The Xen 1.0 CD works fine, and I can start booting the image.gz and xenolinux.gz via Grub if I copy them to my hard drive. However, I need to build a new kernel in order to support the rest of my hardware, also my ReiserFS filesystems (I strongly advise adding ReiserFS to the default for the next Xen release, since it is there by default on SuSE). I have tried oodles of variations, which I''ll list below. The bottom line, though, seems to be that I''m ending up with an invalid kernel that Grub refuses to boot: Error 13: Invalid or unsupported executable format This error happens before I can give the boot command (I''m using the grub shell, launched via MBR, after rebooting my system). Basically: grub> root (hd0,0) grub> kernel /xenimage.gz [args...doesn''t matter which] Error 13: Invalid or unsupported executable format This does not seem to be an LBA problem, since the /boot partition is small, and I''m able to at least start booting the images I copied from the CD. (They won''t boot fully, though, and reboot my system right after saying something like "DOMID is 0".) Grub is installed from the version 0.93 patched source on the CD image. I''ve built kernels using both the 1.0 CD image (copied to my hard drive, of course) and the current development snapshot, built via the supplied xen-clone script as downloaded via BitKeeper. When I built the kernel without any modifications from the CD image, I *still* was unable to get an image that would boot. So, I think this is something with the modified kernel or perhaps something related to my particular system. The kernel build process was otherwise smooth, and built vmlinux as well as arch/xeno/boot/image.gz and the xenolinux in an adjacent directory. This is on a Dell 530 workstation with 2 1.7Ghz Xeon processors and 2GB of RAM. It''s basically a stock SuSE 8.1 installation, with gcc 3.3. Thanks for any tips or leads. I''ve tried things in many, many different ways across the two Xen versions, with different kernel configure options, etc. I tried booting the default vmlinux (before the multi-boot bytes are prepended), but that kernel will not boot either via Grub. At one point, I had a kernel that was too big, but modularized some of the kernel and got it down to size. I was able to make Lilo work (a good way to see if your kernel seems OK), but at boot got: EBDA too big and the boot never started. -- Greg Dr. Gregory B. Newby, Research Faculty, Arctic Region Supercomputing Center University of Alaska Fairbanks. PO Box 756020, Fairbanks, AK 99775 e: newby@arsc.edu v: 907-474-7160 f: 907-474-5494 w: www.arsc.edu/~newby ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> This error happens before I can give the boot > command (I''m using the grub shell, launched via MBR, > after rebooting my system). Basically: > > grub> root (hd0,0) > grub> kernel /xenimage.gz [args...doesn''t matter which] > Error 13: Invalid or unsupported executable formatSo it''s the new build of Xen that doesn''t work? It looks like it''s not recognising the Xen image as a Multiboot OS. What happens if you try using the non-gzipped image (you can take xenimage.gz and just gunzip it, and try that)? If it still fails then we might be able to take a look at it if you can make your image available to us (maybe put it on a web server and send us a URL?). I think someone else reported a similar problem a while ago. Perhaps our build process is broken for certain versions of gcc and/or binutils. Unfortunately it''s not possible to get more useful diagnostics out of GRUB without adding tracing to the source code and recompiling it. -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Mon, Nov 03, 2003 at 10:30:06PM +0000, Keir Fraser wrote:> > This error happens before I can give the boot > > command (I''m using the grub shell, launched via MBR, > > after rebooting my system). Basically: > > > > grub> root (hd0,0) > > grub> kernel /xenimage.gz [args...doesn''t matter which] > > Error 13: Invalid or unsupported executable format > > So it''s the new build of Xen that doesn''t work?I tried it with the 1.0 build from the CD image, and from last Thursday''s "unstable" download via BitKeeper. Both yielded essentially the same results.> It looks like it''s not recognising the Xen image as a Multiboot OS.Correct. The Linux "file" command thinks it''s data. But "file vmlinux" doesn''t recognize it as a Linux kernel, while "file /boot/vmlinuz" (my current non-Xen kernel) does. So, the problem might be happening before the Multiboot bytes are added (in arch/xeno/boot during "make bzImage").> What happens if you try using the non-gzipped image (you can take > xenimage.gz and just gunzip it, and try that)?No difference...same problem.> If it still fails then we might be able to take a look at it if you > can make your image available to us (maybe put it on a web server and > send us a URL?). I think someone else reported a similar problem a > while ago. Perhaps our build process is broken for certain versions of > gcc and/or binutils. Unfortunately it''s not possible to get more > useful diagnostics out of GRUB without adding tracing to the source > code and recompiling it.I put the whole package, new & old & copies of everything in /boot at: http://peabody.arsc.edu/xen One of the (many) images I''ve tried is: http://peabody.arsc.edu/xen/boot/xenimage.gz In the boot/grub/menu.lst , you''ll see the approximate grub commands (I have been typing them in by hand at the grub> prompt, as mentioned earlier). But the exact sequence doesn''t matter: grub complains about /xenimage.gz and won''t proceed with a boot. The image from the CD is just fine. But *both* the multi-boot images (from xen/arch/boot) and "default" vmlinux (from the root kernel compile directory) fail with the same "error 13" from Grub. From *both* the 1.0 and nightly unstable builds. If it would help, I can set you up with access to the machine to try some different configuration tricks. If you think an earlier compiler would help, I can try it. I''m at version 2.14 for my binutils, BTW, and 3.3 for gcc. Thanks! Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> I tried it with the 1.0 build from the CD image, > and from last Thursday''s "unstable" download via BitKeeper. > Both yielded essentially the same results.What I mean is ---- it''s the Xen image that you''re trying and that is failing. You''re not trying to boot the Xenolinux image directly from GRUB, right? That wouldn''t work :-)> Correct. The Linux "file" command thinks it''s data. > But "file vmlinux" doesn''t recognize it as a Linux kernel, > while "file /boot/vmlinuz" (my current non-Xen kernel) does. > > So, the problem might be happening before the Multiboot > bytes are added (in arch/xeno/boot during "make bzImage").Are we talking about Xen, or Xenolinux (the port of Linux to run on top of Xen)? Xen is a Multiboot OS. It has 12 bytes of Multiboot header that must appear in the first 8192 bytes of the uncompressed image (GRUB will uncompress the image before searching for the Multiboot header). Xenolinux is /not/ a multiboot OS. In fact, it''s not a normal Linux image either. It''s just a raw image file with an 8-byte custom header on the front. GRUB will not recognise Xenolinux as a valid kernel image. This is because Xenolinux can only be booted via Xen. GRUB /should/ recognise Xen, as it should find the magic Multiboot header at a multiple-of-4 offset somewhere in the first 8192 bytes.