Hello,
I''m new to Xen world. As the first step I have tried to run debian
stable
which I''m running on my testing machine in domain 0 (no xend started
yet, no
virtual machines...). I''m using kernel 2.4.29, I have copied .config
from my
kernel tree and did "make ARCH=xen oldconfig" etc. Xen version is
2.0.4.
Xen and kernel boots OK, but I''m not able to use openafs (1.2.11).
Module libafs
can be loaded, but afsd generated oops immediately.
Is someone runnign openafs with xen? Should I try 2.6 kernel or create more
simple domain 0 linux and run my version in virtual machine?
Regards
--
Mirek Ruda
-------------------------------------------------------
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=6595&alloc_id=14396&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel
> I''m new to Xen world. As the first step I have tried to run debian stable > which I''m running on my testing machine in domain 0 (no xend started yet, no > virtual machines...). I''m using kernel 2.4.29, I have copied .config from my > kernel tree and did "make ARCH=xen oldconfig" etc. Xen version is 2.0.4. > > Xen and kernel boots OK, but I''m not able to use openafs > (1.2.11). Module libafs can be loaded, but afsd generated oops > immediately. > > Is someone runnign openafs with xen? Should I try 2.6 kernel or create more > simple domain 0 linux and run my version in virtual machine?We have used openafs successfully in the past (with 2.4.x kernels too).>From your somewhat vague description my best guess is that the moduleyou''re loading is not "xen-enabled" and hence is trying to use ring 0 instructions or somesuch. However difficult to see without more info. Can you post the backtrace you get after the module loads? And in fact the boot output you get from Xen and XenLinux? (also: I presume openafs is known working on your configuration when not using Xen?) cheers, S. ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
Steven Hand wrote:>> I''m new to Xen world. As the first step I have tried to run debian stable >>which I''m running on my testing machine in domain 0 (no xend started yet, no >>virtual machines...). I''m using kernel 2.4.29, I have copied .config from my >>kernel tree and did "make ARCH=xen oldconfig" etc. Xen version is 2.0.4. >> >>Xen and kernel boots OK, but I''m not able to use openafs >>(1.2.11). Module libafs can be loaded, but afsd generated oops >>immediately. >> >>Is someone runnign openafs with xen? Should I try 2.6 kernel or create more >>simple domain 0 linux and run my version in virtual machine? > > > We have used openafs successfully in the past (with 2.4.x kernels too). >>From your somewhat vague description my best guess is that the module > you''re loading is not "xen-enabled" and hence is trying to use ring 0 > instructions or somesuch.That may be source of my problem! I have compiled openafs myself, but just with option --with-linux-kernel-headers=/home/ruda/soft-local/xen-2.0/linux-2 .4.29-xen0, which may not be enough, because I didn''t specified equivalent of ARCH=xen anywhere. If someone already compiled openafs with xen kernel, could you send me suggestions what must be changed in openafs building? Looks like they create symlinks to several include directories of linux kernel, including asm->asm-i386. Would it be enough to change it to asm-xen? Also if I want to compile openafs for -xenU kernel, I should make it somehow "xen-enabled", right? BTW, I will experiment with arla too, because I was easily able to set ARCH=xen in Makefiles (they just copy linux subtree, add one subdir and call "make modules".> > Can you post the backtrace you get after the module loads? And > in fact the boot output you get from Xen and XenLinux?I''ll try to do it later, I don''t have serial console now.> (also: I presume openafs is known working on your configuration > when not using Xen?)Yes, this is machine I''m using for my daily work. I just tried to boot xen-enabled kernel (vmlinuz-2.4.29-xen0) instead of my linux-2.4.29. The same root filesystem etc. -- Mirek Ruda ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
Steven Hand wrote:> Can you post the backtrace you get after the module loads? And > in fact the boot output you get from Xen and XenLinux?See attached files, output of kernel and xen when booting title 2.4.29-xen0 root (hd0,0) kernel /boot/xen.gz dom0_mem=262144 com1=38400,8n1 module /boot/vmlinuz-2.4.29-xen0 root=/dev/hde1 ro console=ttyS0,38400 savedefault where /dev/hde1 is my standard (=working) root filesystem. File xen_linux.out is complete output, files ksymoops.X are results of ksymoops command used on two oops reports from this output (openafs and sendmail). Let me know if some other file is needed. The machine I''m using for tests is AMD Athlon, isn''t there some P4 dependency? BTW, I''m still not subscribed to mailing list (I''m waiting for confirmation from maileserver), please resend my email to list too. Thanks -- Mirek Ruda
