| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Wed, 20 Nov 2019 10:30:05 +0900
From: Sergey Senozhatsky <sergey.senozhatsky@...il.com>
To: Petr Mladek <pmladek@...e.com>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>,
Qian Cai <cai@....pw>, Steven Rostedt <rostedt@...dmis.org>,
Sergey Senozhatsky <sergey.senozhatsky@...il.com>,
Michal Hocko <mhocko@...nel.org>,
Eric Dumazet <eric.dumazet@...il.com>, davem@...emloft.net,
netdev@...r.kernel.org, linux-mm@...ck.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] net/skbuff: silence warnings under memory pressure
On (19/11/19 10:41), Petr Mladek wrote:
[..]
> > > I do not like this. As a result, normal printk() will always deadlock
> > > in the scheduler code, including WARN() calls. The chance of the
> > > deadlock is small now. It happens only when there is another
> > > process waiting for console_sem.
> >
> > Why would it *always* deadlock? If this is the case, why we don't *always*
> > deadlock doing the very same wake_up_process() from console_unlock()?
>
> I speak about _normal_ printk() and not about printk_deferred().
>
> wake_up_process() is called in console_unlock() only when
> sem->wait_list is not empty, see up() in kernel/locking/semaphore.c.
> printk() itself uses console_trylock() and does not wait.
Sorry, I should have spent more time on this. So here's what I should
have written instead.
console_sem is not for kthreads or kernel only. User space apps can sleep
on it. And on some setups console_sem becomes really congested, with often
times non-empty console_sem->wait_list. All because of user space, we don't
even need printk()-s. For example, when user space uses tty for I/O.
Significantly trimmed logs from my laptop (a simple patch below). Just
to show that user space sleeps on console_sem (IOW, non-empty ->wait_list).
Notice - ls, bash, journalctl, etc. All competing for console_sem.
[..]
printk: >>> bash WOKEUP console_sem() waiter
printk: task journalctl couldn't lock console_sem and had to schedule out
Call Trace:
dump_stack+0x50/0x70
console_lock.cold+0x21/0x26
con_write+0x28/0x50
n_tty_write+0x12f/0x4c0
? wait_woken+0xa0/0xa0
tty_write+0x1a9/0x2f0
? n_tty_read+0x930/0x930
vfs_write+0xb7/0x1e0
printk: >>> ls WOKEUP console_sem() waiter
printk: task kworker/3:1 couldn't lock console_sem and had to schedule out
Workqueue: events console_callback
Call Trace:
dump_stack+0x50/0x70
console_lock.cold+0x21/0x26
console_callback+0xa/0x150
? _cond_resched+0x15/0x50
process_one_work+0x18b/0x2e0
worker_thread+0x4e/0x3d0
? process_one_work+0x2e0/0x2e0
kthread+0x119/0x130
printk: >>> kworker/3:1 WOKEUP console_sem() waiter
printk: task ls couldn't lock console_sem and had to schedule out
Call Trace:
dump_stack+0x50/0x70
console_lock.cold+0x21/0x26
con_write+0x28/0x50
do_output_char+0x191/0x1f0
n_tty_write+0x17e/0x4c0
? __switch_to_asm+0x24/0x60
? wait_woken+0xa0/0xa0
tty_write+0x1a9/0x2f0
? n_tty_read+0x930/0x930
vfs_write+0xb7/0x1e0
[..]
> I believe that this is the rason why printk_sched() was added
> so late in 2012. It was more than 10 years after adding
> the semaphore into console_unlock(). IMHO, the deadlock
> was rare. Of course, it was also hard to debug but it
> would not take 10 years.
Right. I also think scheduler people do pretty nice work avoiding printk
calls under ->rq lock.
What I tried to say - it's really not that hard to have a non-empty
console_sem ->wait_list, any "wrong" printk() call from scheduler
will deadlock us, because we have something to wake_up().