> > If it still fails then we might be able to take a look at it if you > > can make your image available to us (maybe put it on a web server and > > send us a URL?). I think someone else reported a similar problem a > > while ago. Perhaps our build process is broken for certain versions of > > gcc and/or binutils. Unfortunately it''s not possible to get more > > useful diagnostics out of GRUB without adding tracing to the source > > code and recompiling it. > > I put the whole package, new & old & copies of everything > in /boot at: > http://peabody.arsc.edu/xen > > One of the (many) images I''ve tried is: > http://peabody.arsc.edu/xen/boot/xenimage.gzI''ll take a quick look... -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> The image from the CD is just fine. But *both* the multi-boot > images (from xen/arch/boot) and "default" vmlinux (from > the root kernel compile directory) fail with the same > "error 13" from Grub. From *both* the 1.0 and nightly > unstable builds. > > If it would help, I can set you up with access to > the machine to try some different configuration tricks. > If you think an earlier compiler would help, I can try it. > I''m at version 2.14 for my binutils, BTW, and 3.3 for gcc.Your xenimage.gz is a Xenolinux file, not a Xen file. It should be specified as a module to GRUB. The kernel line should specify a Xen image. Case dismissed. :-) -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Mon, Nov 03, 2003 at 11:08:30PM +0000, Keir Fraser wrote:> > The image from the CD is just fine. But *both* the multi-boot > > images (from xen/arch/boot) and "default" vmlinux (from > > the root kernel compile directory) fail with the same > > "error 13" from Grub. From *both* the 1.0 and nightly > > unstable builds. > > > > If it would help, I can set you up with access to > > the machine to try some different configuration tricks. > > If you think an earlier compiler would help, I can try it. > > I''m at version 2.14 for my binutils, BTW, and 3.3 for gcc. > > Your xenimage.gz is a Xenolinux file, not a Xen file.Aha. Um.> It should be specified as a module to GRUB. The kernel line should > specify a Xen image. > > Case dismissed. :-)Maybe...more likely I just got lost and put the wrong file there. I just re-copied, based on what I wrote below. Please take another quick look: http://peabody.arsc.edu/xen/boot/xenimage.gz After I do a "xeno-clone" or similar, where is the actual bootable image file located? I thought the bootable image is here (which is what I just re-copied to http://peabody.arsc.edu/xen/boot/xenimage.gz): xeno-unstable.bk/xeno-clone/xenolinux-2.4.22/arch/xeno/boot/image.gz and the xenolinux file is: xeno-unstable.bk/xeno-clone/install/boot/xenolinux.gz So if this isn''t right, where the heck is the kernel image??? I already tried to boot from: xeno-unstable.bk/xeno-clone/xenolinux-2.4.22/vmlinux (or vmlinux.gz), but from the "make bzImage" it seems that the image.gz is simply the vmlinux with the multiboot bytes prepended. Neither will work via the "grub> kernel /FILENAME" command I admit to being confused! Thanks for further guidance on this... -- Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> I thought the bootable image is here (which is what I just re-copied > to http://peabody.arsc.edu/xen/boot/xenimage.gz): > xeno-unstable.bk/xeno-clone/xenolinux-2.4.22/arch/xeno/boot/image.gz > > and the xenolinux file is: > xeno-unstable.bk/xeno-clone/install/boot/xenolinux.gz > > So if this isn''t right, where the heck is the kernel image???We chose rather unsuitable names for the image files. In fact both Xen and Xenolinux compile to image.gz within their respective trees, although Xenolinux''s image is renamed when it''s installed in install/boot. I can easily change that for the development trees so I''ll sort that out.... To build Xen, from within the xeno-clone directory you need to: cd xeno-1.0.bk make install You should then find a Xen image.gz file in xeno-clone/install/boot. This same file also resides at xeno-clone/xeno-1.0.bk/xen/image.gz. xeno-xlone/xenolinux-2.4.22/arch/xeno/boot/image.gz is /not/ a Xen image file; it''s a Xenolinux image file (it''s the same as xeno-clone/install/boot/xenolinux.gz, in fact).> I already tried to boot from: > xeno-unstable.bk/xeno-clone/xenolinux-2.4.22/vmlinux (or vmlinux.gz), > but from the "make bzImage" it seems that the image.gz is simply the > vmlinux with the multiboot bytes prepended. Neither will work > via the "grub> kernel /FILENAME" commandvmlinux is not directly bootable or usable -- ignore it. -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Tue, Nov 04, 2003 at 12:52:43AM +0000, Keir Fraser wrote:> > I thought the bootable image is here (which is what I just re-copied > > to http://peabody.arsc.edu/xen/boot/xenimage.gz): > > xeno-unstable.bk/xeno-clone/xenolinux-2.4.22/arch/xeno/boot/image.gz > > > > and the xenolinux file is: > > xeno-unstable.bk/xeno-clone/install/boot/xenolinux.gz > > > > So if this isn''t right, where the heck is the kernel image??? > > We chose rather unsuitable names for the image files. In fact both Xen > and Xenolinux compile to image.gz within their respective trees, > although Xenolinux''s image is renamed when it''s installed in > install/boot. I can easily change that for the development trees so > I''ll sort that out.... > > To build Xen, from within the xeno-clone directory you need to: > cd xeno-1.0.bk > make install > > You should then find a Xen image.gz file in xeno-clone/install/boot. This > same file also resides at xeno-clone/xeno-1.0.bk/xen/image.gz. > > xeno-xlone/xenolinux-2.4.22/arch/xeno/boot/image.gz is /not/ a Xen > image file; it''s a Xenolinux image file (it''s the same as > xeno-clone/install/boot/xenolinux.gz, in fact).Ok, I was able to do this successfully. Thanks! It definitely wasn''t obvious from the README.CD, though I should have noticed the Makefile in the xeno-clone directory. I was basically messing around in the xenolinux-2.4.22 subdirectory, like it was a "regular" kernel build. Now, I''m back at the original problem, which is the same as for the image.gz and xenoimage.gz from the 1.0 CD image: - I''m able to boot image.gz via Grub (I renamed image.gz to xenimage.gz): root (hd0,0) kernel /xenimage.gz dom0_mem=512000 