This is a discussion on UML 2.6.10 on Gentoo x86 system within the Linux Operating System forums, part of the Unix Operating Systems category; --> This is beginning to annoy me, so hopefully someone can give me a clue here, as I get the ...
| |||||||
| FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
| ||||
| This is beginning to annoy me, so hopefully someone can give me a clue here, as I get the feeling that I'm missing something very basic. I have a working 2.6.10 kernel, installed using 'emerge development-sources' in the usual fashion. cd /usr/src cp -rp /usr/src/linux-2.6.10-gentoo-r6 /usr/src/linux-uml-2.6.10-gentoo-r6 cd /usr/src/linux-uml-2.6.10-gentoo-r6 make mrproper make mrproper ARCH=um make menuconfig ARCH=um make linux ARCH=um The compile comes out with a mix of CC and LD lines, but also comes up with raw gcc lines; this looks unclean to me, though it may be a non-problem (I can't tell). The final link barfs out a few lines of the form /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.5/../../../../i686-pc-linux-gnu/bin/ld: BFD 2.15.92.0.2 20040927 assertion fail elf.c:3637 /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.5/../../../../i686-pc-linux-gnu/bin/ld: `__strtol_ul_rem_tab' referenced in section `.text' of /usr/lib/libc.a(strtoul_l.o): defined in discarded section `.gnu.linkonce.r.__strtol_ul_rem_tab' of /usr/lib/libc.a(strtoul_l.o) /usr/lib/gcc-lib/i686-pc-linux-gnu/3.3.5/../../../../i686-pc-linux-gnu/bin/ld: `__strtol_ul_max_tab' referenced in section `.text' of /usr/lib/libc.a(strtoul_l.o): defined in discarded section `.gnu.linkonce.r.__strtol_ul_max_tab' of /usr/lib/libc.a(strtoul_l.o) The resulting executable dies with a segfault; ddd suggests it's dying in ptmalloc_init(), with a corrupted stack. So a few questions: [1] Am I fighting something very ugly in the linker? (GNU ld version 2.15.92.0.2 20040927) Or, even worse, in libc? [2] The file /usr/src/linux-uml-2.6.10-gentoo-r6/arch/um/kernel/uml.lds.S comes up in a scan for modified files after 'make mrproper ARCH=um'. Deleting it, however, borks the build even worse. [3] It's not a 2G/2G problem, as far as I can tell, since System.map of my *host* kernel (which is in another directory) has every entry starting with 'c0...' (except for a very few at the end); presumably with a 2G/2G most entries would instead start with '80...'. [4] What am I missing? (Aside from some of my hair... :-) ) At this point I'm tempted to start with a clean 2.6.10 kernel (as far as I can tell there's no patches that Gentoo knows about) or maybe a 2.6.11 kernel, and hope for better luck. -- #191, ewill3@earthlink.net -- insert random thanks in advance here It's still legal to go .sigless. |
| Thread Tools | |
| Display Modes | |
|
|