openafs does not build with natty kernel 2.6.37, 2.6.38

FWIW, openafs-modules-dkms from debian experimental (1.6.0~pre1-1) builds on 2.6.37-11-generic and 2.6.25-23-generic, but when the module is loaded into 2.6.37-11-generic, it segfaults and pretty much locks up the vfs subsystem. Here's the segfault from kern.log:

(after # /etc/init.d/openafs-client start )

Jan 3 16:02:39 lynx kernel: [ 73.658371] Cache bypass patched libafs module init.
Jan 3 16:02:39 lynx kernel: [ 73.699312] enabling dynamically allocated vcaches
Jan 3 16:02:40 lynx kernel: [ 73.699321] Starting AFS cache scan...found 0 non-empty cache files (0%).
Jan 3 16:02:40 lynx kernel: [ 74.102052] general protection fault: 0000 [#2] SMP
Jan 3 16:02:40 lynx kernel: [ 74.102055] last sysfs file: /sys/devices/platform/applesmc.768/leds/smc::kbd_backlight/brightness
Jan 3 16:02:40 lynx kernel: [ 74.102057] CPU 0
Jan 3 16:02:40 lynx kernel: [ 74.102058] Modules linked in: openafs(P) michael_mic arc4 coda ppdev ipmi_msghandler vboxdrv(+) ip6table_filter ip6_tables xt_CHECK
SUM xt_state ipt_REJECT iptable_filter bridge stp kvm_intel kvm dm_crypt snd_hda_codec_realtek snd_hda_intel xt_mark xt_multiport xt_tcpudp snd_hda_codec nvidia(P)
ipt_MASQUERADE snd_hwdep snd_pcm iptable_mangle btusb iptable_nat nf_nat nf_conntrack_ipv4 snd_seq_midi nf_conntrack nf_defrag_ipv4 lib80211_crypt_tkip bluetooth ip
_tables snd_rawmidi x_tables snd_seq_midi_event wl(P) snd_seq uvcvideo snd_timer snd_seq_device videodev v4l1_compat v4l2_compat_ioctl32 snd soundcore lib80211 snd_
page_alloc shpchp joydev i2c_nforce2 usbhid bcm5974 applesmc input_polldev coretemp nvidia_bl lp parport raid10 raid456 async_pq async_xor xor async_memcpy async_ra
id6_recov hid_apple raid6_pq async_tx raid1 raid0 multipath linear btrfs zlib_deflate libcrc32c uvesafb hid forcedeth ahci libahci [last unloaded: usbhid]
Jan 3 16:02:40 lynx kernel: [ 74.102107]
Jan 3 16:02:40 lynx kernel: [ 74.102109] Pid: 2567, comm: afsd Tainted: P D 2.6.37-11-generic #25-Ubuntu Mac-F2268EC8/MacBookPro5,2
Jan 3 16:02:40 lynx kernel: [ 74.102111] RIP: 0010:[<ffffffffa11c3f1f>] [<ffffffffa11c3f1f>] afs_fill_super+0x5f/0x240 [openafs]
Jan 3 16:02:40 lynx kernel: [ 74.102136] RSP: 0018:ffff880105121dc8 EFLAGS: 00010246
Jan 3 16:02:40 lynx kernel: [ 74.102137] RAX: ffff880143eb4440 RBX: ffff8801033fb800 RCX: 0000000000000000
Jan 3 16:02:40 lynx kernel: [ 74.102139] RDX: ffff880105121fd8 RSI: 0000000000000000 RDI: 0000000000000000
Jan 3 16:02:40 lynx kernel: [ 74.102141] RBP: ffff880105121dd8 R08: ffff88014bc18708 R09: 0000000000000000
Jan 3 16:02:40 lynx kernel: [ 74.102142] R10: 0000000000000000 R11: 00000000000f4490 R12: ffff88010ef416c0
Jan 3 16:02:40 lynx kernel: [ 74.102144] R13: 0000000000000000 R14: ffff8801033fb800 R15: 0000000000000000
Jan 3 16:02:40 lynx kernel: [ 74.102146] FS: 00007f55c9d34700(0000) GS:ffff88007fc00000(0000) knlGS:0000000000000000
Jan 3 16:02:40 lynx kernel: [ 74.102148] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
Jan 3 16:02:40 lynx kernel: [ 74.102150] CR2: 000000000062b668 CR3: 0000000105357000 CR4: 00000000000406f0
Jan 3 16:02:40 lynx kernel: [ 74.102151] DR0: 0000000000000000 DR1: 00000000000...

Is it possible you’re trying to use openafs-modules-dkms 1.6.0~pre1-1 with openafs-client 1.4.12.1+dfsg-3? That probably won’t work.

You can get 1.6.0pre1 packages compiled for Ubuntu at this PPA (upgrade all packages, not just openafs-modules-dkms):
https://launchpad.net/~openafs/+archive/master

Nope... all is matched up.

