linux-kernel - Re: [PATCHv6 6/7] printk: use printk

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: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 23 Dec 2016 10:46:43 +0900
From:   Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>
To:     Petr Mladek <pmladek@...e.com>
Cc:     Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Jan Kara <jack@...e.cz>, Tejun Heo <tj@...nel.org>,
        Calvin Owens <calvinowens@...com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ingo Molnar <mingo@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Andy Lutomirski <luto@...nel.org>,
        Peter Hurley <peter@...leysoftware.com>,
        linux-kernel@...r.kernel.org,
        Sergey Senozhatsky <sergey.senozhatsky@...il.com>
Subject: Re: [PATCHv6 6/7] printk: use printk_safe buffers in printk

On (12/22/16 18:10), Petr Mladek wrote:
...
> There are many callers. I think that such wrappers make sense.
> I would only like to keep naming scheme similar to the classic
> locks. I mean:
> 
> printk_safe_enter_irq()
> printk_safe_exit_irq()
> 
> printk_safe_enter_irqsave(flags)
> printk_safe_exit_irqrestore(flags)

sure.

> and
> 
> logbuf_lock_irq()
> logbuf_unlock_irq()
> 
> logbuf_lock_irqsave(flags)
> logbuf_lock_irqrestore(flags)

ok.

> I actually like this change. It makes it clear that the operation
> has a side effect (disables/enables irq) which was not visible
> from the original name.

agree.

> Well, I wonder how many times we need to call printk_save_enter/exit
> standalone (ouside these macros).

not every switch to printk_safe is "dictated" by logbuf_lock.
down_trylock_console_sem(), for instance, takes semaphore spin_lock
which already may be locked on the same CPU (*), so we need to be
in safe mode:

vprintk_emit()
 down_trylock()
  raw_spin_lock_irqsave(&sem->lock, flags);
  ...
  raw_spin_unlock_irqrestore(&sem->lock, flags);
   spin_dump()
    printk()
     vprintk_emit()
      down_trylock()
       raw_spin_lock_irqsave(&sem->lock, flags)   << deadlock


and so on. IOW, "printk_save_enter()" != "logbuf_lock is acquired".


> The question is if we really need all the variants of
> printk_safe_enter()/exit(). Alternative solution would be
> to handle only the printk_context in pritnk_safe_enter()
> and make sure that it is called with IRQs disabled.
> I mean to define only __printk_safe_enter()/exit()
> and do:
> 
> #define logbuf_lock_irqsave(flags)		\
> 	do {					\
> 		local_irq_save(flags)		\
> 		__printk_safe_enter();		\
> 		raw_spin_lock(&logbuf_lock);	\
> 	} while (0)

won't do the trick for console sem spin_lock.

[..]
> PS: I still think if we could come with a better name than
> printk_safe() but I cannot find one.

well, not that I'm the fan of printk_safe name, but can't think
of anything better. we make printk calls safe (deadlock safe) in
places where previously it was unsafe... quick-&-dirty name that
is implementation-specific -- printk_percpu_enter/exit, or
printk_pcpu_enter/exit... dunno.

	-ss