From: Nikolaus H. <nha...@sc...> - 2003-12-25 16:18:37
|
Hi, Sorry that I write the third mail this week. The former problems are "solved". I've now compiled and installed Kernel 2.6.0 on my host system (the stable one). Now, my virtual machine does not work anymore that had worked on my 2.4.20 kernel! I get this kernel messages: tracing thread pid = 6798 Linux version 2.4.20-8um (niki@nobaq) (gcc version 2.95.4 20011002 (Debian prerelease)) #1 Mit Dez 24 17:27:39 CET 2003 On node 0 totalpages: 8192 zone(0): 8192 pages. zone(1): 0 pages. zone(2): 0 pages. Kernel command line: root=/dev/ubd0 Calibrating delay loop... 307.05 BogoMIPS Memory: 28996k available Dentry cache hash table entries: 4096 (order: 3, 32768 bytes) Inode cache hash table entries: 2048 (order: 2, 16384 bytes) Mount-cache hash table entries: 512 (order: 0, 4096 bytes) Buffer-cache hash table entries: 1024 (order: 0, 4096 bytes) Page-cache hash table entries: 8192 (order: 3, 32768 bytes) Checking for host processor cmov support...Yes Checking for host processor xmm support...No Checking that ptrace can change system call numbers...OK Checking that host ptys support output SIGIO...Yes Checking that host ptys support SIGIO on close...No, enabling workaround POSIX conformance testing by UNIFIX Linux NET4.0 for Linux 2.4 Based upon Swansea University Computer Society NET3.039 Initializing RT netlink socket Starting kswapd VFS: Diskquotas version dquot_6.4.0 initialized Journalled Block Device driver loaded devfs: v1.12c (20020818) Richard Gooch (rg...@at...) devfs: boot_options: 0x1 pty: 256 Unix98 ptys configured RAMDISK driver initialized: 16 RAM disks of 4096K size 1024 blocksize loop: loaded (max 8 devices) Universal TUN/TAP device driver 1.5 (C)1999-2002 Maxim Krasnyansky Initializing software serial port version 1 mconsole (version 2) initialized on /home/niki/.uml/0GDXOz/mconsole Partition check: ubda: unknown partition table Initializing stdio console driver NET4: Linux TCP/IP 1.0 for NET4.0 IP Protocols: ICMP, UDP, TCP IP: routing cache hash table of 512 buckets, 4Kbytes TCP: Hash tables configured (established 2048 bind 2048) NET4: Unix domain sockets 1.0/SMP for Linux NET4.0. VFS: Mounted root (ext2 filesystem) readonly. Mounted devfs on /dev INIT: version 2.84 booting Child 6830 exited with signal 11 So everything seems to be ok, the rootfs mounted, devfs mounted and so on. "Don't panic" - I thought and compiled another UML with the same 2.6.0 kernel sources from my hostfs. It worked without compiling error, but when I call "linux" I got the same error. There is a "solution" on the homepage that tells something with an old filesystem. But I just changed my host kernel. I would very like to not give my new 2.6.0 kernel away. What can I do? I use the UML fs root_fs_toms1.7.205.bz2. Thanks again! Niki |
From: Nikolaus H. <nha...@sc...> - 2003-12-26 16:46:10
|
Hi again, > (...) > VFS: Mounted root (ext2 filesystem) readonly. > Mounted devfs on /dev > INIT: version 2.84 booting > Child 6830 exited with signal 11 I have "additional" information. I have now 3 or 4(!!) versions of an UML linux running and nothing works - seems to be an error on the host. 1.) The original UML which I can download. It produces this error: Kernel panic: Segfault with no mm In idle task - not syncing 2.) My old one (2.4.20 kernel with UML patch) produces this: Child 391 exited with signal 11 3.) A new UML based on the new stable 2.6.0 kernel with patch uml-patch-2.6.0-test9-1.bz2. Produces same error as in 2) 4.) Just a new kernel 2.6.0 source compiled with "make linux ARCH=um". Fails to compile because of a missing header. The strange thing is that when I call ./linux init=/bin/sh then I get a shell. In the internet, I read many descriptions of this error, many of them about a RedHat kernel. But I don't have a Redhat kernel. 1.) Does not work because of the 2G/2G on my system. 