On 26.02.2018 05:04, kernel test robot wrote:
>
> FYI, we noticed the following commit (built with gcc-7):
>
> commit: 37b927536f46b27de240934db9580c3d3eba9937 ("net: Introduce net_sem for
> protection of pernet_list")
> url:
> https://github.com/0day-ci/linux/commits/Kirill-Tkhai/Replacing-net_mutex-with-rw_semaphore/20180215-064145
>
>
> in testcase: trinity
> with following parameters:
>
> runtime: 300s
>
> test-description: Trinity is a linux system call fuzz tester.
> test-url: http://codemonkey.org.uk/projects/trinity/
>
>
> on test machine: qemu-system-x86_64 -enable-kvm -m 512M
>
> caused below changes (please refer to attached dmesg/kmsg for entire
> log/backtrace):
>
>
> +------------------------------------------+------------+------------+
> | | b6d8347bf7 | 37b927536f |
> +------------------------------------------+------------+------------+
> | boot_successes | 8 | 0 |
> | boot_failures | 0 | 8 |
> | kernel_BUG_at_net/core/net_namespace.c | 0 | 8 |
> | invalid_opcode:#[##] | 0 | 8 |
> | RIP:ops_init | 0 | 8 |
> | Kernel_panic-not_syncing:Fatal_exception | 0 | 8 |
> +------------------------------------------+------------+------------+
>
>
>
> [ 0.258559] kernel BUG at net/core/net_namespace.c:73!
> [ 0.260018] invalid opcode: 0000 [#1]
> [ 0.261073] Modules linked in:
> [ 0.262005] CPU: 0 PID: 1 Comm: swapper Not tainted 4.15.0-05871-g37b9275
> #1
> [ 0.263875] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
> 1.10.2-1 04/01/2014
> [ 0.266084] RIP: 0010:ops_init+0x101/0x120
> [ 0.267233] RSP: 0000:ffffc9000000be00 EFLAGS: 00010246
> [ 0.268647] RAX: 0000000000000000 RBX: ffffffff823280a0 RCX:
> 0000000000000000
> [ 0.270000] RDX: 0000000000000000 RSI: 0000000000000000 RDI:
> ffff88001c602180
> [ 0.270000] RBP: ffff88001c602140 R08: ffffffff82323d00 R09:
> ffff88001c602140
> [ 0.270000] R10: ffffffff82323308 R11: ffffffff827ff64d R12:
> ffffffff82323400
> [ 0.270000] R13: 0000000000000003 R14: ffffffff826e17b9 R15:
> 0000000000000000
> [ 0.270000] FS: 0000000000000000(0000) GS:ffffffff82235000(0000)
> knlGS:0000000000000000
> [ 0.270000] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 0.270000] CR2: 0000000000000000 CR3: 0000000002213000 CR4:
> 00000000000006b0
> [ 0.270000] Call Trace:
> [ 0.270000] ? set_debug_rodata+0x11/0x11
> [ 0.270000] register_pernet_operations+0x101/0x1e0
> [ 0.270000] register_pernet_subsys+0x24/0x40
> [ 0.270000] netlink_proto_init+0x143/0x160
> [ 0.270000] ? eth_offload_init+0x14/0x14
> [ 0.270000] do_one_initcall+0x4d/0x190
> [ 0.270000] kernel_init_freeable+0x10e/0x190
> [ 0.270000] ? rest_init+0xa0/0xa0
> [ 0.270000] kernel_init+0xa/0xf0
> [ 0.270000] ret_from_fork+0x3a/0x50
> [ 0.270000] Code: 8d 14 c5 e8 ff ff ff e8 be 14 11 00 4b 89 2c ef 49 8d 7e
> 08 4d 89 bc 24 50 06 00 00 be 08 00 00 00 e8 34 22 95 ff e9 77 ff ff ff <0f>
> 0b 0f 0b b8 f4 ff ff ff eb 91 b8 f4 ff ff ff e9 77 ff ff ff
> [ 0.270000] RIP: ops_init+0x101/0x120 RSP: ffffc9000000be00
> [ 0.270013] ---[ end trace e0045ead2b146f98 ]---
There was the check:
BUG_ON(!mutex_is_locked(&net_mutex));
which was killed in one of next commits of this patchset:
commit 447cd7a0d7d1e5b4486e99cce289654fec9951e3
Author: Kirill Tkhai <ktk...@virtuozzo.com>
Date: Tue Feb 13 12:26:44 2018 +0300
net: Allow pernet_operations to be executed in parallel
So, the whole patchset as an atomic change is safe; and the only thing
is the hunk removing the BUG_ON() had to go in another patch.
Anyway, thanks for testing, Xiaolong.
Kirill
