From nobody Mon Nov 25 02:51:02 2024 Received: from galois.linutronix.de (Galois.linutronix.de [193.142.43.55]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 5DA4D1C1AB3 for ; Thu, 31 Oct 2024 15:46:38 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=193.142.43.55 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1730389602; cv=none; b=APXMs9zDQE7HDr1rjL5A4ekkcIbUFz7qtoI1dMeIgcVYO3gPUgyyhSLjtGMLdqu5Z1lIF5SknuMRqOn8S3/wCHVw79MQ0939JfNNoyYsbzyPRBgMVk+EINli5AWB0SRJ7yZG26FkoEGaEKUoZqgAi5R+POLJrUkND+wlBcpt6IE= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1730389602; c=relaxed/simple; bh=enQyLOaHXZPmgh1pbWtFfEdPlnrVUAoE3+VZxJY+E64=; h=Message-ID:From:To:Cc:Subject:References:MIME-Version: Content-Type:Date; b=GCSf/gpc6z4zkD/hQ4ZVxCfHxtJ/L25q9Y25+pS1xJ1CC4bTbj/LsCeGr+/IjgYtMnuo9bXvsF5+IkTFz/8GsN9d87tZ0S3jgBrzpX8wwlruyGuicb7VE909OBGJ6Yz9tO9r1Hb+zTsWWm1uBiDU/SvL2tPxEDu9ddFET4HjItU= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=linutronix.de; spf=pass smtp.mailfrom=linutronix.de; dkim=pass (2048-bit key) header.d=linutronix.de header.i=@linutronix.de header.b=ePiAFA97; dkim=permerror (0-bit key) header.d=linutronix.de header.i=@linutronix.de header.b=hgv3WtXT; arc=none smtp.client-ip=193.142.43.55 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=linutronix.de Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=linutronix.de Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=linutronix.de header.i=@linutronix.de header.b="ePiAFA97"; dkim=permerror (0-bit key) header.d=linutronix.de header.i=@linutronix.de header.b="hgv3WtXT" Message-ID: <20241031154425.245095838@linutronix.de> DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linutronix.de; s=2020; t=1730389596; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: references:references; bh=Yyy8Ki41ZJBE+KVosF2Hc7+f2cRyz+UaB/PHCswwRHs=; b=ePiAFA97tcQ0fpK//4Xe7OR/IKl9PbpT5/qa3W3w2WddamxDl3b8+wCKe4NGukfYOhKHoY evyI5qVbnc3hGoozwuexPN2g6DCJuy/ITE+EG3SRQdsrYU+ymJg4aGZNOvGIxgElHeFzmU 42FcpwzJEEvuiFQaqWDiTELHjRDDEplr/yvKVAaw7JfTnNlFHU1rVU03bruGOkiKUA19Om 4TI8zFHbWJxKdqZLjIgwOq1R69bR9hC80xHAZ0V/d3Y6HNJb/i1RoPmuaM+tYkjVkFTnXk MbPAkMc9/VjF+JApqBbgGmDaRTtKwRq4f1VH6dJF2WzAKqLnCT8RuPevINaqzQ== DKIM-Signature: v=1; a=ed25519-sha256; c=relaxed/relaxed; d=linutronix.de; s=2020e; t=1730389596; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: references:references; bh=Yyy8Ki41ZJBE+KVosF2Hc7+f2cRyz+UaB/PHCswwRHs=; b=hgv3WtXTsoA7D6zvpTmAU+8cfSc4FMa08p67ckNqLFzDqKoyiqwzVaQm/L5dxTVQWGq2E/ F5GphGwkZ5wY2IAw== From: Thomas Gleixner To: LKML Cc: Anna-Maria Behnsen , Frederic Weisbecker , John Stultz , Peter Zijlstra , Ingo Molnar , Stephen Boyd , Eric Biederman , Oleg Nesterov Subject: [patch v6 11/20] posix-timers: Embed sigqueue in struct k_itimer References: <20241031151625.361697424@linutronix.de> Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Date: Thu, 31 Oct 2024 16:46:36 +0100 (CET) Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" From: Thomas Gleixner To cure the SIG_IGN handling for posix interval timers, the preallocated sigqueue needs to be embedded into struct k_itimer to prevent life time races of all