2.) and 3.) works fine when I boot with my old 2.4.20 kernel. Everything is ok when I set the init to /bin/sh. I have tried everything, searched the web and i found: NOTHING. Is this a bug or something? I have a PII with 190MB of RAM and the 2.6.0 Kernel. Is there a kernel option for the host I've forgotton to set? Are the binaries in the file system compiled and linked with wrong versions? Does it help to take another file system/to build one for myself. Who has UML with a 2.6.0 Kernel (stable) and wants to give me his ".config" file? Maybe I've really forgotten something! Developers: What *could* this be? There's a description in the FAQ I does not understand exactly. Maybe this is a bug? I'd be very happy if anyone has a tip for me or other things I could try or debug. I do not want to fall back to my 2.4.20 kernel. The problem is, that I have removed my ".config" from this kernel and I do not know how, which settings were set. But I would have no recompile it, because the bridge driver does not work (so I would compile it directly into the kernel). thanks a lot!! Niki |
From: BlaisorBlade <bla...@ya...> - 2003-12-26 19:11:03
|
Alle 17:45, venerd=EC 26 dicembre 2003, Nikolaus Hammler ha scritto: > > Child 6830 exited with signal 11 > 1.) The original UML which I can download. It produces this error: > Kernel panic: Segfault with no mm > In idle task - not syncing You mean the RPM of 2.4.19? That RPM stays there only because nobody delete= d=20 it. There is an enormous number of bug with it. > 2.) My old one (2.4.20 kernel with UML patch) produces this: > Child 391 exited with signal 11 Use a new 2.4; this problem has already been solved in later UML patches. > 3.) A new UML based on the new stable 2.6.0 kernel with > patch uml-patch-2.6.0-test9-1.bz2. > Produces same error as in 2) Go here, and download the appropriate fix. Report here any problem as the=20 author says it "should" work: http://toast.debian.net/~may/umlproject/bugs.html > 4.) Just a new kernel 2.6.0 source compiled with > "make linux ARCH=3Dum". Fails to compile because of a missing header. You mean stock 2.6.0 without the 2.6 patches? Doesn't try to build UML, it= =20 won't work and this is known. > 1.) Does not work because of the 2G/2G on my system. You have 2G/2G on the host? I'm especially interested in testing here, sinc= e=20 such host are not frequent and UML doesn't get a lot of testing in this=20 config. By the way, have you any pointer about the host 2G/2G patch(as I=20 don't find it anywhere)? > Is there a kernel option for the host I've forgotton to set? =46or the host the only option that matters is to enable /proc/mm if you ha= ve=20 the SKAS patch on the host. Nothing else. Bye =2D-=20 cat <<EOSIGN Paolo Giarrusso, aka Blaisorblade Linux Kernel 2.4.21/2.6.0-test on an i686; Linux registered user n. 292729 EOSIGN |
From: Peter <pe...@ri...> - 2003-12-26 22:15:40
|
Hi. I've noticed errors similar to the following on 3 UML instances = running on the same host server: Dec 26 18:12:59 kernel: do_io - write failed err =3D 28 fd =3D 14 Dec 26 18:12:59 kernel: end_request: I/O error, dev 62:70 (User-mode = block device), sector 21520 Dec 26 18:12:59 kernel: do_io - write failed err =3D 28 fd =3D 14 Dec 26 18:12:59 kernel: end_request: I/O error, dev 62:70 (User-mode = block device), sector 21568 Dec 26 18:12:59 kernel: do_io - write failed err =3D 28 fd =3D 14 Dec 26 18:12:59 kernel: end_request: I/O error, dev 62:70 (User-mode = block device), sector 21592 There are another 15 or so UML instances on the host server that do not = have these errors. The host server is a 2.4.21 server with the skas patch. The UML instances having the problems are: Linux 2.4.23-6um #1 Wed Dec 3 19:45:36 UTC 2003 i686 Linux 2.4.21-7um #4 Sat Aug 16 10:04:11 UTC 2003 i686 Linux 2.4.23-6um #1 Wed Dec 3 19:45:36 UTC 2003 i686 Linux 2.4.23-1um (2.4.23-6um IIRC is actually 2.4.23 + the 2.4.22-6um patches) Other, non-broken, UMLs on the host server are running the same kernels. The errors seem to pop up when the daily crontabs run. Probably because = that is when IO is going on. From what I can tell the errors indicate drive failure on regular Linux = servers. However the host server is complaining of no such thing. Anyone have any ideas how to better troubleshoot or resolve this one? = Anyone having similar problems? Regards, Peter |
