Skip to contentSkip to conversation

mm/slub: Add a dump_stack() to the unexpected GFP check

PATCH Applied Borislav Petkov 4 participants 11 messages Jan 16, 2017

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mm/slub: Add a dump_stack() to the unexpected GFP check

From: Borislav Petkov <[email protected]>

We wanna know who's doing such a thing. Like slab.c does that.

---
 mm/slub.c | 1 +
 1 file changed, 1 insertion(+)
mm/slub.c +1 -0
index 067598a00849..1b0fa7625d6d 100644
@@ -1623,6 +1623,7 @@ static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
 		flags &= ~GFP_SLAB_BUG_MASK;
 		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
 				invalid_mask, &invalid_mask, flags, &flags);
+		dump_stack();
 	}
 
 	return allocate_slab(s,
Signature · 1 line
2.11.0
#1 lore raw
Quoted 14 lines · Borislav Petkov

From: Borislav Petkov <[email protected]>

We wanna know who's doing such a thing. Like slab.c does that.

---
 mm/slub.c | 1 +
 1 file changed, 1 insertion(+)
mm/slub.c +0 -0
index 067598a00849..1b0fa7625d6d 100644

mm/slub.c static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node) · line 1626

Quoted 4 lines
mm/slub.c +1 -0
@@ -1623,6 +1623,7 @@ static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
 		flags &= ~GFP_SLAB_BUG_MASK;
 		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
 				invalid_mask, &invalid_mask, flags, &flags);
+		dump_stack();

Will it make sense to change these two lines above to WARN(true, .....)?

Quoted 6 lines
mm/slub.c +0 -1
 	}
 
 	return allocate_slab(s,
--

2.11.0

#2 lore raw
Quoted 10 lines · Leon Romanovsky
Quoted 9 lines · Borislav Petkov

From: Borislav Petkov <[email protected]>

We wanna know who's doing such a thing. Like slab.c does that.

---
 mm/slub.c | 1 +
 1 file changed, 1 insertion(+)
Quoted 11 lines

mm/slub.c static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node) · line 1626

Quoted 9 lines
mm/slub.c +1 -0
index 067598a00849..1b0fa7625d6d 100644
@@ -1623,6 +1623,7 @@ static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
 		flags &= ~GFP_SLAB_BUG_MASK;
 		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
 				invalid_mask, &invalid_mask, flags, &flags);
+		dump_stack();

Will it make sense to change these two lines above to WARN(true, .....)?

Should be equivalent.

I'd even go a step further and make this a small inline function, something like warn_unexpected_gfp(flags) or so and call it from both from slab.c and slub.c.

Depending on what mm folks prefer, that is.

Signature · 4 lines
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.
Acked-by on v1 1/1
#9 lore raw
In reply to Borislav Petkov

[Let's add Andrew]

Quoted 3 lines · Borislav Petkov

From: Borislav Petkov <[email protected]>

We wanna know who's doing such a thing. Like slab.c does that.

Yes this was an omission on my side in 72baeef0c271 ("slab: do not panic on invalid gfp_mask").

Quoted 2 lines

Thanks!

Quoted 18 lines
---
 mm/slub.c | 1 +
 1 file changed, 1 insertion(+)
mm/slub.c +1 -0
index 067598a00849..1b0fa7625d6d 100644
@@ -1623,6 +1623,7 @@ static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
 		flags &= ~GFP_SLAB_BUG_MASK;
 		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
 				invalid_mask, &invalid_mask, flags, &flags);
+		dump_stack();
 	}
 
 	return allocate_slab(s,
Signature · 1 line
2.11.0
Signature · 8 lines
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#3 lore raw
In reply to Borislav Petkov
Quoted 23 lines · Borislav Petkov
Quoted 10 lines · Leon Romanovsky
Quoted 9 lines · Borislav Petkov

From: Borislav Petkov <[email protected]>

We wanna know who's doing such a thing. Like slab.c does that.

---
 mm/slub.c | 1 +
 1 file changed, 1 insertion(+)
Quoted 11 lines

mm/slub.c static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node) · line 1626

Quoted 9 lines
mm/slub.c +1 -0
index 067598a00849..1b0fa7625d6d 100644
@@ -1623,6 +1623,7 @@ static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
 		flags &= ~GFP_SLAB_BUG_MASK;
 		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
 				invalid_mask, &invalid_mask, flags, &flags);
+		dump_stack();

Will it make sense to change these two lines above to WARN(true, .....)?

Quoted 1 line

Should be equivalent.

