From: Christoph D. <cdw...@gm...> - 2005-12-07 16:07:29
Attachments:
smime.p7s
|
Hi! I'm trying to build my own UML kernel here. It builds fine, but executing the kernel leads to: alderaan:~/uml# strace ./linux -stderr execve("./linux", ["./linux", "-stderr"], [/* 30 vars */]) =3D 0 uname({sys=3D"Linux", node=3D"alderaan", ...}) =3D 0 --- SIGSEGV (Segmentation fault) @ 0 (0) --- +++ killed by SIGSEGV +++ GDB shows this: Program received signal SIGSEGV, Segmentation fault. 0xa028a2d5 in __libc_setup_tls () Current language: auto; currently asm (gdb) bt #0 0xa028a2d5 in __libc_setup_tls () #1 0xa028a66f in __pthread_initialize_minimal () #2 0xa028a045 in __libc_start_main () #3 0xa0014021 in _start () at ../sysdeps/i386/elf/start.S:119 I've built kernels 2.4.23, 2.4.27, 2.4.28 and 2.6.14 with Blaisorblade's UML patches. The host kernel is 2.6.15-rc5 vanilla, and I'm running Debian sid (updated daily) on x86. Precompiled kernels do run fine! It's not the 3G/1G split problem, I've tried all possible memory configurations of the host kernel and UML. Renaming /lib/tls does not help. I've found quite a few people with the same problem posting on this list, none of the proposed solutions worked for me. Here's my segfaulting linux binary:=20 http://www.informatik.uni-bremen.de/~nermal/linux.bz2 I've spent two days on it now, and I'd appreciate any hints on how to get this working. Thanks! Christoph |
From: Blaisorblade <bla...@ya...> - 2005-12-07 18:10:07
|
On Wednesday 07 December 2005 17:07, Christoph Dwertmann wrote: > Hi! > I'm trying to build my own UML kernel here. It builds fine, but > executing the kernel leads to: > GDB shows this: > Program received signal SIGSEGV, Segmentation fault. > 0xa028a2d5 in __libc_setup_tls () > Current language: auto; currently asm > (gdb) bt > #0 0xa028a2d5 in __libc_setup_tls () > #1 0xa028a66f in __pthread_initialize_minimal () > #2 0xa028a045 in __libc_start_main () > #3 0xa0014021 in _start () at ../sysdeps/i386/elf/start.S:119 Ok, it's due to some weird toolchain-related bug. > I've built kernels 2.4.23, 2.4.27, 2.4.28 and 2.6.14 with > Blaisorblade's UML patches. The host kernel is 2.6.15-rc5 vanilla, and > I'm running Debian sid (updated daily) on x86. Ok, it's a bug in some new binutils/GCC/glibc. While we try to debug the thing, can you post the releases of these 3 components? Also, can you upload somewhere and post the URL of your compressed "ld" binary? Ok, don't need GCC version, but still need binutils: Linux version 2.6.14-bs2 (root@alderaan) (gcc version 4.0.3 20051201 (prerelease) (Debian 4.0.2-5)) #3 Wed Dec 7 16:20:12 CET 2005 > Precompiled kernels do run fine! Likely due to binutils then. (Not necessarily to a bug, but to some change). Btw, to help with the current problem (while fixing the bug), have you any need unsatisified by existing binaries (either on my site or on uml.nagafix.co.uk ?). > Renaming /lib/tls does not help. Even doing that on the host? In this (exceptional) case, the result of that may be useful. > I've found quite a few people with the same problem posting on this > list, none of the proposed solutions worked for me. > Here's my segfaulting linux binary: > http://www.informatik.uni-bremen.de/~nermal/linux.bz2 Downloaded and looking at it, it doesn't run here either and shows the same syntoms. > I've spent two days on it now, and I'd appreciate any hints on how to > get this working. Thanks! -- 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: Christoph D. <cdw...@gm...> - 2005-12-07 18:34:51
Attachments:
smime.p7s
|
On Wed, Dec 07, 2005 at 07:09:47PM +0100, Blaisorblade wrote: > On Wednesday 07 December 2005 17:07, Christoph Dwertmann wrote: > > Hi! >=20 > > I'm trying to build my own UML kernel here. It builds fine, but > > executing the kernel leads to: >=20 > > GDB shows this: >=20 > > Program received signal SIGSEGV, Segmentation fault. > > 0xa028a2d5 in __libc_setup_tls () > > Current language: auto; currently asm > > (gdb) bt > > #0 0xa028a2d5 in __libc_setup_tls () > > #1 0xa028a66f in __pthread_initialize_minimal () > > #2 0xa028a045 in __libc_start_main () > > #3 0xa0014021 in _start () at ../sysdeps/i386/elf/start.S:119 >=20 > Ok, it's due to some weird toolchain-related bug. >=20 > > I've built kernels 2.4.23, 2.4.27, 2.4.28 and 2.6.14 