From: <s-...@rh...> - 2003-12-26 23:18:45
|
On Sat, Dec 27, 2003 at 11:08:22AM +1300, Peter wrote: > Hi. I've noticed errors similar to the following on 3 UML instances running on the same host server: > > Dec 26 18:12:59 kernel: do_io - write failed err = 28 fd = 14 > Dec 26 18:12:59 kernel: end_request: I/O error, dev 62:70 (User-mode block device), sector 21520 You ran out of disk space on the host machine. At least that's what that errno means. -Steve |
From: Nikolaus H. <nha...@sc...> - 2003-12-27 15:39:52
|
Hi, > > > Child 6830 exited with signal 11 > > > > 1.) The original UML which I can download. It produces this error: > > Kernel panic: Segfault with no mm In idle task - not syncing > > You mean the RPM of 2.4.19? That RPM stays there only because > nobody deleted > it. There is an enormous number of bug with it. Hmm, yes, but this was just a try. I've already compiled for myself. > > 2.) My old one (2.4.20 kernel with UML patch) produces > this: Child 391 > > exited with signal 11 > > Use a new 2.4; this problem has already been solved in later > UML patches. Hmm, no I don't think that this is the problem. I think I haven't quoted good enough ;-) so you don't know what I want to say. I have a 2.6.0 as HOST kernel and now the error above comes on a 2.4.0, 2.6.0 AND the RPM UML (with tomfs). But now, it works with the debian fs for some strange reasons. However, I'm sure this is a bug. (Also, with tomfs, it worked with /bin/sh as init). [I've described everything in the last emails]. > > 1.) Does not work because of the 2G/2G on my system. > > You have 2G/2G on the host? I'm especially interested in > testing here, since > such host are not frequent and UML doesn't get a lot of > testing in this > config. By the way, have you any pointer about the host 2G/2G > patch(as I > don't find it anywhere)? I would be interested: On what exactly depends if I have 2G/2G? Is this a kernel option on the host kernel or a hardware issue or something else? The strange thing is, that I don't have to compile with 2G/2G anymore since I have 2.6.0 as host kernel. > > Is there a kernel option for the host I've forgotton to set? > For the host the only option that matters is to enable > /proc/mm if you have > the SKAS patch on the host. Nothing else. Ouh ;-). Now everything is clear... niki -- Nikolaus Hammler SC&C Software, Communication & Consulting GmbH & Co KEG Grottenhofstr. 3, A-8053 Graz Mobile: +43/(0)676 840439-400 Office: +43/(0)316 265-205 Fax: +43/(0)316 265-234 nha...@sc..., http://scc.co.at |
From: BlaisorBlade <bla...@ya...> - 2003-12-27 19:05:55
|
Alle 16:38, sabato 27 dicembre 2003, Nikolaus Hammler ha scritto: > > > 2.) My old one (2.4.20 kernel with UML patch) produces > > > > this: Child 391 > > > > > exited with signal 11 > > > > Use a new 2.4; this problem has already been solved in later > > UML patches. > > Hmm, no I don't think that this is the problem. > I think I haven't quoted good enough ;-) so you don't know > what I want to say. I have a 2.6.0 as HOST kernel and now > the error above comes on a 2.4.0, 2.6.0 AND the RPM UML > (with tomfs). But now, it works with the debian fs > for some strange reasons. However, I'm sure this is a bug. > (Also, with tomfs, it worked with /bin/sh as init). > [I've described everything in the last emails]. I think I've understood correctly. IIRC, if you use 2.6 as host (even 2.6-test), an UML bug appears. That bug is not triggered if you have a 2.4 host; also, it appears randomically (not with /bin/sh, not with the Debian rootfs but could reappear even with them...). The patch has been applied to later 2.4 guest, not to stock 2.6 guest; see the end of this email to find it. > I