sorts. Now that the prerequisites are in place, embed the sigqueue into struct k_itimer and fixup the relevant usage sites. Aside of preparing for proper SIG_IGN handling, this spares an extra allocation. Signed-off-by: Thomas Gleixner Reviewed-by: Frederic Weisbecker Acked-by: Peter Zijlstra (Intel) --- V6: Split out the timer related part from the delivery function to use guard() and avoid gotos --- fs/proc/base.c | 4 - include/linux/posix-timers.h | 23 ++++++++++- kernel/signal.c | 19 +++++---- kernel/time/posix-timers.c | 88 +++++++++++++++++++++++++-------------= ----- 4 files changed, 87 insertions(+), 47 deletions(-) --- --- a/fs/proc/base.c +++ b/fs/proc/base.c @@ -2553,8 +2553,8 @@ static int show_timer(struct seq_file *m =20 seq_printf(m, "ID: %d\n", timer->it_id); seq_printf(m, "signal: %d/%px\n", - timer->sigq->info.si_signo, - timer->sigq->info.si_value.sival_ptr); + 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], (notify & SIGEV_THREAD_ID) ? "tid" : "pid", --- a/include/linux/posix-timers.h +++ b/include/linux/posix-timers.h @@ -39,6 +39,8 @@ static inline int clockid_to_fd(const cl =20 #ifdef CONFIG_POSIX_TIMERS =20 +#include + /** * cpu_timer - Posix CPU timer representation for k_itimer * @node: timerqueue node to queue in the task/sig @@ -166,7 +168,7 @@ static inline void posix_cputimers_init_ * @it_pid: The pid of the process/task targeted by the signal * @it_process: The task to wakeup on clock_nanosleep (CPU timers) * @rcuref: Reference count for life time management - * @sigq: Pointer to preallocated sigqueue + * @sigq: Embedded sigqueue * @it: Union representing the various posix timer type * internals. * @rcu: RCU head for freeing the timer. @@ -190,7 +192,7 @@ struct k_itimer { struct pid *it_pid; struct task_struct *it_process; }; - struct sigqueue *sigq; + struct sigqueue sigq; rcuref_t rcuref; union { struct { @@ -218,6 +220,23 @@ static inline void posixtimer_putref(str if (rcuref_put(&tmr->rcuref)) posixtimer_free_timer(tmr); } + +static inline void posixtimer_sigqueue_getref(struct sigqueue *q) +{ + struct k_itimer *tmr =3D container_of(q, struct k_itimer, sigq); + + WARN_ON_ONCE(!rcuref_get(&tmr->rcuref)); +} + +static inline void posixtimer_sigqueue_putref(struct sigqueue *q) +{ + struct k_itimer *tmr =3D container_of(q, struct k_itimer, sigq); + + posixtimer_putref(tmr); +} +#else /* CONFIG_POSIX_TIMERS */ +static inline void posixtimer_sigqueue_getref(struct sigqueue *q) { } +static inline void posixtimer_sigqueue_putref(struct sigqueue *q) { } #endif /* !CONFIG_POSIX_TIMERS */ =20 #endif --- a/kernel/signal.c +++ b/kernel/signal.c @@ -460,8 +460,10 @@ static struct sigqueue *__sigqueue_alloc =20 static void __sigqueue_free(struct sigqueue *q) { - if (q->flags & SIGQUEUE_PREALLOC) + if (q->flags & SIGQUEUE_PREALLOC) { + posixtimer_sigqueue_putref(q); return; + } if (q->ucounts) { dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING); q->ucounts =3D NULL; @@ -569,11 +571,11 @@ static void collect_signal(int sig, stru copy_siginfo(info, &first->info); =20 /* - * posix-timer signals are preallocated and freed when the - * timer goes away. Either