Re: [Bug 203107] New: Bad page map in process during boot

Re: [Bug 203107] New: Bad page map in process during boot

[Jan Kara]

Switching to email...

On Fri 29-03-19 20:46:22, bugzilla-daemon@bugzilla.kernel.org wrote:
> https://bugzilla.kernel.org/show_bug.cgi?id=203107
> 
>             Bug ID: 203107
>            Summary: Bad page map in process during boot
>            Product: File System
>            Version: 2.5
>     Kernel Version: 5.0.5
>           Hardware: All
>                 OS: Linux
>               Tree: Mainline
>             Status: NEW
>           Severity: normal
>           Priority: P1
>          Component: ext4
>           Assignee: fs_ext4@kernel-bugs.osdl.org
>           Reporter: echto1@gmail.com
>         Regression: No
> 
> Error occurs randomly at boot after upgrading kernel from 5.0.0 to 5.0.4.
> 
> https://justpaste.it/387uf

I don't think this is an ext4 error. Sure this is an error in file mapping
of libblkid.so.1.1.0 (which is handled by ext4) but the filesystem has very
little to say wrt how or which PTEs are installed. And the problem is that
invalid PTE (dead000000000100) is present in page tables. So this looks
more like a problem in MM itself. Adding MM guys to CC.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [Bug 203107] New: Bad page map in process during boot

[Kirill A. Shutemov]

On Tue, Apr 02, 2019 at 12:16:13PM +0200, Jan Kara wrote:
> Switching to email...
> 
> On Fri 29-03-19 20:46:22, bugzilla-daemon@bugzilla.kernel.org wrote:
> > https://bugzilla.kernel.org/show_bug.cgi?id=203107
> > 
> >             Bug ID: 203107
> >            Summary: Bad page map in process during boot
> >            Product: File System
> >            Version: 2.5
> >     Kernel Version: 5.0.5
> >           Hardware: All
> >                 OS: Linux
> >               Tree: Mainline
> >             Status: NEW
> >           Severity: normal
> >           Priority: P1
> >          Component: ext4
> >           Assignee: fs_ext4@kernel-bugs.osdl.org
> >           Reporter: echto1@gmail.com
> >         Regression: No
> > 
> > Error occurs randomly at boot after upgrading kernel from 5.0.0 to 5.0.4.
> > 
> > https://justpaste.it/387uf
> 
> I don't think this is an ext4 error. Sure this is an error in file mapping
> of libblkid.so.1.1.0 (which is handled by ext4) but the filesystem has very
> little to say wrt how or which PTEs are installed. And the problem is that
> invalid PTE (dead000000000100) is present in page tables. So this looks
> more like a problem in MM itself. Adding MM guys to CC.

0xdead000000000100 and 0xdead000000000200 are LIST_POISON1 and
LIST_POISON2 repectively. Have no idea how would they end up in page table.

-- 
 Kirill A. Shutemov

Re: [Bug 203107] New: Bad page map in process during boot

[Vlastimil Babka]

On 4/4/19 3:08 PM, Kirill A. Shutemov wrote:
> On Tue, Apr 02, 2019 at 12:16:13PM +0200, Jan Kara wrote:
>> Switching to email...
>>
>> On Fri 29-03-19 20:46:22, bugzilla-daemon@bugzilla.kernel.org wrote:
>>> https://bugzilla.kernel.org/show_bug.cgi?id=203107
>>>
>>>             Bug ID: 203107
>>>            Summary: Bad page map in process during boot
>>>            Product: File System
>>>            Version: 2.5
>>>     Kernel Version: 5.0.5
>>>           Hardware: All
>>>                 OS: Linux
>>>               Tree: Mainline
>>>             Status: NEW
>>>           Severity: normal
>>>           Priority: P1
>>>          Component: ext4
>>>           Assignee: fs_ext4@kernel-bugs.osdl.org
>>>           Reporter: echto1@gmail.com
>>>         Regression: No
>>>
>>> Error occurs randomly at boot after upgrading kernel from 5.0.0 to 5.0.4.
>>>
>>> https://justpaste.it/387uf
>>
>> I don't think this is an ext4 error. Sure this is an error in file mapping
>> of libblkid.so.1.1.0 (which is handled by ext4) but the filesystem has very
>> little to say wrt how or which PTEs are installed. And the problem is that
>> invalid PTE (dead000000000100) is present in page tables. So this looks
>> more like a problem in MM itself. Adding MM guys to CC.
> 
> 0xdead000000000100 and 0xdead000000000200 are LIST_POISON1 and
> LIST_POISON2 repectively. Have no idea how would they end up in page table.

It's possible that CONFIG_DEBUG_LIST could catch the issue. Between
5.0.0 to 5.0.4 it should be also relatively easy to bisect with the
stable git tree [1], although if it happens randomly, you need to
perform enough attempts to accurately determine which commit is "good".

[1] git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git