netdev - Re: Bug#656802: linux-image-2.6.32-5-amd64: dmesg BUG: scheduling while atomic: ksoftirqd/0/4/0x10000100

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <1327195921.8004.70.camel@deadeye>
Date:	Sun, 22 Jan 2012 01:32:01 +0000
From:	Ben Hutchings <ben@...adent.org.uk>
To:	James Chapman <jchapman@...alix.com>
Cc:	netdev <netdev@...r.kernel.org>, 656802@...s.debian.org,
	alex <alex@....org>
Subject: Re: Bug#656802: linux-image-2.6.32-5-amd64: dmesg BUG: scheduling
 while atomic: ksoftirqd/0/4/0x10000100

On Sat, 2012-01-21 at 22:19 +0200, alex wrote:
[...]
>  [148654.740747] xt_TCPMSS: bad length (41 bytes)
> [148656.341170] BUG: scheduling while atomic: ksoftirqd/0/4/0x10000100
> [148656.341402] Modules linked in: act_police sch_ingress cls_u32 sch_sfq sch_cbq dummy 8021q garp stp xt_TCPMSS ipt_REJECT xt_tcpudp xt_state xt_multiport iptable_filt
> er iptable_nat nf_nat nf_conntrack_ipv4 nf_conntrack nf_defrag_ipv4 ip_tables x_tables xfs exportfs pppoe pppol2tp pptp pppox ppp_generic slhc loop firewire_sbp2 snd_hd
> a_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm snd_timer nouveau ttm drm_kms_helper snd drm i2c_piix4 i2c_algo_bit i2c_core k10temp soundcore edac_core s
> nd_page_alloc edac_mce_amd pcspkr button wmi evdev processor ext3 jbd mbcache ata_generic sd_mod crc_t10dif ohci_hcd pata_via pata_atiixp tg3 libphy ahci ehci_hcd libat
> a ixgbe xhci firewire_ohci dca floppy usbcore thermal scsi_mod firewire_core nls_base crc_itu_t thermal_sys [last unloaded: scsi_wait_scan]
> [148656.341440] Pid: 4, comm: ksoftirqd/0 Not tainted 2.6.32-5-amd64 #1
> [148656.341442] Call Trace:
> [148656.341443]  <IRQ>  [<ffffffff812facb8>] ? schedule+0xc5/0x7b4
> [148656.341453]  [<ffffffff810963c9>] ? handle_edge_irq+0xdd/0x101
> [148656.341457]  [<ffffffff8104aa4c>] ? __cond_resched+0x1d/0x26
> [148656.341459]  [<ffffffff812fb5ab>] ? _cond_resched+0x24/0x2f
> [148656.341463]  [<ffffffff81243945>] ? lock_sock_nested+0x16/0xab
> [148656.341465]  [<ffffffff810114d3>] ? ret_from_intr+0x0/0x11
> [148656.341469]  [<ffffffffa01a6b2a>] ? pppol2tp_tunnel_destruct+0xe7/0x1f7 [pppol2tp]
> [148656.341472]  [<ffffffff8124433f>] ? __sk_free+0x15/0xe8
> [148656.341474]  [<ffffffff81247c06>] ? skb_release_head_state+0x6d/0xc8
> [148656.341476]  [<ffffffff8124797a>] ? __kfree_skb+0x9/0x7d
> [148656.341485]  [<ffffffffa00cf989>] ? ixgbe_poll+0x119/0x1840 [ixgbe]
> [148656.341489]  [<ffffffff81272e08>] ? ip_rcv_finish+0x0/0x38d
> [148656.341493]  [<ffffffff8124fd8e>] ? net_rx_action+0xae/0x1c9
> [148656.341496]  [<ffffffff81053d2b>] ? __do_softirq+0xdd/0x1a6
> [148656.341498]  [<ffffffff81011cac>] ? call_softirq+0x1c/0x30
[...]

It appears that a transmit completion results in dropping the last
reference to a PPPoL2TP tunnel, causing it to be destroyed.
pppo2ltp_tunnel_destruct() calls pppo2ltp_tunnel_closeall() (inlined
here) which calls lock_sock(), and that may sleep.

Although this was observed in 2.6.32, the bug appears to be present
today: l2tp_tunnel_destruct() calls l2tp_tunnel_closeall() calls
pppol2tp_session_close() calls lock_sock().

But maybe this is actually a ref-counting bug and the tunnel should
never actually be destroyed in atomic context.

Ben.

-- 
Ben Hutchings
Knowledge is power.  France is bacon.

Download attachment "signature.asc" of type "application/pgp-signature" (829 bytes)