Re: Bisected 3.9 regression: Corrupted low memory (x86, mm: Move reserving low memory later in initialization)

[H. Peter Anvin]

Yes, I'm sure.  The old memory reservation code kept the checker from working.

On May 21, 2014 10:58:39 AM PDT, "RafaÅ MiÅecki" <zajec5@xxxxxxxxx> wrote:
>On 21 May 2014 19:34, H. Peter Anvin <hpa@xxxxxxxxx> wrote:
>> On 05/21/2014 06:31 AM, RafaÅ MiÅecki wrote:
>>> On 16 May 2014 07:29, RafaÅ MiÅecki <zajec5@xxxxxxxxx> wrote:
>>>> Sorry for the late report, this bug appears on my old notebook I
>don't
>>>> commonly use. Anyway, I've noticed following problem when using
>>>> 3.15-rc1:
>>>>
>>>> Corrupted low memory at ffff88000000be98 (be98 phys) =
>b02a000400000000
>>>
>>> Ping?
>>>
>>
>> It's not a regression, but rather a progression: the checker now does
>> what it claims to do, which is to check for low memory corruption.
>>
>> However, the checker is also rather useless as anything other than a
>> diagnostic tool.  It tells you that your BIOS is corrupting memory,
>and
>> the solution to that is to reserve the memory, which we already do.
>>
>> All the checker tells us is that on your system, yes, we really do
>need
>> to reserve the memory.
>
>What about making checker a bit more user friendly?
>
>First of all, I got many repeated messages like:
>Corrupted low memory at (...)
>In the first e-mail you can see ~10 of them and I really got more.
>
>Secondly, if this is just a confirmation that low memory reservation
>was indeed needed, what about making it debugging only? Right now it
>sounds like an error for end-users. What about making it something
>like
>Low memory reservation prevented corruption at (...)
>
>The last question, just for sure... are you convinced what I'm seeing
>is not any kind of error? After all, it started appearing after commit
>that moves memory reservation. It's not a commit that enabled the
>checker or something like that.
>What it seems to be for someone not experienced is a bit opposite. It
>looks that after commit in question, kernel reserves low memory too
>late and BIOS already corrupts some data placed in it.
>However I'm not on expert, so I'll just listen to your opinion. I
>simply wanted to ask, just to be sure everything is alright.

-- 
Sent from my mobile phone.  Please pardon brevity and lack of formatting.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/