svn commit: r345171 - head/usr.sbin/bhyve
John Baldwin
jhb at FreeBSD.org
Fri Mar 15 16:27:52 UTC 2019
On 3/14/19 10:24 PM, Conrad Meyer wrote:
> On Thu, Mar 14, 2019 at 8:06 PM Andrew Thompson <andy at fud.org.nz> wrote:
>>
>> On Fri, 15 Mar 2019 at 15:11, Chuck Tuffli <chuck at freebsd.org> wrote:
>>> bzero(&pciecap, sizeof(pciecap));
> ...
>>> + pciecap.dev_capabilities = PCIEM_CAP_ROLE_ERR_RPT;
>>
>> If the message you say 'set the bit' but you are overwriting the whole variable, is this intended?
>
> Looks like it was zero before. So yeah, it sets the bit.
It would probably be cleaner for future changes to make it a |=, but that's a
tiny nit. style(9) wants a blank line before the comment as well.
I hadn't approved it yet only because I hadn't gone and dug through my PCIe
books / specs to see what this bit is and confirm it is required.
OTOH, it's not clear to me that bhyve PCI-e devices don't want to just be 1.0a
devices as a lowest common denominator to be as accommodating to as wide variety
of OS's as possible.
One thing I didn't see in a review was a reason for why to make this change?
Does some OS reject devices without this bit set or is it just based on reading
the spec? bhyve doesn't assert any PCI-e errors for virtual devices, so
this bit is pretty meaningless.
--
John Baldwin
More information about the svn-src-all
mailing list