[PATCH] xenconsole: add file lock to xenconsole

[PATCH] xenconsole: add file lock to xenconsole

[Yufang Zhang]

This patch add a file lock to xenconsole for each console id, so
that only one console could be attached to a guest at a time.
Otherwise, consoles would get stuck and print strange outputs.

Signed-off-by: Yufang Zhang <yuzhang@redhat.com>
---
 tools/console/client/main.c |   31 +++++++++++++++++++++++++++++++
 1 files changed, 31 insertions(+), 0 deletions(-)

diff --git a/tools/console/client/main.c b/tools/console/client/main.c
index df3636f..d7461a8 100644
--- a/tools/console/client/main.c
+++ b/tools/console/client/main.c
@@ -46,6 +46,28 @@
 
 static volatile sig_atomic_t received_signal = 0;
 
+static int console_locked(const char *file)
+{
+        int fd;
+
+        fd = open(file, O_RDWR | O_CREAT, S_IRUSR|S_IWUSR);
+        if (fd == -1) {
+                fprintf(stderr,"can't open lock file: %s\n", file);
+                exit(1);
+        }
+
+        if (lockf(fd, F_TLOCK, 0) == -1) {
+            if ( errno == EACCES || errno == EAGAIN ) {
+                close(fd);
+                return (1);
+            }
+            fprintf(stderr,"can't lock file: %s\n", file);
+            exit(1);
+        }
+
+        return (0);
+}
+
 static void sighandler(int signum)
 {
 	received_signal = 1;
@@ -280,6 +302,7 @@ int main(int argc, char **argv)
 	int spty, xsfd;
 	struct xs_handle *xs;
 	char *end;
+        char buf[100];
 	console_type type = CONSOLE_INVAL;
 
 	while((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
@@ -368,6 +391,13 @@ int main(int argc, char **argv)
 		exit(EINVAL);
 	}
 
+        /* Make sure only one console is attached to guest */
+        sprintf(buf, "/tmp/xenconsole-%d-%d", domid, num);
+        if (console_locked(buf) == 1) {
+                fprintf(stderr, "Another console has already been attached to guest\n");
+                exit(EINVAL);
+        }
+
 	/* Set a watch on this domain's console pty */
 	if (!xs_watch(xs, path, ""))
 		err(errno, "Can't set watch for console pty");
@@ -387,6 +417,7 @@ int main(int argc, char **argv)
 	console_loop(spty, xs, path);
 	restore_term(STDIN_FILENO, &attr);
 
+        unlink(buf);
 	free(path);
 	free(dom_path);
 	return 0;
-- 
1.7.4.4

Re: [PATCH] xenconsole: add file lock to xenconsole

[Ian Jackson]

Yufang Zhang writes ("[Xen-devel] [PATCH] xenconsole: add file lock to xenconsole"):
> This patch add a file lock to xenconsole for each console id, so
> that only one console could be attached to a guest at a time.
> Otherwise, consoles would get stuck and print strange outputs.

If only we had a better console protocol, it would be possible to
attach multiple times.  Oh well.  In the meantime your semantic change
is sensible.

However:

> +static int console_locked(const char *file)
> +{
> +        int fd;
> +

You need to use the same indent level and coding style as the
surrounding code.

> +        sprintf(buf, "/tmp/xenconsole-%d-%d", domid, num);

The lockfile should be in /var/run/xen.  You should use snprintf.

What arrangements do you plan to make for cleaning up stale lockfiles
(which I think will be left behind if the xenconsole client crashes) ?
Are we just going to rely on them being removed at reboot and apart
from that let them accumulate a bit ?

Ian.

Re: [PATCH] xenconsole: add file lock to xenconsole

[Yufang Zhang]

On 05/25/2011 01:09 AM, Ian Jackson wrote:
> Yufang Zhang writes ("[Xen-devel] [PATCH] xenconsole: add file lock to xenconsole"):
>> This patch add a file lock to xenconsole for each console id, so
>> that only one console could be attached to a guest at a time.
>> Otherwise, consoles would get stuck and print strange outputs.
> If only we had a better console protocol, it would be possible to
> attach multiple times.  Oh well.  In the meantime your semantic change
> is sensible.
>
> However:
>
>> +static int console_locked(const char *file)
>> +{
>> +        int fd;
>> +
> You need to use the same indent level and coding style as the
> surrounding code.
>
>> +        sprintf(buf, "/tmp/xenconsole-%d-%d", domid, num);
> The lockfile should be in /var/run/xen.  You should use snprintf.
>

Thanks Ian. Version2 patch has been sent, please review.

> What arrangements do you plan to make for cleaning up stale lockfiles
> (which I think will be left behind if the xenconsole client crashes) ?
> Are we just going to rely on them being removed at reboot and apart
> from that let them accumulate a bit ?
>
> Ian.
Considering the lock files are just blank and can be reused, leaving 
them behind is not that bad?

Yufang

Re: [PATCH] xenconsole: add file lock to xenconsole

[Ian Campbell]

On Fri, 2011-05-27 at 04:20 +0100, Yufang Zhang wrote:
> On 05/25/2011 01:09 AM, Ian Jackson wrote:
> > Yufang Zhang writes ("[Xen-devel] [PATCH] xenconsole: add file lock to xenconsole"):
> >> This patch add a file lock to xenconsole for each console id, so
> >> that only one console could be attached to a guest at a time.
> >> Otherwise, consoles would get stuck and print strange outputs.
> > If only we had a better console protocol, it would be possible to
> > attach multiple times.  Oh well.  In the meantime your semantic change
> > is sensible.
> >
> > However:
> >
> >> +static int console_locked(const char *file)
> >> +{
> >> +        int fd;
> >> +
> > You need to use the same indent level and coding style as the
> > surrounding code.
> >
> >> +        sprintf(buf, "/tmp/xenconsole-%d-%d", domid, num);
> > The lockfile should be in /var/run/xen.  You should use snprintf.
> >
> 
> Thanks Ian. Version2 patch has been sent, please review.
> 
> > What arrangements do you plan to make for cleaning up stale lockfiles
> > (which I think will be left behind if the xenconsole client crashes) ?
> > Are we just going to rely on them being removed at reboot and apart
> > from that let them accumulate a bit ?
> >
> Considering the lock files are just blank and can be reused, leaving 
> them behind is not that bad?

Well, even an empty file has a (small) overhead. It'd be nice to avoid
if we can.

Is it permissible to use lockf on a tty device directly?

http://pubs.opengroup.org/onlinepubs/009695399/functions/lockf.html says
        Record locking with lockf() shall be supported for regular files
        and may be supported for other files.
so I guess it's a question of what the OSes we support actually
implement.

Ian.