Re: Context switches & interrupts affecting cache?

["Mohamed Bamakhrama" <bamakhrama@gmail.com>]

On 8/3/07, Ralf Baechle <ralf@linux-mips.org> wrote:
> On Fri, Aug 03, 2007 at 12:59:41PM +0200, Mohamed Bamakhrama wrote:
>
> > Hi all,
> > I have one question regarding context switches between user and kernel
> > modes and interrupts. Do they invalidate the I-cache or D-cache?
>
> Never on MIPS.
>
> I call an architecture that would require a cacheflush for such a
> context switch totally broken and yes, they exist - but nothing from
> the MIPS family.

Thanks Ralf for your reply.
The problem that I face now is that the lock works only when I lock
the I-cache lines *just* before calling the function OR when I lock
the code and execute it with all interrupts disabled (via
local_irq_save() & restore()). I always do the lock with interrupts
disabled, but the strange thing is that it works only when the lock is
*close* enough to the code and in my case "close" means just before
calling it :-(
That's why I was wondering if the kernel is doing something strange
with the I-cache such as invalidation on interrupts or other events.

Greetings,

--
Mohamed