with > > Blaisorblade's UML patches. The host kernel is 2.6.15-rc5 vanilla, and > > I'm running Debian sid (updated daily) on x86. >=20 > Ok, it's a bug in some new binutils/GCC/glibc. While we try to debug the= =20 > thing, can you post the releases of these 3 components? Also, can you upl= oad=20 > somewhere and post the URL of your compressed "ld" binary? Hi, thanks for your reply. Here's my configuration: alderaan:~/uml# gcc --version gcc (GCC) 4.0.3 20051201 (prerelease) (Debian 4.0.2-5) alderaan:~/uml# dpkg -l | grep binutils ii binutils 2.16.1cvs20051117-1 alderaan:~/uml# dpkg -l | grep glibc ii libstdc++2.10-glibc2.2 2.95.4-24=20 (is this the package containing glibc?) alderaan:~/uml# dpkg -l | grep libc6 ii libc6 2.3.5-8.1 ii libc6-dev 2.3.5-8.1 Grab my ld binary here: http://www.informatik.uni-bremen.de/~nermal/ld.bz2 > Btw, to help with the current problem (while fixing the bug), have you an= y=20 > need unsatisified by existing binaries (either on my site or on=20 > uml.nagafix.co.uk ?). yeah I need to apply a few custom kernel patches, and maybe modify them to fit my needs. I really need to be able to compile my own UML kernels. But thanks for the offer! > > Renaming /lib/tls does not help. >=20 > Even doing that on the host? In this (exceptional) case, the result of th= at=20 > may be useful. I just re-checked that: it makes no difference. I noticed that most of the people having this error use a recent Debian or Gentoo system. The cause must be somewhere in my build system, as precompiled binaries run. Thanks, Christoph |
From: Christoph <ne...@gm...> - 2005-12-13 11:45:30
|
On 12/7/05, Christoph Dwertmann <cdw...@gm...> wrote: > On Wed, Dec 07, 2005 at 07:09:47PM +0100, Blaisorblade wrote: > > On Wednesday 07 December 2005 17:07, Christoph Dwertmann wrote: > > > Hi! > > > > > I'm trying to build my own UML kernel here. It builds fine, but > > > executing the kernel leads to: > > > > > GDB shows this: > > > > > Program received signal SIGSEGV, Segmentation fault. > > > 0xa028a2d5 in __libc_setup_tls () > > > Current language: auto; currently asm > > > (gdb) bt > > > #0 0xa028a2d5 in __libc_setup_tls () > > > #1 0xa028a66f in __pthread_initialize_minimal () > > > #2 0xa028a045 in __libc_start_main () > > > #3 0xa0014021 in _start () at ../sysdeps/i386/elf/start.S:119 > > > > Ok, it's due to some weird toolchain-related bug. > > > > > I've built kernels 2.4.23, 2.4.27, 2.4.28 and 2.6.14 with > > > Blaisorblade's UML patches. The host kernel is 2.6.15-rc5 vanilla, an= d > > > I'm running Debian sid (updated daily) on x86. > > > > Ok, it's a bug in some new binutils/GCC/glibc. While we try to debug th= e > > thing, can you post the releases of these 3 components? Also, can you u= pload > > somewhere and post the URL of your compressed "ld" binary? > > Hi, thanks for your reply. Here's my configuration: > > alderaan:~/uml# gcc --version > gcc (GCC) 4.0.3 20051201 (prerelease) (Debian 4.0.2-5) > > alderaan:~/uml# dpkg -l | grep binutils > ii binutils 2.16.1cvs20051117-1 > > alderaan:~/uml# dpkg -l | grep glibc > ii libstdc++2.10-glibc2.2 2.95.4-24 > (is this the package containing glibc?) > > alderaan:~/uml# dpkg -l | grep libc6 > ii libc6 2.3.5-8.1 > ii libc6-dev 2.3.5-8.1 > > Grab my ld binary here: > http://www.informatik.uni-bremen.de/~nermal/ld.bz2 > > > Btw, to help with the current problem (while fixing the bug), have you = any > > need unsatisified by existing binaries (either on my site or on > > uml.nagafix.co.uk ?). > > yeah I need to apply a few custom kernel patches, and maybe modify > them to fit my needs. I really need to be able to compile my own UML > kernels. But thanks for the offer! > > > > Renaming /lib/tls does not help. > > > > Even doing that on the host? In this (exceptional) case, the result of = that > > may be useful. > > I just re-checked that: it makes no difference. > > I noticed that most of the people having this error use a recent > Debian or Gentoo system. The cause must be somewhere in my build > system, as precompiled binaries run. > > Thanks, > > Christoph Hey Blaisorblade and all the others, does anyone have any idea what causes this instant segfault? I'm still stuck here. Thanks, Christoph |
From: Blaisorblade <bla...@ya...> - 2005-12-13 18:34:50
|