ser_baud=115000 noht module /xenolinux.gz root=/dev/sda2 ro console=/dev/tty0 DOMID=0 - It looks like the kernel loads successfully, including SCSI & network drivers - There is a message about "guest OS address is 0x0c000000" then my system reboots. Unfortunately, I cannot catch the exact message because it''s gone too quickly. I''m thinking either something is missing (like an initrd) or it''s looking for programs in /bin or /usr/bin that were not installed (the "make install" in xeno-clone makes some binaries, but just leaves them in the install/bin directories). I''m motivated to try to solve this problem, so I don''t need to buy Virtuozzo ;-) So, I hope you are able to provide some guidance. I''ll try to help you tune the documentation, at the same time.... My application, BTW, is Grid services deployment (I''m co-chair of one of the working groups at http://www.gridforum.org). -- Greg> > I already tried to boot from: > > xeno-unstable.bk/xeno-clone/xenolinux-2.4.22/vmlinux (or vmlinux.gz), > > but from the "make bzImage" it seems that the image.gz is simply the > > vmlinux with the multiboot bytes prepended. Neither will work > > via the "grub> kernel /FILENAME" command > > vmlinux is not directly bootable or usable -- ignore it. > > -- Keir > > > ------------------------------------------------------- > This SF.net email is sponsored by: SF.net Giveback Program. > Does SourceForge.net help you be more productive? Does it > help you create better code? SHARE THE LOVE, and help us help > YOU! Click Here: http://sourceforge.net/donate/ > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/xen-devel------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> Now, I''m back at the original problem, which is the same as > for the image.gz and xenoimage.gz from the 1.0 CD image: > > - I''m able to boot image.gz via Grub (I renamed image.gz to xenimage.gz): > root (hd0,0) > kernel /xenimage.gz dom0_mem=512000 ser_baud=115000 noht > module /xenolinux.gz root=/dev/sda2 ro console=/dev/tty0 DOMID=0 > - There is a message about "guest OS address is 0x0c000000" > then my system reboots. Unfortunately, I cannot catch the > exact message because it''s gone too quickly.Please can you add ''noreboot'' to the xen (kernel) command line so that you''ll be able to see the last lines of debug output. Alternatively, perhaps you could connect a serial console? Just to make sure that what you have as the ''module'' line is actually a xen guest image, please can you do: "zcat xenolinux.gz | strings | head -1" You should get the string "XenoGues". How much memory has the machine got? You''re configuring domain0 with 512MB, and Xen will have taken some for itself, so you should have at least 576MB.> So, I hope you are able to provide some guidance. I''ll try to > help you tune the documentation, at the same time....Thanks. Our current documentation is very poor. There''s a lot of good information that''s been posted in the xen-devel mailing list, but we need to pull it into a FAQ.> My application, BTW, is Grid services deployment (I''m co-chair > of one of the working groups at http://www.gridforum.org).We''re keen to see take-up of Xen by the GRID community... Ian ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Tue, Nov 04, 2003 at 03:55:19AM +0000, Ian Pratt wrote:> > Now, I''m back at the original problem, which is the same as > > for the image.gz and xenoimage.gz from the 1.0 CD image: > > > > - I''m able to boot image.gz via Grub (I renamed image.gz to xenimage.gz): > > root (hd0,0) > > kernel /xenimage.gz dom0_mem=512000 ser_baud=115000 noht > > module /xenolinux.gz root=/dev/sda2 ro console=/dev/tty0 DOMID=0 > > - There is a message about "guest OS address is 0x0c000000" > > then my system reboots. Unfortunately, I cannot catch the > > exact message because it''s gone too quickly. > > > Please can you add ''noreboot'' to the xen (kernel) command line so > that you''ll be able to see the last lines of debug > output. Alternatively, perhaps you could connect a serial console?Done. Here''s what I get: For grub with ''noreboot'' it goes through the whole boot process, identifies the SCSI drives and network: ... Device eth0 opened and ready for use DOM0: Guest OS virtual load address is c0000000 Without the "reboot" option, the system would immediately reboot. The full grub commands are: root (hd0,0) kernel /xenimage.gz dom0_mem=512000 ser_baud=115000 noht module /xenolinux.gz root=/dev/sda2 ro console=/dev/tty0 DOMID=0 As mentioned, /dev/sda2 is ReiserFS. Also as mentioned, I do not think any of the binaries built during xeno-clone were copied to /bin or /usr/bin. Maybe these are the missing piece. In case you''ll suggest building an initrd, it looks like no modules were installed (I then tried "make install" by hand in the kernel build directory, but it still didn''t install any modules. Strange, because I think that some were selected in the kernel''s .config): # mkinitrd -k xenimage.gz -i xeno-initrd using "/dev/sda2" as root device (mounted on "/" as "reiserfs") creating initrd "//boot/xeno-initrd" for kernel "//boot/xenimage.gz" () no version "" modules found> Just to make sure that what you have as the ''module'' line is > actually a xen guest image, please can you do: > > "zcat xenolinux.gz | strings | head -1" > > You should get the string "XenoGues".Yes, no problem: zcat /boot/xenolinux.gz | strings | head -1 XenoGues> How much memory has the machine got? You''re configuring domain0 > with 512MB, and Xen will have taken some for itself, so you > should have at least 576MB.It''s got 2GB. The kernel it usually runs is: # uname -a Linux peabody 2.4.18-64GB-SMP #1 SMP Wed Mar 27 13:58:12 UTC 2002 i686 unknown I could not get high memory (either 4GB or 64GB) to compile with the new kernel, but it should be able to utilize 2GB with no problem.> > So, I hope you are able to provide some guidance. I''ll try to > > help you tune the documentation, at the same time.... > > Thanks. Our current documentation is very poor. There''s a lot of > good information that''s been posted in the xen-devel mailing > list, but we need to pull it into a FAQ. > > > My application, BTW, is Grid services deployment (I''m co-chair > > of one of the working groups at http://www.gridforum.org). > > We''re keen to see take-up of Xen by the GRID community...I will help, definitely! If successful, Xen will be a really great platform for developing/testing Grid applications. -- Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> For grub with ''noreboot'' it goes through the whole boot > process, identifies the SCSI drives and network: > > Device eth0 opened and ready for use > DOM0: Guest OS virtual load address is c0000000This is a weird one -- we haven''t seen anyone else with this problem. Things are going wrong very early in the dom0 boot process. Please could you go into xenolinux-2.4.22/kernel/printk.c and "#if 1" the HYPERVISOR_console_write in printk. hopefully this will produce a bit more debug output before the xenolinux explodes.