would be interested: On what exactly depends if I have 2G/2G? > Is this a kernel option on the host kernel or a hardware issue > or something else? It is a kernel option or patch on the host kernel. I can guess that having 2G/2G is needed if you have at least 900 Mega of Ram on the host. > The strange thing is, that I don't have to compile with 2G/2G > anymore since I have 2.6.0 as host kernel. I have no idea about this. Maybe, you enabled an option on the kernel compile and didn't enable it now. Since you didn't apply a 2G/2G patch, I guess you've set CONFIG_HIGHMEM. > > > Is there a kernel option for the host I've forgotton to set? > > > > For the host the only option that matters is to enable > > /proc/mm if you have > > the SKAS patch on the host. Nothing else. > > Ouh ;-). Now everything is clear... The host normally doesn't compile with SKAS if you don't enable /proc/mm. Also, I don't think you have applied SKAS for 2.6. If you need any UML patch for 2.6, I think you should go here: http://web.tiscali.it/no-redirect-tiscali/blaisorblade/index.html There you'll find SKAS for the host, a fix for the bug above(the G-02... one), fix for modules + other stuff. See instructions you find in that page and in the Readme, and note the warning about the A-04 patch. Bye -- cat <<EOSIGN Paolo Giarrusso, aka Blaisorblade Linux Kernel 2.4.21/2.6.0-test on an i686; Linux registered user n. 292729 EOSIGN |
From: Nikolaus H. <nha...@sc...> - 2003-12-28 14:37:49
|
Hi, > Alle 16:38, sabato 27 dicembre 2003, Nikolaus Hammler ha scritto: > > > > > 2.) My old one (2.4.20 kernel with UML patch) produces > > > > > > this: Child 391 > > > > > > > exited with signal 11 > > > > > > Use a new 2.4; this problem has already been solved in later UML > > > patches. > > > > Hmm, no I don't think that this is the problem. > > I think I haven't quoted good enough ;-) so you don't know > what I want > > to say. I have a 2.6.0 as HOST kernel and now the error > above comes on > > a 2.4.0, 2.6.0 AND the RPM UML (with tomfs). But now, it works with > > the debian fs for some strange reasons. However, I'm sure this is a > > bug. (Also, with tomfs, it worked with /bin/sh as init). > > [I've described everything in the last emails]. > > I think I've understood correctly. IIRC, if you use 2.6 as host (even > 2.6-test), an UML bug appears. That bug is not triggered if > you have a 2.4 > host; also, it appears randomically (not with /bin/sh, not > with the Debian > rootfs but could reappear even with them...). The patch has > been applied to > later 2.4 guest, not to stock 2.6 guest; see the end of this > email to find > it. Ouh, shi**. I think it's clear now. It seems that I've just good luck with the debian-FS :-) > > > > Is there a kernel option for the host I've forgotton to set? > > > > > > For the host the only option that matters is to enable > /proc/mm if > > > you have the SKAS patch on the host. Nothing else. > > > > Ouh ;-). Now everything is clear... > > The host normally doesn't compile with SKAS if you don't > enable /proc/mm. > Also, I don't think you have applied SKAS for 2.6. > > If you need any UML patch for 2.6, I think you should go here: > > http://web.tiscali.it/no-redirect-tiscali/blaisorblade/index.html > > There you'll find SKAS for the host, a fix for the bug above(the G-02... one), > fix for modules + other stuff. See instructions you find in that page and in > the Readme, and note the warning about the A-04 patch. Which patch should I take at the best if I just want to use SKAS and the bugfixes (dor example, there is a patch that tells me that proc/mm isn't needed any more). And what exactly is "vanilla"? And: If I apply the patch, shall I do a make mrproper? Or a make clean? Or something else? And adds the patch an option to make config/menuconfig? I tried to apply 00-Combo-3.patch.bz2. The patch command itself worked without errors but when I compiled, every driver complained about a missing return value somewhere in a header (I think it was mem.h or similar). Finally there was a compiling error at the module of an SCSI device driver. Sorry for my questions but I've never applied patches before I discovered UML :-). I was always afraid that something does not work or would become unstable and somehow I'm now also afraid of patches. Therefor I just want to make sure I do the right thing. And, to make sure again: As UML I use 2.4.20 with exactly this patch uml-patch-2.4.20-8.bz2. And as host I use the normal 2.6.0 (not test or beta or similar). Thanks again for helping! Niki |
