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
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 11 Jan 2024 03:22:05 -0800
From: "Paul E. McKenney" <paulmck@...nel.org>
To: Matthew Wilcox <willy@...radead.org>
Cc: Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
	Will Deacon <will@...nel.org>, Waiman Long <longman@...hat.com>,
	linux-kernel@...r.kernel.org,
	Suren Baghdasaryan <surenb@...gle.com>,
	"Liam R. Howlett" <liam.howlett@...cle.com>
Subject: Re: [RFC] Sleep waiting for an rwsem to be unlocked

On Tue, Jan 09, 2024 at 05:12:06PM +0000, Matthew Wilcox wrote:
> The problem we're trying to solve is a lock-free walk of
> /proc/$pid/maps. If the process is modifying the VMAs at the same time
> the reader is walking them, it can see garbage.  For page faults, we
> handle this by taking the mmap_lock for read and retrying the page fault
> (excluding any further modifications).
> 
> We don't want to take that approach for the maps file.  The monitoring
> task may have a significantly lower process priority, and so taking
> the mmap_lock for read can block it for a significant period of time.
> The obvious answer is to do some kind of backoff+sleep.  But we already
> have a wait queue, so why not use it?
> 
> I haven't done the rwbase version; this is just a demonstration of what
> we could do.  It's also untested other than by compilation.  It might
> well be missing something.
> 
> Signed-off-by: Matthew Wilcox (Oracle) <willy@...radead.org>

At first glance, this is good and sufficient for this use case.

I do have one question that would be important if anyone were to want
to rely on the "This is equivalent to calling down_read(); up_read()"
statement in the header comment, please see below.

							Thanx, Paul

> ---
>  include/linux/rwsem.h  |   6 +++
>  kernel/locking/rwsem.c | 104 ++++++++++++++++++++++++++++++++++++++++-
>  2 files changed, 108 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/rwsem.h b/include/linux/rwsem.h
> index 4f1c18992f76..e7bf9dfc471a 100644
> --- a/include/linux/rwsem.h
> +++ b/include/linux/rwsem.h
> @@ -250,6 +250,12 @@ DEFINE_GUARD_COND(rwsem_write, _try, down_write_trylock(_T))
>   */
>  extern void downgrade_write(struct rw_semaphore *sem);
>  
> +/*
> + * wait for current writer to be finished
> + */
> +void rwsem_wait(struct rw_semaphore *sem);
> +int __must_check rwsem_wait_killable(struct rw_semaphore *sem);
> +
>  #ifdef CONFIG_DEBUG_LOCK_ALLOC
>  /*
>   * nested locking. NOTE: rwsems are not allowed to recurse
> diff --git a/kernel/locking/rwsem.c b/kernel/locking/rwsem.c
> index 2340b6d90ec6..7c8096c5586f 100644
> --- a/kernel/locking/rwsem.c
> +++ b/kernel/locking/rwsem.c
> @@ -332,7 +332,8 @@ EXPORT_SYMBOL(__init_rwsem);
>  
>  enum rwsem_waiter_type {
>  	RWSEM_WAITING_FOR_WRITE,
> -	RWSEM_WAITING_FOR_READ
> +	RWSEM_WAITING_FOR_READ,
> +	RWSEM_WAITING_FOR_RELEASE,
>  };
>  
>  struct rwsem_waiter {
> @@ -511,7 +512,8 @@ static void rwsem_mark_wake(struct rw_semaphore *sem,
>  		if (waiter->type == RWSEM_WAITING_FOR_WRITE)
>  			continue;
>  
> -		woken++;
> +		if (waiter->type == RWSEM_WAITING_FOR_READ)
> +			woken++;
>  		list_move_tail(&waiter->list, &wlist);
>  
>  		/*
> @@ -1401,6 +1403,67 @@ static inline void __downgrade_write(struct rw_semaphore *sem)
>  	preempt_enable();
>  }
>  
> +static inline int __wait_read_common(struct rw_semaphore *sem, int state)
> +{
> +	int ret = 0;
> +	long adjustment = 0;
> +	struct rwsem_waiter waiter;
> +	DEFINE_WAKE_Q(wake_q);
> +
> +	waiter.task = current;
> +	waiter.type = RWSEM_WAITING_FOR_RELEASE;
> +	waiter.timeout = jiffies + RWSEM_WAIT_TIMEOUT;
> +	waiter.handoff_set = false;
> +
> +	preempt_disable();
> +	raw_spin_lock_irq(&sem->wait_lock);
> +	if (list_empty(&sem->wait_list)) {
> +		if (!(atomic_long_read(&sem->count) & RWSEM_WRITER_MASK)) {
> +			/* Provide lock ACQUIRE */
> +			smp_acquire__after_ctrl_dep();
> +			raw_spin_unlock_irq(&sem->wait_lock);
> +			goto done;

If we take this path, we are ordered against the prior writer's release
courtesy of the acquire ordering on ->count.  But we are not ordered
against the next writer's acquisition if that writer takes the fastpath
because rwsem_write_trylock() only does acquire semantics.

