Created attachment 417899 [details]
all patches applied to the devel tree
Description of problem:
There are several patches are needed to get kdump working in rawhide,
* Sat May 29 2010 CAI Qian <email@example.com> - 2.0.0-35
- Forward-port from F13
- Fixed kernel text area search in kcore (bz 587750)
* Sat May 29 2010 CAI Qian <firstname.lastname@example.org> - 2.0.0-34
- Massive forward-port from RHEL6
- Update kexec-kdump-howto.txt
- Update docs to reflect use of ext4
- Update mkdumprd to pull in all modules needed
- Fix mkdumprd typo
- Removed universal add of ata_piix from mkdumprd
- Fix infinite loop from modprobe changes
- Fixed kexec-kdump-howto.doc for RHEL6
- Update makedumpfile to 1.3.5
- Improved mkdumprd run time
- Cai's fix for broken regex
- Fixing crashkernel syntax parsing
- Fix initscript to return proper LSB return codes
- Fixed bad call to resolve_dm_name
- Added poweroff option to mkdumprd
- Fixed readlink issue
- Fixed x86_64 page_offset specifictaion
- Fixed lvm setup loop to not hang
- Added utsname support to makedumpfile for 2.6.32
- Fix critical_disks list to exclude cciss/md
- Add help info for -b option
- Add ability to handle firmware hotplug events
- Update mkdumprd to deal with changes in busybox fsck
- Vitaly's fix to detect need for 64 bit elf
- Fix major/minor numbers on /dev/rtc
- Fix ssh id propogation w/ selinux
- Add blacklist feature to kdump.conf
- Removed rhpl code from firstboot
- Fixed firstboot enable sense
- Remove bogus debug comment from mkdumprd.
- Handle SPARSEMEM properly
- Fix scp monitoring script
- Fix firstboot to find grub on EFI systems
- Fixed mkdumprd to remove dup insmod
- Fixed kdump fsck pause
- Fixed kdump option handling
- fixed raid5 module detection
All patches have been applied/compiled/tested successfully to the devel tree. Please review.
Version-Release number of selected component (if applicable):
Is there any plan to use Dracut for creating kdump initramfs in the future ?
yes, there is actually infrastructure in place to allow dracut to be used in the creation of initramfs. It works, but is not overly user friendly yet. Feel free to experiment
Cai, your summary comment indicates that you've tested 'all patched' against the rawhide tree, but your attachment is a tarball consisting of the contents of what appears to be the cvs repository for RHEL6. If you have a set of patches that you've applied independently, list/attach those instead please. Some may not be needed for upstream, and I'd like to look at them individually.
This is the tree using Fedora CVS's devel directory. Patches have been included, and spec file has been modified against Fedora. It is the basically the new version created for rawhide that you can review and commit it directly.
Created attachment 419403 [details]
This patch was generated between the the above devel tarball and the original devel directory.
The following needs to happen by someone with the commit privilege,
* upload makedumpfile-1.3.5.tar.gz to,
* apply the patch against kexec-tools-2.0.0-33.fc14
Neil: I needed working mkdumprd today and after applying this patch and verifying that it fixed kdump for me and given kdump mkdumprd nonfunctional for a rather long time I thought it is a good idea to commit the resulting package.
I somehow did not notice your comment about your intent to review the patches separately. I'm sorry if I caused any trouble; keeping this open so that this does not disappear from your radar and you can amend the fixes if necessary.
yeah, I just saw that you committed them all. I would have liked to review all the patches, but whats done is done, and given that, while kexec did work in many cases, it was also broken in many, its probably just as well that this is done now. Please take care not to just commit stuff in the future, but I think what you did was likely for the best here. Thanks.
I'll close this as the work is done, and we can fix anything that typical use presents to us a problem as we come to it.