Message-ID: <20090122084133.505daeb5@barsoom.rdu.redhat.com>
Date:	Thu, 22 Jan 2009 08:41:33 -0500
From:	Jeff Layton <jlayton@...hat.com>
To:	Bernhard Schmidt <berni@...kenwald.de>
Cc:	linux-cifs-client@...ts.samba.org, linux-kernel@...r.kernel.org,
	Suresh Jayaraman <sjayaraman@...e.de>
Subject: Re: [linux-cifs-client] BUG: Possible cifs+IPv6-Regression 2.6.27.4
 -> 2.6.27.9

On Mon, 19 Jan 2009 22:38:20 +0100
Bernhard Schmidt <berni@...kenwald.de> wrote:

> On Mon, Jan 19, 2009 at 04:07:25PM -0500, Jeff Layton wrote:
> 
> > That said, there's a lot of jumping around in this assembly code and
> > it's not completely clear to me how it got to the point where it crashed.
> > We'll probably need to see if this can be independently reproduced. Can
> > you email along the details of how you're reproducing this? In
> > particular:
> > 
> > 1) all mount options being used
> 
> Not more than the ones shown in my original mail. The exact mount
> command is *checkingfirewall*
> 
> mount -t cifs -o user=berni,ip=2001:a60:f001:1::69 //fileserv/pub /pub
> 
> > 2) details on the server (what OS, what version of samba, etc)
> > 3) version of mount.cifs being used
> 
> The server is a Debian Lenny based i386 system running a self-compiled
> vanilla 2.6.28 kernel and Debian packaged Samba 3.2.5-3.
> 
> The client is Ubuntu Intrepid (8.10) with intrepid-security,
> intrepid-updates and intrepid-proposed enabled. mount.cifs -V does not
> work misteriously (displays the usage instruction), the package version
> is 3.2.3-1ubuntu3.4.
> 
> All those programs are not changed. After booting into kernel
> 2.6.27-7.14 (intrepid release) or 2.6.27-9.19 (intrepid-security,
> intrepid-updates) the above noted mount command works fine, after
> booting 2.6.27-11.23 (intrepid-proposed) it's broken. I can change
> between those client kernel versions and reproduce it.
> 
> The changes between the working and the non-working version should
> be as noted in
> 
> https://launchpad.net/ubuntu/intrepid/+source/linux/2.6.27-10.20
> https://launchpad.net/ubuntu/intrepid/+source/linux/2.6.27-11.23
> 
> And I've just discovered something. Remind you, I'm trying to mount
> 2001:a60:f001:1::69. However, after entering the password the system
> tries to reach another address
> 
> 22:31:24.336790 IP6 2001:a60:f001:1:219:66ff:fe8b:a6e > ff02::1:ff00:69:
> ICMP6, neighbor solicitation, who has 2001:a60:f001:1:55:abe3:0:69,
> length 32
> 
> The (wrong) address changes slightly between consecutive tries, I've seen 
> 2001:a60:f001:1:4f:abe3:0:69
> 2001:a60:f001:1:58:abe3:0:69
> as well.
> 
> Also, when I try some other (legal, but unused) address it bails out at
> another code location, e.g.
> 
> mount -t cifs -o user=berni,ip=2001:a60:f001:1:cc::69 //fileserv/pub
> /pub
> 
> results in an immediate (!)
> 
> [  486.470983] BUG: unable to handle kernel paging request at 0000cc0c
> [  486.470992] IP: [<f9c0a3d6>] :cifs:ipv6_connect+0x46/0x1a0
> [  486.471008] *pde = 00000000
> [  486.471012] Oops: 0000 [#13] SMP
> [  486.471016] Modules linked in: nls_cp437 cifs af_packet binfmt_misc rfcomm bridge stp bnep sco l2cap bluetooth kvm_amd kvm ppdev tun ipv6 pci_slot container sbs sbshc video output battery
> iptable_filter ip_tables x_tables ac parport_pc lp parport serio_raw pcspkr psmouse snd_hda_intel snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_dummy snd_seq_oss snd_seq_midi k8temp snd_rawmidi
> snd_seq_midi_event snd_seq snd_timer snd_seq_device snd soundcore snd_page_alloc i2c_piix4 i2c_core evdev dm_multipath scsi_dh pl2303 usbserial fglrx(P) agpgart wmi button shpchp pci_hotplug ext3 jbd
