bcache: BUG in fuzz testing without devices

Hi all,

It seems that trying to fuzz bcache without any devices triggers a BUG:

sh-4.2# echo 1 > /sys/fs/bcache/fuzz

[   89.700458] bio: create slab <bio-1> at 1
[   89.704543] ------------[ cut here ]------------
[   89.710043] kernel BUG at drivers/md/bcache/debug.c:447!
[   89.710043] invalid opcode: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
[   89.710043] Modules linked in:
[   89.710043] CPU 2
[   89.710043] Pid: 6885, comm: sh Tainted: G        W    3.9.0-rc4-next-20130328-sasha-00017-g1463000 #321
[   89.710043] RIP: 0010:[<ffffffff8324c005>]  [<ffffffff8324c005>] btree_fuzz+0x95/0x720
[   89.710043] RSP: 0018:ffff880035279be8  EFLAGS: 00010246
[   89.710043] RAX: ffff880034720000 RBX: ffff880034721030 RCX: ffff8800353af400
[   89.710043] RDX: 0000000000000000 RSI: ffff880034723b18 RDI: ffff88003477a960
[   89.710043] RBP: ffff880035279e88 R08: ffff88003477a900 R09: ffff88003d801e00
[   89.710043] R10: 000000000048d002 R11: 0000000000000000 R12: ffff880035279c18
[   89.710043] R13: ffff880034721030 R14: ffff880034720000 R15: ffff880035279c30
[   89.710043] FS:  00007f591db62700(0000) GS:ffff88003e000000(0000) knlGS:0000000000000000
[   89.710043] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   89.710043] CR2: 00007f591db8e000 CR3: 0000000035bf0000 CR4: 00000000000406e0
[   89.710043] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   89.710043] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[   89.710043] Process sh (pid: 6885, threadinfo ffff880035278000, task ffff880037060000)
[   89.710043] Stack:
[   89.710043]  ffffffff84c1b104 ffffffff83e20320 000000000000d8e0 ffff880035279ca8
[   89.710043]  ffff88003e00a000 ffff880035278000 ffff880035279fd8 ffff88003e1d7cd0
[   89.710043]  ffff880035279c58 0000000000000000 0000000000000000 0000000000000000
[   89.710043] Call Trace:
[   89.710043]  [<ffffffff812ffd92>] ? sysfs_write_file+0xe2/0x150
[   89.710043]  [<ffffffff819fff5f>] kobj_attr_store+0xf/0x20
[   89.710043]  [<ffffffff812ffdaa>] sysfs_write_file+0xfa/0x150
[   89.710043]  [<ffffffff8127eff0>] vfs_write+0xb0/0x180
[   89.710043]  [<ffffffff8127f480>] SyS_write+0x50/0xa0
[   89.710043]  [<ffffffff83db7e58>] tracesys+0xe1/0xe6
[   89.710043] Code: c0 49 89 c6 0f 84 2c 06 00 00 4c 8d a5 90 fd ff ff 4c 8d a8 30 10 00 00 4d 89 e7 49 83 c7 18 49 8b 9e 30 10
00 00 4c 39 eb 75 0b <0f> 0b 66 0f 1f 84 00 00 00 00 00 48 8d 83 a0 fd ff ff 48 89 df
[   89.710043] RIP  [<ffffffff8324c005>] btree_fuzz+0x95/0x720
[   89.710043]  RSP <ffff880035279be8>
[   89.755259] ---[ end trace d6647580839fba22 ]---

That BUG looks very intentional there, any reason for adding it instead of exiting if there aren't any devices?

Thanks,
Sasha

The fuzz tester is meant to stress test some of the btree code - it's a purely in memory test. But it looks like it's rotted - thanks for the bug report, I'll disable it for now.