=== zookofamilytime is now known as zooko === jorgp3 [n=jorgp@bnet-dial-47.bartnet.net] has joined #ubuntu-kernel === crimsun [n=crimsun@sh.nu] has joined #ubuntu-kernel [03:34] I'm trying to track down Malone bug #379. The reporter indicates that when a USB drive is attached to the system, it fails to mount, and in the error message is the partition map. I've encountered this with my own /dev/hdc1 (/boot) and suspect a kernel issue rather than an issue with gnome-volume-manager, but wanted to solicit confirmation from someone who understood the kernel better than I. === lamont [n=lamont@15.238.5.82] has joined #ubuntu-kernel === persia [n=persia@p3101-ipbf412marunouchi.tokyo.ocn.ne.jp] has joined #ubuntu-kernel [06:12] BenC: ping? [06:13] pong [06:14] BenC: it is indeed a security patch that changes ABI on sparc [06:14] - Add patch stolen-from-head_CAN-2005-2490.dpatch. [06:14] this on [06:14] e [06:14] -extern int cmsghdr_from_user_compat_to_kern(struct msghdr *, unsigned char *, [06:14] +extern int cmsghdr_from_user_compat_to_kern(struct msghdr *, struct sock *, unsigned char *, [06:14] int); [06:14] that wasn't from me, was it? [06:15] no no [06:15] it was from upstream to fix a CAN [06:15] where did it come from? [06:15] what i don't understand yet is why it does affect sparc only [06:15] BenC: upstream crack via pitti [06:15] either the patch is not complete [06:16] or i am missing something obvious [06:16] iirc amd64 uses CONFIG_COMPAT [06:16] that should pull it net/compat.h too [06:16] but it doesn't break the abi [06:17] it's not a big deal [06:17] we can workaround it easily [06:17] nobody run sparc except me and a few peoplw [06:17] people even [06:17] so even if we change the abi without bumping it [06:18] it's not a big effect like killing i386 [06:18] plus, it's an unsupported arch, and I don't think there are any external modules for it [06:18] ok, odd though [06:18] amd64 and ppc64 should be pulling in compat layer [06:19] i am pretty sure they do [06:19] but than.. why didn't they get the abi change? [06:19] not sure, I checked the ppc64 directly [06:19] did a compile here [06:20] arch/sparc64/kernel/sparc64_ksyms.c:#include [06:20] arch/sparc64/kernel/sys_sunos32.c:#include [06:20] arch/sparc64/solaris/socket.c:#include [06:20] I did an 8.12 compile for amd64 abi too, so the buildd would have failed with 8.13 auto build [06:20] nope [06:20] it's only sparc using it [06:20] net/compat.c:#include [06:20] net/core/scm.c:#include [06:20] net/socket.c:#include [06:21] that explain... [06:22] so for net/compat.h just include net/sock.h [06:22] it's a sparc only change that won't hurt anybody [06:22] i am going to give you the new ABI files soon [06:22] that we will need to put in the 8.13 directory [06:22] so that 8.14 will match [06:22] ok [06:23] but i would like to test build 8.14 before you release [06:23] no problem [06:23] you need to tell me when it is ready for that :) [06:30] should be early next week [06:32] ok.. i am checking out so i can add stuff directly if you want me to [06:33] in a few hours i should get the new workstations :) [06:33] X again! [06:34] working on console isn't exactly my favourite when i could have 3 heads :) [07:03] I'm tracking down an issue with Malone bug #379. This is a report that sometimes usb drives do not mount. I have experienced this problem on an IDE drive on my workstation with the same error message. This behavior is new to Breezy: Hoary did not have this issue, and the Hoary livecd does not today. Would anyone be willing to verify for me whether this is a kernel or userspace bug? [07:05] persia: full url to the bug pleasE? [07:11] fabbione: https://launchpad.net/malone/bugs/379. This does not include my local experience with /dev/hdc1. [07:12] userland problem [07:12] the dmesg shows clearly that the kernel has recognized the device [07:13] Apr 13 10:18:57 localhost scsi.agent[19294] : disk at /devices/pci0000:00/0000:00:0c.0/usb1/1-2/1-2:1.0/host5/target5:0:0/5:0:0:0 [07:13] this shows that the kernel did inform userland about the new device [07:13] after that it's not kernel business anymore [07:14] fabbione: My local experience with mount is that when it works it is silent, and when it fails, the /dev/scsi/host5/bus0/target0/lun0: p1 (or in my case /dev/ide/...) line prints to console. Is this just userland failing to catch the report? [07:15] possibly [07:15] it depends from the error [07:16] your problem might not be related to 379 at all [07:17] For me, I get no error reported