linux-mips
[Top] [All Lists]

Re: [PATCH] various sibyte 2.6.x bugfixes

To: Dave Johnson <djohnson+linuxmips@sw.starentnetworks.com>
Subject: Re: [PATCH] various sibyte 2.6.x bugfixes
From: Andy Isaacson <adi@hexapodia.org>
Date: Wed, 22 Jun 2005 16:24:46 -0700
Cc: "Maciej W. Rozycki" <macro@linux-mips.org>, linux-mips@linux-mips.org
In-reply-to: <17081.42083.756755.92799@cortez.sw.starentnetworks.com>
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <17081.32401.574987.337795@cortez.sw.starentnetworks.com> <Pine.LNX.4.61L.0506221616020.4849@blysk.ds.pg.gda.pl> <17081.42083.756755.92799@cortez.sw.starentnetworks.com>
Sender: linux-mips-bounce@linux-mips.org
User-agent: Mutt/1.4.2i
Great minds think alike...

On Wed, Jun 22, 2005 at 01:48:19PM -0400, Dave Johnson wrote:
> Now that I look at it more that check isn't needed at all.
> 
> zero bits wont make it past irq_affinity_write_proc() so that's not
> needed.
> 
> multiple bits are valid (it's also the default) so just using the
> first bit that is set should be fine.

Multiple bits are valid at the API, but the current implementation of
sb1250_set_affinity can only handle setting affinity to a single CPU, so
the test should remain.  Unless I'm missing something, which is entirely
possible.

>       /* Convert logical CPU to physical CPU */
> -     cpu = cpu_logical_map(i);
> +     cpu = cpu_logical_map(first_cpu(mask));

Ah, but that's not right, is it?  If I as the admin say "please bind
IRQs to CPUs 0 and 1" I'll be annoyed to have them all hitting 0.  And
with the BCM1480 it's no longer an academic question.

-andy

<Prev in Thread] Current Thread [Next in Thread>