Re: [PATCH 5.4 000/191] 5.4.8-stable review
From: Mike Kravetz
Date: Fri Jan 03 2020 - 12:34:54 EST
On 1/3/20 7:56 AM, Arnd Bergmann wrote:
> On Fri, Jan 3, 2020 at 4:45 PM Greg Kroah-Hartman
> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>> On Fri, Jan 03, 2020 at 04:29:56PM +0100, Arnd Bergmann wrote:
>>> On Fri, Jan 3, 2020 at 4:25 PM Arnd Bergmann <arnd@xxxxxxxx> wrote:
>>>> On Fri, Jan 3, 2020 at 4:03 PM Naresh Kamboju <naresh.kamboju@xxxxxxxxxx> wrote:
>>>>> On Fri, 3 Jan 2020 at 03:42, Greg Kroah-Hartman
>>>>> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>>>> -ENOENT is what you get when hugetlbfs is not mounted, so this hints to
>>>> 8fc312b32b2 mm/hugetlbfs: fix error handling when setting up mounts
>>> I see that Mike Kravetz suggested not putting this patch into stable in
>>> but it was picked through the autosel mechanism later.
>> So does that mean that Linus's tree shows this LTP failure as well?
> Yes, according to
> mainline has the same testcase failure, it started happening between
> v5.4-10135-gc3bfc5dd73c6 and v5.4-10271-g596cf45cbf6e, when the patch
> was originally merged into 5.5-rc1.
>> This does seem to fix a real issue, as shown by the LTP test noticing
>> it, so should the error code value be fixed in Linus's tree?
> No idea what to conclude from the testcase failure, let's see if Mike has
> any suggestions.
Thanks for isolating to this patch!
There are dependencies between arch specific code and arch independent code
during the setup of hugetlb sizes/mounts. Let me take a closer look at the
arm64 code and get access to a system for debug.