Re: [PATCH 0/2] ACPI: Re-factor and remove ./drivers/acpi/atomicio.[ch]

[Thomas Renninger <trenn@suse.de>]

On Friday 04 November 2011 01:56:04 Huang Ying wrote:
> On Fri, 2011-11-04 at 00:31 +0800, Thomas Renninger wrote:
> > On Thursday, September 29, 2011 11:59:08 PM Myron Stowe wrote:
> > > Late last year I submitted a patch series that re-factored some existing
> > > work that Huang Ying introduced adding support for accessing ACPI
> > > generic registers backed by Memory Mapped I/O (MMIO) while within
> > > interrupt context:
> > >   Huang Ying's commit 15651291a2f8c11e7e6a42d8bfde7a213ff13262,
> > >   My series: http://marc.info/?l=linux-acpi&m=128769263327206&w=2.
> > 
> > Ying: What is your opinion about this patchset?
> 
> I am OK with the patchset.
Great, thanks.
> We just need some testing.
Hm, that's what release candidates (-rcX) are for.

> > One major change seem to be to use a mutex instead of a spinlock
> > which looks like a fix as the pre-mapping should never happen in
> > irq context.
> 
> Sorry, where is the mutex?
you used a spinlock in atomicio.c:
static DEFINE_SPINLOCK(acpi_iomaps_lock);
in osl.c a mutex is used:
static DEFINE_MUTEX(acpi_ioremap_lock);

> > I expect the rest is more or less the same, but double checking
> > by someone who is more involved in these code paths would be
> > great.

   Thomas