Re: [PATCH] ceph: fix multiple unsafe decodes in decode_locker()

[Viacheslav Dubeyko]

On Sat, 2026-05-23 at 04:59 -0400, Pavitra Jha wrote:
> decode_locker() in cls_lock_client.c contains three unsafe decode
> operations that allow a malicious or compromised OSD to trigger
> slab-out-of-bounds reads:
> 
> 1. ceph_decode_copy() at the locker_id_t name field has no preceding
>    bounds check. With p == end after ceph_start_decoding() accepts
>    struct_len=0, this reads sizeof(ceph_entity_name) = 9 bytes past
>    the validated buffer boundary.
> 
> 2. *p += sizeof(struct ceph_timespec) after the locker_info_t header
>    is an unchecked pointer advance. A malicious OSD can position p
>    past end, causing all subsequent _safe checks to pass against a
>    bogus boundary.
> 
> 3. len = ceph_decode_32(p) has no preceding bounds check, and the
>    immediately following *p += len is uncapped. A malicious OSD can
>    send len=0xffffffff, advancing p gigabytes past end and escaping
>    the decode window entirely.
> 
> Fix all three by replacing bare operations with their safe variants:
>   ceph_decode_copy   -> ceph_decode_copy_safe
>   *p += sizeof(...)  -> ceph_decode_skip_n
>   ceph_decode_32(p)  -> ceph_decode_32_safe
>   *p += len          -> ceph_decode_skip_n
> 
> A new bad: label is added to return -EINVAL on any bounds violation.
> 
> KASAN report (kernel 7.0.0-rc7, QEMU/x86_64, KASLR disabled):
> 
>   [   26.183969] ceph_oob4_poc: buf=ffff888009e31000 end=ffff888009e31fa0
>   [   26.186087] ceph_oob4_poc: struct_v=1 struct_len=0 p==end: 1
>   [   26.186738] ceph_oob4_poc: triggering bare ceph_decode_32 past slab boundary...
>   [   26.187679] ==================================================================
>   [   26.188236] BUG: KASAN: slab-out-of-bounds in ceph_oob4_init+0x22b/0xff0 [ceph_oob4_poc]
>   [   26.188236] Read of size 4 at addr ffff888009e31fa0 by task insmod/59
>   [   26.188236] CPU: 0 UID: 0 PID: 59 Comm: insmod Tainted: G           O        7.0.0-rc7-g9c2abf69da83-dirty #15 PREEMPT(lazy)
>   [   26.188236] Tainted: [O]=OOT_MODULE
>   [   26.188236] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.17.0-debian-1.17.0-1 04/01/2014
>   [   26.188236] Call Trace:
>   [   26.188236]  <TASK>
>   [   26.188236]  dump_stack_lvl+0x4d/0x70
>   [   26.188236]  print_report+0x170/0x4f3
>   [   26.188236]  ? __pfx__raw_spin_lock_irqsave+0x10/0x10
>   [   26.188236]  kasan_report+0xda/0x110
>   [   26.188236]  ? ceph_oob4_init+0x22b/0xff0 [ceph_oob4_poc]
>   [   26.188236]  ? ceph_oob4_init+0x22b/0xff0 [ceph_oob4_poc]
>   [   26.188236]  ? __pfx_ceph_oob4_init+0x10/0x10 [ceph_oob4_poc]
>   [   26.188236]  ceph_oob4_init+0x22b/0xff0 [ceph_oob4_poc]
>   [   26.188236]  do_one_initcall+0x9a/0x3a0
>   [   26.188236]  ? __pfx_do_one_initcall+0x10/0x10
>   [   26.188236]  do_init_module+0x27c/0x790
>   [   26.188236]  load_module+0x4a9a/0x6350
>   [   26.188236]  init_module_from_file+0x15c/0x180
>   [   26.188236]  idempotent_init_module+0x21f/0x750
>   [   26.188236]  __x64_sys_finit_module+0xba/0x120
>   [   26.188236]  do_syscall_64+0xe2/0x570
>   [   26.188236]  entry_SYSCALL_64_after_hwframe+0x77/0x7f
>   [   26.188236]  </TASK>
>   [   26.188236] The buggy address belongs to the object at ffff888009e31000
>   [   26.188236]  which belongs to the cache kmalloc-4k of size 4096
>   [   26.188236] The buggy address is located 0 bytes to the right of
>   [   26.188236]  allocated 4000-byte region [ffff888009e31000, ffff888009e31fa0)
>   [   26.188236] Memory state around the buggy address:
>   [   26.188236]  ffff888009e31f00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>   [   26.188236] >ffff888009e31f80: 00 00 00 00 fc fc fc fc fc fc fc fc fc fc fc fc
>   [   26.188236]                                ^
>   [   26.188236]  ffff888009e32000: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>   [   26.188236] ==================================================================
>   [   26.255513] ceph_oob4_poc: len=0xcccccccc (OOB garbage from KASAN redzone)
> 
>   0xCCCCCCCC is KASAN redzone poison, confirming the read landed in
>   the slab redzone immediately past the 4000-byte allocation.
> 
> Attacker model: a malicious or compromised OSD in a multi-tenant Ceph
> deployment can trigger this against any kernel client that issues the
> lock.get_info class method (e.g. during RBD exclusive lock acquisition)
> without any further privileges beyond OSD session establishment.
> 
> Fixes: d4ed4a530562 ("libceph: support for lock.lock_info")
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Pavitra Jha <jhapavitra98@xxxxxxxxx>
> ---
>  net/ceph/cls_lock_client.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/net/ceph/cls_lock_client.c b/net/ceph/cls_lock_client.c
> index 78276273c..00f0309a6 100644
> --- a/net/ceph/cls_lock_client.c
> +++ b/net/ceph/cls_lock_client.c
> @@ -259,7 +259,7 @@ static int decode_locker(void **p, void *end, struct ceph_locker *locker)
>  	if (ret)
>  		return ret;
>  
> -	ceph_decode_copy(p, &locker->id.name, sizeof(locker->id.name));
> +	ceph_decode_copy_safe(p, end, &locker->id.name, sizeof(locker->id.name), bad);

