Re: [PATCH] RFC: console: hack up console_lock more v3

[Chris Wilson]

Quoting Daniel Vetter (2019-05-09 13:09:03)
> console_trylock, called from within printk, can be called from pretty
> much anywhere. Including try_to_wake_up. Note that this isn't common,
> usually the box is in pretty bad shape at that point already. But it
> really doesn't help when then lockdep jumps in and spams the logs,
> potentially obscuring the real backtrace we're really interested in.
> One case I've seen (slightly simplified backtrace):
> 
>  Call Trace:
>   <IRQ>
>   console_trylock+0xe/0x60
>   vprintk_emit+0xf1/0x320
>   printk+0x4d/0x69
>   __warn_printk+0x46/0x90
>   native_smp_send_reschedule+0x2f/0x40
>   check_preempt_curr+0x81/0xa0
>   ttwu_do_wakeup+0x14/0x220
>   try_to_wake_up+0x218/0x5f0
>   pollwake+0x6f/0x90
>   credit_entropy_bits+0x204/0x310
>   add_interrupt_randomness+0x18f/0x210
>   handle_irq+0x67/0x160
>   do_IRQ+0x5e/0x130
>   common_interrupt+0xf/0xf
>   </IRQ>
> 
> This alone isn't a problem, but the spinlock in the semaphore is also
> still held while waking up waiters (up() -> __up() -> try_to_wake_up()
> callchain), which then closes the runqueue vs. semaphore.lock loop,
> and upsets lockdep, which issues a circular locking splat to dmesg.
> Worse it upsets developers, since we don't want to spam dmesg with
> clutter when the machine is dying already.
> 
> Fix this by creating a prinkt_safe_up() which calls wake_up_process
> outside of the spinlock. This isn't correct in full generality, but
> good enough for console_lock:
> 
> - console_lock doesn't use interruptible or killable or timeout down()
>   calls, hence an up() is the only thing that can wake up a process.
>   Hence the process can't get woken and killed and reaped while we try
>   to wake it up too.
> 
> - semaphore.c always updates the waiter list while under the spinlock,
>   so there's no other races. Specifically another process that races
>   with a quick console_lock/unlock while we've dropped the spinlock
>   already won't see our own waiter.
> 
> Note that we only have to break the recursion for the semaphore.lock
> spinlock of the console_lock. Recursion within various scheduler
> related locks is already prevented by the printk_safe_enter/exit pair
> in __up_console_sem().
> 
> Also cc'ing John Ogness since perhaps his printk rework fixes this all
> properly.
> 
> v2: Ditch attempt to fix console_trylock.
> 
> v3: Add a comment explaining why the taks we're waking won't
> disappear (Chris), and improve commit message to address review
> questions.
> 
> Signed-off-by: Daniel Vetter <daniel.vetter@xxxxxxxxx>
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: Will Deacon <will.deacon@xxxxxxx>
> Cc: Petr Mladek <pmladek@xxxxxxxx>
> Cc: Sergey Senozhatsky <sergey.senozhatsky@xxxxxxxxx>
> Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
> Cc: Daniel Vetter <daniel.vetter@xxxxxxxx>
> Cc: John Ogness <john.ogness@xxxxxxxxxxxxx>
> Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> Signed-off-by: Daniel Vetter <daniel.vetter@xxxxxxxx>

I'm a bit nervous about that this is only safe for the precisely
controlled conditions, but then again that it is called printk_safe
should deter any other users.

The logic checks out, and you convinced me that the dereference is
protected, so
Reviewed-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
-Chris