Created attachment 404984 [details] gzip of strace.out from "strace -f -o strace.out rpm -ivh kernel-2.6.33.1-24.fc13.x86_64.rpm" Description of problem: Running "rpm -ivh kernel-2.6.33.1-24.fc13.x86_64.rpm" fails to create initramfs because of a crash in /bin/bash invoked from grubby. Version-Release number of selected component (if applicable): dracut-004-5.fc13.noarch grubby-7.0.13-1.fc13.x86_64 How reproducible: every time Steps to Reproduce: 1. rpm --erase kernel-2.6.33.1-24.fc13.x86_64.rpm 2. rpm -ivh kernel-2.6.33.1-24.fc13.x86_64.rpm 3. Actual results: ----- text console during rpm /sbin/new-kernel-pkg: line 289: 2329 Segmentation fault (core dumped) $tool mkinitrd failed warning: %posttrans(kernel-2.6.33.1-24.fc13.x86_64) scriptlet failed, exit status 1 ----- /var/log/messages Apr 7 07:43:04 f13a64 abrt[2406]: saved core dump of pid 2329 (/bin/bash) to /var/cache/abrt/ccpp-1270651384-2329.new/coredump (647168 bytes) Apr 7 07:43:04 f13a64 abrtd: Directory 'ccpp-1270651384-2329' creation detected Apr 7 07:43:04 f13a64 abrtd: New crash /var/cache/abrt/ccpp-1270651384-2329, processing Apr 7 07:43:04 f13a64 abrtd: RunApp('/var/cache/abrt/ccpp-1270651384-2329','test x"`cat component`" = x"xorg-x11-server-Xorg" && cp /var/log/Xorg.0.log .') ----- Expected results: success Additional info: abrt noticed, although I had to look in /var/log/messages to see the report; there was no notification icon in the Gnome desktop menu bar at top of screen. Will attach an archive of the abrt directory. Also will attach 'strace -f' of rpm.
Created attachment 404985 [details] "tar cvfz" of abrt crash directory
hmm, ok, so bash crashed... and why is this a dracut bug then?
It was filed under dracut because dracut is the user-visible piece that failed reproducibly when I was encouraged to file a bug report; see http://lists.fedoraproject.org/pipermail/devel/2010-April/134498.html Subsequent investigation suggests a hardware error. The point of failure was ----- => 0x42b079 <make_command_string_internal+345>: mov 0x12aa16d(%rip),%eax # 0x16d51ec ----- but a re-install of bash now says ----- => 0x42b079 <make_command_string_internal+345>: mov 0x2aa16d(%rip),%eax # 0x6d51ec <indentation_amount> ----- which has a 1-bit difference in the high byte of the offset. Memtest86+ reports no errors. It could be a flaky ICH10 south bridge-to-SATA I/O error.