> The full grub commands are: > > root (hd0,0) > kernel /xenimage.gz dom0_mem=512000 ser_baud=115000 noht > module /xenolinux.gz root=/dev/sda2 ro console=/dev/tty0 DOMID=0 > > As mentioned, /dev/sda2 is ReiserFS.This looks sane -- the ReiserFS shouldn''t make any difference. I forget: what happens when you boot a standard default configuration. It will obviously fail to mount the reiserfs rootfs, but does it get that far? Adding a "console=xenocons0" might generate more debug.> Also as mentioned, I do not think any of the binaries built > during xeno-clone were copied to /bin or /usr/bin. Maybe these > are the missing piece.You shouldn''t need the binaries until you begin starting other domains. It''s important that you use the xi_* binaries that were built with the appropriate Xen version.> In case you''ll suggest building an initrd, it looks like > no modules were installed (I then tried "make install" by hand > in the kernel build directory, but it still didn''t install any > modules. Strange, because I think that some were selected > in the kernel''s .config):You shouldn''t need an initrd.> I could not get high memory (either 4GB or 64GB) to compile with the > new kernel, but it should be able to utilize 2GB with no problem.We haven''t currently got either Highmem (4GB) or PAE36 (64GB) support. Adding the former should be easy. The latter requires changes to Xen as well, and is a fair amount of work. We haven''t any machines with >4GB, so it hasn''t been a priority. Best, Ian ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Tue, Nov 04, 2003 at 11:02:25PM +0000, Ian Pratt wrote:> > > For grub with ''noreboot'' it goes through the whole boot > > process, identifies the SCSI drives and network: > > > > Device eth0 opened and ready for use > > DOM0: Guest OS virtual load address is c0000000 > > This is a weird one -- we haven''t seen anyone else with this problem. > > Things are going wrong very early in the dom0 boot process.If it would help, I can try to capture the earlier part of the process via serial console....but it''s a bit of a hassle.> Please could you go into xenolinux-2.4.22/kernel/printk.c and > "#if 1" the HYPERVISOR_console_write in printk. hopefully this > will produce a bit more debug output before the xenolinux > explodes.I will do this, but won''t get to test it until tomorrow. More:> > The full grub commands are: > > > > root (hd0,0) > > kernel /xenimage.gz dom0_mem=512000 ser_baud=115000 noht > > module /xenolinux.gz root=/dev/sda2 ro console=/dev/tty0 DOMID=0 > > > > As mentioned, /dev/sda2 is ReiserFS. > > This looks sane -- the ReiserFS shouldn''t make any difference. I > forget: what happens when you boot a standard default > configuration. It will obviously fail to mount the reiserfs > rootfs, but does it get that far?When I use the image.gz and xenolinux.gz from the CD (copied to my /boot), the exact same outcome happens -- a reboot after "DOM0: Guest OS virtual load address is c0000000" However, I *can* boot directly from the CD with no problem. I didn''t experiment much, and never tried to get networking running, but it seemed like all the hardware was recognized.> Adding a "console=xenocons0" might generate more debug.Ok.> > Also as mentioned, I do not think any of the binaries built > > during xeno-clone were copied to /bin or /usr/bin. Maybe these > > are the missing piece. > > You shouldn''t need the binaries until you begin starting other > domains. It''s important that you use the xi_* binaries that were > built with the appropriate Xen version.Good, that eliminates some possibilities. Thanks again! -- Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> > Things are going wrong very early in the dom0 boot process. > > If it would help, I can try to capture the earlier > part of the process via serial console....but it''s a bit > of a hassle.I don''t think this is going to help -- Xen seems happy, its dom0 that''s exploding.> > Please could you go into xenolinux-2.4.22/kernel/printk.c and > > "#if 1" the HYPERVISOR_console_write in printk. hopefully this > > will produce a bit more debug output before the xenolinux > > explodes. > > I will do this, but won''t get to test it until tomorrow. More:Thanks. I think this is going to be revealing.> When I use the image.gz and xenolinux.gz from the CD > (copied to my /boot), the exact same outcome happens > -- a reboot after "DOM0: Guest OS virtual load address is c0000000" > > However, I *can* boot directly from the CD with no problem. I > didn''t experiment much, and never tried to get networking > running, but it seemed like all the hardware was recognized.Hmm, this is important. Please could you try booting with dom0_mem=100000 I wander if there''s something odd about the memory layout of your machine. What version of GRUB are you using? Could you install the one from the CD? I''m wandering whether it''s failing to load the module into memory correctly. Thanks, Ian ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> > > For grub with ''noreboot'' it goes through the whole boot > > process, identifies the SCSI drives and network: > > > > Device eth0 opened and ready for use > > DOM0: Guest OS virtual load address is c0000000 > > This is a weird one -- we haven''t seen anyone else with this problem. > > Things are going wrong very early in the dom0 boot process. > > Please could you go into xenolinux-2.4.22/kernel/printk.c and > "#if 1" the HYPERVISOR_console_write in printk. hopefully this > will produce a bit more debug output before the xenolinux > explodes.Do the following: 1. Change ''console=tty0'' to ''console=xencons0'' 2. ''#if 1'' the HYPERVISOR_console_write line in xenolinux/kernel/printk.c 3. Edit xen/common/kernel.c:putchar_console() to remove the initial if ( !opt_console ) return; You should then get useful debug info! -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> > When I use the image.gz and xenolinux.gz from the CD > > (copied to my /boot), the exact same outcome happens > > -- a reboot after "DOM0: Guest OS virtual load address is c0000000" > > > > However, I *can* boot directly from the CD with no problem. I > > didn''t experiment much, and never tried to get networking > > running, but it seemed like all the hardware was recognized. > > Hmm, this is important. > > Please could you try booting with dom0_mem=100000 > I wander if there''s something odd about the memory layout of > your machine. > > What version of GRUB are you using? Could you install the one > from the CD? I''m wandering whether it''s failing to load the > module into memory correctly.Actually that''s very interesting. Xen currently assumes that physical memory is contiguous (ie. there are no memory holes above 