Are you sure that this line not longer than 80 symbols?

>  	s = ceph_extract_encoded_string(p, end, NULL, GFP_NOIO);
>  	if (IS_ERR(s))
>  		return PTR_ERR(s);
> @@ -270,19 +270,21 @@ static int decode_locker(void **p, void *end, struct ceph_locker *locker)
>  	if (ret)
>  		return ret;
>  
> -	*p += sizeof(struct ceph_timespec); /* skip expiration */
> +	ceph_decode_skip_n(p, end, sizeof(struct ceph_timespec), bad); /* skip expiration */

I don't think that it makes sense to keep comment at the end of line now. Let's
do it in such way:

/* skip expiration */
ceph_decode_skip_n(p, end, sizeof(struct ceph_timespec), bad);

>  
>  	ret = ceph_decode_entity_addr(p, end, &locker->info.addr);
>  	if (ret)
>  		return ret;
>  
> -	len = ceph_decode_32(p);
> -	*p += len; /* skip description */
> +	ceph_decode_32_safe(p, end, len, bad);
> +	ceph_decode_skip_n(p, end, len, bad); /* skip description */

Ditto.

/* skip description */
ceph_decode_skip_n(p, end, len, bad);

>  
>  	dout("%s %s%llu cookie %s addr %s\n", __func__,
>  	     ENTITY_NAME(locker->id.name), locker->id.cookie,
>  	     ceph_pr_addr(&locker->info.addr));
>  	return 0;
> +bad:

You are not consistent. Now it's bad name but not e_inval. :)

> +	return -EINVAL;

I still think that we have not input arguments here. I am not fully sure that
EINVAL is proper error code here.

Thanks,
Slava.

> 
> 

>  }
>  
>  static int decode_lockers(void **p, void *end, u8 *type, char **tag,