From: Ben Greear <greearb@xxxxxxxxxxxxxxx> I see repeatable nfs server crashes in 2.6.39.1 and 3.0.0-rc3+. I have not yet tested older builds. The problem appears to be corrupted xpt_ready list entries: WARNING: at /home/greearb/git/linux-2.6.linus/lib/list_debug.c:30 __list_add+0x68/0x80() Hardware name: X7DBU list_add corruption. prev->next should be next (ffff880222777e38), but was 6b6b6b6b6b6b6b6b. (prev=ffff88021cb8d190). Modules linked in: xt_addrtype xt_TPROXY nf_tproxy_core xt_socket nf_defrag_ipv6 xt_set ip_set nfnetlink xt_connlimit macvlan pktgen fuse iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi ip6table_filter ip6_tables ebtable_nat ebtables stp llc nfsd nfs lockd fscache auth_rpcgss nfs_acl sunrpc ipv6 kvm uinput i5k_amb i5000_edac iTCO_wdt iTCO_vendor_support i2c_i801 ioatdma pcspkr edac_core e1000e shpchp microcode dca floppy radeon ttm drm_kms_helper drm hwmon i2c_algo_bit i2c_core [last unloaded: xt_connmark] Pid: 3347, comm: nfsd Not tainted 3.0.0-rc3+ #1 Call Trace: <IRQ> [<ffffffff81049f72>] warn_slowpath_common+0x80/0x98 [<ffffffff8104a01e>] warn_slowpath_fmt+0x41/0x43 [<ffffffff81232ea4>] __list_add+0x68/0x80 [<ffffffffa029958a>] ? svc_xprt_enqueue+0x91/0x1c0 [sunrpc] [<ffffffffa029969b>] svc_xprt_enqueue+0x1a2/0x1c0 [sunrpc] [<ffffffff8140e170>] ? tcp_rcv_state_process+0x89e/0x8de [<ffffffffa028f2b4>] svc_tcp_listen_data_ready+0x4c/0x8c [sunrpc] [<ffffffff81416b3c>] tcp_child_process+0x61/0x109 [<ffffffff814150c1>] tcp_v4_do_rcv+0x2bf/0x33c [<ffffffff81415f20>] ? tcp_v4_rcv+0x2f3/0x7a2 [<ffffffff814160e4>] tcp_v4_rcv+0x4b7/0x7a2 [<ffffffff813f8502>] ? ip_local_deliver_finish+0x46/0x246 [<ffffffff813f866d>] ip_local_deliver_finish+0x1b1/0x246 [<ffffffff813f8502>] ? ip_local_deliver_finish+0x46/0x246 [<ffffffff813f84bc>] ? xfrm4_policy_check.clone.0+0x5c/0x5c [<ffffffff813f874e>] NF_HOOK.clone.1+0x4c/0x53 [<ffffffff813f879e>] ip_local_deliver+0x49/0x4d [<ffffffff813f82db>] ip_rcv_finish+0x330/0x35a [<ffffffff813f7fab>] ? skb_dst+0x41/0x41 [<ffffffff813f874e>] NF_HOOK.clone.1+0x4c/0x53 [<ffffffff813f89d2>] ip_rcv+0x230/0x25e [<ffffffff813bfed7>] __netif_receive_skb+0x502/0x561 [<ffffffff813c5356>] netif_receive_skb+0x7c/0x83 [<ffffffff813baadb>] ? __netdev_alloc_skb+0x1d/0x3a [<ffffffff813c5381>] napi_skb_finish+0x24/0x3b [<ffffffff813c5b60>] napi_gro_receive+0x2a/0x2f [<ffffffffa01bb92f>] e1000_receive_skb+0x50/0x5b [e1000e] [<ffffffffa01bdbb1>] e1000_clean_rx_irq+0x1fb/0x28d [e1000e] [<ffffffffa01be067>] e1000_clean+0x75/0x23b [e1000e] [<ffffffff813c560a>] net_rx_action+0xc6/0x235 [<ffffffff810507fa>] __do_softirq+0x117/0x280 [<ffffffff810754aa>] ? tick_dev_program_event+0x37/0xf8 [<ffffffff8106a879>] ? hrtimer_interrupt+0x12e/0x1c0 [<ffffffff81485d1c>] call_softirq+0x1c/0x30 [<ffffffff8100bbfe>] do_softirq+0x46/0x9e [<ffffffff810500b3>] irq_exit+0x4e/0xba [<ffffffff81023291>] smp_apic_timer_interrupt+0x85/0x93 [<ffffffffa0299a6e>] ? svc_delete_xprt+0x90/0xf4 [sunrpc] [<ffffffff814854d3>] apic_timer_interrupt+0x13/0x20 <EOI> [<ffffffff81050025>] ? _local_bh_enable_ip+0xc6/0xec [<ffffffff81050054>] local_bh_enable_ip+0x9/0xb [<ffffffff8147e643>] _raw_spin_unlock_bh+0x39/0x3e [<ffffffffa0299a6e>] svc_delete_xprt+0x90/0xf4 [sunrpc] [<ffffffffa0299af7>] svc_close_all+0x25/0x3e [sunrpc] [<ffffffffa028e194>] svc_destroy+0x77/0x136 [sunrpc] [<ffffffffa028e2eb>] svc_exit_thread+0x98/0xa1 [sunrpc] [<ffffffffa0364836>] ? nfsd_svc+0x183/0x183 [nfsd] [<ffffffffa036495a>] nfsd+0x124/0x13e [nfsd] [<ffffffff81066e10>] kthread+0x7d/0x85 [<ffffffff81485c24>] kernel_thread_helper+0x4/0x10 [<ffffffff8147ee18>] ? retint_restore_args+0x13/0x13 [<ffffffff81066d93>] ? __init_kthread_worker+0x56/0x56 [<ffffffff81485c20>] ? gs_change+0x13/0x13 ---[ end trace 6a1345b5521d6905 ]--- The test case is: Create 300 mounts to the server (unique clients), have 150 writing, 150 reading at steady state. We have each reader/writer going at about 50kbps, 2k read/write size. Then, repeatedly: /etc/init.d/nfs restart on the server. We normally see crash within 2-5 restarts. With this patch, I did 20 successful restarts. Signed-off-by: Ben Greear <greearb@xxxxxxxxxxxxxxx> --- NOTE: This needs review by someone who understands this code better. The code I re-added was first removed in around 2006 it seems. Maybe the race is hard to hit in most cases, or maybe something exacerbated it recently and that should be fixed instead of the patch below. :100644 100644 ab86b79... 178716f... M net/sunrpc/svc_xprt.c net/sunrpc/svc_xprt.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c index ab86b79..178716f 100644 --- a/net/sunrpc/svc_xprt.c +++ b/net/sunrpc/svc_xprt.c @@ -901,6 +901,7 @@ void svc_delete_xprt(struct svc_xprt *xprt) spin_lock_bh(&serv->sv_lock); if (!test_and_set_bit(XPT_DETACHED, &xprt->xpt_flags)) list_del_init(&xprt->xpt_list); + list_del_init(&xprt->xpt_ready); /* * We used to delete the transport from whichever list * it's sk_xprt.xpt_ready node was on, but we don't actually -- 1.7.3.4 -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
