2019-02-13 02:39:32

by Qian Cai

Subject: [PATCH] slub: untag object before slab end

get_freepointer() could return NULL if there is no more free objects in
the slab. However, it could return a tagged pointer (like
0x2200000000000000) with KASAN_SW_TAGS which would escape the NULL
object checking in check_valid_pointer() and trigger errors below, so
untag the object before checking for a NULL object there.

[ 35.797667] BUG kmalloc-256 (Not tainted): Freepointer corrupt
[ 35.803584] -----------------------------------------------------------------------------
[ 35.803584]
[ 35.813368] Disabling lock debugging due to kernel taint
[ 35.818766] INFO: Allocated in build_sched_domains+0x28c/0x495c age=92 cpu=0 pid=1
[ 35.826443] __kmalloc_node+0x4ac/0x508
[ 35.830343] build_sched_domains+0x28c/0x495c
[ 35.834764] sched_init_domains+0x184/0x1d8
[ 35.839012] sched_init_smp+0x38/0xd4
[ 35.842732] kernel_init_freeable+0x67c/0x1104
[ 35.847243] kernel_init+0x18/0x2a4
[ 35.850790] ret_from_fork+0x10/0x18
[ 35.854423] INFO: Freed in destroy_sched_domain+0xa0/0xcc age=11 cpu=0 pid=1
[ 35.861569] destroy_sched_domain+0xa0/0xcc
[ 35.865814] cpu_attach_domain+0x304/0xb34
[ 35.869971] build_sched_domains+0x4654/0x495c
[ 35.874480] sched_init_domains+0x184/0x1d8
[ 35.878724] sched_init_smp+0x38/0xd4
[ 35.882443] kernel_init_freeable+0x67c/0x1104
[ 35.886953] kernel_init+0x18/0x2a4
[ 35.890495] ret_from_fork+0x10/0x18
[ 35.894128] INFO: Slab 0x(____ptrval____) objects=85 used=0 fp=0x(____ptrval____) flags=0x7ffffffc000200
[ 35.903733] INFO: Object 0x(____ptrval____) @offset=38528 fp=0x(____ptrval____)
[ 35.903733]
[ 35.912637] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[ 35.922155] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[ 35.931672] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[ 35.941190] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[ 35.950707] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[ 35.960224] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[ 35.969741] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[ 35.979258] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
[ 35.988776] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 35.998206] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.007636] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.017065] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.026494] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.035923] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.045353] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.054783] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.064212] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.073642] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.083071] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.092501] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.101930] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.111359] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.120788] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
[ 36.130218] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b a5 kkkkkkkkkkkkkkk.
[ 36.139647] Redzone (____ptrval____): bb bb bb bb bb bb bb bb ........
[ 36.148462] Padding (____ptrval____): 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
[ 36.157979] Padding (____ptrval____): 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
[ 36.167496] Padding (____ptrval____): 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
[ 36.177021] CPU: 0 PID: 1 Comm: swapper/0 Tainted: G B 5.0.0-rc6+ #41
[ 36.184854] Call trace:
[ 36.187328] dump_backtrace+0x0/0x450
[ 36.191032] show_stack+0x20/0x2c
[ 36.194385] __dump_stack+0x20/0x28
[ 36.197911] dump_stack+0xa0/0xfc
[ 36.201265] print_trailer+0x1a8/0x1bc
[ 36.205057] object_err+0x40/0x50
[ 36.208408] check_object+0x214/0x2b8
[ 36.212111] __free_slab+0x9c/0x31c
[ 36.215638] discard_slab+0x78/0xa8
[ 36.219165] kfree+0x918/0x980
[ 36.222259] destroy_sched_domain+0xa0/0xcc
[ 36.226489] cpu_attach_domain+0x304/0xb34
[ 36.230631] build_sched_domains+0x4654/0x495c
[ 36.235125] sched_init_domains+0x184/0x1d8
[ 36.239357] sched_init_smp+0x38/0xd4
[ 36.243060] kernel_init_freeable+0x67c/0x1104
[ 36.247555] kernel_init+0x18/0x2a4
[ 36.251083] ret_from_fork+0x10/0x18

Signed-off-by: Qian Cai <[email protected]>
---

Depends on slub-fix-slab_consistency_checks-kasan_sw_tags.patch.

mm/slub.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/slub.c b/mm/slub.c
index 4a61959e1887..2fd1cf39914c 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -503,11 +503,11 @@ static inline int check_valid_pointer(struct kmem_cache *s,
{
void *base;

+ object = kasan_reset_tag(object);
if (!object)
return 1;

base = page_address(page);
- object = kasan_reset_tag(object);
object = restore_red_left(s, object);
if (object < base || object >= base + page->objects * s->size ||
(object - base) % s->size) {
--
2.17.2 (Apple Git-113)

2019-02-13 10:05:59

by Pekka Enberg

Subject: Re: [PATCH] slub: untag object before slab end

On 13/02/2019 4.05, Qian Cai wrote:
> get_freepointer() could return NULL if there is no more free objects in
> the slab. However, it could return a tagged pointer (like
> 0x2200000000000000) with KASAN_SW_TAGS which would escape the NULL
> object checking in check_valid_pointer() and trigger errors below, so
> untag the object before checking for a NULL object there.

