Re: [PATCH] x86: don't change affinity with interrupt unmasked

["Jan Beulich" <JBeulich@suse.com>]

>>> On 23.03.15 at 19:46, <konrad.wilk@oracle.com> wrote:
> On Fri, Mar 20, 2015 at 03:40:02PM +0000, Jan Beulich wrote:
>> With ->startup unmasking the IRQ, setting the affinity afterwards
>> without masking the IRQ again is invalid namely for MSI (which can't
>> have their affinity updated atomically).
> 
> That took a bit of verification :-)
> 
> Could you include this in the commit please:
> 
> Per 6.8.3.5. Per-vector Masking and Function Masking from
> https://www.pcisig.com/specifications/conventional/msi-x_ecn.pdf 
> ".. anytime software unmasks a currently masked MSI-X
> Table entry either by clearing its Mask bit or by clearing the Function Mask 
> bit, the
> function must update any Address or Data values that it cached from that 
> entry. If
> software changes the Address or Data value of an entry while the entry is 
> unmasked, the 
> result is undefined."

I'd rather not, as that may result in the wrong impression that only
MSI-X is affected, while MSI is as much. Apart from the potential
caching is only one part of the problem, the other is (as said in the
description) the non-atomic nature of the address/data updates
for both variants of MSI. I guess I'll simply extend what is in
parentheses at the end to include the caching issue.

Jan