Thread: [Ani-users] Interrupted transfers from samba share
Status: Beta
Brought to you by:
substring-sfn
Menu
▾
▴
ani-users
|
From: Alex G. <al...@vi...> - 2005-04-11 13:47:14
|
Hi everybody, i use ANI for the first time, have setup a tftp-server and a samba share where my files are located on. I boot my client, the client boots pxe, runs the linux kernel, does a dhcp request and mounts the share. After transferring some files the client software die()s with a fatal error message from MountParts::part(8192). I tried to search mailing list archives, but didn't find a similar error. Any hints on fixing this. I'm familiar to perl so i could debug the problem further if somebody would give me some first hints ;) cheers, alex -- alexander goller al...@vi... |
|
From: Alex G. <al...@vi...> - 2005-04-11 14:06:51
|
On Mon, Apr 11, 2005 at 03:47:10PM +0200, Alex Goller wrote: > Any hints on fixing this. I'm familiar to perl so i could debug the > problem further if somebody would give me some first hints ;) Ok, this was my fault. After the copying stopped i simply restarted /etc/main.pl without unmounting the filesystems. My original error that was trapped before i did that was: dirCopy::copy:(/mnt/inst/i386 -> /mnt/img) Illegal seek (256) So, any hints on that one? The disks are SATA drives, maybe this causes a failure. cheers, alex -- alexander goller al...@vi... |
|
From: Till W. <t.w...@su...> - 2005-04-11 14:12:29
|
Hi Alex!
Alex Goller wrote:
> My original error that was trapped before i did that was:
>
> dirCopy::copy:(/mnt/inst/i386 -> /mnt/img) Illegal seek (256)
>
> So, any hints on that one?
> The disks are SATA drives, maybe this causes a failure.
Oh yes it does! SATA is not yet supported :( We will change this with
the next kernel update. If you're lucky, your BIOS offers an IDE
emulation mode...
Sorry for this bad news.
bg
Till
>
> cheers, alex
--
-------------------------
Substring GmbH
Tscharnerstrasse 39a
CH-3007 Bern
Tel. ++41 (0)31 382 51 12
Fax. ++41 (0)31 382 51 87
http://www.substring.ch
SMS via
http://www.substring.ch/sms.php?sendto=twimmer
|
|
From: Alex G. <al...@vi...> - 2005-04-11 14:22:00
|
Hi Till, thanks for the quick response, i will check if ide emulation is available. On Mon, Apr 11, 2005 at 04:12:14PM +0200, Till Wimmer wrote: > Oh yes it does! SATA is not yet supported :( We will change this with > the next kernel update. If you're lucky, your BIOS offers an IDE > emulation mode... The other workstations we want to install have SCSI drives, and i have the same issues than somebody else had with the hardcoded drive hda in /etc/ani.cfg in the pxelinux image :) cheers, alex -- alexander goller al...@vi... |
|
From: Alexander S. <a.s...@su...> - 2005-04-11 22:01:01
|
Hi Alex, Just some observation, don't think it is the case with you, but nevertheless... If this "illegal seek" stuff happens while having already copied some files, it mostly has todo with windows encountring some strange files - like a backup of files being edited with vi or some other file which has a name diffrent from the 8.3-DOS-format (8 letter +9 dot + 3 letters suffix) like a .bashr or whatever :-). BIOS-settings: look also for other silly description of IDE-emulation of SATA drive in the BIOS. We have successfully installed some DELL machines with SATA-drives, after having enabled this, but it was called: "compatibility with old operating systems" or something like that. Xa Alex Goller wrote: > Hi Till, > > thanks for the quick response, i will check if ide emulation is > available. > > On Mon, Apr 11, 2005 at 04:12:14PM +0200, Till Wimmer wrote: > >>Oh yes it does! SATA is not yet supported :( We will change this with >>the next kernel update. If you're lucky, your BIOS offers an IDE >>emulation mode... > > The other workstations we want to install have SCSI drives, and i have > the same issues than somebody else had with the hardcoded drive hda in > /etc/ani.cfg in the pxelinux image :) > > cheers, alex |
|
From: Carsten L. <cal...@us...> - 2005-04-11 15:54:33
|
