Re: more signal locking bugs?

[Manfred Spraul <manfred@colorfullife.com>]

Linus Torvalds wrote:

>>What about this minimal patch? The performance critical operation is 
>>signal delivery - we should fix the synchronization between signal 
>>delivery and exec first.
>>    
>>
>
>The patch looks ok, although I'd also remove the locking and testing from
>collect_sigign_sigcatch() once it is done at a higher level.
>
>And yeah, what about signal delivery? Put back the same lock there?
>  
>
I don't know.
Taking read_lock(&tasklist_lock) for send_specific_sig_info might hurt 
scalability. Ingo?

If we do not want a global lock, then we have two options:
- make task_lock an interrupt spinlock.
- add a second spinlock to the task structure, for the signal stuff.

Design question - what's worse? Memory bloat or a few additional 
local_irq_{en,dis}able().
I don't care - no performance critical codepaths.
Additionally many task_lock()/task_unlock users could be replaced with 
spin_unlock_wait(&task->alloc_lock), which would not need the 
local_irq_disable().

--
    Manfred