1MB). This is because it was hassle to implement the proper checking that is required, and I wouldn''t be able to bug-test it because I have no access to a machine with a memory hole. Also, I suspect that our Xenoboot CD (our flexible alternative to GRUB) would need upgrading to supply the required ''e820 memory-areas list''. Can you send us the output from ''dmesg'' for a normal Linux (not Xenolinux!) running on the problem machine? Actually all we need are the lines: BIOS-provided physical RAM map: BIOS-e820: ... BIOS-e820: ... ... Cheers, Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Wed, Nov 05, 2003 at 09:17:41AM +0000, Keir Fraser wrote:> > > When I use the image.gz and xenolinux.gz from the CD > > > (copied to my /boot), the exact same outcome happens > > > -- a reboot after "DOM0: Guest OS virtual load address is c0000000" > > > > > > However, I *can* boot directly from the CD with no problem. I > > > didn''t experiment much, and never tried to get networking > > > running, but it seemed like all the hardware was recognized. > > > > Hmm, this is important. > > > > Please could you try booting with dom0_mem=100000 > > I wander if there''s something odd about the memory layout of > > your machine. > > > > What version of GRUB are you using? Could you install the one > > from the CD? I''m wandering whether it''s failing to load the > > module into memory correctly. > > Actually that''s very interesting. Xen currently assumes that physical > memory is contiguous (ie. there are no memory holes above 1MB).I don''t think this is the problem, but supplied some details below.> This is because it was hassle to implement the proper checking that is > required, and I wouldn''t be able to bug-test it because I have no > access to a machine with a memory hole. Also, I suspect that our > Xenoboot CD (our flexible alternative to GRUB) would need upgrading to > supply the required ''e820 memory-areas list''. > > Can you send us the output from ''dmesg'' for a normal Linux (not > Xenolinux!) running on the problem machine? Actually all we need are > the lines: > BIOS-provided physical RAM map: > BIOS-e820: ... > BIOS-e820: ... > ...That line didn''t appear in my dmesg output. Here is /proc/meminfo, running 2.4.18-64GB-SMP: total: used: free: shared: buffers: cached: Mem: 2110263296 1106849792 1003413504 0 61300736 242073600 Swap: 3800055808 0 3800055808 MemTotal: 2060804 kB MemFree: 979896 kB MemShared: 0 kB Buffers: 59864 kB Cached: 236400 kB SwapCached: 0 kB Active: 84292 kB Inactive: 218564 kB HighTotal: 1179100 kB HighFree: 923992 kB LowTotal: 881704 kB LowFree: 55904 kB SwapTotal: 3710992 kB SwapFree: 3710992 kB Physically, this is 4 RIMMs, 512MB each. I am using the grub from your CD (recompiled from source). More soon... -- Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
Just a quick note: I noticed that a top-level "make clean" in the xeno-clone directory does not remove xenolinux.gz, and a top-level "make" does not rebuild xenolinux.gz (destination is the xeno-clone/install/boot directory) Since I cannot find it documented elsewhere, here is the procedure I''m using to build from scratch. 1. mkdir xen-new ; cd xen-new (or whatever...) 2. Grab source: bk clone bk://xen.bkbits.net/xeno-1.1.bk 3. Grab pristine kernel, put it with links so it will be found under xen-new and the newly-created xeno-unstable.bk: linux-2.4.22.tar.gz 4. cd xeno-1.1.bk 5. Build everything, merging the Xen source w/ the pristine kernel: tools/misc/xen-clone 6. The xenolinux file was built: cp xeno-clone/install/boot/xenolinux /boot/xenolinux.gz 7. But there was no image.gz as expected in xeno-clone/install/boot/, it was here: xeno-clone/xenolinux-2.4.22/arch/xeno/boot/image.gz and not bootable. (Kier wrote: "You should then find a Xen image.gz file in xeno-clone/install/boot. This same file also resides at xeno-clone/xeno-1.0.bk/xen/image.gz.") There is no other "image" or "image.gz" anywhere. 8. I went to the kernel directory, and built the kernel from scratch: cd xeno-clone/xenolinux-2.4.22 (edited .config to add ReiserFS support) make dep make make modules (as root) make modules_install (but there were none) make bzImage 9. Then, back up to the xeno-clone directory (cd ../..) make install 10. NOW, there was image.gz in xen-new/install/boot/image.gz also the same in xen-new/xeno-1.1.bk/xen/image.gz (and image) which was bootable -- Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Wed, Nov 05, 2003 at 08:35:51AM +0000, Keir Fraser wrote:> > > > > For grub with ''noreboot'' it goes through the whole boot > > > process, identifies the SCSI drives and network: > > > > > > Device eth0 opened and ready for use > > > DOM0: Guest OS virtual load address is c0000000 > > > > This is a weird one -- we haven''t seen anyone else with this problem. > > > > Things are going wrong very early in the dom0 boot process. > > > > Please could you go into xenolinux-2.4.22/kernel/printk.c and > > "#if 1" the HYPERVISOR_console_write in printk. hopefully this > > will produce a bit more debug output before the xenolinux > > explodes.Ok....you saw the details of how I re-built in my prior message in this thread. I added: 8. I went to the kernel directory, and built the kernel from scratch: cd xeno-clone/xenolinux-2.4.22 (edited .config to add ReiserFS support) Did the following:> Do the following: > 1. Change ''console=tty0'' to ''console=xencons0'' > 2. ''#if 1'' the HYPERVISOR_console_write line in > xenolinux/kernel/printk.c > 3. Edit xen/common/kernel.c:putchar_console() to remove the initial > if ( !opt_console ) return;then: make dep make make modules (as root) make modules_install (but there were none) make bzImage cd ../.. (to xeno-clone) make install then copied in the new image.gz (xenolinux.gz was unchanged) My grub boot (same as before, but with console=xencons0) grub> root (hd0,0) grub> kernel /xenimage.gz dom0_mem=100000 ser_baud=115200 noht noreboot [output saying "yep, looks like a Multiboot ELF"] grub> module /xenolinux.gz root=/dev/sda2 ro console=xencons0 DOMID=0 [output saying, "yep, it''s a module"] The result was the exact same boot I had before -- no extra debugging. The final lines before death: Device eth0 opened and ready for use DOM0: Guest OS virtual load address is c0000000 Other ideas are welcome! -- Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> Just a quick note: I noticed that a top-level "make clean" in the > xeno-clone directory does not remove xenolinux.gz, and a top-level > "make" does not rebuild xenolinux.gz > > (destination is the xeno-clone/install/boot directory) > > Since I cannot find it documented elsewhere, here is the > procedure I''m using to build from scratch.