$ dpkg -l | grep openafs

ii libpam-openafs-session 1.7-1build1 PAM module to obtain PAGs and AFS tokens (transitional package)
ii openafs-client 1.6.0~pre1-1 AFS distributed filesystem client support
ii openafs-krb5 1.6.0~pre1-1 AFS distributed filesystem Kerberos 5 integration
ii openafs-modules-dkms 1.6.0~pre1-1 AFS distributed filesystem kernel module DKMS source

I pulled the packages from the PPA and latest 2.6.37-11-generic (~lucid1) kernel, all works fine! No Oops. Of course, it needs broader testing since this is roughly RC quality... would be good to release 1.4.14 for the less adventurous and close this bug specifically. Let me know if you want any additional testing help.

# dpkg -l | egrep openafs\|linux-image

ii linux-image 2.6.37.11.13 Generic Linux kernel image.
ii linux-image-2.6.35-23-generic 2.6.35-23.41 Linux kernel image for version 2.6.35 on x86/x86_64
ii linux-image-2.6.37-11-generic 2.6.37-11.25~lucid1 Linux kernel image for version 2.6.37 on x86/x86_64
ii linux-image-generic 2.6.37.11.13 Generic Linux kernel image
ii linux-image-generic-lts-backport-natty 2.6.37.11.11 Generic Linux kernel image
ii openafs-client 1.6.0~pre1-1~ppa0~lucid1 AFS distributed filesystem client support
ii openafs-krb5 1.6.0~pre1-1~ppa0~lucid1 AFS distributed filesystem Kerberos 5 integration
ii openafs-modules-dkms 1.6.0~pre1-1~ppa0~lucid1 AFS distributed filesystem kernel module DKMS source

copied from dupe:
"A similar bug was reported in Gentoo:
http://bugs.gentoo.org/show_bug.cgi?format=multiple&id=350767

With a proposed fix:
http://bugs.gentoo.org/attachment.cgi?id=259000&action=view "

looks like moving to 1.4.14 is a good way to fix http://dl.openafs.org/dl/openafs/1.4.14/RELNOTES-1.4.14

The proposed fix is not enough. 1.4.14 is the proper solution here.

openafs 1.4.14+dfsg-1 landed in debian.
http://packages.qa.debian.org/o/openafs/news/20110210T190313Z.html

Yeah, but 1.4.14 still doesn’t build with natty’s kernel 2.6.38. Upstream is stlil working on this; we’ll probably have to wait for 1.6.0pre2.

Do we really want 1.6.X... in natty? Is it/will it be stable enough to be in natty?

Here are OpenAFS 1.6.0pre2 packages that work on natty:
https://launchpad.net/~openafs/+archive/master

I think 1.6.0 is what we’ll want in natty. I’ve been using the 1.5.x/1.6.x series on my laptop for almost a year (except for the last month where it didn’t compile with kernel 2.6.38), and it’s worked well enough for me. It also gets much more active attention from upstream. I don’t know if kernel 2.6.38 support will end up in the 1.4.x branch at all.

I submitted a 1.6.0pre2 package for sponsorship (bug #723062).

I also backported kernel 2.6.38 support to the 1.4.x branch; that’s waiting on upstream review:
  http://gerrit.openafs.org/3992
  http://gerrit.openafs.org/3993
  http://gerrit.openafs.org/3994
  http://gerrit.openafs.org/3995
  http://gerrit.openafs.org/3996
  http://gerrit.openafs.org/3997
  http://gerrit.openafs.org/3998
  http://gerrit.openafs.org/3999
but I still think I would prefer to go with 1.6.0 for natty, because upstream doesn’t seem to love 1.4.x anymore.

This bug was fixed in the package openafs - 1.4.14+dfsg-1+ubuntu1

---------------
openafs (1.4.14+dfsg-1+ubuntu1) natty; urgency=low

  * Apply upstream deltas for kernel 2.6.38 support (LP: #675768):
    - [02f2c7cb] Linux: Add autoconf macro for structure checks
    - [a89d6b02] Linux: Add general autoconf macro for Linux kernel
    - [08bb83d9] linux: 2.6.38: New d_op handling
    - [52556d50] linux: 2.6.38: Make d_revalidate RCU-walk aware
    - [ca38c954] Linux: allow compile flags to be passed to AC_CHECK_LINUX_BUILD
    - [5bcc0ea7] Linux: 2.6.38: Adjust for permission inode operation changes
    - [2eca7aef] Linux: 2.6.38: deal with dcache_lock removal
    - [281f5bf5] Linux: 2.6.38: dentry->d_count is not an atomic
  * Apply upstream deltas:
    - [c4537f04] Don't count root session keyrings against quota (LP: #723481)
    - [dfcd3fca] Linux: Reduce key_alloc flags confusion
      (from http://gerrit.openafs.org/4033)
    - [01c6e6bb] Linux: install_session_keyring: key_alloc flags are
      unsigned long
      (from http://gerrit.openafs.org/4034)
 -- Anders Kaseorg <email address hidden> Tue, 22 Feb 2011 20:17:04 -0500

Was going to open a new bug, but the title does cover 2.6.38 as well...

