[syzbot] [bpf?] general protection fault in bpf_prog_offload_verifier_prep

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit: fa09bc40b21a igb: disable virtualization features on 82580
git tree: net
console+strace: https://syzkaller.appspot.com/x/log.txt?x=13382fa8680000
kernel config: https://syzkaller.appspot.com/x/.config?x=634e05b4025da9da
dashboard link: https://syzkaller.appspot.com/bug?extid=291100dcb32190ec02a8
compiler: gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40
syz repro: https://syzkaller.appspot.com/x/repro.syz?x=1529c448680000
C reproducer: https://syzkaller.appspot.com/x/repro.c?x=15db0248680000

Downloadable assets:
disk image: https://storage.googleapis.com/syzbot-assets/7ab461d84992/disk-fa09bc40.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/3ac6d43ab2db/vmlinux-fa09bc40.xz
kernel image: https://storage.googleapis.com/syzbot-assets/778d096a134e/bzImage-fa09bc40.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+291100...@syzkaller.appspotmail.com

general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007]
CPU: 1 PID: 5055 Comm: syz-executor625 Not tainted 6.5.0-syzkaller-04012-gfa09bc40b21a #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/26/2023
RIP: 0010:bpf_prog_offload_verifier_prep+0xaa/0x170 kernel/bpf/offload.c:295
Code: 00 fc ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 0f 85 a1 00 00 00 48 b8 00 00 00 00 00 fc ff df 4c 8b 65 10 4c 89 e2 48 c1 ea 03 <80> 3c 02 00 0f 85 93 00 00 00 48 b8 00 00 00 00 00 fc ff df 4d 8b
RSP: 0018:ffffc900039ff7f8 EFLAGS: 00010246
RAX: dffffc0000000000 RBX: ffffc9000156e000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffffffff81a8cf76 RDI: ffff888021b25f10
RBP: ffff888021b25f00 R08: 0000000000000001 R09: fffffbfff195203d
R10: ffffffff8ca901ef R11: 0000000000000000 R12: 0000000000000000
R13: 0000000000000005 R14: 0000000000000003 R15: ffffc9000156e060
FS: 0000555556071380(0000) GS:ffff8880b9900000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000020000100 CR3: 0000000022f6b000 CR4: 00000000003506e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
<TASK>
bpf_check+0x52f3/0xabd0 kernel/bpf/verifier.c:19762
bpf_prog_load+0x153a/0x2270 kernel/bpf/syscall.c:2708
__sys_bpf+0xbb6/0x4e90 kernel/bpf/syscall.c:5335
__do_sys_bpf kernel/bpf/syscall.c:5439 [inline]
__se_sys_bpf kernel/bpf/syscall.c:5437 [inline]
__x64_sys_bpf+0x78/0xc0 kernel/bpf/syscall.c:5437
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
RIP: 0033:0x7f7c0df78ea9
Code: 28 00 00 00 75 05 48 83 c4 28 c3 e8 d1 19 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 b8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ffde3592128 EFLAGS: 00000246 ORIG_RAX: 0000000000000141
RAX: ffffffffffffffda RBX: 0000000000000003 RCX: 00007f7c0df78ea9
RDX: 0000000000000090 RSI: 0000000020000940 RDI: 0000000000000005
RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000100000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
</TASK>
Modules linked in:
---[ end trace 0000000000000000 ]---
RIP: 0010:bpf_prog_offload_verifier_prep+0xaa/0x170 kernel/bpf/offload.c:295
Code: 00 fc ff df 48 89 fa 48 c1 ea 03 80 3c 02 00 0f 85 a1 00 00 00 48 b8 00 00 00 00 00 fc ff df 4c 8b 65 10 4c 89 e2 48 c1 ea 03 <80> 3c 02 00 0f 85 93 00 00 00 48 b8 00 00 00 00 00 fc ff df 4d 8b
RSP: 0018:ffffc900039ff7f8 EFLAGS: 00010246
RAX: dffffc0000000000 RBX: ffffc9000156e000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffffffff81a8cf76 RDI: ffff888021b25f10
RBP: ffff888021b25f00 R08: 0000000000000001 R09: fffffbfff195203d
R10: ffffffff8ca901ef R11: 0000000000000000 R12: 0000000000000000
R13: 0000000000000005 R14: 0000000000000003 R15: ffffc9000156e060
FS: 0000555556071380(0000) GS:ffff8880b9900000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000020000100 CR3: 0000000022f6b000 CR4: 00000000003506e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
----------------
Code disassembly (best guess), 3 bytes skipped:
0: df 48 89 fisttps -0x77(%rax)
3: fa cli
4: 48 c1 ea 03 shr $0x3,%rdx
8: 80 3c 02 00 cmpb $0x0,(%rdx,%rax,1)
c: 0f 85 a1 00 00 00 jne 0xb3
12: 48 b8 00 00 00 00 00 movabs $0xdffffc0000000000,%rax
19: fc ff df
1c: 4c 8b 65 10 mov 0x10(%rbp),%r12
20: 4c 89 e2 mov %r12,%rdx
23: 48 c1 ea 03 shr $0x3,%rdx
* 27: 80 3c 02 00 cmpb $0x0,(%rdx,%rax,1) <-- trapping instruction
2b: 0f 85 93 00 00 00 jne 0xc4
31: 48 b8 00 00 00 00 00 movabs $0xdffffc0000000000,%rax
38: fc ff df
3b: 4d rex.WRB
3c: 8b .byte 0x8b


