Jarek Poplawski wrote:
James Chapman wrote, On 02/11/2008 10:22 AM:
Fix locking issues in the pppol2tp driver which can cause a kernel
crash on SMP boxes when hundreds of L2TP sessions are created/deleted
simultaneously (ISP environment). The driver was violating read_lock()
and write_lock() scheduling rules so we now consistently use the _irq
variants of the lock functions.
...
Hi,
Could you explain what exactly scheduling rules do you mean here,
and why disabling interrupts is the best solution for this?
Below is example output from lockdep. The oops is reproducible when
creating/deleting lots of sessions while passing data. The lock is being
acquired for read and write in softirq contexts.
Is there a better way to fix this?
=================================
[ INFO: inconsistent lock state ]
2.6.24-core2 #1
---------------------------------
inconsistent {in-softirq-R} -> {softirq-on-W} usage.
openl2tpd/3215 [HC0[0]:SC0[0]:HE1:SE1] takes:
(&tunnel->hlist_lock){---?}, at: [<f8eea157>]
pppol2tp_connect+0x517/0x6d0 [pppol2tp]
{in-softirq-R} state was registered at:
[<c014edaf>] __lock_acquire+0x6bf/0x10a0
[<c03ee75b>] fn_hash_lookup+0x1b/0xe0
[<c014f804>] lock_acquire+0x74/0xa0
[<f8ee859f>] pppol2tp_session_find+0x1f/0x80 [pppol2tp]
[<c040427a>] _read_lock+0x2a/0x40
[<f8ee859f>] pppol2tp_session_find+0x1f/0x80 [pppol2tp]
[<f8ee859f>] pppol2tp_session_find+0x1f/0x80 [pppol2tp]
[<f8ee8dc8>] pppol2tp_recv_core+0xd8/0x960 [pppol2tp]
[<f8d3f72a>] ipt_do_table+0x23a/0x500 [ip_tables]
[<f8ee967e>] pppol2tp_udp_encap_recv+0x2e/0x70 [pppol2tp]
[<c0403fb4>] _read_unlock+0x14/0x20
[<c03dd696>] udp_queue_rcv_skb+0x106/0x2a0
[<c03ddc5a>] __udp4_lib_rcv+0x42a/0x7e0
[<f8d57090>] ipt_hook+0x0/0x20 [iptable_filter]
[<c03bc2da>] ip_local_deliver_finish+0xca/0x1c0
[<c03bc23e>] ip_local_deliver_finish+0x2e/0x1c0
[<c03bbfaf>] ip_rcv_finish+0xff/0x360
[<c03bc6dc>] ip_rcv+0x20c/0x2a0
[<c03bbeb0>] ip_rcv_finish+0x0/0x360
[<c039ad87>] netif_receive_skb+0x317/0x4b0
[<c039ab70>] netif_receive_skb+0x100/0x4b0
[<f8d9627a>] e1000_clean_rx_irq_ps+0x28a/0x560 [e1000]
[<f8d95ff0>] e1000_clean_rx_irq_ps+0x0/0x560 [e1000]
[<f8d9384d>] e1000_clean+0x5d/0x290 [e1000]
[<c039d580>] net_rx_action+0x1a0/0x2a0
[<c039d43f>] net_rx_action+0x5f/0x2a0
[<c0131e72>] __do_softirq+0x92/0x120
[<c0131f78>] do_softirq+0x78/0x80
[<c010b15a>] do_IRQ+0x4a/0xa0
[<c0108dcc>] common_interrupt+0x24/0x34
[<c0108dd6>] common_interrupt+0x2e/0x34
[<c01062d6>] mwait_idle_with_hints+0x46/0x60
[<c0106550>] mwait_idle+0x0/0x20
[<c0106694>] cpu_idle+0x74/0xe0
[<c0536a9a>] start_kernel+0x30a/0x3a0
[<c0536150>] unknown_bootoption+0x0/0x1f0
[<ffffffff>] 0xffffffff
irq event stamp: 275
hardirqs last enabled at (275): [<c0132317>] local_bh_enable_ip+0xa7/0x120
hardirqs last disabled at (273): [<c01322a6>] local_bh_enable_ip+0x36/0x120
softirqs last enabled at (274): [<f8eab8bc>]
ppp_register_channel+0xdc/0xf0 [ppp_generic]
softirqs last disabled at (272): [<c040410b>] _spin_lock_bh+0xb/0x40
--
James Chapman
Katalix Systems Ltd
http://www.katalix.com
Catalysts for your Embedded Linux software development
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
