udev 16, two copies of udevd

udev 16, two copies of udevd

[Jon Smirl]

When I boot I'm getting two copies of udevd running.  If I kill them, then
insert/remove a device I only get one udevd. Maybe a race condition when first
starting udevd?

==Jon Smirl
jonsmirl@yahoo.com

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free web site building tool. Try it!
http://webhosting.yahoo.com/ps/sb/


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Kay Sievers]

On Tue, Feb 03, 2004 at 05:11:17PM -0800, Jon Smirl wrote:
> When I boot I'm getting two copies of udevd running.  If I kill them, then
> insert/remove a device I only get one udevd. Maybe a race condition when first
> starting udevd?

Are you mounting a tmpfs over udev?
The lockfile is placed in there, so it is lost if you mount it over.


Kay


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Jon Smirl]

Yes, I'm mounting a tmpfs. I'm using the tmpfs becuase I've been having trouble
with stale nodes in my udev dir when I crash the kernel.

How is this going to work in final form? Will I need an initial /dev with mknod
for the 4-5 minimal devices needed to boot, and then can I mount a tmpfs over
this to hold udev devices?

--- Kay Sievers <kay.sievers@vrfy.org> wrote:
> On Tue, Feb 03, 2004 at 05:11:17PM -0800, Jon Smirl wrote:
> > When I boot I'm getting two copies of udevd running.  If I kill them, then
> > insert/remove a device I only get one udevd. Maybe a race condition when
> first
> > starting udevd?
> 
> Are you mounting a tmpfs over udev?
> The lockfile is placed in there, so it is lost if you mount it over.
> 
> 
> Kay
> 
> 
> -------------------------------------------------------
> The SF.Net email is sponsored by EclipseCon 2004
> Premiere Conference on Open Tools Development and Integration
> See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
> http://www.eclipsecon.org/osdn
> _______________________________________________
> Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
> Linux-hotplug-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel


==Jon Smirl
jonsmirl@yahoo.com

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free web site building tool. Try it!
http://webhosting.yahoo.com/ps/sb/


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Greg KH]

On Tue, Feb 03, 2004 at 05:30:24PM -0800, Jon Smirl wrote:
> Yes, I'm mounting a tmpfs. I'm using the tmpfs becuase I've been having trouble
> with stale nodes in my udev dir when I crash the kernel.
> 
> How is this going to work in final form? Will I need an initial /dev with mknod
> for the 4-5 minimal devices needed to boot, and then can I mount a tmpfs over
> this to hold udev devices?

I don't know, what do you think it should look like?

:)

greg k-h


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Kay Sievers]

On Thu, Feb 05, 2004 at 09:25:05AM -0800, Greg KH wrote:
> On Tue, Feb 03, 2004 at 05:30:24PM -0800, Jon Smirl wrote:
> > Yes, I'm mounting a tmpfs. I'm using the tmpfs becuase I've been having trouble
> > with stale nodes in my udev dir when I crash the kernel.
> > 
> > How is this going to work in final form? Will I need an initial /dev with mknod
> > for the 4-5 minimal devices needed to boot, and then can I mount a tmpfs over
> > this to hold udev devices?
> 
> I don't know, what do you think it should look like?

Jon, please try the current bk tree. We switched away from the socket file,
that gets missing with the tmpfs mount.
It should work now, cause we can't listen twice on the same socket in
abstract namespace.

Greg, it seems we can remove the lock file code completely.
Anyone willing to try? Just comment out the one_and_only() and look if
you can produce any situation with two udevd's running. It works for me
nice without it, also on tmpfs.


thanks,
Kay


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Jon Smirl]

I just tried current bk. I still get two copies of udevd running.

--- Kay Sievers <kay.sievers@vrfy.org> wrote:
> On Thu, Feb 05, 2004 at 09:25:05AM -0800, Greg KH wrote:
> > On Tue, Feb 03, 2004 at 05:30:24PM -0800, Jon Smirl wrote:
> > > Yes, I'm mounting a tmpfs. I'm using the tmpfs becuase I've been having
> trouble
> > > with stale nodes in my udev dir when I crash the kernel.
> > > 
> > > How is this going to work in final form? Will I need an initial /dev with
> mknod
> > > for the 4-5 minimal devices needed to boot, and then can I mount a tmpfs
> over
> > > this to hold udev devices?
> > 
> > I don't know, what do you think it should look like?
> 
> Jon, please try the current bk tree. We switched away from the socket file,
> that gets missing with the tmpfs mount.
> It should work now, cause we can't listen twice on the same socket in
> abstract namespace.
> 
> Greg, it seems we can remove the lock file code completely.
> Anyone willing to try? Just comment out the one_and_only() and look if
> you can produce any situation with two udevd's running. It works for me
> nice without it, also on tmpfs.
> 
> 
> thanks,
> Kay


==Jon Smirl
jonsmirl@yahoo.com

__________________________________
Do you Yahoo!?
Yahoo! Finance: Get your refund fast by filing online.
http://taxes.yahoo.com/filing.html


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Jon Smirl]

--- Jon Smirl <jonsmirl@yahoo.com> wrote:
> I just tried current bk. I still get two copies of udevd running.
> 
Not true. My pull went into PENDING so I ran the old code.

