salvager: Do not fork for single VG salvage

Currently we always fork a child in the salvager in order to salvage a
volume group. I believe this is in order to protect SEGV, exit(), etc
in one salvage operation from preventing salvaging anything else. When
salvaging a single volume group, though, there appears to be little
benefit.

In addition, we need to keep the VG salvaging code in the same process
as the cleanup code for single-volume salvages, so we can know which
volumes were deleted by SalvageVolumeGroup, so we know which volumes
to bring back online. So, do not fork for the singleVolumeNumber case.

Note that for DAFS, we already never fork for the entire salvage
operation when salvaging an individual volume group. So, this is
effectively a non-DAFS-only change.

Change-Id: I2b646b9ff932a673c714bc27ac32ed41cc3e22e6
Reviewed-on: http://gerrit.openafs.org/6787
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
This commit is contained in:
Andrew Deason 2012-02-23 13:02:13 -06:00 committed by Derrick Brashear
parent 2a1719faa8
commit 46e85dd468

View File

@ -859,6 +859,16 @@ SalvageFileSys1(struct DiskPartition64 *partP, VolumeId singleVolumeNumber)
goto retry;
}
if (singleVolumeNumber) {
/* If we delete a volume during the salvage, we indicate as such by
* setting the volsummary->deleted field. We need to know if we
* deleted a volume or not in order to know which volumes to bring
* back online after the salvage. If we fork, we will lose this
* information, since volsummary->deleted will not get set in the
* parent. So, don't fork. */
canfork = 0;
}
for (i = j = 0, vsp = salvinfo->volumeSummaryp, esp = vsp + salvinfo->nVolumes;
i < salvinfo->nVolumesInInodeFile; i = j) {
VolumeId rwvid = salvinfo->inodeSummary[i].RWvolumeId;