---
This report is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzk...@googlegroups.com.

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

If the bug is already fixed, let syzbot know by replying with:
#syz fix: exact-commit-title

If you want syzbot to run the reproducer, reply with:
#syz test: git://repo/address.git branch-or-commit-hash
If you attach or paste a git patch, syzbot will apply it before testing.

If you want to overwrite bug's subsystems, reply with:
#syz set subsystems: new-subsystem
(See the list of subsystem names on the web dashboard)

If the bug is a duplicate of another bug, reply with:
#syz dup: exact-subject-of-another-report

If you want to undo deduplication, reply with:
#syz undup

[Eduard Zingerman]

I have an explanation of why this error occurs, but I need an advice
on how to fix it.

Then NULL pointer deference occurs in the following function from offload.c:

int bpf_prog_offload_verifier_prep(struct bpf_prog *prog)
{
struct bpf_prog_offload *offload;
int ret = -ENODEV;

down_read(&bpf_devs_lock);
offload = prog->aux->offload;
if (offload) {
ret = offload->offdev->ops->prepare(prog);
^^^^^^
this pointer is NULL
offload->dev_state = !ret;
}
up_read(&bpf_devs_lock);

return ret;
}

# Short explanation

(a) call chain bpf_prog_load -> bpf_prog_dev_bound_init -> __bpf_prog_dev_bound_init
-> __bpf_offload_dev_netdev_register
might insert an instance of struct bpf_offload_netdev with {.offdev == NULL}
into hash table offload.c:offdevs;
(b) call chain bpf_prog_load -> bpf_check -> bpf_prog_offload_verifier_prep
assumes that from (prog->aux->offload != NULL)
follows (prog->aux->offload->offdev != NULL)
which is not the case because of (a).

# Long explanation

The reproducer generated by testbot has the following structure:
- in a loop call function execute_one(), which does the following
system calls in sequence:
- socket(AF_INET6, SOCK_RAW, IPPROTO_IGMP) = <some fd>
- ioctl(3, SIOCGIFINDEX, {ifr_name="batadv_slave_1"}) = 0
- bpf(BPF_PROG_LOAD,
{prog_type=BPF_PROG_TYPE_XDP, ... prog_flags=0x40, prog_ifindex=29, ...}) = -1 EINVAL
(referred to as program #1 below)
- socket(AF_INET6, SOCK_RAW, IPPROTO_IGMP) = <some fd>
- ioctl(4, SIOCGIFINDEX, {ifr_name="batadv_slave_1"}) = 0
- bpf(BPF_PROG_LOAD,
{prog_type=BPF_PROG_TYPE_XDP, ... prog_flags=0, ... prog_ifindex=29}) = -1 EINVAL
(referred to as program #2 below)

