Re: Linux 6.10 regression resulting in a crash when using an ext4 filesystem

Re: Linux 6.10 regression resulting in a crash when using an ext4 filesystem

[Thorsten Leemhuis]

On 23.07.24 06:11, Theodore Ts'o wrote:
> On Mon, Jul 22, 2024 at 12:06:59AM -0700, Kees Cook wrote:
>>> Is strscpy_pad appropriate if the @src parameter itself is a fixed
>>> length char[16] which isn't null terminated when the label itself is 16
>>> chars long?
>>
>> Nope; it needed memtostr_pad(). I sent the fix back at the end of May, but it only just recently landed:
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=be27cd64461c45a6088a91a04eba5cd44e1767ef
> 
> Yeah, sorry, I was on vacation for 3.5 weeks starting just before
> Memorial day, and it took me a while to get caught up.  Unfortunately,
> I missed the bug in the strncpy extirpation patch, and it was't
> something that our regression tests caught.  (Sometimes, the
> old/deprecated ways are just more reliable; all of ext4's strncpy()
> calls were working and had been correct for decades.  :-P )
> 
> Anyway, Kees's bugfix is in Linus's tree, and it should be shortly be
> making its way to -stable.

Adding Greg and the stable list to the list of recipients: given that we
already have two reports about trouble due to this[1] he might want to
fast-track the fix (be27cd64461c45 ("ext4: use memtostr_pad() for
s_volume_name")) to 6.10.y, as it's not queued yet -- at least afaics
from looking at
https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/

Ciao, Thorsten

[1] https://bugzilla.kernel.org/show_bug.cgi?id=219072 and
https://bugzilla.kernel.org/show_bug.cgi?id=219078

Re: Linux 6.10 regression resulting in a crash when using an ext4 filesystem

[Greg KH]

On Tue, Jul 23, 2024 at 12:04:41PM +0200, Thorsten Leemhuis wrote:
> On 23.07.24 06:11, Theodore Ts'o wrote:
> > On Mon, Jul 22, 2024 at 12:06:59AM -0700, Kees Cook wrote:
> >>> Is strscpy_pad appropriate if the @src parameter itself is a fixed
> >>> length char[16] which isn't null terminated when the label itself is 16
> >>> chars long?
> >>
> >> Nope; it needed memtostr_pad(). I sent the fix back at the end of May, but it only just recently landed:
> >> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=be27cd64461c45a6088a91a04eba5cd44e1767ef
> > 
> > Yeah, sorry, I was on vacation for 3.5 weeks starting just before
> > Memorial day, and it took me a while to get caught up.  Unfortunately,
> > I missed the bug in the strncpy extirpation patch, and it was't
> > something that our regression tests caught.  (Sometimes, the
> > old/deprecated ways are just more reliable; all of ext4's strncpy()
> > calls were working and had been correct for decades.  :-P )
> > 
> > Anyway, Kees's bugfix is in Linus's tree, and it should be shortly be
> > making its way to -stable.
> 
> Adding Greg and the stable list to the list of recipients: given that we
> already have two reports about trouble due to this[1] he might want to
> fast-track the fix (be27cd64461c45 ("ext4: use memtostr_pad() for
> s_volume_name")) to 6.10.y, as it's not queued yet -- at least afaics
> from looking at
> https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/

Now queued up.  And as it was not explicitly marked for stable
inclusion, thank you for asking for it to be added.

I'll go push out a 6.10.1-rc1 in a short bit with this important fix.

thanks,

greg k-h