On 01/17/2012 05:48 AM, Jes.Sorensen@xxxxxxxxxx wrote:
> From: Jes Sorensen <Jes.Sorensen@xxxxxxxxxx>
>
> Hi,
>
> In order to resolve the problem with reboots hanging on systems with /
> on an IMSM RAID, systemd needs to be able to see that a given mdadm
> and mdmon process was launched from the initramfs, and allow it to
> survive until it gets back to there, so it can unmount the root
> filesystem safely.
>
> These patches introduces the --initrd argument to mdadm which makes it
> change the first character of argv[0] to '@' to match the convention
> set by systemd.
>
> In addition I also changed mdmon to use getopt_long to make it easier
> to add more command line arguments to it. Note that I went through
> great effort to preserve the behavior, even though the old code did
> some somewhat dodgy stuff using changing the arvg buffer pointing to
> the '--all' argument.
>
> Comments?
>
> Thanks,
> Jes
>
> Jes Sorensen (4):
> mdmon: Use getopt_long() to parse command line options
> Add --initrd argument to mdadm
> Add --initrd argument to mdmon
> Spawn mdmon with --initrd if mdadm was launched with --initrd
>
> ReadMe.c | 3 ++
> mdadm.c | 11 ++++++++++
> mdadm.h | 3 ++
> mdmon.c | 64 +++++++++++++++++++++++++++++++++++++++++++++++++++----------
> util.c | 17 ++++++++++++---
> 5 files changed, 83 insertions(+), 15 deletions(-)
>
Patch set looks reasonable to me, although you missed adding the new
option to the mdadm man page.
Do we have a specific systemd version for which we know that --initrd is
needed, or a version before which it is the wrong thing to do?
Ditto for dracut, do we know what the minimum version of dracut is
before it will use the --initrd parameter on its mdadm calls?
--
Doug Ledford <dledford@xxxxxxxxxx>
GPG KeyID: 0E572FDD
http://people.redhat.com/dledford
Attachment:
signature.asc
Description: OpenPGP digital signature