Reviewed-by: Pekka Enberg <[email protected]>

2019-02-13 13:45:15

by Andrey Konovalov

Subject: Re: [PATCH] slub: untag object before slab end

On Wed, Feb 13, 2019 at 3:06 AM Qian Cai <[email protected]> wrote:
>
> get_freepointer() could return NULL if there is no more free objects in
> the slab. However, it could return a tagged pointer (like
> 0x2200000000000000) with KASAN_SW_TAGS which would escape the NULL
> object checking in check_valid_pointer() and trigger errors below, so
> untag the object before checking for a NULL object there.

I think this solution is just masking the issue. get_freepointer()
shouldn't return tagged NULLs. Apparently when we save a freelist
pointer, the object where the pointer gets written is tagged
differently, than this same object when the pointer gets read. I found
one case where this happens (the last patch out my 5 patch series),
but apparently there are more.

>
> [ 35.797667] BUG kmalloc-256 (Not tainted): Freepointer corrupt
> [ 35.803584] -----------------------------------------------------------------------------
> [ 35.803584]
> [ 35.813368] Disabling lock debugging due to kernel taint
> [ 35.818766] INFO: Allocated in build_sched_domains+0x28c/0x495c age=92 cpu=0 pid=1
> [ 35.826443] __kmalloc_node+0x4ac/0x508
> [ 35.830343] build_sched_domains+0x28c/0x495c
> [ 35.834764] sched_init_domains+0x184/0x1d8
> [ 35.839012] sched_init_smp+0x38/0xd4
> [ 35.842732] kernel_init_freeable+0x67c/0x1104
> [ 35.847243] kernel_init+0x18/0x2a4
> [ 35.850790] ret_from_fork+0x10/0x18
> [ 35.854423] INFO: Freed in destroy_sched_domain+0xa0/0xcc age=11 cpu=0 pid=1
> [ 35.861569] destroy_sched_domain+0xa0/0xcc
> [ 35.865814] cpu_attach_domain+0x304/0xb34
> [ 35.869971] build_sched_domains+0x4654/0x495c
> [ 35.874480] sched_init_domains+0x184/0x1d8
> [ 35.878724] sched_init_smp+0x38/0xd4
> [ 35.882443] kernel_init_freeable+0x67c/0x1104
> [ 35.886953] kernel_init+0x18/0x2a4
> [ 35.890495] ret_from_fork+0x10/0x18
> [ 35.894128] INFO: Slab 0x(____ptrval____) objects=85 used=0 fp=0x(____ptrval____) flags=0x7ffffffc000200
> [ 35.903733] INFO: Object 0x(____ptrval____) @offset=38528 fp=0x(____ptrval____)
> [ 35.903733]
> [ 35.912637] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
> [ 35.922155] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
> [ 35.931672] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
> [ 35.941190] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
> [ 35.950707] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
> [ 35.960224] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
> [ 35.969741] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
> [ 35.979258] Redzone (____ptrval____): bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb ................
> [ 35.988776] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 35.998206] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.007636] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.017065] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.026494] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.035923] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.045353] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.054783] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.064212] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.073642] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.083071] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.092501] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.101930] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.111359] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.120788] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk
> [ 36.130218] Object (____ptrval____): 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b a5 kkkkkkkkkkkkkkk.
> [ 36.139647] Redzone (____ptrval____): bb bb bb bb bb bb bb bb ........
> [ 36.148462] Padding (____ptrval____): 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
> [ 36.157979] Padding (____ptrval____): 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
> [ 36.167496] Padding (____ptrval____): 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
> [ 36.177021] CPU: 0 PID: 1 Comm: swapper/0 Tainted: G B 5.0.0-rc6+ #41
> [ 36.184854] Call trace:
> [ 36.187328] dump_backtrace+0x0/0x450
> [ 36.191032] show_stack+0x20/0x2c
> [ 36.194385] __dump_stack+0x20/0x28
> [ 36.197911] dump_stack+0xa0/0xfc
> [ 36.201265] print_trailer+0x1a8/0x1bc
> [ 36.205057] object_err+0x40/0x50
> [ 36.208408] check_object+0x214/0x2b8
> [ 36.212111] __free_slab+0x9c/0x31c
> [ 36.215638] discard_slab+0x78/0xa8
> [ 36.219165] kfree+0x918/0x980
> [ 36.222259] destroy_sched_domain+0xa0/0xcc
> [ 36.226489] cpu_attach_domain+0x304/0xb34
> [ 36.230631] build_sched_domains+0x4654/0x495c
> [ 36.235125] sched_init_domains+0x184/0x1d8
> [ 36.239357] sched_init_smp+0x38/0xd4
> [ 36.243060] kernel_init_freeable+0x67c/0x1104
> [ 36.247555] kernel_init+0x18/0x2a4
> [ 36.251083] ret_from_fork+0x10/0x18
>
> Signed-off-by: Qian Cai <[email protected]>
> ---
>
> Depends on slub-fix-slab_consistency_checks-kasan_sw_tags.patch.
>
> mm/slub.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index 4a61959e1887..2fd1cf39914c 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -503,11 +503,11 @@ static inline int check_valid_pointer(struct kmem_cache *s,
> {
> void *base;
>
> + object = kasan_reset_tag(object);
> if (!object)
> return 1;
>
> base = page_address(page);
> - object = kasan_reset_tag(object);
> object = restore_red_left(s, object);
> if (object < base || object >= base + page->objects * s->size ||
> (object - base) % s->size) {
> --
> 2.17.2 (Apple Git-113)
>