<snipped> Try the following instead: 1. mkdir xen-new ; cd xen-new 2. bk clone bk://xen.bkbits/net/xeno-1.1.bk 3. ./xeno-1.1.bk/tools/misc/xen-clone bk://xen.bkbits/net/xeno-1.1.bk This should work better --- without explicitly telling the clone script which repository to use it defaults to xeno-1.0.bk. You may have found it therefore went and downloaded xeno-1.0.bk and built that! This might also explain any boot problems you''ve had with Xenolinux --- the clone script may have built a 1.0 Xenolinux rather than a 1.1 Xenolinux! -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> 5. Build everything, merging the Xen source w/ the pristine kernel: > tools/misc/xen-cloneThe xen-clone script isn''t designed to be used like this -- it''s designed for pulling and building a repository from scratch rather than rebuilding an existing one. It will actually do what you want if you run it If you''ve got a repository as created by xen-clone, you can rebuild and install xen and the various tools just by cd''ing into the xeno.XX.bk directory and doing a ''make install'' To rebuild xenolinux, cd into xenolinux-2.4.22 and ''make dep modules bzImage install''. This will build xenolinux-2.4.22/arch/xeno/boot/image.gz (which the install target copies to install/boot)> 6. The xenolinux file was built: > cp xeno-clone/install/boot/xenolinux /boot/xenolinux.gz > > 7. But there was no image.gz as expected in xeno-clone/install/boot/, > it was here: > xeno-clone/xenolinux-2.4.22/arch/xeno/boot/image.gz > and not bootable.This is actually what becomes xenolinux.gz, not the xen image.gz which should be in xeno.XX.bk/xen/image.gz (and will have been copied in to install/boot) Ian ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> Ok....you saw the details of how I re-built in my prior message in > this thread.I suspect that running xen-clone in the directory where you did will actually have just pulled and rebuilt a brand new repository rather than rebuilding the one you''d edited :-(> My grub boot (same as before, but with console=xencons0) > > grub> root (hd0,0) > grub> kernel /xenimage.gz dom0_mem=100000 ser_baud=115200 noht noreboot > > [output saying "yep, looks like a Multiboot ELF"] > > grub> module /xenolinux.gz root=/dev/sda2 ro console=xencons0 DOMID=0 > > [output saying, "yep, it''s a module"] > > The result was the exact same boot I had before -- no extra debugging. > The final lines before death: > > Device eth0 opened and ready for use > DOM0: Guest OS virtual load address is c0000000So to summarise: If you do the exact same sequence of grub steps with the xen and xenolinux images from the CD it works fine, but fails with the ones you build yourself. Please could you pull and try these two images from the unstable tree that I''ve just built for you: http://www.cl.cam.ac.uk/users/iap10/temp/xenolinux.gz http://www.cl.cam.ac.uk/users/iap10/temp/image.gz If these fail, please can you do as Keir suggests and boot a standard Linux kernel and send us the boot output, in particular anything relating to e820 messages. Thanks, Ian ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> > Can you send us the output from ''dmesg'' for a normal Linux (not > > Xenolinux!) running on the problem machine? Actually all we need are > > the lines: > > BIOS-provided physical RAM map: > > BIOS-e820: ... > > BIOS-e820: ... > > ... > > That line didn''t appear in my dmesg output. Here is /proc/meminfo, > running 2.4.18-64GB-SMP:If those lines don''t appear in your dmesg output then your memory does not contain any holes, so that is not the problem. I suspect that you may have tried to boot a Xenolinux 1.0 image on a Xen 1.1 or Xen 1.2 system. Unfortunately there is no versioning on the guest-OS interface at the moment (I really ought to add some as it''s rather a waste of people''s time if they make this mistake). As Ian says: try running xeno-clone ''from scratch'', getting it to pull and build all images for you. By default it will pull and build Xen 1.0. If you want 1.1 or the unstable 1.2 trees then try: xen-clone bk://xen.bkbits.net/xeno-1.1.bk or xen-clone bk://xen.bkbits.net/xeno-unstable.bk -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Thu, Nov 06, 2003 at 07:45:28AM +0000, Keir Fraser wrote:> > Just a quick note: I noticed that a top-level "make clean" in the > > xeno-clone directory does not remove xenolinux.gz, and a top-level > > "make" does not rebuild xenolinux.gz > > > > (destination is the xeno-clone/install/boot directory) > > > > Since I cannot find it documented elsewhere, here is the > > procedure I''m using to build from scratch. > > <snipped> > > Try the following instead: > > 1. mkdir xen-new ; cd xen-new > > 2. bk clone bk://xen.bkbits/net/xeno-1.1.bk > > 3. ./xeno-1.1.bk/tools/misc/xen-clone bk://xen.bkbits/net/xeno-1.1.bkOk, I''ll try this. Two important questions: 1. At what point can I customize the kernel, (in my case, I need to add ReiserFS support)? Should I run "make dep" etc. afterwards, or if not what command comes after I edit the kernel''s .config? 2. Do I need to be root for anything other than copying in the images (which I''ve been doing with regular "cp") to /boot? So far, I''ve not been running as root until I need it. Thanks!! Greg> This should work better --- without explicitly telling the clone > script which repository to use it defaults to xeno-1.0.bk. You may > have found it therefore went and downloaded xeno-1.0.bk and built > that! > > This might also explain any boot problems you''ve had with Xenolinux > --- the clone script may have built a 1.0 Xenolinux rather than a 1.1 > Xenolinux! > > -- Keir------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> 1. At what point can I customize the kernel, > (in my case, I need to add ReiserFS support)? Should I > run "make dep" etc. afterwards, or if not what command > comes after I edit the kernel''s .config?After xen-clone finishes, try something like: cd xen-new/xenolinux-2.4.22 make xconfig make clean make dep make bzImage cp arch/xeno/boot/image.gz ../install/boot/xenolinux.gz Beware of adding non-default "character device" drivers. Apart from that, most other options oughgt to work but may not have been tested.