Again, this does not matter for your use case, and it all just works on
strongly ordered systems such as x86.

Assuming I am not just confused here, as far as I am concerned, this could
be fixed by adjusting the guarantees in the rwsem_wait_killable() function's
header comment.

But it might be good to avoid the sharp edges that would be provided
by weakening that guarantee.

To that end, I -think- that a fix that would save that header
comment's current wording would insert an smp_mb() before the above
atomic_long_read(), but I could easily be wrong.  Plus there might well
need to be similar adjustments later in the code.  (I don't immediately
see any, but it has been a good long while since I have stared at
this code.)

Thoughts from people more familiar with this code?

> +		}
> +		adjustment = RWSEM_FLAG_WAITERS;
> +	}
> +	rwsem_add_waiter(sem, &waiter);
> +	if (adjustment) {
> +		long count = atomic_long_add_return(adjustment, &sem->count);
> +		rwsem_cond_wake_waiter(sem, count, &wake_q);
> +	}
> +	raw_spin_unlock_irq(&sem->wait_lock);
> +
> +	if (!wake_q_empty(&wake_q))
> +		wake_up_q(&wake_q);
> +
> +	for (;;) {
> +		set_current_state(state);
> +		if (!smp_load_acquire(&waiter.task)) {
> +			/* Matches rwsem_mark_wake()'s smp_store_release(). */
> +			break;
> +		}
> +		if (signal_pending_state(state, current)) {
> +			raw_spin_lock_irq(&sem->wait_lock);
> +			if (waiter.task)
> +				goto out_nolock;
> +			raw_spin_unlock_irq(&sem->wait_lock);
> +			/* Ordered by sem->wait_lock against rwsem_mark_wake(). */
> +			break;
> +		}
> +		schedule_preempt_disabled();
> +	}
> +
> +	__set_current_state(TASK_RUNNING);
> +done:
> +	preempt_enable();
> +	return ret;
> +out_nolock:
> +	rwsem_del_wake_waiter(sem, &waiter, &wake_q);
> +	__set_current_state(TASK_RUNNING);
> +	ret = -EINTR;
> +	goto done;
> +}
> +
>  #else /* !CONFIG_PREEMPT_RT */
>  
>  #define RT_MUTEX_BUILD_MUTEX
> @@ -1500,6 +1563,11 @@ static inline void __downgrade_write(struct rw_semaphore *sem)
>  	rwbase_write_downgrade(&sem->rwbase);
>  }
>  
> +static inline int __wait_read_killable(struct rw_semaphore *sem)
> +{
> +	return rwbase_wait_lock(&sem->rwbase, TASK_KILLABLE);
> +}
> +
>  /* Debug stubs for the common API */
>  #define DEBUG_RWSEMS_WARN_ON(c, sem)
>  
> @@ -1643,6 +1711,38 @@ void downgrade_write(struct rw_semaphore *sem)
>  }
>  EXPORT_SYMBOL(downgrade_write);
>  
> +/**
> + * rwsem_wait_killable - Wait for current write lock holder to release lock
> + * @sem: The semaphore to wait on.
> + *
> + * This is equivalent to calling down_read(); up_read() but avoids the
> + * possibility that the thread will be preempted while holding the lock
> + * causing threads that want to take the lock for writes to block.  The
> + * intended use case is for lockless readers who notice an inconsistent
> + * state and want to wait for the current writer to finish.
> + */
> +int rwsem_wait_killable(struct rw_semaphore *sem)
> +{
> +	might_sleep();
> +
> +	rwsem_acquire_read(&sem->dep_map, 0, 0, _RET_IP_);
> +	rwsem_release(&sem->dep_map, _RET_IP_);
> +
> +	return __wait_read_common(sem, TASK_KILLABLE);
> +}
> +EXPORT_SYMBOL(rwsem_wait_killable);
> +
> +void rwsem_wait(struct rw_semaphore *sem)
> +{
> +	might_sleep();
> +
> +	rwsem_acquire_read(&sem->dep_map, 0, 0, _RET_IP_);
> +	rwsem_release(&sem->dep_map, _RET_IP_);
> +
> +	__wait_read_common(sem, TASK_UNINTERRUPTIBLE);
> +}
> +EXPORT_SYMBOL(rwsem_wait);
> +
>  #ifdef CONFIG_DEBUG_LOCK_ALLOC
>  
>  void down_read_nested(struct rw_semaphore *sem, int subclass)
> -- 
> 2.43.0
> 

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