Miroslav Ruda wrote:> Steven Hand wrote: > >> Can you post the backtrace you get after the module loads? And in fact >> the boot output you get from Xen and XenLinux?Short update - openafs with 2.4.29-xenU has the same problem. afsd is killed immediately when started. I''m able to use latest versions of arla (0.38) both with 2.6.10-xen and 2.4.29-xen kernels (both xen0 and xenU), but I would still prefere possibility to use openafs too. Is someone runnign openafs with xen? -- Mirek Ruda ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> Short update - openafs with 2.4.29-xenU has the same problem. > afsd is killed > immediately when started. > > I''m able to use latest versions of arla (0.38) both with > 2.6.10-xen and > 2.4.29-xen kernels (both xen0 and xenU), but I would still > prefere possibility > to use openafs too. Is someone runnign openafs with xen?We''ve used openafs with 2.4.29-xen0 without problems. I seem to recall that getting it to build against xen/i386 rather than plain i386 was somewhat tricky, as setting ARCH=xen was insufficient due to a busted Makefile. I think I resorted to renaming arch/xen to arch/i386 and the same for include/asm-xen. Ian ------------------------------------------------------- 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_ide95&alloc_id396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
Ian Pratt wrote:> We''ve used openafs with 2.4.29-xen0 without problems. I seem to recall > that getting it to build against xen/i386 rather than plain i386 was > somewhat tricky, as setting ARCH=xen was insufficient due to a busted > Makefile. I think I resorted to renaming arch/xen to arch/i386 and the > same for include/asm-xen.I wish it''s so easy. However, renaming of include/asm-xen to include/asm-i386 doesn''t work, because some include files from asm-xen (param.h for example) include ../asm-i386/ files too :-) Moreover, my guess is that it''s still not sufficient, see bellow. Openafs'' Makefile creates symlinks to several kernel-source/include directories including asm-i386 (others are include/net, include/linux, include/asm). I was easily able to change this makefile to use asm-xen instead and I was able to compile openafs with such setup. And this version creates oops I have sent to xen mailing list. However, I don''t see any other usage of kernel source and therefore I don''t understand what is missing to create "xen-enabled" binaries. -- Mirek Ruda ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
I haven''t looked at it in detail yet (though I will), but I expect that OpenAFS on Xen is going to need its own sysname target, rather than being able to just use the i386 one. I had to do the same when I did the port to UML (which was actually fairly easy). So, the possible options are that a new sysname target is needed (e.g. it could be something like i386_xenlinux24, and so on), or the target for regular i386 linux may need to be extended to also operate for Xen, but that is less likely to work well I think. Kris On Wed, Mar 09, 2005 at 04:49:25PM +0100, Miroslav Ruda wrote:> Ian Pratt wrote: > >We''ve used openafs with 2.4.29-xen0 without problems. I seem to recall > >that getting it to build against xen/i386 rather than plain i386 was > >somewhat tricky, as setting ARCH=xen was insufficient due to a busted > >Makefile. I think I resorted to renaming arch/xen to arch/i386 and the > >same for include/asm-xen. > > I wish it''s so easy. However, renaming of include/asm-xen to > include/asm-i386 > doesn''t work, because some include files from asm-xen (param.h for example) > include ../asm-i386/ files too :-) Moreover, my guess is that it''s still > not sufficient, see bellow. > > Openafs'' Makefile creates symlinks to several kernel-source/include > directories > including asm-i386 (others are include/net, include/linux, include/asm). > I was easily able to change this makefile to use asm-xen instead and I was > able to compile openafs with such setup. And this version > creates oops I have sent to xen mailing list. However, I don''t see any > other usage of kernel source and therefore I don''t understand what is > missing to create "xen-enabled" binaries. > > -- > Mirek Ruda > > > ------------------------------------------------------- > 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=6595&alloc_id=14396&op=click > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/xen-devel-- Never underestimate a Mage with: - the Intelligence to cast Magic Missile, - the Constitution to survive the first hit, and - the Dexterity to run fast enough to avoid being hit a second time. ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> > We''ve used openafs with 2.4.29-xen0 without problems. I > seem to recall > > that getting it to build against xen/i386 rather than plain i386 was > > somewhat tricky, as setting ARCH=xen was insufficient due > to a busted > > Makefile. I think I resorted to renaming arch/xen to > arch/i386 and the > > same for include/asm-xen. > > I wish it''s so easy. However, renaming of include/asm-xen to > include/asm-i386 > doesn''t work, because some include files from asm-xen > (param.h for example) > include ../asm-i386/ files too :-) Moreover, my guess is that > it''s still not > sufficient, see bellow.I think it was for openafs that I created the tools/misc/fakei386 script. Ian ------------------------------------------------------- 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_ide95&alloc_id396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