From: BlaisorBlade <bla...@ya...> - 2003-12-30 17:59:01
|
> Which patch should I take at the best if I just want to use SKAS > and the bugfixes (dor example, there is a patch that tells me that > proc/mm > isn't needed any more). Yes, the readme is probably more for developers and experienced users than for end users. However, what you need to download is the Combo (the number 4, since the previous ones were broken) and the A-04... patch. While answering to you, I added this explaination to that URL. Apply the combo patch, configure the kernel for UML (enable /proc/mm; if this gives problem, disable it and retry, but this shouldn't happen). Then build it. After that, if you want to compile a 2.6 host kernel with the SKAS patch, revert the A-04 patch: patch -p1 -R < A-04-MM-compile-fix.patch (note the -R: this "unapplies" a patch. This one is included in the combo, but should be disabled for the host kernel). Now configure the kernel(without ARCH=um, this is an host kernel), enable /proc/mm under "Processor type and features" menu, and build it. > And what exactly is "vanilla"? When you have a kernel from kernel.org, without any patches applied, that is called a vanilla kernel. > And: If I apply the patch, shall I do a make mrproper? Or a make clean? > Or something else? With 2.4 you must do a make clean; with 2.6 you shouldn't need it. > And adds the patch an option to make > config/menuconfig? For the host, only the /proc/mm one(i.e. the host-skas support). > I tried to apply 00-Combo-3.patch.bz2. The patch command itself worked > without > errors but when I compiled, every driver complained about a missing > return > value somewhere in a header (I think it was mem.h or similar). > Finally there was a compiling error at the module of an SCSI device > driver. Hmmm... err.... that patch was a lot bogus. Sorry for this... it's my fault (I hadn't had the time to test that in the 4 configurations where it must work : for i386 and UML, with and without /proc/mm). The number 4 combo is tested, however. As a general rule: if you have a compile error, post the exact text. I was answering to another thread and I noticed you posted it, too. So I put the answer here: > However, I read that hostfs is incompatible with kernel 2.6 Are you sure? I don't think so. > 4.) Is the SKAS mode integrated in kernel 2.6.0 stable? ??? In the official 2.6.0, there is not SKAS support. I can't guarantee that my patches are "production-ready". > If not, which patch shall I take? My host system is a "productive" > system where I do not want to have unstable software. If you have a production system, you shouldn't run on it 2.6.0. It is called stable, but no kernel developer would say "it's production ready". Rather, it has been released so "people will test it and send us bug reports". Also, some security related bug fixes have been applied to 2.4 but not to 2.6. Read something from lwn.net or kerneltrap.org if you still think 2.6.0 is stable. However, it's far more stable than 2.4.n where n is <16 . > Is the > patch unstable? What's happening when I apply "an official" patch > from www.kernel.org to my host sources? Does this work? I don't understand your question. If you use patches from kernel.org to update to a new version(i.e. patch-2.4/6.NN.bz2), they are ok. If you find other patches at kernel.org while browsing the FTP, they can be stable, unstable or completely broken. Bye -- cat <<EOSIGN Paolo Giarrusso, aka Blaisorblade Linux Kernel 2.4.21/2.6.0-test on an i686; Linux registered user n. 292729 EOSIGN |