This still isn't working for me. Getting a protection fault when calling afs_fill_super. Both 1.4.14 and 1.6.0-pre2 are giving me this error when AFS starts, linux-image-2.6.38-6-generic...

Anybody else seeing this?

Mar 10 15:26:55 lynx kernel: [ 23.530821] enabling dynamically allocated vcaches
Mar 10 15:26:56 lynx kernel: [ 23.530823] Starting AFS cache scan...found 5803 non-empty cache files (15%).
Mar 10 15:26:56 lynx kernel: [ 24.020503] general protection fault: 0000 [#1] SMP
Mar 10 15:26:56 lynx kernel: [ 24.030005] last sysfs file: /sys/devices/system/cpu/cpu1/cpufreq/scaling_setspeed
Mar 10 15:26:56 lynx kernel: [ 24.030005] CPU 0
Mar 10 15:26:56 lynx kernel: [ 24.030005] Modules linked in: openafs(P) ipmi_msghandler vboxnetadp vboxnetflt vboxdrv ip6table_filter ip6_tables xt_CHECKSUM xt_state ipt_REJECT iptable_filter bridge stp kvm_intel kvm parport_pc ppdev dm_crypt snd_hda_codec_realtek rfcomm sco bnep l2cap nvidia(P) snd_hda_intel snd_hda_codec snd_hwdep btusb snd_pcm bluetooth uvcvideo snd_seq_midi videodev v4l2_compat_ioctl32 xt_mark lib80211_crypt_tkip xt_multiport xt_tcpudp snd_rawmidi ipt_MASQUERADE wl(P) snd_seq_midi_event iptable_mangle iptable_nat nf_nat nf_conntrack_ipv4 snd_seq joydev nf_conntrack nf_defrag_ipv4 ip_tables x_tables snd_timer snd_seq_device lib80211 i2c_nforce2 shpchp usbhid bcm5974 snd soundcore snd_page_alloc applesmc input_polldev coretemp nvidia_bl lp parport raid10 raid456 async_pq async_xor xor async_memcpy async_raid6_recov raid6_pq async_tx raid1 raid0 multipath linear btrfs zlib_deflate libcrc32c uvesafb hid_apple hid ahci firewire_ohci firewire_core crc_itu_t libahci forcedeth [last unloaded: us
Mar 10 15:26:56 lynx kernel: bhid]
Mar 10 15:26:56 lynx kernel: [ 24.030005]
Mar 10 15:26:56 lynx kernel: [ 24.030005] Pid: 1969, comm: afsd Tainted: P 2.6.38-6-generic #34-Ubuntu Apple Inc. MacBookPro5,2/Mac-F2268EC8
Mar 10 15:26:56 lynx kernel: [ 24.030005] RIP: 0010:[<ffffffffa08fb1bf>] [<ffffffffa08fb1bf>] afs_fill_super+0x5f/0x250 [openafs]
Mar 10 15:26:56 lynx kernel: [ 24.030005] RSP: 0018:ffff8801444e3da8 EFLAGS: 00010246
Mar 10 15:26:56 lynx kernel: [ 24.030005] RAX: ffff8801445116c0 RBX: ffff88014248b400 RCX: 0000000000000000
Mar 10 15:26:56 lynx kernel: [ 24.030005] RDX: ffff8801444e3fd8 RSI: 0000000000000000 RDI: ffffffffa0919920
Mar 10 15:26:56 lynx kernel: [ 24.030005] RBP: ffff8801444e3db8 R08: ffff88014bc0b708 R09: 0000000000000000
Mar 10 15:26:56 lynx kernel: [ 24.030005] R10: 0000000000000000 R11: 00000000000f2098 R12: ffff880144515b00
Mar 10 15:26:56 lynx kernel: [ 24.030005] R13: 0000000000000000 R14: ffff88014248b400 R15: 0000000000000000
Mar 10 15:26:56 lynx kernel: [ 24.030005] FS: 00007fce7bf30700(0000) GS:ffff88007fc00000(0000) knlGS:0000000000000000
Mar 10 15:26:56 lynx kernel: [ 24.030005] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
Mar 10 15:26:56 lynx kernel: [ 24.030005] CR2: 000000000062b020 CR3: 00000001438e0000 CR4: 00000000000406f0
Mar 10 15:26:56 lynx kernel: [ 24.030005] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000...

Michael: no, I don’t think that’s a known problem. Can you go ahead and open a separate bug report for that? This one’s about the build failure, and you’re seeing something different. Thanks for your help!