2019-02-14 08:35:02

by Qian Cai

Subject: Re: [PATCH] slub: untag object before slab end

On Wed, 2019-02-13 at 11:31 +0100, Andrey Konovalov wrote:
> On Wed, Feb 13, 2019 at 3:06 AM Qian Cai <[email protected]> wrote:
> >
> > get_freepointer() could return NULL if there is no more free objects in
> > the slab. However, it could return a tagged pointer (like
> > 0x2200000000000000) with KASAN_SW_TAGS which would escape the NULL
> > object checking in check_valid_pointer() and trigger errors below, so
> > untag the object before checking for a NULL object there.
>
> I think this solution is just masking the issue. get_freepointer()
> shouldn't return tagged NULLs. Apparently when we save a freelist
> pointer, the object where the pointer gets written is tagged
> differently, than this same object when the pointer gets read. I found
> one case where this happens (the last patch out my 5 patch series),
> but apparently there are more.

Well, the problem is that,

__free_slab
for_each_object(p, s, page_address(page) [1]
check_object(s, page, p ...)
get_freepointer(s, p)

[1]: p += s->size

page_address() tags the address using page_kasan_tag(page), so each "p" here has
that tag.

However, at beginning in allocate_slab(), it tags each object with a random tag,
and then calls set_freepointer(s, p, NULL)

As the result, get_freepointer() returns a tagged NULL because it never be able
to obtain the original tag of the object anymore, and this calculation is now
wrong.

return (void *)((unsigned long)ptr ^ s->random ^ ptr_addr);

This also explain why this patch also works, as it unifies the tags.

https://marc.info/?l=linux-mm&m=154955366113951&w=2

2019-02-14 10:04:12

by Andrey Konovalov

Subject: Re: [PATCH] slub: untag object before slab end

On Wed, Feb 13, 2019 at 10:12 PM Qian Cai <[email protected]> wrote:
>
> On Wed, 2019-02-13 at 11:31 +0100, Andrey Konovalov wrote:
> > On Wed, Feb 13, 2019 at 3:06 AM Qian Cai <[email protected]> wrote:
> > >
> > > get_freepointer() could return NULL if there is no more free objects in
> > > the slab. However, it could return a tagged pointer (like
> > > 0x2200000000000000) with KASAN_SW_TAGS which would escape the NULL
> > > object checking in check_valid_pointer() and trigger errors below, so
> > > untag the object before checking for a NULL object there.
> >
> > I think this solution is just masking the issue. get_freepointer()
> > shouldn't return tagged NULLs. Apparently when we save a freelist
> > pointer, the object where the pointer gets written is tagged
> > differently, than this same object when the pointer gets read. I found
> > one case where this happens (the last patch out my 5 patch series),
> > but apparently there are more.
>
> Well, the problem is that,
>
> __free_slab
> for_each_object(p, s, page_address(page) [1]
> check_object(s, page, p ...)
> get_freepointer(s, p)
>
> [1]: p += s->size
>
> page_address() tags the address using page_kasan_tag(page), so each "p" here has
> that tag.

Ah, I see what the issue is. With those 5 patches page_address()
should return 0xff-tagged pointer here, but when we set_freepointer()
the tag indeed might be different. OK, I think that patch that you
linked below is the better way to deal with this. I've added a
detailed comment to it and sent it.

Thanks!

>
> However, at beginning in allocate_slab(), it tags each object with a random tag,
> and then calls set_freepointer(s, p, NULL)
>
> As the result, get_freepointer() returns a tagged NULL because it never be able
> to obtain the original tag of the object anymore, and this calculation is now
> wrong.
>
> return (void *)((unsigned long)ptr ^ s->random ^ ptr_addr);
>
> This also explain why this patch also works, as it unifies the tags.
>
> https://marc.info/?l=linux-mm&m=154955366113951&w=2
>
>