through dmesg: just many repeats of "/dev/ide/host0/bus1/target0/lun0: p1" with timestamps. It may not be related, but when it was first discussed on ubuntu-bugs, it was suggested I ask here, both about the bug, and about my issue, and whether they were related. [07:19] Just for fun, my latest run of `sudo mount /boot` generated the following in dmesg "[ 9297.537247] /dev/ide/host0/bus1/target0/lun0: p1". [07:23] persia: i need to see the full dmesg [07:23] that kind of msg is in dmesg too [07:23] /join #ubuntu [07:24] sorry: misplaced space...looking for pastebibn... [07:25] BenC: i committed the sparc fixes to baz. i still need to hack the abi files [07:25] waiting to complete the build [07:25] Full dmesg is available from http://paste.ubuntulinux.nl/2243. The tun/tap is from a vpnc run earlier. [07:29] persia: what device is connected to hdc ? [07:29] It's an IDE->CF converter with a flash card installed. [07:30] The manufacturer claims no drivers are required, regardless of OS. [07:31] so you have /boot on the flash? [07:31] in theory there is no need of any driver.. that's correct.. [07:31] Yep. /boot is on flash. Worked perfectly under Hoary. Problems started around 2.6.12-4-?, but I don't remember exactly. Hoary livecd works perfectly today. [07:32] livecd doesn't use /boot from flash.. [07:32] but the machine boots fine with that setup, right? [07:32] i mean.. you can boot and you can mount /boot ? [07:32] No, but the livecd can mount the drive every time. With current Breezy I have success about 1 time in 10. [07:32] oh i see... [07:33] does the boot stop because it cannot mount /boot, right? [07:33] The machine boots fine, and /boot rarely mounts. I usually go through a cycle of `ls /boot`, `sudo mount ./boot` to get it to mount. [07:33] ok [07:33] can you try something for me? [07:33] Sure. What shall I try? [07:34] you first need to boot in breezy [07:34] Done. Shall I reboot? [07:34] no [07:34] not yet :) [07:34] give me a second that i need to search what you need to change exactly. [07:35] as root go to /usr/sbin [07:35] cp mkinitramfs mkinitramfs.backup [07:35] edit mkinitramfs [07:35] meh no [07:35] sorry.. wrong file :) [07:36] `sudo su -`; `cd /usr/sbin`; `cp mkinitramfs mkinitramfs.backup`. OK. deleting the backup... [07:36] cd /usr/share/initramfs-tools/scripts [07:36] cp functions functions.backup [07:36] (just in case :) [07:37] OK. [07:37] edit functions [07:37] go to line 226 [07:37] Error: no "edit" mailcap rules found for type "application/*" [07:37] and look for ide_boot_events [07:37] I'm using vi instead. [07:37] yeah edit = use your favourite text editor :) [07:37] vi is fine ;) [07:38] around line 226 [07:38] ide_boot_events() { [ -e /proc/ide ] || return ... [07:38] ide_boot_events [07:38] second.. i have an old file.. [07:39] I'm on line 146 on the current file. [07:39] 226 please [07:39] There's also an entry on line 226. [07:39] 226 it is. [07:40] the line after ide_boot_events [07:40] add sleep 10 [07:40] so that looks like: [07:40] Done. [07:40] ide_boot_events [07:40] sleep 10 [07:40] scsi_boot_events [07:41] I have a space between the sleep line and the scsi_boot_events line. It shouldn't matter, should it? [07:41] no it doesn't matter [07:41] than as root: [07:42] dpkg-reconfigure linux-kernel-whatever? [07:42] mkinitramfs -o /boot/initrd.img-`uname -r` `uname -r` [07:42] Ah. Processing in the same root session... [07:43] you need to have /boot mounted to do that [07:43] Hold on - it'll take me a little while to get it mounted... [07:43] sure [07:44] OK. Mounted. Remounted RW, mkinitramfs run. [07:44] once it is finished, you should reboot the machine [07:45] and see if it can mount /boot [07:45] OK. It's the machine on which I'm here, so I'll be back in a bit. [07:45] basically the sleep 10 will give more time to the device to settle after initizializa [07:45] if this work, i will tell you what to do next [07:45] If the device sometimes doesn't settle for over 10000 seconds, I'm not sure 10 will help, but trying anyway... [07:46] it will take a few seconds more to boot === persia [n=persia@p3101-ipbf412marunouchi.tokyo.ocn.ne.jp] has joined #ubuntu-kernel [07:48] The mount of /boot failed. BTW, thanks for looking into this now. [07:49] persia: one second.. [07:49] can you give me ls -las /boot ? [07:49] No worries. I'm not in a hurry. [07:50] sleep 10 might not be enough... [07:50] total 2 [07:50] 1 drwxr-xr-x 2 root root 1024 2005-08-21 19:08 . [07:50] 1 drwxr-xr-x 23 root root 1024 2005-09-13 19:37 .. [07:50] well