directly or by clearing - * SIGQUEUE_PREALLOC so that the next delivery will free - * them. Spare the extra round through __sigqueue_free() - * which is ignoring preallocated signals. + * posix-timer signals are preallocated and freed when the last + * reference count is dropped in posixtimer_deliver_signal() or + * immediately on timer deletion when the signal is not pending. + * Spare the extra round through __sigqueue_free() which is + * ignoring preallocated signals. */ if (unlikely((first->flags & SIGQUEUE_PREALLOC) && (info->si_code =3D=3D= SI_TIMER))) *timer_sigq =3D first; @@ -1989,7 +1991,7 @@ static inline struct task_struct *posixt =20 int posixtimer_send_sigqueue(struct k_itimer *tmr) { - struct sigqueue *q =3D tmr->sigq; + struct sigqueue *q =3D &tmr->sigq; int sig =3D q->info.si_signo; struct task_struct *t; unsigned long flags; @@ -2020,9 +2022,12 @@ int posixtimer_send_sigqueue(struct k_it =20 ret =3D 0; if (unlikely(!list_empty(&q->list))) { + /* This holds a reference count already */ result =3D TRACE_SIGNAL_ALREADY_PENDING; goto out; } + + posixtimer_sigqueue_getref(q); posixtimer_queue_sigqueue(q, t, tmr->it_pid_type); result =3D TRACE_SIGNAL_DELIVERED; out: --- a/kernel/time/posix-timers.c +++ b/kernel/time/posix-timers.c @@ -250,15 +250,40 @@ static void common_hrtimer_rearm(struct hrtimer_restart(timer); } =20 +static bool __posixtimer_deliver_signal(struct kernel_siginfo *info, struc= t k_itimer *timr) +{ + guard(spinlock)(&timr->it_lock); + + /* + * Check if the timer is still alive or whether it got modified + * since the signal was queued. In either case, don't rearm and + * drop the signal. + */ + if (timr->it_signal_seq !=3D info->si_sys_private || WARN_ON_ONCE(!timr->= it_signal)) + return false; + + if (!timr->it_interval || WARN_ON_ONCE(timr->it_status !=3D POSIX_TIMER_R= EQUEUE_PENDING)) + return true; + + timr->kclock->timer_rearm(timr); + timr->it_status =3D POSIX_TIMER_ARMED; + timr->it_overrun_last =3D timr->it_overrun; + timr->it_overrun =3D -1LL; + ++timr->it_signal_seq; + info->si_overrun =3D timer_overrun_to_int(timr); + return true; +} + /* * This function is called from the signal delivery code. It decides - * whether the signal should be dropped and rearms interval timers. + * whether the signal should be dropped and rearms interval timers. The + * timer can be unconditionally accessed as there is a reference held on + * it. */ bool posixtimer_deliver_signal(struct kernel_siginfo *info, struct sigqueu= e *timer_sigq) { - struct k_itimer *timr; - unsigned long flags; - bool ret =3D false; + struct k_itimer *timr =3D container_of(timer_sigq, struct k_itimer, sigq); + bool ret; =20 /* * Release siglock to ensure proper locking order versus @@ -266,28 +291,11 @@ bool posixtimer_deliver_signal(struct ke */ spin_unlock(¤t->sighand->siglock); =20 - timr =3D lock_timer(info->si_tid, &flags); - if (!timr) - goto out; - - if (timr->it_signal_seq !=3D info->si_sys_private) - goto out_unlock; - - if (timr->it_interval && !WARN_ON_ONCE(timr->it_status !