[PATCH] security/keys: fix slab-out-of-bounds in key_task_permission

[Jarkko Sakkinen]

On Wed Sep 18, 2024 at 10:30 AM EEST, Chen Ridong wrote:
>
>
> On 2024/9/15 21:59, Jarkko Sakkinen wrote:
> > On Sun Sep 15, 2024 at 3:55 AM EEST, Chen Ridong wrote:
> >>
> >>
> >> On 2024/9/14 19:33, Jarkko Sakkinen wrote:
> >>> On Fri Sep 13, 2024 at 10:09 AM EEST, Chen Ridong wrote:
> >>>> We meet the same issue with the LINK, which reads memory out of bounds:
> >>>
> >>> Nit: don't use "we" anywhere".
> >>>
> >>> Tbh, I really don't understand the sentence above. I don't what
> >>> "the same issue with the LINK" really is.
> >>>
> >>
> >> Hello, Jarkko.
> >> I apologize for any confusion caused.
> >>
> >> I've encountered a bug reported by syzkaller. I also found the same bug
> >> reported at this LINK:
> >> https://syzkaller.appspot.com/bug?id=68a5e206c2a8e08d317eb83f05610c0484ad10b9.
> >>
> >>>> BUG: KASAN: slab-out-of-bounds in __kuid_val include/linux/uidgid.h:36
> >>>> BUG: KASAN: slab-out-of-bounds in uid_eq include/linux/uidgid.h:63 [inline]
> >>>> BUG: KASAN: slab-out-of-bounds in key_task_permission+0x394/0x410
> >>>> security/keys/permission.c:54
> >>>> Read of size 4 at addr ffff88813c3ab618 by task stress-ng/4362
> >>>>
> >>>> CPU: 2 PID: 4362 Comm: stress-ng Not tainted 5.10.0-14930-gafbffd6c3ede #15
> >>>> Call Trace:
> >>>>    __dump_stack lib/dump_stack.c:82 [inline]
> >>>>    dump_stack+0x107/0x167 lib/dump_stack.c:123
> >>>>    print_address_description.constprop.0+0x19/0x170 mm/kasan/report.c:400
> >>>>    __kasan_report.cold+0x6c/0x84 mm/kasan/report.c:560
> >>>>    kasan_report+0x3a/0x50 mm/kasan/report.c:585
> >>>>    __kuid_val include/linux/uidgid.h:36 [inline]
> >>>>    uid_eq include/linux/uidgid.h:63 [inline]
> >>>>    key_task_permission+0x394/0x410 security/keys/permission.c:54
> >>>>    search_nested_keyrings+0x90e/0xe90 security/keys/keyring.c:793
> >>>>    keyring_search_rcu+0x1b6/0x310 security/keys/keyring.c:922
> >>>>    search_cred_keyrings_rcu+0x111/0x2e0 security/keys/process_keys.c:459
> >>>>    search_process_keyrings_rcu+0x1d/0x310 security/keys/process_keys.c:544
> >>>>    lookup_user_key+0x782/0x12e0 security/keys/process_keys.c:762
> >>>>    keyctl_invalidate_key+0x20/0x190 security/keys/keyctl.c:434
> >>>>    __do_sys_keyctl security/keys/keyctl.c:1978 [inline]
> >>>>    __se_sys_keyctl+0x1de/0x5b0 security/keys/keyctl.c:1880
> >>>>    do_syscall_64+0x30/0x40 arch/x86/entry/common.c:46
> >>>>    entry_SYSCALL_64_after_hwframe+0x67/0xd1
> >>>>
> >>>> However, we can't reproduce this issue.
> >>>
> >>> "The issue cannot be easily reproduced but by analyzing the code
> >>> it can be broken into following steps:"
> >>
> >> Thank you for your correction.
> >> Does this patch address the issue correctly? Is this patch acceptable?
> > 
> > I only comment new patch versions so not giving any promises but I can
> > say that it is I think definitely in the correct direction :-)
> > 
> > BR, Jarkko
>
> Hello, Jarkko. I have reproduced this issue. It can be reproduced by 
> following these steps:
>
> 1. Add the helper patch.
>
> @@ -205,6 +205,9 @@ static void hash_key_type_and_desc(struct 
> keyring_index_key *index_key)
>          else if (index_key->type == &key_type_keyring && (hash & 
> fan_mask) != 0)
>                  hash = (hash + (hash << level_shift)) & ~fan_mask;
>          index_key->hash = hash;
> +       if ((index_key->hash & 0xff) == 0xe6) {
> +                       pr_err("hash_key_type_and_desc: type %s %s 
> 0x%x\n",  index_key->type->name, index_key->description, index_key->hash);
> +       }
>   }
>
> 2. Pick up the inputs whose hash is xxe6 using the following cmd. If a 
> key's hash is xxe6, it will be printed.
>
> for ((i=0; i<=10000; i++)); do ./test_key user user$i "Some payload"; done
>
> You have complile test_key whith following code.
>
> #include <sys/types.h>
> #include <keyutils.h>
> #include <stdint.h>
> #include <stdio.h>
> #include <stdlib.h>
> #include <string.h>
>
> int
> main(int argc, char *argv[])
> {
>     key_serial_t key;
>
>     if (argc != 4) {
> 	   fprintf(stderr, "Usage: %s type description payload\n",
> 			   argv[0]);
> 	   exit(EXIT_FAILURE);
>     }
>
>     key = add_key(argv[1], argv[2], argv[3], strlen(argv[3]),
> 			   KEY_SPEC_SESSION_KEYRING);
>     if (key == -1) {
> 	   perror("add_key");
> 	   exit(EXIT_FAILURE);
>     }
>
>     printf("Key ID is %jx\n", (uintmax_t) key);
>
>     exit(EXIT_SUCCESS);
> }
>
>
> 3. Have more than 32 inputs now. their hashes are xxe6.
> eg.
> hash_key_type_and_desc: type user user438 0xe3033fe6
> hash_key_type_and_desc: type user user526 0xeb7eade6
> ...
> hash_key_type_and_desc: type user user9955 0x44bc99e6
>
> 4. Reboot and add the keys obtained from step 3.
> When adding keys to the ROOT that their hashes are all xxe6, and up to 
> 16, the ROOT has keys with hashes that are not xxe6 (e.g., slot 0), so 
> the keys are dissimilar. The ROOT will then split NODE A without using a 
> shortcut. When NODE A is filled with keys that have hashes of xxe6, the 
> keys are similar. NODE A will split with a shortcut.
>
> As my analysis, if a slot of the root is a shortcut(slot 6), it may be 
> mistakenly be transferred to a key*, leading to an read out-of-bounds read.
>
>                        NODE A
>                +------>+---+
>        ROOT    |       | 0 | xxe6
>        +---+   |       +---+
>   xxxx | 0 | shortcut  :   : xxe6
>        +---+   |       +---+
>   xxe6 :   :   |       |   | xxe6
>        +---+   |       +---+
>        | 6 |---+       :   : xxe6
>        +---+           +---+
>   xxe6 :   :           | f | xxe6
>        +---+           +---+
>   xxe6 | f |
>        +---+
>
> 5. cat /proc/keys. and the issue is reproduced.

Hi, I'll try to run through your procedure next week and give my comments. 
Thanks for doing this.

BR, Jarkko