_task_zombie(p, options, infop, stat_addr, ru);
/*
* It's stopped or running now, so it might
* later continue, exit, or stop again.
*/
*notask_error = 0;
if (task_is_stopped_or_traced(p))
return wait_task_stopped(ptrace, p, options,
infop, stat_addr, ru);
return wait_task_continued(p, options, infop, stat_addr, ru);
}
/*
* Do the work of do_wait() for one thread in the group, @tsk.
*
* -ECHILD should be in *@notask_error before the first call.
* Returns nonzero for a final return, when we have unlocked tasklist_lock.
* Returns zero if the search for a child should continue; then
* *@notask_error is 0 if there were any eligible children,
* or another error from security_task_wait(), or still -ECHILD.
*/
static int do_wait_thread(struct task_struct *tsk, int *notask_error,
enum pid_type type, struct pid *pid, int options,
struct siginfo __user *infop, int __user *stat_addr,
struct rusage __user *ru)
{
struct task_struct *p;
list_for_each_entry(p, &tsk->children, sibling) {
/*
* Do not consider detached threads.
*/
if (!task_detached(p)) {
int ret = wait_consider_task(tsk, 0, p, notask_error,
type, pid, options,
infop, stat_addr, ru);
if (ret)
return ret;
}
}
return 0;
}
static int ptrace_do_wait(struct task_struct *tsk, int *notask_error,
enum pid_type type, struct pid *pid, int options,
struct siginfo __user *infop, int __user *stat_addr,
struct rusage __user *ru)
{
struct task_struct *p;
/*
* Traditionally we see ptrace'd stopped tasks regardless of options.
*/
options |= WUNTRACED;
list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
int ret = wait_consider_task(tsk, 1, p, notask_error,
type, pid, options,
infop, stat_addr, ru);
if (ret)
return ret;
}
return 0;
}
static long do_wait(enum pid_type type, struct pid *pid, int options,
struct siginfo __user *infop, int __user *stat_addr,
struct rusage __user *ru)
{
DECLARE_WAITQUEUE(wait, current);
struct task_struct *tsk;
int retval;
trace_sched_process_wait(pid);
add_wait_queue(&current->signal->wait_chldexit,&wait);
repeat:
/*
* If there is nothing that can match our critiera just get out.
* We will clear @retval to zero if we see any child that might later
* match our criteria, even if we are not able to reap it yet.
*/
retval = -ECHILD;
if ((type < PIDTYPE_MAX) && (!pid || hlist_empty(&pid->tasks[type])))
goto end;
current->state = TASK_INTERRUPTIBLE;
read_lock(&tasklist_lock);
tsk = current;
do {
int tsk_result = do_wait_thread(tsk, &retval,
type, pid, options,
infop, stat_addr, ru);
if (!tsk_result)
tsk_result = ptrace_do_wait(tsk, &retval,
type, pid, options,
infop, stat_addr, ru);
if (tsk_result) {
/*
* tasklist_lock is unlocked and we have a final result.
*/
retval = tsk_result;
goto end;
}
if (options & __WNOTHREAD)
break;
tsk = next_thread(tsk);
BUG_ON(tsk->signal != current->signal);
} while (tsk != current);
read_unlock(&tasklist_lock);
if (!retval && !(options & WNOHANG)) {
retval = -ERESTARTSYS;
if (!signal_pending(current)) {
schedule();
goto repeat;
}
}
end:
current->state = TASK_RUNNING;
remove_wait_queue(&current->signal->wait_chldexit,&wait);
if (infop) {
if (retval > 0)
retval = 0;
else {
/*
* For a WNOHANG return, clear out all the fields
* we would set so the user can easily tell the
* difference.
*/
if (!retval)
retval = put_user(0, &infop->si_signo);
if (!retval)
retval = put_user(0, &infop->si_errno);
if (!retval)
retval = put_user(0, &infop->si_code);
if (!retval)
retval = put_user(0, &infop->si_pid);
if (!retval)
retval = put_user(0, &infop->si_uid);
if (!retval)
retval = put_user(0, &infop->si_status);
}
}
return retval;
}
asmlinkage long sys_waitid(int which, pid_t upid,
struct siginfo __user *infop, int options,
struct rusage __user *ru)
{
struct pid *pid = NULL;
enum pid_type type;
long ret;
if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
return -EINVAL;
if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
return -EINVAL;
switch (which) {
case P_ALL:
type = PIDTYPE_MAX;
break;
case P_PID:
type = PIDTYPE_PID;
if (upid <= 0)
return -EINVAL;
break;
case P_PGID:
type = PIDTYPE_PGID;
if (upid <= 0)
return -EINVAL;
break;
default:
return -EINVAL;
}
if (type < PIDTYPE_MAX)
pid = find_get_pid(upid);
ret = do_wait(type, pid, options, infop, NULL, ru);
put_pid(pid);
/* avoid REGPARM breakage on x86: */
asmlinkage_protect(5, ret, which, upid, infop, options, ru);
return ret;
}
asmlinkage long sys_wait4(pid_t upid, int __user *stat_addr,
int options, struct rusage __user *ru)
{
struct pid *pid = NULL;
enum pid_type type;
long ret;
if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
__WNOTHREAD|__WCLONE|__WALL))
return -EINVAL;
if (upid == -1)
type = PIDTYPE_MAX;
else if (upid < 0) {
type = PIDTYPE_PGID;
pid = find_get_pid(-upid);
} else if (upid == 0) {
type = PIDTYPE_PGID;
pid = get_pid(task_pgrp(current));
} else /* upid > 0 */ {
type = PIDTYPE_PID;
pid = find_get_pid(upid);
}
ret = do_wait(type, pid, options | WEXITED, NULL, stat_addr, ru);
put_pid(pid);
/* avoid REGPARM breakage on x86: */
asmlinkage_protect(4, ret, upid, stat_addr, options, ru);
return ret;
}
#ifdef __ARCH_WANT_SYS_WAITPID
/*
* sys_waitpid() remains for compatibility. waitpid() should be
* implemented by calling sys_wait4() from libc.a.
*/
asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
{
return sys_wait4(pid, stat_addr, options, NULL);
}
#endif