The error occurs when second bpf call is processed.
Interestingly, if sleep(1) is inserted somewhere between first and
second bpf calls error does not occur:

@@ -1246,6 +1246,7 @@ void execute_one(void)
*(uint32_t*)0x200009cc = 4;
syscall(__NR_bpf, /*cmd=*/5ul, /*arg=*/0x20000940ul, /*size=*/0x90ul);
res = syscall(__NR_socket, /*domain=*/0xaul, /*type=*/3ul, /*proto=*/2);
+ // sleep(1); /* uncomment to hide the error */
if (res != -1)
r[2] = res;
memcpy((void*)0x20000100, "batadv_slave_1\000\000", 16);

## Control flow when error occurs

For program #1:
- bpf_prog_load():
- bpf_prog_is_dev_bound(prog->aux) is true
- bpf_prog_dev_bound_init
- prog->aux->offload_requested is 0 (because of 0x40 prog_flags)
- __bpf_prog_dev_bound_init
- netdev is "batadv_slave_1"
- bpf_offload_find_netdev(offload->netdev) == NULL,
(this is a lookup in hash table offload.c:offdevs)
which triggers a call to __bpf_offload_dev_netdev_register
- __bpf_offload_dev_netdev_register(NULL, offload->netdev)
registers struct bpf_offload_netdev with {.offdev = NULL}
for netdev "batadv_slave_1" in offload.c:offdevs hash table.

For program #2:
- bpf_prog_load():
- bpf_prog_is_dev_bound(prog->aux) is true
- bpf_prog_dev_bound_init
- prog->aux->offload_requested is 1 (because of 0x0 prog_flags)
- __bpf_prog_dev_bound_init
- netdev is "batadv_slave_1"
- bpf_offload_find_netdev(offload->netdev) != NULL,
this is struct bpf_offload_netdev with {.offdev = NULL}
created for program #1
- prog->aux->offload = struct bpf_prog_offload {.offload -> {.offdev = NULL}},
The bpf_prog_offload remembered for prog points to bpf_offload_netdev
with .offdev == NULL.
- ...
- bpf_check
- bpf_prog_offload_verifier_prep
- prog->aux->offload != NULL, but prog->aux->offload->offdev == NULL
=> null pointer deference.

## Control flow when error does not occur

For program #1:
- ... all as in the previous case ...

Some worker thread:
- kernel/bpf/core.c:bpf_prog_free_deferred, registered for program #1:
- bpf_prog_is_dev_bound(aux) is true
- bpf_prog_dev_bound_destroy
- netdev is "batadv_slave_1"
- (!ondev->offdev && list_empty(&ondev->progs)) is true
- __bpf_offload_dev_netdev_unregister
this removes struct bpf_offload_netdev with {.offdev = NULL}
from offload.c:offdevs hash table.

For program #2:
- bpf_prog_load():
- bpf_prog_is_dev_bound(prog->aux) is true
- bpf_prog_dev_bound_init
- prog->aux->offload_requested is 1 (because of 0x0 prog_flags)
- __bpf_prog_dev_bound_init
- netdev is "batadv_slave_1"
- bpf_offload_find_netdev(offload->netdev) == NULL
- bpf_prog_is_offloaded(prog->aux) is true
- -EINVAL is returned.

[Eduard Zingerman]

I think the fix should look as follows:

diff --git a/kernel/bpf/offload.c b/kernel/bpf/offload.c
index 3e4f2ec1af06..302e38bffffa 100644
--- a/kernel/bpf/offload.c
+++ b/kernel/bpf/offload.c
@@ -199,12 +199,11 @@ static int __bpf_prog_dev_bound_init(struct bpf_prog *prog, struct net_device *n
offload->netdev = netdev;

ondev = bpf_offload_find_netdev(offload->netdev);
+ if (bpf_prog_is_offloaded(prog->aux) && (!ondev || !ondev->offdev)) {
+ err = -EINVAL;
+ goto err_free;
+ }
if (!ondev) {
- if (bpf_prog_is_offloaded(prog->aux)) {
- err = -EINVAL;
- goto err_free;
- }
-
/* When only binding to the device, explicitly
* create an entry in the hashtable.
*/

With the following reasoning: for offloaded programs offload device
should exist and it should not be a fake device create in !ondev branch.

