summaryrefslogtreecommitdiff
path: root/fs/proc/base.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2025-03-25 10:33:23 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2025-03-25 10:33:23 -0700
commitd5048d1176b8e76e687fc145df785118424e1ec2 (patch)
tree69d905e72c3fd88c8bdbae14a3b29ac9f6211e37 /fs/proc/base.c
parent0ae2062ee3ebf11981f9bdb6198fa734d8c0f46c (diff)
parente40d3709c0225f5f681fd300f65a65ac63b10f83 (diff)
Merge tag 'timers-core-2025-03-23' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull timer core updates from Thomas Gleixner: - Fix a memory ordering issue in posix-timers Posix-timer lookup is lockless and reevaluates the timer validity under the timer lock, but the update which validates the timer is not protected by the timer lock. That allows the store to be reordered against the initialization stores, so that the lookup side can observe a partially initialized timer. That's mostly a theoretical problem, but incorrect nevertheless. - Fix a long standing inconsistency of the coarse time getters The coarse time getters read the base time of the current update cycle without reading the actual hardware clock. NTP frequency adjustment can set the base time backwards. The fine grained interfaces compensate this by reading the clock and applying the new conversion factor, but the coarse grained time getters use the base time directly. That allows the user to observe time going backwards. Cure it by always forwarding base time, when NTP changes the frequency with an immediate step. - Rework of posix-timer hashing The posix-timer hash is not scalable and due to the CRIU timer restore mechanism prone to massive contention on the global hash bucket lock. Replace the global hash lock with a fine grained per bucket locking scheme to address that. - Rework the proc/$PID/timers interface. /proc/$PID/timers is provided for CRIU to be able to restore a timer. The printout happens with sighand lock held and interrupts disabled. That's not required as this can be done with RCU protection as well. - Provide a sane mechanism for CRIU to restore a timer ID CRIU restores timers by creating and deleting them until the kernel internal per process ID counter reached the requested ID. That's horribly slow for sparse timer IDs. Provide a prctl() which allows CRIU to restore a timer with a given ID. When enabled the ID pointer is used as input pointer to read the requested ID from user space. When disabled, the normal allocation scheme (next ID) is active as before. This is backwards compatible for both kernel and user space. - Make hrtimer_update_function() less expensive. The sanity checks are valuable, but expensive for high frequency usage in io/uring. Make the debug checks conditional and enable them only when lockdep is enabled. - Small updates, cleanups and improvements * tag 'timers-core-2025-03-23' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (27 commits) selftests/timers: Improve skew_consistency by testing with other clockids timekeeping: Fix possible inconsistencies in _COARSE clockids posix-timers: Drop redundant memset() invocation selftests/timers/posix-timers: Add a test for exact allocation mode posix-timers: Provide a mechanism to allocate a given timer ID posix-timers: Dont iterate /proc/$PID/timers with sighand:: Siglock held posix-timers: Make per process list RCU safe posix-timers: Avoid false cacheline sharing posix-timers: Switch to jhash32() posix-timers: Improve hash table performance posix-timers: Make signal_struct:: Next_posix_timer_id an atomic_t posix-timers: Make lock_timer() use guard() posix-timers: Rework timer removal posix-timers: Simplify lock/unlock_timer() posix-timers: Use guards in a few places posix-timers: Remove SLAB_PANIC from kmem cache posix-timers: Remove a few paranoid warnings posix-timers: Cleanup includes posix-timers: Add cond_resched() to posix_timer_add() search loop posix-timers: Initialise timer before adding it to the hash table ...
Diffstat (limited to 'fs/proc/base.c')
-rw-r--r--fs/proc/base.c48
1 files changed, 20 insertions, 28 deletions
diff --git a/fs/proc/base.c b/fs/proc/base.c
index 61526420d0ee..5538c4aee8fa 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -2494,11 +2494,9 @@ static const struct file_operations proc_map_files_operations = {
#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
struct timers_private {
- struct pid *pid;
- struct task_struct *task;
- struct sighand_struct *sighand;
- struct pid_namespace *ns;
- unsigned long flags;
+ struct pid *pid;
+ struct task_struct *task;
+ struct pid_namespace *ns;
};
static void *timers_start(struct seq_file *m, loff_t *pos)
@@ -2509,54 +2507,48 @@ static void *timers_start(struct seq_file *m, loff_t *pos)
if (!tp->task)
return ERR_PTR(-ESRCH);
- tp->sighand = lock_task_sighand(tp->task, &tp->flags);
- if (!tp->sighand)
- return ERR_PTR(-ESRCH);
-
- return seq_hlist_start(&tp->task->signal->posix_timers, *pos);
+ rcu_read_lock();
+ return seq_hlist_start_rcu(&tp->task->signal->posix_timers, *pos);
}
static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
{
struct timers_private *tp = m->private;
- return seq_hlist_next(v, &tp->task->signal->posix_timers, pos);
+
+ return seq_hlist_next_rcu(v, &tp->task->signal->posix_timers, pos);
}
static void timers_stop(struct seq_file *m, void *v)
{
struct timers_private *tp = m->private;
- if (tp->sighand) {
- unlock_task_sighand(tp->task, &tp->flags);
- tp->sighand = NULL;
- }
-
if (tp->task) {
put_task_struct(tp->task);
tp->task = NULL;
+ rcu_read_unlock();
}
}
static int show_timer(struct seq_file *m, void *v)
{
- struct k_itimer *timer;
- struct timers_private *tp = m->private;
- int notify;
static const char * const nstr[] = {
- [SIGEV_SIGNAL] = "signal",
- [SIGEV_NONE] = "none",
- [SIGEV_THREAD] = "thread",
+ [SIGEV_SIGNAL] = "signal",
+ [SIGEV_NONE] = "none",
+ [SIGEV_THREAD] = "thread",
};
- timer = hlist_entry((struct hlist_node *)v, struct k_itimer, list);
- notify = timer->it_sigev_notify;
+ struct k_itimer *timer = hlist_entry((struct hlist_node *)v, struct k_itimer, list);
+ struct timers_private *tp = m->private;
+ int notify = timer->it_sigev_notify;
+
+ guard(spinlock_irq)(&timer->it_lock);
+ if (!posixtimer_valid(timer))
+ return 0;
seq_printf(m, "ID: %d\n", timer->it_id);
- seq_printf(m, "signal: %d/%px\n",
- timer->sigq.info.si_signo,
+ seq_printf(m, "signal: %d/%px\n", timer->sigq.info.si_signo,
timer->sigq.info.si_value.sival_ptr);
- seq_printf(m, "notify: %s/%s.%d\n",
- nstr[notify & ~SIGEV_THREAD_ID],
+ seq_printf(m, "notify: %s/%s.%d\n", nstr[notify & ~SIGEV_THREAD_ID],
(notify & SIGEV_THREAD_ID) ? "tid" : "pid",
pid_nr_ns(timer->it_pid, tp->ns));
seq_printf(m, "ClockID: %d\n", timer->it_clock);