> 2. Do I need to be root for anything other than copying > in the images (which I''ve been doing with regular > "cp") to /boot? So far, I''ve not been running as > root until I need it.No. -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> Please could you pull and try these two images from the unstable > tree that I''ve just built for you: > > http://www.cl.cam.ac.uk/users/iap10/temp/xenolinux.gz > http://www.cl.cam.ac.uk/users/iap10/temp/image.gzI''ve just updated the prebuilt xenolinux image at this location to include ReiserFS support. [ However, there was a trivial build problem due to a missing include in the asm-xeno/page.h -- did you see this when adding reiserfs to the config? ] If these images work, it''s definitely a build problem. They boot on my system, but I haven''t been able to check whether reiserfs works, though I''d expect it to. Best, Ian ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Thu, Nov 06, 2003 at 06:38:21PM +0000, Ian Pratt wrote:> > > Please could you pull and try these two images from the unstable > > tree that I''ve just built for you: > > > > http://www.cl.cam.ac.uk/users/iap10/temp/xenolinux.gz > > http://www.cl.cam.ac.uk/users/iap10/temp/image.gz > > I''ve just updated the prebuilt xenolinux image at this location > to include ReiserFS support. > > [ However, there was a trivial build problem due to a missing > include in the asm-xeno/page.h -- did you see this when adding > reiserfs to the config? ](I don''t think I saw this error)> If these images work, it''s definitely a build problem. They boot > on my system, but I haven''t been able to check whether reiserfs > works, though I''d expect it to.Ok - this boots! Yay! Yes, this could be something with my build procedures, but remember that the first time I tried it was with the kernel & xeno images copied from the CD (they bombed before the domain 0 was started). Here is the current status with domain 0 filesystem troubles: - Only the first filesystems mount (note that the ReiserFS on / mounted fine): Filesystem 1k-blocks Used Available Use% Mounted on /dev/sda2 16016308 9988792 6027516 63% / /dev/sda1 46636 18682 25546 43% /boot shmfs 48400 0 48400 0% /dev/shm - But these did not mount, and I don''t know why (my other FS'' are commented out in /etc/fstab): /dev/sdb1 /tmp reiserfs defaults 0 0 /dev/sdc2 /globus reiserfs defaults 0 0 /dev/sdh1 /home reiserfs defaults 0 0 /dev/sdm1 /data/web02files reiserfs defaults 0 0 They did not mount at boot time, and the errors went by too fast to see (they''re not in dmesg or the boot logs in /var/log). When I try to mount a reiserfs filesystem (i.e., "mount /tmp") from the running system, the mount just hangs in an IO wait and never comes back (I can''t even kill -9). - Here''s an ext2: /dev/sdd1 /mnt/d ext2 defaults 0 1 When I try to mount after the system is running: # mount /mnt/d mount: wrong fs type, bad option, bad superblock on /dev/sdd1, or too many mounted file systems (could this be the IDE device where you in fact use ide-scsi so that sr0 or sda or so is needed?) When I tried to boot with /mnt/d in /etc/fstab, I got an error about fsck needing to run (which is false), then supposedly got dumped into single-user mode to run fsck, but in fact the system hung...no prompt, hard reboot needed. Yes, these are all SCSI drives, and were configured correctly (right speed, bus, Mhz...) when the kernel initialized. I will try to add a domain, but wanted to write here about the filesystems. (Also, I now need to reboot to the old kernel, to get the filesystems back with the domain tools for 1.1). If it would help, I could put a different distribution on this system. It''s SuSE 8.1, which is generally good but doesn''t even include ext3. RH9 or Debian would probably do fine. Thanks again for working on this, and I hope all my analysis is helpful! -- Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> - But these did not mount, and I don''t know why (my other FS'' are > commented out in /etc/fstab): > > /dev/sdb1 /tmp reiserfs defaults 0 0 > /dev/sdc2 /globus reiserfs defaults 0 0 > /dev/sdh1 /home reiserfs defaults 0 0 > /dev/sdm1 /data/web02files reiserfs defaults 0 0Are these all attached to the same SCSI card as sda? One possibility is that you''re probably entering new territory by trying to access any SCSI disc other than sda. It''s possible that something in Xen''s SCSI code is broken -- the non-sda case may never have been tested. In this case what''s really needed is serial line output, and perhaps some extra tracing in Xen. I may test this out myself tomorrow -- we have machines that can handle multiple SCSI discs, but I think we''ve only ever booted with one connected (or, at least, we''ve only ever mounted from partitions on the first SCSI disc). -- Keir ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Thu, Nov 06, 2003 at 10:04:34PM +0000, Keir Fraser wrote:> > - But these did not mount, and I don''t know why (my other FS'' are > > commented out in /etc/fstab): > > > > /dev/sdb1 /tmp reiserfs defaults 0 0 > > /dev/sdc2 /globus reiserfs defaults 0 0 > > /dev/sdh1 /home reiserfs defaults 0 0 > > /dev/sdm1 /data/web02files reiserfs defaults 0 0 > > Are these all attached to the same SCSI card as sda?Card, yes. Bus, no. It''s an Adaptec 39160 which is Ultra 160 SCSI with dual channels. /dev/sda through /dev/sdf (I think) are on Channel A. The rest are on channel B.> One possibility is that you''re probably entering new territory by > trying to access any SCSI disc other than sda. It''s possible that > something in Xen''s SCSI code is broken -- the non-sda case may never > have been tested. > > In this case what''s really needed is serial line output, and perhaps > some extra tracing in Xen.I will work on serial output, but maybe not until tomorrow since I don''t think I have a null modem here at school.> I may test this out myself tomorrow -- we have machines that can > handle multiple SCSI discs, but I think we''ve only ever booted with > one connected (or, at least, we''ve only ever mounted from partitions > on the first SCSI disc).Both partitions from /dev/sda mounted, and none from the other devices (regardless of where they are on the SCSI bus). So, this could be the common feature. -- Greg ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Thu, Nov 06, 2003 at 06:38:21PM +0000, Ian Pratt wrote:> > > Please could you pull and try these two images from the unstable > > tree that I''ve just built for you: > > > > http://www.cl.cam.ac.uk/users/iap10/temp/xenolinux.gz > > http://www.cl.cam.ac.uk/users/iap10/temp/image.gz > > I''ve just updated the prebuilt xenolinux image at this location > to include ReiserFS support. > > [ However, there was a trivial build problem due to a missing > include in the asm-xeno/page.h -- did you see this when adding > reiserfs to the config? ]I''m building correctly now, and saw the build error. I think I''ve been staring at this for too long, because I can''t figure out what to fix. It''s complaining in fs/reiserfs/hashes.c , but adding #include <asm/hypervisor-ifs/hyperfisor-if.h> doesn''t fix the error. -- Greg> If these images work, it''s definitely a build problem. They boot > on my system, but I haven''t been able to check whether reiserfs > works, though I''d expect it to. > > Best, > Ian------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> I''m building correctly now, and saw the build error. > > I think I''ve been staring at this for too long, because I > can''t figure out what to fix. It''s complaining in > fs/reiserfs/hashes.c , but adding #include <asm/hypervisor-ifs/hyperfisor-if.h> > doesn''t fix the error.I fixed this by adding #include <asm/hypervisor-ifs/hyperfisor-if.h> to