i need the mounted /boot [07:50] not the empty one ;) [07:50] Sometimes it won't mount after a day of settling in, so I'm not sure sleep is the right fix. [07:51] hmm in that case no [07:51] OK. Trying to mount (repeated tries seem to help). [07:51] did you ever try such device in windows? [07:51] or with another OS? [07:52] The device worked fine with Hoary, and also when I first tried the Breezy repositories (my joystick requires 2.6.13) [07:52] i see... [07:52] ok last test... [07:52] still as root [07:52] mkinitrd -o /boot/initrd.img-`uname -r` `uname -r` [07:53] of course you still need /boot mounted [07:53] after that do another reboot [07:53] `ls -las /boot` output for mounted /boot is in http://paste.ubuntulinux.nl/2245 [07:54] Do you want to look at that before I do the mkinitrd? [07:54] done.. it's fine [07:54] i had to check only the timestamps on a file [07:54] please try the mkinitrd [07:55] I don't have mkinitrd installed. Let me find it in the archive. Apologies for the delay. [07:55] apt-get install mkinitrd-tools should do [07:55] I use aptitude, but yep... [07:56] OK. initrd created. Rebooting again. [07:56] ok === persia [n=persia@p3101-ipbf412marunouchi.tokyo.ocn.ne.jp] has joined #ubuntu-kernel [07:59] With the initrd, /boot mounted. Does this indicate a clear userspace issue, or have I just gotten lucky this time? [08:00] you need to try it a few times :( [08:00] if it works it's an issue with initrams [08:00] OK. I'll be back in a while. [08:00] initramfs [08:00] ok take your time === persia [n=persia@p3101-ipbf412marunouchi.tokyo.ocn.ne.jp] has joined #ubuntu-kernel [08:05] I can repeatedly mount /boot with the initrd. Does this indicate the kernel is happy and userspace is broken? [08:06] yeps [08:06] now go to bugzilla.ubuntu.com [08:06] file a bug to component initramfs-tools [08:06] explain all the setup in details [08:06] and you can copy&paste from this IRC session [08:06] so that jbailey can look into it asap [08:07] severity is major [08:07] Well thanks a lot for helping me understand that. I'm still chasing Malone #379, so I'll try to find someone to hit both of them as part of Bug Day before I add to the 3000+ entries in bugzilla, but it will get documented sometime soon. [08:07] 379 is something else tho [08:07] persia: i still suggest you file a bug asap [08:08] nobody in #ubuntu-bugs other than jbailey will ever upload initramgs [08:08] initramfs [08:08] i can tell you that :) [08:08] Are you sure? It looks like the same "mount is not getting the device issue". When it was discussed about 10 hours ago, we thought they were the same. I'll put something in bugzilla now just in case, but I'm still not sure it's not something about mount rather than just initramfs. [08:08] Thanks again :) [08:08] persia: i am sure it's not the same :) [08:09] because the usb is already with a booted system [08:09] the device is recognized byt the kernel [08:09] it is not automounted [08:09] in your case, it's not mounted even if it should (given the entry in fstab) [08:09] so trust me, they are different bugs [08:09] OK. The issue I have with a fresh boot is different, and being logged to bugzilla. The issue I had mounting it later appears the same to me, but I'm not worried about it, and am trusting you. [08:10] mounting it later is not the same [08:10] the 379 is talking about automount [08:10] that is different from mounting it later manually [08:10] bbl [08:11] i need to wake up my wife [08:11] persia: thanks a lot for your testing [08:11] pretty much appreciated [08:15] fabbione: It's bugzillla bug 15637, for your reference. Have a good day. === persia [n=persia@p3101-ipbf412marunouchi.tokyo.ocn.ne.jp] has left #ubuntu-kernel [] === doko_ [n=doko@dsl-084-059-088-000.arcor-ip.net] has joined #ubuntu-kernel === jorgp2 [n=jorgp@bnet-dial-122.bartnet.net] has joined #ubuntu-kernel [04:36] Did we get the ndiswrapper amd64 build fix thing sorted? === jorgp3 [n=jorgp@bnet-dial2-249.bartnet.net] has joined #ubuntu-kernel === airox [n=Gijs`@unaffiliated/airox] has joined #ubuntu-kernel [06:41] Hi [06:42] I seem to have a kernel panic after updating my SMP kernel on breezy preview yesterday. Known issue ? === airox [n=Gijs`@unaffiliated/airox] has left #ubuntu-kernel ["Leaving"] [07:05] airox: what does the panic say? [07:05] damn, missed him === Seveas [n=seveas@seveas.demon.nl] has joined #ubuntu-kernel === Seveas [n=seveas@seveas.demon.nl] has joined #ubuntu-kernel === Seveas [n=seveas@seveas.demon.nl] has joined #ubuntu-kernel