From: Nikolaus H. <nha...@sc...> - 2003-12-31 00:46:27
|
Hi, > > Which patch should I take at the best if I just want to use > SKAS and > > the bugfixes (dor example, there is a patch that tells me > that proc/mm > > isn't needed any more). > > Yes, the readme is probably more for developers and > experienced users than for > end users. However, what you need to download is the Combo > (the number 4, > since the previous ones were broken) and the A-04... patch. > While answering > to you, I added this explaination to that URL. Thank you :-) I will try that solution... > Apply the combo patch, configure the kernel for UML (enable > /proc/mm; if this > gives problem, disable it and retry, but this shouldn't > happen). Then build > it. > After that, if you want to compile a 2.6 host kernel with the > SKAS patch, > revert the A-04 patch: > patch -p1 -R < A-04-MM-compile-fix.patch > (note the -R: this "unapplies" a patch. This one is included > in the combo, but > should be disabled for the host kernel). > > Now configure the kernel(without ARCH=um, this is an host > kernel), enable > /proc/mm under "Processor type and features" menu, and build it. Thank you very much for this exact explanation. I'm just trying it :-) > I was answering to another thread and I noticed you posted > it, too. So I put > the answer here: > > > However, I read that hostfs is incompatible with kernel 2.6 > > Are you sure? I don't think so. No, I'm not sure but I think that I've read this on the sf-Page. > > 4.) Is the SKAS mode integrated in kernel 2.6.0 stable? > ??? In the official 2.6.0, there is not SKAS support. I can't > guarantee that > my patches are "production-ready". Nobody can do that, except developers of well-tested drivers :) > > If not, which patch shall I take? My host system is a "productive" > > system where I do not want to have unstable software. > > If you have a production system, you shouldn't run on it > 2.6.0. It is called > stable, but no kernel developer would say "it's production > ready". Rather, it > has been released so "people will test it and send us bug > reports". Also, some security related bug fixes have been > applied to 2.4 but not to 2.6. Read something from lwn.net or > kerneltrap.org if you still think 2.6.0 is > stable. However, it's far more stable than 2.4.n where n is <16 . This is a very good hint. Thank you. I thought I'll run 2.6.0 because it's stable now and then I have lost my .config file for my old 2.4.20 kernel and there were also problem compiling the bridge driver. But I will give the good old 2.4.20 another try. It's just "important" that the host system runs stable - the uml is not that problem (therefor I want to use UML ;-) > > Is the > > patch unstable? What's happening when I apply "an official" > patch from > > www.kernel.org to my host sources? Does this work? > > I don't understand your question. If you use patches from > kernel.org to update > to a new version(i.e. patch-2.4/6.NN.bz2), they are ok. If > you find other > patches at kernel.org while browsing the FTP, they can be > stable, unstable or > completely broken. No, this are seperate questions. The first one is that these patches are considered to be stable or to be developer patches. The second one is, what happens when I have applied such a UML patch (which is not official) and then I apply an official patch from kernel.org. When I the same files are modified then I will have a problem. Somewhere I read that all patches from kernel.org will work well as long as I leave the source code untouched. And that's my question: Can I excect any problems in this case? niki |
From: Jeff D. <jd...@ad...> - 2004-01-06 02:39:40
|
nha...@sc... said: > Linux version 2.4.20-8um (niki@nobaq) (gcc version 2.95.4 20011002 > Child 6830 exited with signal 11 Try a newer UML. This bug is fixed in UML 2.4.22-2. Jeff |