On Thu, 03 Nov 2011 17:55:33 +0100 Adam Kwolek <adam.kwolek@xxxxxxxxx> wrote:
> When array is not clean dismounted directory /dev/.mdadm is not cleaned up.
> On array re-assembly read pid is not valid and it is not possible
> to connect to monitor. This causes mdmon to exit and array remains
> not monitored.
> Problem is introduced by fix:
> mdmon(): Error out if failing to connect to victim monitor
> 819c158866f466075a1c719f0dc496deb2fb3814
>
> This is critical for container reshape when mdmon is should finish reshape.
> when reshape is not finished, array is reshaped again by mdadm.
>
> Signed-off-by: Adam Kwolek <adam.kwolek@xxxxxxxxx>
> ---
>
> mdmon.c | 15 ++++++++++-----
> 1 files changed, 10 insertions(+), 5 deletions(-)
>
> diff --git a/mdmon.c b/mdmon.c
> index bdcda0e..5ac7cd6 100644
> --- a/mdmon.c
> +++ b/mdmon.c
> @@ -458,11 +458,16 @@ static int mdmon(char *devname, int devnum, int must_fork, int takeover)
>
> victim = mdmon_pid(container->devnum);
> if (victim >= 0) {
> - victim_sock = connect_monitor(container->devname);
> - if (victim_sock < 0) {
> - fprintf(stderr, "mdmon: %s unable to connect monitor\n",
> - container->devname);
> - exit(3);
> + /* It is possible that mdmon that wrote pid file was killed.
> + * check if read pid is valid/mdmon is running
> + */
> + if (mdmon_running(victim)) {
> + victim_sock = connect_monitor(container->devname);
> + if (victim_sock < 0) {
> + fprintf(stderr, "mdmon: %s unable to connect "
> + "monitor\n", container->devname);
> + exit(3);
> + }
> }
> }
>
Thanks for the patch.
I decided to revert the patch that originally caused the problem instead - it
really isn't needed.
I then added a patch to make sure we never use victim_sock when it is -1.
The places were we might have used it we not dangerous at all, but it is
cleaner to check.
Thanks,
NeilBrown
Attachment:
signature.asc
Description: PGP signature