> mbcache sr_mod cdrom pata_acpi sd_mod crc_t10dif pata_atiixp sg usbhid hid usb_storage libusual ata_generic ahci ohci_hcd ehci_hcd libata usbcore scsi_mod dock r8169 mii dm_mirror dm_log dm_snapshot
> dm_mod thermal processor fan fbcon tileblit font bitblit softcursor fuse
> [  486.471076]
> [  486.471079] Pid: 6607, comm: mount.cifs Tainted: P      D   (2.6.27-11-generic #1)
> [  486.471082] EIP: 0060:[<f9c0a3d6>] EFLAGS: 00010246 CPU: 0
> [  486.471092] EIP is at ipv6_connect+0x46/0x1a0 [cifs]
> [  486.471095] EAX: 0000cc00 EBX: f9c3b094 ECX: 0000001c EDX: c4ecfe44
> [  486.471097] ESI: c4ecfe44 EDI: c4ecfe54 EBP: c4ecfd94 ESP: c4ecfd7c
> [  486.471100]  DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
> [  486.471103] Process mount.cifs (pid: 6607, ti=c4ece000 task=ce5abed0 task.ti=c4ece000)
> [  486.471105] Stack: c4ecfd94 f9c1708b 00000000 f9c3b094 c4eaa000 f9c3b094 c4ecfe64 f9c0bf5c
> [  486.471112]        c10b1030 c4eac000 c4ecfdcc c0189c8d c4eab000 000280d0 f9c3b094 000a100c
> [  486.471118]        00000001 caa0e3c0 dbf3e200 0000000c 00000000 00000020 c4ecfe4c c4ecfe0c
> [  486.471124] Call Trace:
> [  486.471127]  [<f9c1708b>] ? cifs_inet_pton+0x7b/0x80 [cifs]
> [  486.471141]  [<f9c0bf5c>] ? cifs_mount+0x46c/0xda0 [cifs]
> [  486.471151]  [<c0189c8d>] ? prep_new_page+0xdd/0x160
> [  486.471161]  [<c024e132>] ? idr_get_empty_slot+0xe2/0x270
> [  486.471167]  [<f9bfddf3>] ? cifs_read_super+0x93/0x1e0 [cifs]
> [  486.471176]  [<c01b38f0>] ? set_anon_super+0x0/0xd0
> [  486.471181]  [<f9bfdfa9>] ? cifs_get_sb+0x69/0xc0 [cifs]
> [  486.471192]  [<c01b468e>] ? vfs_kern_mount+0x5e/0x130
> [  486.471197]  [<c01b47be>] ? do_kern_mount+0x3e/0xe0
> [  486.471201]  [<c01cccff>] ? do_new_mount+0x6f/0x90
> [  486.471206]  [<c01cd242>] ? do_mount+0x1d2/0x1f0
> [  486.471210]  [<c01ca95d>] ? exact_copy_from_user+0x4d/0xa0
> [  486.471214]  [<c01caf6e>] ? copy_mount_options+0x6e/0xd0
> [  486.471218]  [<c01cd2f1>] ? sys_mount+0x91/0xc0
> [  486.471222]  [<c0103f7b>] ? sysenter_do_call+0x12/0x2f
> [  486.471227]  =======================
> [  486.471228] Code: 8b 02 85 c0 0f 84 eb 00 00 00 66 83 7e 02 00 66 c7 06 0a 00 66 c7 45 f2 00 00 75 59 66 c7 46 02 01 bd 8b 07 b9 1c 00 00 00 89 f2 <8b> 58 0c c7 04 24 00 00 00 00 ff 53 10 85 c0 89
> c3 78 67 8b 07
> [  486.471257] EIP: [<f9c0a3d6>] ipv6_connect+0x46/0x1a0 [cifs] SS:ESP 0068:c4ecfd7c
> [  486.471269] ---[ end trace cbe7d886432a8981 ]---#
> 
> This looks like an address corruption in the memory, maybe the original
> BUG was there before and did not trigger because the server was reachable.
>

Unfortunately, I'm still unable to reproduce this.

Here's a small debug patch that may be helpful. Essentially it extends
one of the existing cifsFYI printk's to make it also dump the address
to which we're going to connect. This would allow us to verify that the
address parser is working correctly.

Basically you'll need to build a new cifs.ko (or entire kernel) with
this patch, and then turn up cifsFYI debug messages:

# insmod /location/of/new/cifs.ko
# echo 7 > /proc/fs/cifs/cifsFYI
# mount -t cifs -o user=berni,ip=2001:a60:f001:1:cc::69 //fileserv/pub /pub

Once it crashes, please send the entire set of messages dumped to the
console after the mount attempt, including the oops message. Hopefully
that will allow us to verify whether address corruption is in the
parsing phase.

Thanks,
-- 
Jeff Layton <jlayton@...hat.com>

View attachment "cifs-ipv6-debug.patch" of type "text/x-patch" (629 bytes)

Powered by blists - more mailing lists