Carsten Lehmann wrote: > Alex Goller wrote: > >> >> The other workstations we want to install have SCSI drives, and i have >> the same issues than somebody else had with the hardcoded drive hda in >> /etc/ani.cfg in the pxelinux image :) > > > You could comment out the INSTDEV line in ani.cfg. Then it should be > possible to change the pxelinux.cfg/default file on the tftp Server to > > ------------------------------------------------------------------- > > default local > timeout 40 > prompt 1 > > > # To protect against accidentally pressing F12 or similar > # on boot.. > # > # So you must be quick to enter 'new' on bootprompt ! > > label local > LOCALBOOT 0 > > # this one is really booting from LAN > label new > kernel vmlinuz > append root=/dev/ram0 init=linuxrc INITDEV=pxe INSTDEV=/dev/hda > initrd=initrd_pxe.gz rw > > label update > kernel vmlinuz > append root=/dev/ram0 init=linuxrc INITDEV=pxe INSTDEV=/dev/hda > SETUP=1 initrd=initrd_pxe.gz rw > > label scsi > kernel vmlinuz > append root=/dev/ram0 init=linuxrc INITDEV=pxe INSTDEV=/dev/sda > initrd=initrd_pxe.gz rw > > ------------------------------------------------------------------- > > Now it should be possible (everything not tested yet !) to boot from > both ide or scsi... > Sorry, it should be INSTDEV=hda resp. INSTDEV=sda ! |
|
From: Alex G. <al...@vi...> - 2005-04-13 09:01:41
|
On Mon, Apr 11, 2005 at 05:18:52PM +0200, Thomas Seiler wrote: > > Is the ANI kernel patched in any way in order to work or is it an > > vanilla kernel? > > As far as I can tell its vanilla 2.4.25. the original .config is > available here: > http://cvs.sourceforge.net/viewcvs.py/ani/W2K/sources/AniInst/vmlinuz-FD/ > > > I can do tests, sure. My problem is that i have a LSI logic controller > > and drives won't get recognized at all at boot time. > Autsch, so you need to compile a custom kernel. I compiled a 2.4.30 kernel and i am successfully booting that from pxe. I have encountered some problems with ANI's autoprobe command which doesn't seem to work on 2.4.30 anymore. I had a look in the sources, but had not time for further debugging. Autoprobe outputs nothing when i point it to my /lib/modules/2.4.30/modules.pcimap (this could be my fault because i may not have followed all conventions or assumptions autoprobe makes). So instead of autoprobing i have hardcoded probing for the networks cards we use into /linuxrc. Mounting the samba share works fine, the files get transferred, but i still get the error message of my original post (illegal seek) at some point. I will now try to turn on debugging in the main installer and Ani.pm, in order to get more details when or at which point the error occurs. cheers, alex -- alexander goller al...@vi... |
|
From: Till W. <t.w...@su...> - 2005-04-13 13:58:37
|
Hi Alex Alex Goller schrieb: > >I compiled a 2.4.30 kernel and i am successfully booting that from >pxe. I have encountered some problems with ANI's autoprobe command >which doesn't seem to work on 2.4.30 anymore. I had a look in the >sources, but had not time for further debugging. Autoprobe outputs >nothing when i point it to my /lib/modules/2.4.30/modules.pcimap (this >could be my fault because i may not have followed all conventions or >assumptions autoprobe makes). >So instead of autoprobing i have hardcoded probing for the networks >cards we use into /linuxrc. > > Autoprobe maybe does not work when you have verbose pci info enabled. Autoprobe parses /proc/pci and it can only handle the non-textual format of this file. Another reason could be that the format of /proc/pci has changed in 2.4.30 but this is rather implausible. >Mounting the samba share works fine, the files get transferred, but i >still get the error message of my original post (illegal seek) at some >point. > > Can you say for sure that the partitons are correctly mounted when Ani.pm starts to copy the files? What does the /mnt/tmp look like after Ani.pm failed? Is there enough space on the concerning partition? Have a look in /ani.log. Maybe there is a more verbose error message from the cp command... Good luck Till -- Where do you want to go today? - ANIwhere! http://ani.sourceforge.net |
|
From: Alex G. <al...@vi...> - 2005-04-14 14:56:29
|
