Re: [PATCH v2] mm: slub: call WARN() instead of pr_err on slab_fix.

[David Rientjes <rientjes@google.com>]

On Wed, 5 Feb 2025, Vlastimil Babka wrote:

> On 2/5/25 18:10, David Rientjes wrote:
> > On Wed, 5 Feb 2025, Hyesoo Yu wrote:
> > 
> >> If a slab object is corrupted or an error occurs in its internal
> >> value, continuing after restoration may cause other side effects.
> >> At this point, it is difficult to debug because the problem occurred
> >> in the past. It is better to use WARN() instead of pr_err to catch
> >> errors at the point of issue because WARN() could trigger panic for
> >> system debugging when panic_on_warn is enabled. WARN() should be
> >> called prior to fixing the value because when a panic is triggered by WARN(),
> >> it allows us to check corrupted data.
> >> 
> > 
> > I think this makes sense, but it doesn't document why the other changes 
> > are being made, like moving the setting of *freelist to NULL.  This is 
> > presumably something that you want in the crash dump when 
> > kernel.panic_on_warn is enabled.  Probably best to call that out, but to 
> > also indicate what you're relying on in the crash dump to make forward 
> > progress on in diagnosing the issue.
> 
> Well the last sentence of the changelog above says exactly that, no?
> 

Sorry, I should have been more clear.  It's unclear in the code why 
choosing WARN() here is helpful given the stack would be known.  It makes 
sense to enable kernel.panic_on_warn this way for debugging purposes, but 
thought it should also carry a comment in the code on the rationale (and 
the state we're trying to capture in a crash dump) so a future change 
doesn't go and unravel this for us again.

> >> Changes in v2:
> >> - Replace direct calling with BUG_ON with the use of WARN in slab_fix.
> >> 
> >> Signed-off-by: Hyesoo Yu <hyesoo.yu@samsung.com>
> >> ---
> >>  mm/slub.c | 10 +++++-----
> >>  1 file changed, 5 insertions(+), 5 deletions(-)
> >> 
> >> diff --git a/mm/slub.c b/mm/slub.c
> >> index 1f50129dcfb3..ea956cb4b8be 100644
> >> --- a/mm/slub.c
> >> +++ b/mm/slub.c
> >> @@ -1043,7 +1043,7 @@ static void slab_fix(struct kmem_cache *s, char *fmt, ...)
> >>  	va_start(args, fmt);
> >>  	vaf.fmt = fmt;
> >>  	vaf.va = &args;
> >> -	pr_err("FIX %s: %pV\n", s->name, &vaf);
> >> +	WARN(1, "FIX %s: %pV\n", s->name, &vaf);
> >>  	va_end(args);
> >>  }
> >>  
> >> @@ -1106,8 +1106,8 @@ static bool freelist_corrupted(struct kmem_cache *s, struct slab *slab,
> >>  	if ((s->flags & SLAB_CONSISTENCY_CHECKS) &&
> >>  	    !check_valid_pointer(s, slab, nextfree) && freelist) {
> >>  		object_err(s, slab, *freelist, "Freechain corrupt");
> >> -		*freelist = NULL;
> >>  		slab_fix(s, "Isolate corrupted freechain");
> >> +		*freelist = NULL;
> >>  		return true;
> >>  	}
> >>  
> >> @@ -1445,9 +1445,9 @@ static int on_freelist(struct kmem_cache *s, struct slab *slab, void *search)
> >>  				set_freepointer(s, object, NULL);
> >>  			} else {
> >>  				slab_err(s, slab, "Freepointer corrupt");
> >> +				slab_fix(s, "Freelist cleared");
> >>  				slab->freelist = NULL;
> >>  				slab->inuse = slab->objects;
> >> -				slab_fix(s, "Freelist cleared");
> >>  				return 0;
> >>  			}
> >>  			break;
> >> @@ -1464,14 +1464,14 @@ static int on_freelist(struct kmem_cache *s, struct slab *slab, void *search)
> >>  	if (slab->objects != max_objects) {
> >>  		slab_err(s, slab, "Wrong number of objects. Found %d but should be %d",
> >>  			 slab->objects, max_objects);
> >> -		slab->objects = max_objects;
> >>  		slab_fix(s, "Number of objects adjusted");
> >> +		slab->objects = max_objects;
> >>  	}
> >>  	if (slab->inuse != slab->objects - nr) {
> >>  		slab_err(s, slab, "Wrong object count. Counter is %d but counted were %d",
> >>  			 slab->inuse, slab->objects - nr);
> >> -		slab->inuse = slab->objects - nr;
> >>  		slab_fix(s, "Object count adjusted");
> >> +		slab->inuse = slab->objects - nr;
> >>  	}
> >>  	return search == NULL;
> >>  }
> >> -- 
> >> 2.48.0
> >> 
> >> 
> 
>