[Skiboot] [PATCH] external/gard: Fix displaying 'cleared' gard records
stewart at linux.vnet.ibm.com
Fri Jan 8 11:21:13 AEDT 2016
Cyril Bur <cyril.bur at au1.ibm.com> writes:
> On Thu, 07 Jan 2016 16:08:09 +1100
> Stewart Smith <stewart at linux.vnet.ibm.com> wrote:
>> Cyril Bur <cyril.bur at au1.ibm.com> writes:
>> > When a garded component is replaced hostboot detects this and updates the
>> > gard partition.
>> > What hostboot does is set the record_id field to 0xFFFFFFFF but leaves the
>> > rest of the flash untouched, this has caused issues with the gard tool the
>> > thinking was that an entire record of all 0xFF bytes would signal not a
>> > valid record.
>> > This patch add rectifies this issue and `gard list` will no longer show any
>> > record with an id of 0xFFFFFFFF.
>> > Signed-off-by: Cyril Bur <cyril.bur at au1.ibm.com>
>> > ---
>> > Stewart: Fixes BZ 131707
>> I gather this should also head towards stable ?
> Yes sorry, should I have explicitly stated that?
Yeah... theoretically I say how it doc/stable-skiboot-rules.txt - but
I've yet to make everyone obey that... perhaps I either need to buy more
carrots or sticks.
>> It looks like that functionality isn't
>> removed, is it still correct?
> Short answer, no idea. Long answer, when flash is wiped it will be all 0xFF so
> to avoid reading a lot more than necessary I put that check in, there can't
> possibly be a valid record that is all 0xFF (confirmed now with this clearing
> business) and while I can't be sure, I expect it to be fairly safe that once I
> hit a 0xFF record I've hit erased flash. What I would like to know and haven't
> quite had the will to look through hostboot is, will it write gard records
> after one that it has 'cleared', I've written this assuming it will but it
> would make life easier for the tool if I can be sure it won't.
I think this means you get to write up the spec of the format :)
OPAL Architect, IBM.
More information about the Skiboot