=3D POSIX_TIMER_R= EQUEUE_PENDING)) { - timr->kclock->timer_rearm(timr); + ret =3D __posixtimer_deliver_signal(info, timr); =20 - timr->it_status =3D POSIX_TIMER_ARMED; - timr->it_overrun_last =3D timr->it_overrun; - timr->it_overrun =3D -1LL; - ++timr->it_signal_seq; - - info->si_overrun =3D timer_overrun_to_int(timr); - } - ret =3D true; + /* Drop the reference which was acquired when the signal was queued */ + posixtimer_putref(timr); =20 -out_unlock: - unlock_timer(timr, flags); -out: spin_lock(¤t->sighand->siglock); =20 /* Don't expose the si_sys_private value to userspace */ @@ -404,17 +412,17 @@ static struct pid *good_sigevent(sigeven } } =20 -static struct k_itimer * alloc_posix_timer(void) +static struct k_itimer *alloc_posix_timer(void) { struct k_itimer *tmr =3D kmem_cache_zalloc(posix_timers_cache, GFP_KERNEL= ); =20 if (!tmr) return tmr; - if (unlikely(!(tmr->sigq =3D sigqueue_alloc()))) { + + if (unlikely(!posixtimer_init_sigqueue(&tmr->sigq))) { kmem_cache_free(posix_timers_cache, tmr); return NULL; } - clear_siginfo(&tmr->sigq->info); rcuref_init(&tmr->rcuref, 1); return tmr; } @@ -422,7 +430,8 @@ static struct k_itimer * alloc_posix_tim void posixtimer_free_timer(struct k_itimer *tmr) { put_pid(tmr->it_pid); - sigqueue_free(tmr->sigq); + if (tmr->sigq.ucounts) + dec_rlimit_put_ucounts(tmr->sigq.ucounts, UCOUNT_RLIMIT_SIGPENDING); kfree_rcu(tmr, rcu); } =20 @@ -484,13 +493,13 @@ static int do_timer_create(clockid_t whi goto out; } new_timer->it_sigev_notify =3D event->sigev_notify; - new_timer->sigq->info.si_signo =3D event->sigev_signo; - new_timer->sigq->info.si_value =3D event->sigev_value; + new_timer->sigq.info.si_signo =3D event->sigev_signo; + new_timer->sigq.info.si_value =3D event->sigev_value; } else { new_timer->it_sigev_notify =3D SIGEV_SIGNAL; - new_timer->sigq->info.si_signo =3D SIGALRM; - memset(&new_timer->sigq->info.si_value, 0, sizeof(sigval_t)); - new_timer->sigq->info.si_value.sival_int =3D new_timer->it_id; + new_timer->sigq.info.si_signo =3D SIGALRM; + memset(&new_timer->sigq.info.si_value, 0, sizeof(sigval_t)); + new_timer->sigq.info.si_value.sival_int =3D new_timer->it_id; new_timer->it_pid =3D get_pid(task_tgid(current)); } =20 @@ -499,8 +508,8 @@ static int do_timer_create(clockid_t whi else new_timer->it_pid_type =3D PIDTYPE_TGID; =20 - new_timer->sigq->info.si_tid =3D new_timer->it_id; - new_timer->sigq->info.si_code =3D SI_TIMER; + new_timer->sigq.info.si_tid =3D new_timer->it_id; + new_timer->sigq.info.si_code =3D SI_TIMER; =20 if (copy_to_user(created_timer_id, &new_timer_id, sizeof (new_timer_id)))= { error =3D -EFAULT; @@ -584,7 +593,14 @@ static struct k_itimer *__lock_timer(tim * 1) Set timr::it_signal to NULL with timr::it_lock held * 2) Release timr::it_lock * 3) Remove from the hash under hash_lock - * 4) Call RCU for removal after the grace period + * 4) Put the reference count. + * + * The reference count might not drop to zero if timr::sigq is + * queued. In that case the signal delivery or flush will put the + * last reference count. + * + * When the reference count reaches zero, the timer is scheduled + * for RCU removal after the grace period. * * Holding rcu_read_lock() accross the lookup ensures that * the timer cannot be freed.