#69 Downed soft mounts can panic

[David Zafman]

0xc232c000 263493 263447 1 1 R 0xc232c420 *find
EBP EIP Function (args)
0xc232df2c 0xc023bf45 cfs_reqlist_free+0x55 (0xc32ff760,
0xc365f000, 0xc232c000,
0xc365f050)
kernel .text
0xc0100000 0xc023bef0 0xc023bf80
0xc232df44 0xc022279f cfs_put_super+0x1f (0xc365f000,
0xc0511360, 0xc05113cc, 0x
c127cc00, 0xc365f000)
kernel .text
0xc0100000 0xc0222780 0xc02227d0
0xc232df64 0xc0158773 kill_super+0x193 (0xc365f000,
0xc127cc00, 0xc36104e0, 0xc36c2720, 0x0)
kernel .text
0xc0100000 0xc01585e0 0xc01587a0
0xc232df80 0xc016df50 __mntput+0x20 (0xc127cc00, 0xc232c000,
0x100, 0x1)
kernel .text
0xc0100000 0xc016df30 0xc016df60
0xc232df98 0xc0127b46 do_exit+0x216 (0x100, 0xc232c000,
0x42130910, 0x1)
kernel .text
0xc0100000 0xc0127930 0xc0127ce0
0xc232dfb0 0xc0127e23 do_group_exit+0x83 (0x100)
kernel .text
0xc0100000 0xc0127da0 0xc0127e50
0xc232dfbc 0xc0127e61 sys_exit_group+0x11 (0x1, 0x1,
0x4212e130, 0x42130910, 0x1
)
kernel .text
0xc0100000 0xc0127e50 0xc0127e70
0xc010b3f3 system_call+0x33
kernel .text
0xc0100000 0xc010b3c0 0xc010b400

The super_block here has an a flags 0x81 which is MI_SB_DOWN |
MI_CLEANED. This means that the routine super_clean_node() has
been run because it is a soft mount. It has mi_server of 1 which is
the node that just went down. The super block may have already
been killed at this point which may mean that the code isn't
handling refcounts properly.

[Roger Tsang]

Logged In: YES
user_id=1246761
Originator: NO

Bug report is outdated (w.r.t. SSI-1.9) since we moved to newer base kernel-2.6.

Before we close this bug we need to verify whether this problem exists in SSI-1.9 / 2.0.0preX