From: Mattia D. <mal...@li...> - 2006-05-12 20:12:26
|
Hello UML developers, I received the following bugreport for 2.6.16 UML running on a 2.6.16 host (both are debian packages). I found a similar report on a french site[1], it seems to be the same problem. Can you help putting some light there? Thanks in advance [1]: http://forum.hardware.fr/hardwarefr/OSAlternatifs/UML-Debian-sujet-57017-1.htm On Fri, May 12, 2006 at 01:50:19PM +1000, Paul Dwerryhouse wrote: > Package: user-mode-linux > Version: 2.6.16-1um-2 > Severity: normal > > > When user-mode linux is used with a host kernel of > linux-image-2.6.16-1-k7 (ie, the package in sid), it panics. When the > host kernel is linux-image-2.6.15-1-k7 (also from sid), it works > perfectly. > > I've tested this with a number of guest images; debian sid, ubuntu > breezy & dapper. Doesn't appear to be an NPTL issue; it crashes under > 2.6.16 whether the /lib/tls libraries are present or not. > > > The boot messages when it fails: > > $ linux ubd0=uml1.img con=pty con0=fd:0,fd:1 mem=256M eth0=daemon,,unix,/var/run/vde/tap1.ctl eth1=daemon,,unix,/var/run/vde/tap1.ctl > Checking that ptrace can change system call numbers...OK > Checking syscall emulation patch for ptrace...OK > Checking advanced syscall emulation patch for ptrace...OK > Checking PROT_EXEC mmap in /tmp...OK > Checking for the skas3 patch in the host: > - /proc/mm...not found > - PTRACE_FAULTINFO...not found > - PTRACE_LDT...not found > UML running in SKAS0 mode > Linux version 2.6.16 (root@inferi) (gcc version 4.0.4 20060422 (prerelease) (Debian 4.0.3-2)) #2 Sat Apr 29 23:36:42 CEST 2006 > Built 1 zonelists > Kernel command line: ubd0=uml1.img con=pty con0=fd:0,fd:1 mem=256M eth0=daemon,,unix,/var/run/vde/tap1.ctl eth1=daemon,,unix,/var/run/vde/tap1.ctl root=98:0 > PID hash table entries: 2048 (order: 11, 32768 bytes) > > [...] > > kjournald starting. Commit interval 5 seconds > EXT3-fs: mounted filesystem with ordered data mode. > VFS: Mounted root (ext3 filesystem) readonly. > mapping mmap stub failed, errno = 12 > Kernel panic - not syncing: start_userspace : expected SIGSTOP, got status = 256 > > EIP: 0073:[<ffffe410>] CPU: 0 Not tainted ESP: 007b:a7e29fa8 EFLAGS: 00200246 Not tainted > EAX: 00000000 EBX: 000013f5 ECX: 00000013 EDX: 000013f5 > ESI: 000013f1 EDI: 00000011 EBP: a7e29fc4 DS: 007b ES: 007b > 08267af8: [<0806f080>] show_regs+0xb4/0xb9 > 08267b24: [<0805e08c>] panic_exit+0x23/0x3d > 08267b38: [<0807e495>] notifier_call_chain+0x1a/0x31 > 08267b58: [<08072ffb>] panic+0x4f/0xca > 08267b74: [<0806c165>] start_userspace+0xce/0x11a > 08267b94: [<0805e31a>] init_new_context_skas+0xfb/0x13e > 08267bb4: [<080ad3f0>] do_execve+0x96/0x1df > 08267bd8: [<080599b8>] execve1+0x28/0x52 > 08267bfc: [<080599f5>] um_execve+0x13/0x3f > 08267c18: [<0805936a>] run_init_process+0x34/0x4f > 08267c38: [<08059471>] init+0xec/0x117 > 08267c48: [<0806955b>] run_kernel_thread+0x42/0x4a > 08267cf8: [<0805e49f>] new_thread_handler+0x7f/0xa9 > 08267d1c: [<ffffe420>] _etext+0xf7e66406/0x0 > > Thanks, > > Paul. > > > > > -- System Information: > Debian Release: testing/unstable > APT prefers unstable > APT policy: (500, 'unstable') > Architecture: i386 (i686) > Shell: /bin/sh linked to /bin/bash > Kernel: Linux 2.6.16-1-k7 > Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968) > > Versions of packages user-mode-linux depends on: > ii uml-utilities 20060323-3 User-mode Linux (utility programs) > > user-mode-linux recommends no packages. > > -- no debconf information > > > -- mattia :wq! |
