From: Neil Brown <neilb@suse.de>
To: Luca Berra <bluca@comedia.it>
Cc: "linux-raid@vger.kernel.org" <linux-raid@vger.kernel.org>
Subject: Re: [mdadm PATCH 8/9] mdmon: rework startup and takeover logic
Date: Wed, 3 Mar 2010 12:52:39 +1100 [thread overview]
Message-ID: <20100303125239.52694981@notabene.brown> (raw)
In-Reply-To: <20100228144545.GG24781@maude.comedia.it>
On Sun, 28 Feb 2010 15:45:45 +0100
Luca Berra <bluca@comedia.it> wrote:
> rework startup and takeover logic in order to make it more robust
With the other things I dropped, this no longer applied, and the I didn't
feel up to picking it apart to see what you were really trying to do.
Again, if you feel something here is still needed, please explain.
Thanks,
NeilBrown
>
> Signed-off-by: Luca Berra <bluca@comedia.it>
> ---
> mdmon.c | 45 +++++++++++++++++++++------------------------
> 1 files changed, 21 insertions(+), 24 deletions(-)
>
> diff --git a/mdmon.c b/mdmon.c
> index eef4bfa..b823a8c 100644
> --- a/mdmon.c
> +++ b/mdmon.c
> @@ -71,7 +71,7 @@ int mon_tid, mgr_tid;
>
> int sigterm;
>
> -static char *pid_dir = VAR_RUN;
> +static char *pid_dir = NULL;
>
> int run_child(void *v)
> {
> @@ -189,9 +189,6 @@ void remove_pidfile(char *devname)
> unlink(buf);
> sprintf(buf, "%s/%s.sock", pid_dir, devname);
> unlink(buf);
> - if (strcmp(pid_dir, ALT_RUN) == 0)
> - /* try to clean up when we are finished with this dir */
> - rmdir(pid_dir);
> }
>
> static int make_control_sock(char *devname)
> @@ -443,26 +440,20 @@ static int mdmon(char *devname, int devnum, int must_fork, int takeover)
> sigaction(SIGPIPE, &act, NULL);
>
> if (takeover) {
> - victim = mdmon_pid(container->devnum, pid_dir);
> - if (victim < 0) {
> - pid_dir = ALT_RUN;
> - victim = mdmon_pid(container->devnum, pid_dir);
> + /* Check the parent process by socket, not pid, which could be
> + * stale, then get the pid from the same pid_dir */
> + char * victim_pid_dir = VAR_RUN;
> + victim_sock = connect_monitor(container->devname, victim_pid_dir);
> + if (victim_sock < 0) {
> + victim_pid_dir = ALT_RUN;
> + victim_sock = connect_monitor(container->devname, victim_pid_dir);
> }
> - if (victim >= 0)
> - victim_sock = connect_monitor(container->devname, pid_dir);
> + if (victim_sock >= 0)
> + victim = mdmon_pid(container->devnum, victim_pid_dir);
> }
>
> ignore = chdir("/");
> if (victim < 0) {
> - pid_dir = ALT_RUN;
> - if (ping_monitor(container->devname) == 0) {
> - fprintf(stderr, "mdmon: %s already managed\n",
> - container->devname);
> - if (!takeover)
> - fprintf(stderr, "\trun mdmon --takeover instead\n");
> - exit(3);
> - }
> - pid_dir = VAR_RUN;
> if (ping_monitor(container->devname) == 0) {
> fprintf(stderr, "mdmon: %s already managed\n",
> container->devname);
> @@ -479,14 +470,20 @@ static int mdmon(char *devname, int devnum, int must_fork, int takeover)
> close(mdfd);
>
> /* Ok, this is close enough. We can say goodbye to our parent now.
> + * take care to remove all possible pid files and sockets.
> */
> - if (victim > 0)
> + if (mkdir(ALT_RUN, 0600) >= 0 || (errno == EEXIST && access(ALT_RUN, W_OK) >= 0)) {
> + pid_dir = ALT_RUN;
> remove_pidfile(devname);
> - if (mkdir(VAR_RUN, 0600) >= 0 || (errno == EEXIST && access(VAR_RUN, W_OK) >= 0))
> + }
> + if (mkdir(VAR_RUN, 0600) >= 0 || (errno == EEXIST && access(VAR_RUN, W_OK) >= 0)) {
> + /* try to clean up when we are finished with ALT_RUN dir */
> + if (pid_dir != NULL)
> + rmdir(pid_dir);
> pid_dir = VAR_RUN;
> - else if (mkdir(ALT_RUN, 0600) >= 0 || (errno == EEXIST && access(ALT_RUN, W_OK) >= 0))
> - pid_dir = ALT_RUN;
> - else {
> + remove_pidfile(devname);
> + }
> + if (pid_dir == NULL) {
> fprintf(stderr, "mdmon: Neither %s nor %s are writable\n"
> " cannot create .pid or .sock files. Aborting\n",
> VAR_RUN, ALT_RUN);
next prev parent reply other threads:[~2010-03-03 1:52 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-02-27 15:05 [mdadm PATCH 0/2] *** mdmon fixes *** Luca Berra
2010-02-27 15:07 ` [mdadm PATCH 1/2] fix mdmon option parsing Luca Berra
2010-02-27 15:21 ` Luca Berra
2010-03-02 5:40 ` Neil Brown
2010-03-02 7:42 ` Luca Berra
2010-02-27 15:07 ` [mdadm PATCH 2/2] fix mdmon takeover Luca Berra
2010-03-03 0:52 ` Neil Brown
2010-02-28 14:41 ` [mdadm PATCH 0/2] *** mdmon fixes *** Luca Berra
2010-02-28 14:42 ` [mdadm PATCH 3/9] add missing newline to mdmon usage message Luca Berra
2010-02-28 14:42 ` [mdadm PATCH 4/9] mdmon: check select a writable pid_dir Luca Berra
2010-03-03 1:07 ` Neil Brown
2010-02-28 14:43 ` [mdadm PATCH 5/9] mdmon: mdmon_pid should return pid from either dir Luca Berra
2010-03-03 1:50 ` Neil Brown
2010-03-03 6:44 ` Luca Berra
2010-03-03 6:55 ` Neil Brown
2010-03-03 7:09 ` Luca Berra
2010-02-28 14:44 ` [mdadm PATCH 6/9] mdmon: connect_monitor should use socket " Luca Berra
2010-03-03 1:51 ` Neil Brown
2010-02-28 14:45 ` [mdadm PATCH 7/9] mdmon: move pid_dir to mdmon.c Luca Berra
2010-02-28 14:45 ` [mdadm PATCH 8/9] mdmon: rework startup and takeover logic Luca Berra
2010-03-03 1:52 ` Neil Brown [this message]
2010-03-03 6:48 ` Luca Berra
2010-02-28 14:46 ` [mdadm PATCH 9/9] allow redefinition of VAR_RUN Luca Berra
2010-03-03 1:53 ` Neil Brown
2010-02-28 15:41 ` [mdadm PATCH 0/2] *** mdmon fixes *** Bill Davidsen
2010-02-28 17:37 ` Luca Berra
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20100303125239.52694981@notabene.brown \
--to=neilb@suse.de \
--cc=bluca@comedia.it \
--cc=linux-raid@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).