On Wed, Apr 13, 2005 at 03:58:17PM +0200, Till Wimmer wrote: > Can you say for sure that the partitons are correctly mounted when > Ani.pm starts to copy the files? Ok, the partitions are mounted as msdos filesystems, /mnt/img is having enough space. > What does the /mnt/tmp look like after Ani.pm failed? Is there enough > space on the concerning partition? > Have a look in /ani.log. Maybe there is a more verbose error message > from the cp command... Ok, cp is setting errno $! and the exit code $? ($! = "Illegal seek", $? = 256). That's all information i get. One thing i found interesting is the smbfs bug you're mentioning in the sourcecode of Ani.pm, could this cause the sort of problems i have? (i found out that it happens when using my scsi box with a MPT Fusion controller as well as a Fujitsu Siemens Celsius workstation with SATA disks, 2.4.30 should handle those fine and probes the controller and disks as ICH5-SATA, so i think this is not a SATA problem). I still can't copy all the files over. Will provide a more verbose log later or tomorrow. Another thing was pretty crazy this morning. I have changed file in manix.bz2 and repacked the archive with tar cjvf manix.bz2 <mymanixfiles>. The tar on the installroot couldn't create the block and character devices in /mnt/tmp/dev, so the chroot failed (prepare.sh). I wasn't aware of such tar incompatibilities and i still need to find a tar that's working. Right now i'm copying the devices over to the new root. cheers, alex -- alexander goller al...@vi... |
|
From: <h.m...@su...> - 2005-04-14 16:10:56
|
> > Another thing was pretty crazy this morning. I have changed file in > manix.bz2 and repacked the archive with tar cjvf manix.bz2 > <mymanixfiles>. The tar on the installroot couldn't create the block > and character devices in /mnt/tmp/dev, so the chroot failed > (prepare.sh). I wasn't aware of such tar incompatibilities and i still > need to find a tar that's working. Right now i'm copying the devices > over to the new root. You have to logon as root if you want to repack the archive. Otherwise the block devices can't be created. -- hagen |
|
From: Alex G. <al...@vi...> - 2005-04-14 16:14:30
|
On Thu, Apr 14, 2005 at 06:10:40PM +0200, h.m...@su... wrote: > > > > Another thing was pretty crazy this morning. I have changed file in > > manix.bz2 and repacked the archive with tar cjvf manix.bz2 > > <mymanixfiles>. The tar on the installroot couldn't create the block > > and character devices in /mnt/tmp/dev, so the chroot failed > > (prepare.sh). I wasn't aware of such tar incompatibilities and i still > > need to find a tar that's working. Right now i'm copying the devices > > over to the new root. > > You have to logon as root if you want to repack the archive. Otherwise the > block devices can't be created. Of course i was doing this as root. I can send you the modified manix.bz2 file and you can try if your prepare.sh on a box of yours can unpack that archive, i assume prepare.sh runs as uid 0. I was using gnu tar on Debian/testing. cheers, alex -- alexander goller al...@vi... |
|
From: Alex G. <al...@vi...> - 2005-04-15 09:45:27
|
Hi Hagen, On Thu, Apr 14, 2005 at 06:14:27PM +0200, Alex Goller wrote: > > You have to logon as root if you want to repack the archive. Otherwise the > > block devices can't be created. > > Of course i was doing this as root. I can send you the modified > manix.bz2 file and you can try if your prepare.sh on a box of yours > can unpack that archive, i assume prepare.sh runs as uid 0. > I was using gnu tar on Debian/testing. I have verified that this is a tar problem. I tried busybox tar for a start and the devices in dev/ get unpacked. I will try other tar versions (FreeBSD, IRIX, AIX) to see if any of those shows similar behaviour. I have also written a bunch of Makefiles to assemble initrd_pxe, filesys.bz2, modules.bz2, manix.bz2 from a set of directories, write those files to an output directory and copy them over to the SMB share, tftproot. They're quite generic so i guess other people could use that too. if anyone is interested i can put the stuff on a webserver. cheers, alex -- alexander goller al...@vi... |
|
From: Carsten L. <cal...@us...> - 2005-04-15 09:58:09
|
Alex Goller wrote: > I have also written a bunch of Makefiles to assemble initrd_pxe, > filesys.bz2, modules.bz2, manix.bz2 from a set of directories, write > those files to an output directory and copy them over to the SMB > share, tftproot. They're quite generic so i guess other people could > use that too. if anyone is interested i can put the stuff on a > webserver. Could you send them to me ? Maybe we should place them on sourceforge ... Carsten |
|
From: Till W. <t.w...@su...> - 2005-04-15 10:45:42
|