Stanislav, could you please take a look? I think this is related to commit:
2b3486bc2d23 ("bpf: Introduce device-bound XDP programs")

[Daniel Borkmann]

[ Also adding Simon to Cc ]

[Jakub Kicinski]

On Wed, 06 Sep 2023 16:50:23 +0300 Eduard Zingerman wrote:
> diff --git a/kernel/bpf/offload.c b/kernel/bpf/offload.c
> index 3e4f2ec1af06..302e38bffffa 100644
> --- a/kernel/bpf/offload.c
> +++ b/kernel/bpf/offload.c
> @@ -199,12 +199,11 @@ static int __bpf_prog_dev_bound_init(struct bpf_prog *prog, struct net_device *n
> offload->netdev = netdev;
>
> ondev = bpf_offload_find_netdev(offload->netdev);
> + if (bpf_prog_is_offloaded(prog->aux) && (!ondev || !ondev->offdev)) {
> + err = -EINVAL;
> + goto err_free;
> + }
> if (!ondev) {
> - if (bpf_prog_is_offloaded(prog->aux)) {
> - err = -EINVAL;
> - goto err_free;
> - }
> -
> /* When only binding to the device, explicitly
> * create an entry in the hashtable.
> */

LGTM, FWIW.

[Eduard Zingerman]

On Wed, 2023-09-06 at 07:57 -0700, Jakub Kicinski wrote:
> On Wed, 06 Sep 2023 16:50:23 +0300 Eduard Zingerman wrote:
> > diff --git a/kernel/bpf/offload.c b/kernel/bpf/offload.c
> > index 3e4f2ec1af06..302e38bffffa 100644
> > --- a/kernel/bpf/offload.c
> > +++ b/kernel/bpf/offload.c
> > @@ -199,12 +199,11 @@ static int __bpf_prog_dev_bound_init(struct bpf_prog *prog, struct net_device *n
> > offload->netdev = netdev;
> >
> > ondev = bpf_offload_find_netdev(offload->netdev);
> > + if (bpf_prog_is_offloaded(prog->aux) && (!ondev || !ondev->offdev)) {
> > + err = -EINVAL;
> > + goto err_free;
> > + }
> > if (!ondev) {
> > - if (bpf_prog_is_offloaded(prog->aux)) {
> > - err = -EINVAL;
> > - goto err_free;
> > - }
> > -
> > /* When only binding to the device, explicitly
> > * create an entry in the hashtable.
> > */
>
> LGTM, FWIW.

Thanks, I'll wrap it up as a proper patch with a test.

[Stanislav Fomichev]

LGTM as well, thanks!

[syzbot]

syzbot has bisected this issue to:

commit 47a71c1f9af0a334c9dfa97633c41de4feda4287
Author: Andrii Nakryiko <and...@kernel.org>
Date: Thu Apr 6 23:41:58 2023 +0000

bpf: Add log_true_size output field to return necessary log buffer size

bisection log: https://syzkaller.appspot.com/x/bisect.txt?x=132a4268680000
start commit: fa09bc40b21a igb: disable virtualization features on 82580
git tree: net
final oops: https://syzkaller.appspot.com/x/report.txt?x=10aa4268680000
console output: https://syzkaller.appspot.com/x/log.txt?x=172a4268680000

kernel config: https://syzkaller.appspot.com/x/.config?x=634e05b4025da9da
dashboard link: https://syzkaller.appspot.com/bug?extid=291100dcb32190ec02a8

syz repro: https://syzkaller.appspot.com/x/repro.syz?x=1529c448680000
C reproducer: https://syzkaller.appspot.com/x/repro.c?x=15db0248680000

Reported-by: syzbot+291100...@syzkaller.appspotmail.com
Fixes: 47a71c1f9af0 ("bpf: Add log_true_size output field to return necessary log buffer size")

For information about bisection process see: https://goo.gl/tpsmEJ#bisection