From: Jeff D. <jd...@ad...> - 2006-05-15 03:44:35
|
On Fri, May 12, 2006 at 10:08:06PM +0200, Mattia Dongili wrote: > > kjournald starting. Commit interval 5 seconds > > EXT3-fs: mounted filesystem with ordered data mode. > > VFS: Mounted root (ext3 filesystem) readonly. > > mapping mmap stub failed, errno = 12 > > Kernel panic - not syncing: start_userspace : expected SIGSTOP, got status = 256 Hummmm, is this a 2G/2G host, by any chance? The French report is complaining about this happening after a host kernel upgrade, and I would assume that this wouldn't spring a 2G/2G split on you without asking. UML is trying to map a page at the top of its address space (which it assumes to be 0xc0000000 unless you are running a 2.6.16-rc4 UML and enabled host 2G/2G support). We need to figure out why that doesn't work. If it's not the host split, then I don't have any other ideas off the top of my head. Jeff |
From: Mattia D. <mal...@li...> - 2006-05-15 10:09:09
Attachments:
2.6.15-2.6.16-k7.config.diff
|
On Mon, May 15, 2006 5:44 am, Jeff Dike said: > On Fri, May 12, 2006 at 10:08:06PM +0200, Mattia Dongili wrote: >> > kjournald starting. Commit interval 5 seconds >> > EXT3-fs: mounted filesystem with ordered data mode. >> > VFS: Mounted root (ext3 filesystem) readonly. >> > mapping mmap stub failed, errno =3D 12 >> > Kernel panic - not syncing: start_userspace : expected SIGSTOP, got >> status =3D 256 > > Hummmm, is this a 2G/2G host, by any chance? It doesn't seem so: CONFIG_VMSPLIT_3G_OPT=3Dy (from debian kernel's 2.6.16-k7 binaries) > The French report is complaining about this happening after a host > kernel upgrade, and I would assume that this wouldn't spring a 2G/2G > split on you without asking. > > UML is trying to map a page at the top of its address space (which it > assumes to be 0xc0000000 unless you are running a 2.6.16-rc4 UML and did you mean .17-rc4? Debian kernels contain very little modifications, and usually track upstream very closely. > enabled host 2G/2G support). We need to figure out why that doesn't Oh, I found CONFIG_PAGE_OFFSET=3D0xB0000000 may this be the problem? I found reports telling it broke other SW: http://lists.debian.org/debian-kernel/2006/04/msg00042.html --=20 mattia :wq! |
From: Jeff D. <jd...@ad...> - 2006-05-15 12:39:52
|
On Mon, May 15, 2006 at 12:08:58PM +0200, Mattia Dongili wrote: > Oh, I found > CONFIG_PAGE_OFFSET=0xB0000000 > > may this be the problem? I found reports telling it broke other SW: > http://lists.debian.org/debian-kernel/2006/04/msg00042.html Yup, that's the problem. They did spring a split change on you. On rc4, UML will run if you enable host 2G/2G support. Jeff |
From: Mattia D. <mal...@li...> - 2006-05-15 20:13:31
|
On Mon, May 15, 2006 at 08:39:54AM -0400, Jeff Dike wrote: > On Mon, May 15, 2006 at 12:08:58PM +0200, Mattia Dongili wrote: > > Oh, I found > > CONFIG_PAGE_OFFSET=0xB0000000 > > > > may this be the problem? I found reports telling it broke other SW: > > http://lists.debian.org/debian-kernel/2006/04/msg00042.html > > Yup, that's the problem. They did spring a split change on you. > > On rc4, UML will run if you enable host 2G/2G support. Does this mean that UML will run on a non-standard-vm-split _only_ if the host has CONFIG_VMSPLIT_2G=y ? If so is this considered a bug or a feature? :) Thanks a lot -- mattia :wq! |
From: Jeff D. <jd...@ad...> - 2006-05-16 01:22:52
|
