External metadata updates and other fixes for 3.0.3
Get series with b4
b4 am [email protected]Versions newest first
- v1 Latest12 patches 1 replyIn review
Patches in v1 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/12 | imsm: cleanup disk status testssuper-intel.c | +29 | 0 replies | |
| 2/12 | imsm: kill close() of component devicesuper-intel.c | +1 | 0 replies | |
| 3/12 | imsm: disambiguate family_numsuper-intel.c | +448 | 0 replies | |
| 4/12 | imsm: fix spare record writeout racesuper-intel.c | +35 | 0 replies | |
| 5/12 | imsm: fix/support --updatesuper-intel.c | +26 | 0 replies | |
| 6/12 | ddf: prevent superblock being zeroed on --updatesuper-ddf.c | +19 | 0 replies | |
| 7/12 | imsm: add --update=uuid supportAssemble.c mdadm.h super-intel.c | +47 | 0 replies | |
| 8/12 | imsm: regression test for prodigal array member scenariotests/ tests/ | +78 | 0 replies | |
| 9/12 | Detail: export MD_UUID from mapfileDetail.c mdadm.h util.c | +14 | 0 replies | |
| 10/12 | mdmon: avoid writes in the startup path for mdmon on root arraysmanagemon.c mdmon.c | +38 | 0 replies | |
| 11/12 | mdmon: exec(2) when the switchroot argument is not "/"mdmon.c | +23 | 1 reply | |
| 12/12 | mdmon: preserve socket over chrootmdadm.c mdadm.h mdmon.c +3 more | +26 | 0 replies |
Reviews trailers per patch
No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.
Cover letter [mdadm PATCH 00/12] External metadata updates and other fixes for 3.0.3
Hi Neil,
The following changes are available at:
git://github.com/djbw/mdadm.git
The stand out change in the diffstat is the incorporation of the Option-ROM logic for handling arrays with different 'family numbers'. The scheme the Option-ROM uses to resolve conflicts is not completely amenable to mdadm's container scheme. The end result is a corner case that requires user intervention. The user will need to assign new container membership ids to prodigal disks that are returned to their system of origin. More details in the commit log and code comments.
Also notable are the changes to cleanup mdmon instance management. I kept the status quo in deciding against implementing an "abstract namespace socket" for the control path. Initramfs environments can take two paths:
1/ Use the built-in 'newroot' option to hand off monitoring to an
mdmon binary launched from the target file system. Currently Dracut will not be able to use this method as its 'switch_root' command does not allow an mdmon instance to be launched between $NEWROOT/dev and $NEWROOT/sys being available and the initramfs exiting.
2/ Keep the mdmon instance from the initramfs running until it can be
killed by the final-root initscripts. The socket will not be available so this should be done while the file system is still read only and any writes from journal recovery have quiesced.
In testing this I noticed that wait_for() (util.c) occasionally timed out on Fedora 11, but not Fedora 9. Increasing the loop count to 50 got around the issue, but need to investigate what is holding up our requested symlink.
Thanks,
Dan
Dan Williams (12):
imsm: cleanup disk status tests
imsm: kill close() of component device
imsm: disambiguate family_num
imsm: fix spare record writeout race
imsm: fix/support --update
ddf: prevent superblock being zeroed on --update
imsm: add --update=uuid support
imsm: regression test for prodigal array member scenario
Detail: export MD_UUID from mapfile
mdmon: avoid writes in the startup path for mdmon on root arrays
mdmon: exec(2) when the switchroot argument is not "/"
mdmon: preserve socket over chroot
Assemble.c | 3 Detail.c | 5 managemon.c | 6 mdadm.c | 4 mdadm.h | 8 mdmon.c | 107 +++--- msg.c | 14 + msg.h | 1 super-ddf.c | 27 +- super-intel.c | 799 ++++++++++++++++++++++++++++++++------------- sysfs.c | 5 tests/09imsm-assemble | 46 +++ tests/env-09imsm-assemble | 32 ++ util.c | 12 - 14 files changed, 780 insertions(+), 289 deletions(-) create mode 100644 tests/09imsm-assemble create mode 100644 tests/env-09imsm-assemble