///////////////////////////////////////////////////////////////////////
// A simple patch
---
diff --git a/include/linux/semaphore.h b/include/linux/semaphore.h
index 6694d0019a68..c80a631f5f96 100644
--- a/include/linux/semaphore.h
+++ b/include/linux/semaphore.h
@@ -35,11 +35,11 @@ static inline void sema_init(struct semaphore *sem, int val)
lockdep_init_map(&sem->lock.dep_map, "semaphore->lock", &__key, 0);
}
-extern void down(struct semaphore *sem);
+extern bool down(struct semaphore *sem);
extern int __must_check down_interruptible(struct semaphore *sem);
extern int __must_check down_killable(struct semaphore *sem);
extern int __must_check down_trylock(struct semaphore *sem);
extern int __must_check down_timeout(struct semaphore *sem, long jiffies);
-extern void up(struct semaphore *sem);
+extern bool up(struct semaphore *sem);
#endif /* __LINUX_SEMAPHORE_H */
diff --git a/kernel/locking/semaphore.c b/kernel/locking/semaphore.c
index d9dd94defc0a..daefcbe08147 100644
--- a/kernel/locking/semaphore.c
+++ b/kernel/locking/semaphore.c
@@ -50,16 +50,20 @@ static noinline void __up(struct semaphore *sem);
* Use of this function is deprecated, please use down_interruptible() or
* down_killable() instead.
*/
-void down(struct semaphore *sem)
+bool down(struct semaphore *sem)
{
unsigned long flags;
+ bool ret = false;
raw_spin_lock_irqsave(&sem->lock, flags);
if (likely(sem->count > 0))
sem->count--;
- else
+ else {
__down(sem);
+ ret = true;
+ }
raw_spin_unlock_irqrestore(&sem->lock, flags);
+ return ret;
}
EXPORT_SYMBOL(down);
@@ -175,16 +179,20 @@ EXPORT_SYMBOL(down_timeout);
* Release the semaphore. Unlike mutexes, up() may be called from any
* context and even by tasks which have never called down().
*/
-void up(struct semaphore *sem)
+bool up(struct semaphore *sem)
{
unsigned long flags;
+ bool ret = false;
raw_spin_lock_irqsave(&sem->lock, flags);
- if (likely(list_empty(&sem->wait_list)))
+ if (likely(list_empty(&sem->wait_list))) {
sem->count++;
- else
+ } else {
__up(sem);
+ ret = true;
+ }
raw_spin_unlock_irqrestore(&sem->lock, flags);
+ return ret;
}
EXPORT_SYMBOL(up);
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index ca65327a6de8..42ffaaa01859 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -214,12 +214,21 @@ int devkmsg_sysctl_set_loglvl(struct ctl_table *table, int write,
/* Number of registered extended console drivers. */
static int nr_ext_console_drivers;
+static void ____down(void)
+{
+ bool sched_out = down(&console_sem);
+ if (sched_out) {
+ pr_err("task %s couldn't lock console_sem and had to schedule out\n", current->comm);
+ dump_stack();
+ }
+}
+
/*
* Helper macros to handle lockdep when locking/unlocking console_sem. We use
* macros instead of functions so that _RET_IP_ contains useful information.
*/
#define down_console_sem() do { \
- down(&console_sem);\
+ ____down(); \
mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\
} while (0)
@@ -244,15 +253,19 @@ static int __down_trylock_console_sem(unsigned long ip)
}
#define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_)
-static void __up_console_sem(unsigned long ip)
+static bool __up_console_sem(unsigned long ip)
{
unsigned long flags;
+ bool ret;
mutex_release(&console_lock_dep_map, 1, ip);
printk_safe_enter_irqsave(flags);
- up(&console_sem);
+ ret = up(&console_sem);
+ if (ret)
+ pr_err(">>> %s WOKEUP console_sem() waiter\n", current->comm);
printk_safe_exit_irqrestore(flags);
+ return ret;
}
#define up_console_sem() __up_console_sem(_RET_IP_)
Powered by blists - more mailing lists