Alex Goller schrieb: >On Wed, Apr 13, 2005 at 03:58:17PM +0200, Till Wimmer wrote: > > > >>Have a look in /ani.log. Maybe there is a more verbose error message >>from the cp command... >> >> > >Ok, cp is setting errno $! and the exit code $? ($! = "Illegal seek", >$? = 256). That's all information i get. > > Did you find the ani.log file? It should contain the output from cp command, not the one of the perl script. Did you try to copy the share by hand? Maybe you find the reason for your troubles by that way... >One thing i found interesting is the smbfs bug you're mentioning in >the sourcecode of Ani.pm, could this cause the sort of problems i have? > > No i don't think so. The bug makes files on the mouted samba share disapear sometimes. Cp hasn't any problem with this, but winnt.exe does ;) Yes it's a awful thing.... (It's a bug in the smbfs-part of the kernel concerning the caching of the directory listing. It only appears when share with huge directories are mounted. I tried to fix this some years ago but i gave up because of bad cooperation between the samba and kernel team. Maybe that's different today?!) >(i found out that it happens when using my scsi box with a MPT Fusion >controller as well as a Fujitsu Siemens Celsius workstation with SATA >disks, 2.4.30 should handle those fine and probes the controller and >disks as ICH5-SATA, so i think this is not a SATA problem). >I still can't copy all the files over. Will provide a more verbose log >later or tomorrow. > > Yes please... We are very interested in that! >Another thing was pretty crazy this morning. I have changed file in >manix.bz2 and repacked the archive with tar cjvf manix.bz2 ><mymanixfiles>. The tar on the installroot couldn't create the block >and character devices in /mnt/tmp/dev, so the chroot failed >(prepare.sh). I wasn't aware of such tar incompatibilities and i still >need to find a tar that's working. Right now i'm copying the devices >over to the new root. > > I'm using tar (GNU tar) 1.13.25 at the moment without any problems. Does tar need some kernel headers to compile (file system or something)? Maybe there is some incompatibility... bg Till -- Where do you want to go today? - ANIwhere! http://ani.sourceforge.net |
|
From: <h.m...@su...> - 2005-04-13 14:02:50
|
I agree with Xa: Normally the illegal seek message occurs, if a file on the installation share is opened by another application, e.g. if runinst.tpl is still opened by vi... -- hagen > On Mon, Apr 11, 2005 at 05:18:52PM +0200, Thomas Seiler wrote: >> > Is the ANI kernel patched in any way in order to work or is it an >> > vanilla kernel? >> >> As far as I can tell its vanilla 2.4.25. the original .config is >> available here: >> http://cvs.sourceforge.net/viewcvs.py/ani/W2K/sources/AniInst/vmlinuz-= FD/ >> >> > I can do tests, sure. My problem is that i have a LSI logic controll= er >> > and drives won't get recognized at all at boot time. >> Autsch, so you need to compile a custom kernel. > > I compiled a 2.4.30 kernel and i am successfully booting that from > pxe. I have encountered some problems with ANI's autoprobe command > which doesn't seem to work on 2.4.30 anymore. I had a look in the > sources, but had not time for further debugging. Autoprobe outputs > nothing when i point it to my /lib/modules/2.4.30/modules.pcimap (this > could be my fault because i may not have followed all conventions or > assumptions autoprobe makes). > So instead of autoprobing i have hardcoded probing for the networks > cards we use into /linuxrc. > Mounting the samba share works fine, the files get transferred, but i > still get the error message of my original post (illegal seek) at some > point. > I will now try to turn on debugging in the main installer and Ani.pm, > in order to get more details when or at which point the error occurs. > > cheers, alex > -- > alexander goller al...@vi... > > > ------------------------------------------------------- > SF email is sponsored by - The IT Product Guide > Read honest & candid reviews on hundreds of IT Products from real users. > Discover which products truly live up to the hype. Start reading now. > http://ads.osdn.com/?ad_id=3D6595&alloc_id=3D14396&op=3Dclick > _______________________________________________ > Ani-users mailing list > Ani...@li... > https://lists.sourceforge.net/lists/listinfo/ani-users > --=20 Hagen M=FCnch ------------------------- Substring GmbH Tscharnerstrasse 39a CH-3007 Bern Tel. +41 (0)31 382 51 12 Fax. +41 (0)31 382 51 87 http://www.substring.ch SMS via http://www.substring.ch/sms.php?sendto=3Dhmuench |
Thanks for helping keep SourceForge clean.
X