On Tuesday 13 December 2005 12:45, Christoph wrote: > On 12/7/05, Christoph Dwertmann <cdw...@gm...> wrote: > > On Wed, Dec 07, 2005 at 07:09:47PM +0100, Blaisorblade wrote: > > > On Wednesday 07 December 2005 17:07, Christoph Dwertmann wrote: > Hey Blaisorblade and all the others, > does anyone have any idea what causes this instant segfault? I'm still > stuck here. Not had time yet to work on this (or on anything else), sorry. -- 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: Philip S. H. <um...@li...> - 2005-12-13 18:30:04
|
Christoph wrote: <chomp> > >> stoph >> > > Hey Blaisorblade and all the others, > > does anyone have any idea what causes this instant segfault? I'm still > stuck here. > > Thanks, > > Christoph > > > The most updated version of cpp seems to have fixed this.. for me.. I have successfully built a image that does not segfault on my AMD64 box.. Not to find if I can get an image to boot with the built in on AMD64 tls... sheesh |
From: Christoph <ne...@gm...> - 2005-12-19 20:32:59
|
On 12/13/05, Philip S. Hempel <um...@li...> wrote: > Christoph wrote: > > <chomp> > > > >> stoph > >> > > > > Hey Blaisorblade and all the others, > > > > does anyone have any idea what causes this instant segfault? I'm still > > stuck here. > > > > Thanks, > > > > Christoph > > > > > > > The most updated version of cpp seems to have fixed this.. for me.. I > have successfully built a image that does not segfault on my AMD64 box.. > > Not to find if I can get an image to boot with the built in on AMD64 > tls... sheesh I also upgraded my Debian/sid and now it's working for me as well. 2.6.x kernels run as UML, but 2.4.x still show a segfault. I upgraded my binutils to: alderaan:~/uml# dpkg -l | grep binutils ii binutils 2.16.1cvs20051214-1 This must have done the trick with kernel 2.6.x. |
From: Blaisorblade <bla...@ya...> - 2005-12-19 20:43:08
|
On Monday 19 December 2005 21:32, Christoph wrote: > On 12/13/05, Philip S. Hempel <um...@li...> wrote: > > Christoph wrote: > > > > <chomp> > > > > >> stoph > > > > > > Hey Blaisorblade and all the others, > > > > > > does anyone have any idea what causes this instant segfault? I'm still > > > stuck here. > > > > > > Thanks, > > > > > > Christoph > > > > The most updated version of cpp seems to have fixed this.. for me.. I > > have successfully built a image that does not segfault on my AMD64 box.. > > > > Not to find if I can get an image to boot with the built in on AMD64 > > tls... sheesh > > I also upgraded my Debian/sid and now it's working for me as well. > 2.6.x kernels run as UML, but 2.4.x still show a segfault. That could well be a different thing - I wish I could diagnose "a segfault", but it's a bit generic ;-). Also probably you can try if 2.4.28-bs2 (applies up to 2.4.31) works better. > I upgraded > my binutils to: > > alderaan:~/uml# dpkg -l | grep binutils > ii binutils 2.16.1cvs20051214-1 > > This must have done the trick with kernel 2.6.x. -- 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! Messenger: chiamate gratuite in tutto il mondo http://it.messenger.yahoo.com |
From: Christoph <ne...@gm...> - 2005-12-21 18:14:16
|
On 12/19/05, Blaisorblade <bla...@ya...> wrote: > On Monday 19 December 2005 21:32, Christoph wrote: > > On 12/13/05, Philip S. Hempel <um...@li...> wrote: > > > Christoph wrote: > > > > > > <chomp> > > > > > > >> stoph > > > > > > > > Hey Blaisorblade and all the others, > > > > > > > > does anyone have any idea what causes this instant segfault? I'm st= ill > > > > stuck here. > > > > > > > > Thanks, > > > > > > > > Christoph > > > > > > The most updated version of cpp seems to have fixed this.. for me.. I > > > have successfully built a image that does not segfault on my AMD64 bo= x.. > > > > > > Not to find if I can get an image to boot with the built in on AMD64 > > > tls... sheesh > > > > I also upgraded my Debian/sid and now it's working for me as well. > > 2.6.x kernels run as UML, but 2.4.x still show a segfault. > That could well be a different thing - I wish I could diagnose "a segfaul= t", > but it's a bit generic ;-). > > Also probably you can try if 2.4.28-bs2 (applies up to 2.4.31) works bett= er. I've built 2.4.28-bs2 and now I can boot it without any segfaulting. The solution was basically to wait for a fixed binutils package in Debian sid. |