lkml.org 
[lkml]   [2019]   [Sep]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v2 2/2] mm, sl[aou]b: guarantee natural alignment for kmalloc(power-of-two)
On Tue, Sep 03, 2019 at 08:13:45PM +0000, Christopher Lameter wrote:
> On Sat, 31 Aug 2019, Matthew Wilcox wrote:
>
> > > The current behavior without special alignment for these caches has been
> > > in the wild for over a decade. And this is now coming up?
> >
> > In the wild ... and rarely enabled. When it is enabled, it may or may
> > not be noticed as data corruption, or tripping other debugging asserts.
> > Users then turn off the rare debugging option.
>
> Its enabled in all full debug session as far as I know. Fedora for
> example has been running this for ages to find breakage in device drivers
> etc etc.

Are you telling me nobody uses the ramdisk driver on fedora? Because
that's one of the affected drivers.

> > > If there is an exceptional alignment requirement then that needs to be
> > > communicated to the allocator. A special flag or create a special
> > > kmem_cache or something.
> >
> > The only way I'd agree to that is if we deliberately misalign every
> > allocation that doesn't have this special flag set. Because right now,
> > breakage happens everywhere when these debug options are enabled, and
> > the very people who need to be helped are being hurt by the debugging.
>
> That is customarily occurring for testing by adding "slub_debug" to the
> kernel commandline (or adding debug kernel options) and since my
> information is that this is done frequently (and has been for over a
> decade now) I am having a hard time believing the stories of great
> breakage here. These drivers were not tested with debugging on before?
> Never ran with a debug kernel?

Whatever is being done is clearly not enough to trigger the bug. So how
about it? Create an option to slab/slub to always return misaligned
memory.

\
 
 \ /
  Last update: 2019-09-03 22:54    [W:0.076 / U:11.996 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site