lkml.org 
[lkml]   [2015]   [Nov]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH 20/40] staging: lustre: fix kernel crash when network failed to start
Date
From: Amir Shehata <amir.shehata@intel.com>

When loading Lustre modules without proper network configuration,
it always hit the following kernel panic:
LNetError: 105-4: Error -100 starting up LNI tcp
LNetError: 2145:0:(api-ni.c:823:lnet_unprepare())
ASSERTION( list_empty(&the_lnet.ln_nis) ) failed:
NetError: 2145:0:(api-ni.c:823:lnet_unprepare()) LBUG
Pid: 2145, comm: modprobe
x0aCall Trace:
[<ffffffffa044f853>] libcfs_debug_dumpstack+0x53/0x80 [libcfs]
[<ffffffffa044fdf5>] lbug_with_loc+0x45/0xc0 [libcfs]
[<ffffffffa04f3267>] lnet_unprepare+0x297/0x340 [lnet]
[<ffffffffa04f3b5c>] LNetNIInit+0x25c/0x3e0 [lnet]
[<ffffffff81061bc6>] ? put_online_cpus+0x56/0x80
[<ffffffffa0983000>] ? init_module+0x0/0x1000 [ptlrpc]
[<ffffffffa081310c>] ptlrpc_ni_init+0x2c/0x1a0 [ptlrpc]
[<ffffffffa0983000>] ? init_module+0x0/0x1000 [ptlrpc]
[<ffffffffa0813291>] ptlrpc_init_portals+0x11/0xf0 [ptlrpc]
[<ffffffffa0983000>] ? init_module+0x0/0x1000 [ptlrpc]
[<ffffffffa09831c4>] init_module+0x1c4/0x1000 [ptlrpc]
[<ffffffff810020e2>] do_one_initcall+0xe2/0x190
[<ffffffff810ca7fb>] load_module+0x129b/0x1a90
[<ffffffff812da590>] ? ddebug_dyndbg_module_param_cb+0x0/0x60
[<ffffffff810c7133>] ? copy_module_from_fd.isra.43+0x53/0x150
[<ffffffff810cb1a6>] SyS_finit_module+0xa6/0xd0
[<ffffffff815f2119>] system_call_fastpath+0x16/0x1b
...
This is because in lnet_startup_lndnis(), we may add list items to
@the_lnet.ln_nis and @the_lnet.ln_nis_cpt before it failed. But in
lnet_startup_lndis() failure path,it did not cleanup list thus
causing assertion in lnet_unprepare().

Fix the assertion by cleaning up using lnet_shutdown_lndnis()
if the startup fails.

In a future enahancement the ni startup API will be modified to
cleanup after itself in case of failure.

Signed-off-by: Amir Shehata <amir.shehata@intel.com>
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5568
Reviewed-on: http://review.whamcloud.com/12512
Reviewed-by: Liang Zhen <liang.zhen@intel.com>
Reviewed-by: Isaac Huang <he.huang@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
---
drivers/staging/lustre/lnet/lnet/api-ni.c | 6 +++++-
1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/drivers/staging/lustre/lnet/lnet/api-ni.c b/drivers/staging/lustre/lnet/lnet/api-ni.c
index 4c4e6d3..bfc1f13 100644
--- a/drivers/staging/lustre/lnet/lnet/api-ni.c
+++ b/drivers/staging/lustre/lnet/lnet/api-ni.c
@@ -1246,6 +1246,10 @@ lnet_shutdown_lndni(__u32 net)
return 0;
}

+/*
+ * Callers of lnet_startup_lndnis need to clean up using
+ * lnet_shutdown_lndnis if startup fails
+ */
static int
lnet_startup_lndnis(struct list_head *nilist, __s32 peer_timeout,
__s32 peer_cr, __s32 peer_buf_cr, __s32 credits,
@@ -1554,7 +1558,7 @@ LNetNIInit(lnet_pid_t requested_pid)

rc = lnet_startup_lndnis(&net_head, -1, -1, -1, -1, &ni_count);
if (rc != 0)
- goto failed1;
+ goto failed2;

if (the_lnet.ln_eq_waitni && ni_count > 1) {
lnd_type = the_lnet.ln_eq_waitni->ni_lnd->lnd_type;
--
1.7.1


\
 
 \ /
  Last update: 2015-11-21 01:01    [W:0.245 / U:0.192 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site