Almost, except one point - pr_warn and dump_stack have different log levels. There is a chance that user won't see pr_warn message above, but dump_stack will be always present.

For WARN_XXX, users will always see message and stack at the same time.

Quoted 12 lines

I'd even go a step further and make this a small inline function, something like warn_unexpected_gfp(flags) or so and call it from both from slab.c and slub.c.

Depending on what mm folks prefer, that is.

--
Regards/Gruss,
Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.

#4 lore raw
Quoted 1 line · Leon Romanovsky

Almost, except one point - pr_warn and dump_stack have different log

Actually, Michal pointed out on IRC a more relevant difference:

WARN() taints the kernel and we don't want that for GFP flags misuse. Also, from looking at __warn(), it checks panic_on_warn and we explode if set.

So no, we probably don't want WARN() here.

Signature · 4 lines
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.
#8 lore raw
In reply to Leon Romanovsky
Quoted 24 lines · Leon Romanovsky
Quoted 23 lines · Borislav Petkov
Quoted 21 lines · Leon Romanovsky
Quoted 14 lines · Borislav Petkov

From: Borislav Petkov <[email protected]>

We wanna know who's doing such a thing. Like slab.c does that.

---
 mm/slub.c | 1 +
 1 file changed, 1 insertion(+)
mm/slub.c +0 -0
index 067598a00849..1b0fa7625d6d 100644

mm/slub.c static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node) · line 1626

Quoted 4 lines
mm/slub.c +1 -0
@@ -1623,6 +1623,7 @@ static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
 		flags &= ~GFP_SLAB_BUG_MASK;
 		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
 				invalid_mask, &invalid_mask, flags, &flags);
+		dump_stack();

Will it make sense to change these two lines above to WARN(true, .....)?

Quoted 7 lines
Quoted 1 line

Should be equivalent.

Almost, except one point - pr_warn and dump_stack have different log levels. There is a chance that user won't see pr_warn message above, but dump_stack will be always present.

For WARN_XXX, users will always see message and stack at the same time.

On the other hand WARN* will taint the kernel and this sounds a bit overreacting for something like a wrong gfp mask which is perfectly recoverable. Not to mention users who care configured to panic on warning.

So while I do not have a strong opinion on this I would rather stay with the dump_stack.

Signature · 8 lines
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Acked-by on v1 1/1
#10 lore raw
In reply to Borislav Petkov
Quoted 5 lines · Borislav Petkov

From: Borislav Petkov <[email protected]>

We wanna know who's doing such a thing. Like slab.c does that.

Quoted 10 lines
---
 mm/slub.c | 1 +
 1 file changed, 1 insertion(+)
mm/slub.c +0 -0
index 067598a00849..1b0fa7625d6d 100644

mm/slub.c static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node) · line 1629

Quoted 7 lines
mm/slub.c +1 -0
@@ -1623,6 +1623,7 @@ static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
 		flags &= ~GFP_SLAB_BUG_MASK;
 		pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
 				invalid_mask, &invalid_mask, flags, &flags);
+		dump_stack();
 	}
 
 	return allocate_slab(s,

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#5 lore raw
In reply to Borislav Petkov
Quoted 6 lines · Borislav Petkov
Quoted 1 line · Leon Romanovsky

Almost, except one point - pr_warn and dump_stack have different log

Actually, Michal pointed out on IRC a more relevant difference:

WARN() taints the kernel and we don't want that for GFP flags misuse.

And doesn't dump_stack do the same? It pollutes the log too.

Quoted 2 lines

Also, from looking at __warn(), it checks panic_on_warn and we explode if set.

Right, it is very valid point.

Quoted 2 lines

So no, we probably don't want WARN() here.

I understand, Thanks.

Quoted 6 lines

--
Regards/Gruss,
Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.

#6 lore raw
Quoted 1 line · Leon Romanovsky

And doesn't dump_stack do the same? It pollutes the log too.

It is not about polluting the log - it is about tainting.

__warn()->add_taint(taint, LOCKDEP_STILL_OK);

Signature · 4 lines
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.
#7 lore raw
Quoted 6 lines · Borislav Petkov
Quoted 1 line · Leon Romanovsky

And doesn't dump_stack do the same? It pollutes the log too.

It is not about polluting the log - it is about tainting.

__warn()->add_taint(taint, LOCKDEP_STILL_OK);

Thanks,
I had something different in mind for word "taint". Sorry for that.

Quoted 6 lines

--
Regards/Gruss,
Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.