the top of include/asm-xeno/page.h, just below the #include <linux/config.h> It builds OK for me. I''ve pushed the change to bkbits.net xeno-unstable.bk The failing to work with SCSI drives other than sda is not totally surprising. I''m pretty sure it used to work, but I bet it hasn''t been tested in months. It''ll be an easy fix... Do you happen to notice whether Xen spots the second drive as it boots? When xenolinux boots, do you get a "XenoLinux Virtual Block Device Driver installed [device: 0810]" message as well as [device: 0800] ? Ian ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> It''s an Adaptec 39160 which is Ultra 160 SCSI with > dual channels. /dev/sda through /dev/sdf (I think) are on > Channel A. The rest are on channel B. > > > One possibility is that you''re probably entering new territory by > > trying to access any SCSI disc other than sda. It''s possible that > > something in Xen''s SCSI code is broken -- the non-sda case may never > > have been tested.Indeed -- this was the case. I''ve just checked in a fix which should hopefully allow you to access all your disks; let me know if there are any problems. S. ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On Fri, Nov 07, 2003 at 05:08:21PM +0000, Steven Hand wrote:> > > It''s an Adaptec 39160 which is Ultra 160 SCSI with > > dual channels. /dev/sda through /dev/sdf (I think) are on > > Channel A. The rest are on channel B. > > > > > One possibility is that you''re probably entering new territory by > > > trying to access any SCSI disc other than sda. It''s possible that > > > something in Xen''s SCSI code is broken -- the non-sda case may never > > > have been tested. > > Indeed -- this was the case. > > I''ve just checked in a fix which should hopefully allow you to access > all your disks; let me know if there are any problems.Yes, this worked, thanks! I''ll describe my build process, for the benefit of others. In my next message, I''ll describe the next stage of uncertainty.... Also, below, is a memory limitation I encountered. 1. mkdir xen-new ; cd xen-new 2. bk clone bk://xen.bkbits.net/xeno-unstable.bk 3. ./xeno-unstable.bk/tools/misc/xen-clone bk://xen.bkbits.net/xeno-unstable.bk I noted that ReiserFS was not in this kernel, so: 4. cd xeno-clone/xenolinux-2.4.22 5. edited .config to add ReiserFS support (could "make menuconfig") 6. make dep; make bzImage 7. cp arch/xeno/boot/image.gz ../install/boot/xenolinux.gz Then, copied the images to /boot: 8. cd ../install/boot 9. (as root) cp image.gz /boot/ ; cp xenolinux.gz /boot/ My (working) /boot/grub/menu.lst is: title xen root (hd0,0) kernel /image.gz dom0_mem=768000 ser_baud=115200 noht module /xenolinux.gz root=/dev/sda2 ro console=xencons0 DOMID=0 This works fine with a 768MB memory limit. However, when I tried it with "dom0_mem=1024000" I got an immediate reboot after the kernel loaded. As mentioned earlier, this machine has 2GB of physical RAM, which seems to be mapped contiguously. After re-adding items to my fstab, "df -k" is: Filesystem 1k-blocks Used Available Use% Mounted on /dev/sda2 16016308 9989552 6026756 63% / /dev/sda1 46636 20740 23488 47% /boot /dev/sdb1 16064472 32840 16031632 1% /tmp /dev/sdc2 71647708 32840 71614868 1% /globus /dev/sdh1 71679804 40623744 31056060 57% /home /dev/sdm1 71679804 32840 71646964 1% /data/web02files shmfs 378152 0 378152 0% /dev/shm /dev/sdd1 70557052 20 66972936 1% /mnt/d /dev/sde1 71671804 732140 70939664 2% /htdocs /dev/sdg1 71679804 32840 71646964 1% /mnt/g /dev/sdi2 71647708 32840 71614868 1% /mnt/i /dev/sdj1 71671804 32840 71638964 1% /mnt/j /dev/sdk1 71679804 600724 71079080 1% /mnt/k /dev/sdl1 71679804 10808300 60871504 16% /mnt/l /dev/sdn1 71667120 8 68083416 1% /mnt/n Thanks again for this help! On to virtual domains, next... -- Greg ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> Yes, this worked, thanks!Phew! It''s been a hard slog...> 1. mkdir xen-new ; cd xen-new > 2. bk clone bk://xen.bkbits.net/xeno-unstable.bk > 3. ./xeno-unstable.bk/tools/misc/xen-clone bk://xen.bkbits.net/xeno-unstable.bkWhat you''ve done here is actually a bit odd and likely to lead to confusion: the xen-clone is going to pull another copy of the repository you''re going to end up with two copies. I''d advise just taking a local copy of the xen-clone script (or downloading it from the website) rather than cloning the repository just to get it (it doesn''t change frequently). tip: you can then edit it to create your own site-specific section such that the install/boot directory is a link to /boot, hence ''make install'' will just work without you having to do the copy manually.> I noted that ReiserFS was not in this kernel, so: > > 4. cd xeno-clone/xenolinux-2.4.22 > 5. edited .config to add ReiserFS support (could "make menuconfig") > 6. make dep; make bzImage > 7. cp arch/xeno/boot/image.gz ../install/boot/xenolinux.gzYou can do this by using the ''make install'' target. If you wanted to rebuild xen and the xi_* tools, just cd into ''xeno.unstable.bk'' and do ''make install''. [memo to self: perhaps we should create a "tools/misc/xen-build-all" script to do both of the above?]> Then, copied the images to /boot: > 8. cd ../install/boot > 9. (as root) cp image.gz /boot/ ; cp xenolinux.gz /boot/ > > My (working) /boot/grub/menu.lst is: > > title xen > root (hd0,0) > kernel /image.gz dom0_mem=768000 ser_baud=115200 noht > module /xenolinux.gz root=/dev/sda2 ro console=xencons0 DOMID=0Fine.> This works fine with a 768MB memory limit. However, when > I tried it with "dom0_mem=1024000" I got an immediate reboot > after the kernel loaded. As mentioned earlier, this machine > has 2GB of physical RAM, which seems to be mapped contiguously.The current xenolinux config does not support CONFIG_HIGHMEM4G, hence you''re limited to ~800MB (the kernel virtual address space contains a linear mapping of _all_ physical memory, hence the limit). Adding CONFIG_HIGHMEM4G support (which will give you up to 4GB of RAM per guest) should be straightforward but we haven''t done this as no-one has needed it before. You GRID guys have too much money ;-)) Adding support for CONFIG_HIGHMEM64G would be more of a pain, as we''d have to implement PAE36 support in Xen. Rather than trying to start one really big guest, can you start a couple of 800MB ones? It''s unfortunate that domain0 exits rather than just mapping all of the memory it can handle (and ideally giving the rest back). We can fix this. Adding CONFIG_HIGHMEM4G should be on the todo list. Making it work should mainly be a case of putting code that''s in arch/i386 back into arch/xeno (it was removed in the early days for simplicity, and can now safely go back).> Thanks again for this help! On to virtual domains, next...It''s good to have people pushing the operational envelope... Ian ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel