Re: 2.5.64-mm7 - dies on smp with raid

[Neil Brown <neilb@cse.unsw.edu.au>]

On Saturday March 15, akpm@digeo.com wrote:
> 
> A lot of md updates went into Linus's tree overnight.  Can you get some more
> details for Neil?
> 
> Here is a wild guess:
> 
> diff -puN drivers/md/md.c~a drivers/md/md.c
> --- 25/drivers/md/md.c~a	2003-03-15 12:02:04.000000000 -0800
> +++ 25-akpm/drivers/md/md.c	2003-03-15 12:02:14.000000000 -0800
> @@ -2818,6 +2818,8 @@ int md_thread(void * arg)
>  
>  void md_wakeup_thread(mdk_thread_t *thread)
>  {
> +	if (!thread)
> +		return;
>  	dprintk("md: waking up MD thread %p.\n", thread);
>  	set_bit(THREAD_WAKEUP, &thread->flags);
>  	wake_up(&thread->wqueue);
> 

Looks like a good guess to me.

I hadn't considered raid0/linear properly in that last change suite.
They don't have a thread so there is nothing to wake up.

There are two places where the wrong thing will happen:
  do_md_run where it also calls md_update_sb which doesn't
    hurt but isn't really needed (there is never any point
    updating the superblock metadata for raid0/linear).
  restart_array where we switch back to read/write and wakeup
    the thread to see if there is anything to do.

We either need this "if(!thread)" test inside md_wakeup_thread
or at those two call sites, in which case we can avoid md_update_sb
as well.

I send one to Linus later...

Thanks,
NeilBrown
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org">aart@kvack.org</a>