On Mon, May 15, 2006 at 09:55:09PM +0200, Mattia Dongili wrote: > Does this mean that UML will run on a non-standard-vm-split _only_ if > the host has CONFIG_VMSPLIT_2G=y ? The question is not clear. The host will only have a nonstandard split if it has CONFIG_VMSPLIT_2G or equivalent enabled. Do you instead mean whether UML will run on a nonstandard split host if IT (not the host) has CONFIG_HOST_2G_2G (not CONFIG_VMSPLIT_2G) enabled? If so, the answer for now is yes. > If so is this considered a bug or a feature? :) More a bug. We are considering ways of detecting the host vmsplit at run time and adapting to it, but added back the CONFIG_HOST_2G_2G as an interim measure. Jeff |
From: Mattia D. <mal...@li...> - 2006-05-16 18:09:38
|
On Mon, May 15, 2006 at 09:23:19PM -0400, Jeff Dike wrote: > On Mon, May 15, 2006 at 09:55:09PM +0200, Mattia Dongili wrote: > > Does this mean that UML will run on a non-standard-vm-split _only_ if > > the host has CONFIG_VMSPLIT_2G=y ? > > The question is not clear. The host will only have a nonstandard > split if it has CONFIG_VMSPLIT_2G or equivalent enabled. > > Do you instead mean whether UML will run on a nonstandard split host > if IT (not the host) has CONFIG_HOST_2G_2G (not CONFIG_VMSPLIT_2G) > enabled? > > If so, the answer for now is yes. Aah! thanks for clarifing and sorry for the confusion on my side.. > > > If so is this considered a bug or a feature? :) > > More a bug. We are considering ways of detecting the host vmsplit at > run time and adapting to it, but added back the CONFIG_HOST_2G_2G as > an interim measure. Thanks, will enable it in the .17 release of the debian package. -- mattia :wq! |
From: Jeff D. <jd...@ad...> - 2006-05-16 19:09:11
|
On Tue, May 16, 2006 at 08:09:39PM +0200, Mattia Dongili wrote: > > More a bug. We are considering ways of detecting the host vmsplit at > > run time and adapting to it, but added back the CONFIG_HOST_2G_2G as > > an interim measure. > > Thanks, will enable it in the .17 release of the debian package. The option is not without cost. It reduces the address space available to UML processes. Since the B0000000 split seems to be Debian-specific, you might add a corresponding Debian-specific patch to UML which defines the top of the address space to B0000000, rather than A0000000, which CONFIG_HOST_2G_2G does. Jeff |
From: Jeff D. <jd...@ad...> - 2006-05-16 01:18:56
|
On Mon, May 15, 2006 at 12:08:58PM +0200, Mattia Dongili wrote: > did you mean .17-rc4? Debian kernels contain very little modifications, > and usually track upstream very closely. Yes I did, sorry. Jeff |
From: Blaisorblade <bla...@ya...> - 2006-05-15 20:51:51
|
On Friday 12 May 2006 22:08, Mattia Dongili wrote: > Hello UML developers, > > I received the following bugreport for 2.6.16 UML running on a 2.6.16 > host (both are debian packages). I found a similar report on a french > site[1], it seems to be the same problem. > > Can you help putting some light there? > Thanks in advance > > [1]: > http://forum.hardware.fr/hardwarefr/OSAlternatifs/UML-Debian-sujet-57017-1. >htm I also get a problem specific to host 2.6.16 kernels, but that's when I run x64 guests on x64 2.6.16 hosts (x86 32-bit guests work normally); I've not tested 32-bit 2.6.16 hosts. The error message is different. However, I can't exclude there is the same root cause acting in a different way, but I've now debugged this a bit and this seems unlikely. EXT3-fs: mounted filesystem with ordered data mode. VFS: Mounted root (ext3 filesystem) readonly. Warning: unable to open an initial console. Kernel panic - not syncing: handle_trap - failed to wait at end