Ian Pratt wrote:> > > We''ve used openafs with 2.4.29-xen0 without problems. I > >>seem to recall >> >>>that getting it to build against xen/i386 rather than plain i386 was >>>somewhat tricky, as setting ARCH=xen was insufficient due >> >>to a busted >> >>>Makefile. I think I resorted to renaming arch/xen to >> >>arch/i386 and the >> >>>same for include/asm-xen. >> >>I wish it''s so easy. However, renaming of include/asm-xen to >>include/asm-i386 >>doesn''t work, because some include files from asm-xen >>(param.h for example) >>include ../asm-i386/ files too :-) Moreover, my guess is that >>it''s still not >>sufficient, see bellow. > > > I think it was for openafs that I created the tools/misc/fakei386 > script.Great, that was missing piece! I''m running openafs-1.2.11 compiled with faked linux-2.4.29-xen0 and until now everything works fine. Many thanks! -- Mirek Ruda ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> > Great, that was missing piece! I''m running openafs-1.2.11 > compiled with faked > linux-2.4.29-xen0 and until now everything works fine.Please can you write a good bug report for the openafs maintainers to get them to fix their Makefiles. Thanks, Ian ------------------------------------------------------- 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_ide95&alloc_id396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Wed, Mar 09, 2005 at 04:58:26PM -0000, Ian Pratt wrote:> > Great, that was missing piece! I''m running openafs-1.2.11 > > compiled with faked > > linux-2.4.29-xen0 and until now everything works fine. > > Please can you write a good bug report for the openafs maintainers to > get them to fix their Makefiles.Make sure to send the bug report to openafs-bugs@openafs.org. Please also explain the way that you are currently working around what you consider to be a problem in the OpenAFS makefiles. That will assist everyone looking at the bug reports to evaluate the appropriate action that is needed to make this all work fine. Kris ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
Ian Pratt wrote:>>Great, that was missing piece! I''m running openafs-1.2.11 >>compiled with faked >>linux-2.4.29-xen0 and until now everything works fine. > > > Please can you write a good bug report for the openafs maintainers to > get them to fix their Makefiles.Are you sure that this is bug in their Makefile? Even if I created correct symlinks to asm-xen instead of asm-i386 (not using your script), compilation included some i386 header instead of xen-enabled. I can imagine several reasons: - some header file from /usr/include/ was used (because openafs Makefile creates only several faked (=symlink to kernel source) include directories) -> this is bug in openafs - openafs code included some xen header file, which is not "expected" to be included directly and therefore this xen header file included wrong i386 version - if for example asm-xen is just subset of asm-i386, attempt to include some header existing only in i386 version should lead to usage of system header from /usr/include/asm-i386 I''ll try to investigate it (first step will be trace usage of headers pre-installed to /usr/include), but I''m not yet convinced what''s real source of problems. Mirek Ruda ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
How are you compiling OpenAFS, and where? OpenAFS expects to have a configured linux source tree available, or at a minimum the kernel headers from a configured kernel source tree. It uses that to determine various things during the compilation (like the sysname, architecture, etc...), and to be able to compile the OpenAFS kernel module. If the kernel headers are not consistent (e.g. they claim to be for i386, but asm is linked to some other architecture directory), things will fail badly. Either way, I wouldn''t call any of this problem a ''bug'' given that OpenAFS never claimed to support Xen (yet). So, it''s more a case of Xen simply not being supported yet. It is being worked on, though (at least from my side) at a somewhat slower pace due to other work. Any patches to add support for Xen to OpenAFS would be welcome. But either way, it will come to pass :) Kris On Wed, Mar 09, 2005 at 11:39:42PM +0100, Miroslav Ruda wrote:> Are you sure that this is bug in their Makefile? Even if I created correct > symlinks to asm-xen instead of asm-i386 (not using your script), > compilation included some i386 header instead of xen-enabled. I can imagine > several reasons: > - some header file from /usr/include/ was used (because openafs Makefile > creates only several faked (=symlink to kernel source) include directories) > -> this is bug in openafs > - openafs code included some xen header file, which is not "expected" to be > included directly and therefore this xen header file included wrong i386 > version > - if for example asm-xen is just subset of asm-i386, attempt to include some > header existing only in i386 version should lead to usage of system header > from /usr/include/asm-i386 > > I''ll try to investigate it (first step will be trace usage of headers > pre-installed to /usr/include), but I''m not yet convinced what''s real > source of problems. > > Mirek Ruda > > > ------------------------------------------------------- > 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=6595&alloc_id=14396&op=click > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/xen-devel-- Never underestimate a Mage with: - the Intelligence to cast Magic Missile, - the Constitution to survive the first hit, and - the Dexterity to run fast enough to avoid being hit a second time. ------------------------------------------------------- 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=6595&alloc_id=14396&op=click _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel