Re: + down_write-preserve-local-irqs.patch added to -mm tree

Re: + down_write-preserve-local-irqs.patch added to -mm tree

[Arjan van de Ven]

On Wed, 2006-12-20 at 02:38 -0800, akpm@osdl.org wrote:
> The patch titled
>      down_write(): preserve local irqs
> has been added to the -mm tree.  Its filename is
>      down_write-preserve-local-irqs.patch
> 
> See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find
> out what to do about this
> 
> ------------------------------------------------------
> Subject: down_write(): preserve local irqs

excuse me? Am I missing something here?

down_write() is a sleeping function, right? what business does it have
to be *ever* called with irqs off?

if the answer is "none whatsoever", what good is saving irq state then?

Re: + down_write-preserve-local-irqs.patch added to -mm tree

[Andrew Morton]

On Wed, 20 Dec 2006 11:54:23 +0100
Arjan van de Ven <arjan@infradead.org> wrote:

> On Wed, 2006-12-20 at 02:38 -0800, akpm@osdl.org wrote:
> > The patch titled
> >      down_write(): preserve local irqs
> > has been added to the -mm tree.  Its filename is
> >      down_write-preserve-local-irqs.patch
> > 
> > See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find
> > out what to do about this
> > 
> > ------------------------------------------------------
> > Subject: down_write(): preserve local irqs
> 
> excuse me? Am I missing something here?

Lots.

> down_write() is a sleeping function, right? what business does it have
> to be *ever* called with irqs off?
> 
> if the answer is "none whatsoever", what good is saving irq state then?
> 

a) this patch is for testers to confirm that this fixes the hangs

b) there's plenty of precendent for this - powerpc won't boot if
   mutex_lock() enables interrupts, for example.  We just assume that
   during early boot these sleeping locks aren't contended.

yes, it does suck.