| Summary: | grub causes 'Segmentation Fault' when run from boot CD in chroot | ||
|---|---|---|---|
| Product: | Gentoo Release Media | Reporter: | Adam Bolte <adam> |
| Component: | Everything | Assignee: | Gentoo Kernel Bug Wranglers and Kernel Maintainers <kernel> |
| Status: | RESOLVED FIXED | ||
| Severity: | blocker | CC: | johnm, robmoss, x86-kernel |
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | x86 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
OK. Rebooting into Gentoo to complete the installation using LILO, I re-attempted emerging grub 0.92-r1. It worked perfectly. Now I can unmerge LILO (phew). Just confirming that the same segfault happens on my IBM Thinkpad T30 during final stages of installation. The machine was previously - just couple of hours ago - running RH 9 with grub 0.93 without any problems. CFLAGS used during installation are "-mcpu=pentium4 -march=pentium3 -O3 -pipe -fomit-frame-pointer" Happens on my vaio laptop as well. Latest version of the Live CD, grub 0.93.20030118 I get to the part in the install where, in a chroot'd environment I type "grub", and it dies segfaults almost immediately. If I type 'grub' again, it just hangs. Hardware: Pentium III 320Megs Ram CFLAGS="-O2 -march=i686 -funroll-loops -pipe" adding kernel team in case someone knows something 1) Can you please point me to the exact LiveCD ISO image so that I can debug that stack trace... 2) Can you please tell us the results of 'emerge info' 3) Can you try using GRUB when compiled with CFLAGS=""? I have a similar kernel segmentation fault, but not with grub - should I report a new bug instead? ---------------------------------------------------------- Unable to handle kernel NULL pointer dereference at virtual address 00000000 printing eip: 00000000 *pde = 00000000 Oops: 0000 via-rhine mii serial isa-pnp cloop usb-storage hid uhci ehci-hcd usbcore CPU: 0 EIP: 0010:[<00000000>] Tainted: GF EFLAGS: 00010202 eax: c033b1e0 ebx: c13c6610 ecx: c12be80c edx: c127e3d4 esi: c12be80c edi: cfef3168 ebp: 0000192f esp: c634fdac ds: 0018 es: 0018 ss: 0018 Process bash (pid: 1268, stackpage = c634f000) Stack: --[a lot of numbers]-- Call Trace: --[a lot of numbers]-- Code: Bad EIP value ---------------------------------------------------------- I boot up a machine (AMD Athlon XP 2200+) with Gentoo LiveCD (I think I'm using x86-1.4-20030911-cd1.iso). The segmentation faults occured on very different occasions (one time when executed "passwd"; two times when executed "df -h" in a chroot; one time when pressed Ctrl-C when bootstrap.sh was compiling). The system is not usable after the segmentation fault any more and I have to reboot. I also tried different kernel bootup options (framebuffer on/off; acpi on/off).. Linux has previously run on this machine (RedHat 8.0). Alar: The stack traces and callbacks would be quite useful for debugging purposes. Thanks. I had this problem as well on a Toshiba Satellite Pro with a pentium 4. I had used stage 3 to install so I had my CFLAGS and CXXFLAGS equal to "-03 -march=pentium4 -funroll-loops -fprefetch-loop-arrays -pipe". To solve this I set my CFLAGS and CXXFLAGS blank. To do this I used "CFLAGS="" CXXFLAGS="" emerge grub". Now when I started grub, it worked perfectly. Looks like we might need to filter flags then - can somebody please confirm whether that particular flag combination causes it or one of the specific flags does? I'm still waiting for a stack trace so that I can debug this. Thanks. moving to the kernel team. Well, to be honest, we shouldn't be letting this pick up the CFLAGS from make.conf anyway. Indeed, it should use -march=i386 as long as you're on x86. I suspect, however (yet to test) that this is down to -funroll-loops. Can someone please close this? grub now uses "unset CFLAGS", so -funroll-loops will be filtered. Please reopen if this is still an issue, which I suspect will not be the case. Ok, closing this for now. I anyone experiences similar problems with grub-0.94-r1, please reopen/comment. |
Hardware: AMD Athlon XP Asus A7N8X Deluxe (nVidia nForce 2) PowerColor ATI Radeon 8500 Evil Master II Multi-Display Edition 512MB RAM Kingston DDR333 Grub has previously worked on this PC (not sure which version). Now, neither 0.92-r1 or grub-0.93-20030118 work. Not sure yet if it's just because of running 'grub' in the chrooted environment or not, but previous UT2003 Gentoo versions worked fine with the same setup. Even exactly the same make.conf file (I backed it up and reused it). Reproducible: Always Steps to Reproduce: 1. Type 'grub' as the last stage of the install from Live CD 1.4RC4 2. 3. Actual Results: >grub Probing devices to guess BIOS drives. This may take a long time. Unable to handle kernel paging request at virtual address 00410000 printing eip: 00410000 *pde = 00000000 Oops: 0000 CPU: 0 EUP: 0010:[<00410000>] Tainted: PF ELFAGS: 00010206 eax: 00410000 ebx: ddd85d80 ecx: df90482c edx: e48be41a esi: df904810 edi: df90482c dbp: dffc0200 esp: d777ff1c Stack: c0138e0e ddd85d80 df31bc80 00000000 ddd85d80 df31bc80 ffffffe9 dffc6390 c0138ec0 df904810 ddd85d80 df31bc80 ddd85d80 df31bc80 ddd85d80 c0132818 ddd85d80 df31bc80 00008000 c6388000 00000008 bfffe758 c013274e dc205d60 Call Trace: [<c0138e0e>] [<c0138ec0>] [<c0132818>] [<c013274e>] [<c0132a91>] [<c0108a07>] Code: Bad EIP value. Segmentation fault. Does this make sense to anyone? Expected Results: Having the software actually run would have been nice. :) Now I'm using LILO. :( LILO worked fine. I can't possibly imagine why Samba was a dependency for LILO. Sure, it's in my USE flag, but so what? Probably not the right place for this discussion.