svn commit: r332070 - in head/sys/geom: . label raid

[Conrad Meyer]

Yep, exactly — md_intel was tasting the md(4) device.  md_intel
invokes g_raid_md_get_label() early on in its taste routine.

On Thu, Apr 5, 2018 at 6:29 PM, Allan Jude <allanjude at freebsd.org> wrote:
> On 2018-04-05 21:20, Conrad Meyer wrote:
>> On Thu, Apr 5, 2018 at 6:08 PM, Allan Jude <allanjude at freebsd.org> wrote:
>>> On 2018-04-05 09:56, Sean Bruno wrote:
>>>> Author: sbruno
>>>> Date: Thu Apr  5 13:56:40 2018
>>>> New Revision: 332070
>>>> URL: https://svnweb.freebsd.org/changeset/base/332070
>>>>
>>>> Log:
>>>>   Squash error from geom by sizing ident strings to DISK_IDENT_SIZE.
>>>>
>>>> ...
>>> I might be missing something here, but it seems like you didn't actually
>>> change the side of
>>> ...
>>> struct md_s
>>
>> The md side can be as small as it likes.  It's the caller who needs to
>> provide the large buffer, for other providers of the GEOM::ident
>> attribute.
>>
>>> And I don't see where the destination buffer went from 24 bytes to a
>>> larger value.
>>
>> Check the change in md_intel.c.
>>
>> Best,
>> Conrad
>>
>
> Right, but md_intel.c is for intel BIOS raid.
>
> I don't see the change that increased the size of the buffer in the
> caller from 24 bytes to something more useful?
>
> Or was it md_intel tasting the md(4) device?
>
> --
> Allan Jude