of syscall, errno = 0, status = 2943 Modules linked in: Pid: 1, comm: init Not tainted 2.6.16.9-bs2 RIP: 0033:[<00000000400101ba>] RSP: 0000007f7f9ec358 EFLAGS: 00000246 RAX: 000000000050a000 RBX: 0000000000000000 RCX: ffffffffffffffff RDX: 0000000000000026 RSI: 0000000000000000 RDI: 0000000000000000 RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000040000000 R10: 0000000040000a18 R11: 0000000000000246 R12: 0000000000000000 R13: 0000000040001dd0 R14: 0000000000000008 R15: 0000000000400040 Call Trace: 6031bc48: [<6000fe8c>] panic_exit+0x27/0x42 6031bc58: [<60034f15>] notifier_call_chain+0x1f/0x2f 6031bc78: [<6002784a>] panic+0xcf/0x165 6031bcb8: [<601791f9>] __libc_waitpid+0x39/0xc0 6031bcd8: [<6001e7f6>] set_signals+0x26/0x53 6031bd08: [<60021640>] move_registers+0x45/0x7e 6031bd58: [<6002295f>] userspace+0x213/0x265 6031bdc0: [<6000a17f>] init+0x0/0x13d 6031bdd8: [<60010410>] new_thread_handler+0x121/0x12f 6031be38: [<6016c9a0>] __restore_rt+0x0/0x10 6031bee8: [<6016cc59>] kill+0x9/0x20 > > When user-mode linux is used with a host kernel of > > linux-image-2.6.16-1-k7 (ie, the package in sid), it panics. When the > > host kernel is linux-image-2.6.15-1-k7 (also from sid), it works > > perfectly. > > mapping mmap stub failed, errno = 12 > > Kernel panic - not syncing: start_userspace : expected SIGSTOP, got > > status = 256 > > EIP: 0073:[<ffffe410>] CPU: 0 Not tainted ESP: 007b:a7e29fa8 EFLAGS: > > 00200246 Not tainted EAX: 00000000 EBX: 000013f5 ECX: 00000013 EDX: > > 000013f5 > > ESI: 000013f1 EDI: 00000011 EBP: a7e29fc4 DS: 007b ES: 007b > > 08267af8: [<0806f080>] show_regs+0xb4/0xb9 > > 08267b24: [<0805e08c>] panic_exit+0x23/0x3d > > 08267b38: [<0807e495>] notifier_call_chain+0x1a/0x31 > > 08267b58: [<08072ffb>] panic+0x4f/0xca > > 08267b74: [<0806c165>] start_userspace+0xce/0x11a > > 08267b94: [<0805e31a>] init_new_context_skas+0xfb/0x13e > > 08267bb4: [<080ad3f0>] do_execve+0x96/0x1df > > 08267bd8: [<080599b8>] execve1+0x28/0x52 > > 08267bfc: [<080599f5>] um_execve+0x13/0x3f > > 08267c18: [<0805936a>] run_init_process+0x34/0x4f > > 08267c38: [<08059471>] init+0xec/0x117 > > 08267c48: [<0806955b>] run_kernel_thread+0x42/0x4a > > 08267cf8: [<0805e49f>] new_thread_handler+0x7f/0xa9 > > 08267d1c: [<ffffe420>] _etext+0xf7e66406/0x0 > > > > Thanks, > > > > Paul. > > > > > > > > > > -- System Information: > > Debian Release: testing/unstable > > APT prefers unstable > > APT policy: (500, 'unstable') > > Architecture: i386 (i686) > > Shell: /bin/sh linked to /bin/bash > > Kernel: Linux 2.6.16-1-k7 > > Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968) > > > > Versions of packages user-mode-linux depends on: > > ii uml-utilities 20060323-3 User-mode Linux (utility > > programs) > > > > user-mode-linux recommends no packages. > > > > -- no debconf information -- Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!". Paolo Giarrusso, aka Blaisorblade (Skype ID "PaoloGiarrusso", ICQ 215621894) http://www.user-mode-linux.org/~blaisorblade ___________________________________ Yahoo! Mail: gratis 1GB per i messaggi e allegati da 10MB http://mail.yahoo.it |
From: Jeff D. <jd...@ad...> - 2006-05-16 19:14:16
|
On Mon, May 15, 2006 at 12:39:48PM +0200, Blaisorblade wrote: > The error message is different. However, I can't exclude there is the same > root cause acting in a different way, but I've now debugged this a bit and > this seems unlikely. Have you learned anything? See the register dumping patch I just posted and see if that tells you anything new. Jeff |