Fixing that problem now I only have one copy of udevd running.

==Jon Smirl
jonsmirl@yahoo.com

__________________________________
Do you Yahoo!?
Yahoo! Finance: Get your refund fast by filing online.
http://taxes.yahoo.com/filing.html


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Kay Sievers]

On Thu, Feb 05, 2004 at 03:56:37PM -0800, Jon Smirl wrote:
> I just tried current bk. I still get two copies of udevd running.

What says "lsof |grep udevd"?

Kay

> --- Kay Sievers <kay.sievers@vrfy.org> wrote:
> > Jon, please try the current bk tree. We switched away from the socket file,
> > that gets missing with the tmpfs mount.
> > It should work now, cause we can't listen twice on the same socket in
> > abstract namespace.


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Kay Sievers]

On Thu, Feb 05, 2004 at 04:01:39PM -0800, Jon Smirl wrote:
> --- Jon Smirl <jonsmirl@yahoo.com> wrote:
> > I just tried current bk. I still get two copies of udevd running.
> > 
> Not true. My pull went into PENDING so I ran the old code.
> 
> Fixing that problem now I only have one copy of udevd running.

Ahh, nice. I feel better now :)
Looks promising, next thing to kill is the lock file.

Kay


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Greg KH]

On Fri, Feb 06, 2004 at 12:49:46AM +0100, Kay Sievers wrote:
> 
> Jon, please try the current bk tree. We switched away from the socket file,
> that gets missing with the tmpfs mount.
> It should work now, cause we can't listen twice on the same socket in
> abstract namespace.
> 
> Greg, it seems we can remove the lock file code completely.

I agree, it should be safe to remove that now, which is a very good
thing.

Care to send a patch?  Or I can do it.

thanks,

greg k-h


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: udev 16, two copies of udevd

[Kay Sievers]

[-- Attachment #1: Type: text/plain, Size: 559 bytes --]

On Thu, Feb 05, 2004 at 04:32:16PM -0800, Greg KH wrote:
> On Fri, Feb 06, 2004 at 12:49:46AM +0100, Kay Sievers wrote:
> > 
> > Jon, please try the current bk tree. We switched away from the socket file,
> > that gets missing with the tmpfs mount.
> > It should work now, cause we can't listen twice on the same socket in
> > abstract namespace.
> > 
> > Greg, it seems we can remove the lock file code completely.
> 
> I agree, it should be safe to remove that now, which is a very good
> thing.
> 
> Care to send a patch?  Or I can do it.

Here we go...



[-- Attachment #2: 01-kill-lockfile.patch --]
[-- Type: text/plain, Size: 2121 bytes --]

===== Makefile 1.103 vs edited =====
--- 1.103/Makefile	Thu Feb  5 02:41:19 2004
+++ edited/Makefile	Fri Feb  6 01:39:07 2004
@@ -241,7 +241,6 @@
 	@echo \#define UDEV_PERMISSION_FILE	\"$(configdir)\udev.permissions\" >> $@
 	@echo \#define UDEV_BIN		\"$(DESTDIR)$(sbindir)/udev\" >> $@
 	@echo \#define UDEVD_BIN	\"$(DESTDIR)$(sbindir)/udevd\" >> $@
-	@echo \#define UDEVD_LOCK	\"$(udevdir)/\.udevd.lock\" >> $@
 
 # config files automatically generated
 GEN_CONFIGS =	$(LOCAL_CFG_DIR)/udev.conf
===== udev-remove.c 1.17 vs edited =====
--- 1.17/udev-remove.c	Thu Jan 15 16:22:25 2004
+++ edited/udev-remove.c	Fri Feb  6 01:24:11 2004
@@ -34,7 +34,6 @@
 #include "logging.h"
 #include "namedev.h"
 #include "udevdb.h"
-#include "libsysfs/libsysfs.h"
 
 static int delete_path(char *path)
 {
===== udevd.c 1.13 vs edited =====
--- 1.13/udevd.c	Thu Feb  5 10:32:51 2004
+++ edited/udevd.c	Fri Feb  6 01:37:52 2004
@@ -21,7 +21,6 @@
 
 #include <pthread.h>
 #include <stddef.h>
-#include <sys/types.h>
 #include <sys/wait.h>
 #include <signal.h>
 #include <unistd.h>
@@ -30,7 +29,6 @@
 #include <stdlib.h>
 #include <string.h>
 #include <time.h>
-#include <fcntl.h>
 #include <sys/types.h>
 #include <sys/socket.h>
 #include <sys/un.h>
@@ -324,7 +322,6 @@
 	switch (signum) {
 		case SIGINT:
 		case SIGTERM:
-			unlink(UDEVD_LOCK);
 			exit(20 + signum);
 			break;
 		default:
@@ -332,28 +329,6 @@
 	}
 }
 
-static int one_and_only(void)
-{
-	char string[50];
-	int lock_file;
-
-	lock_file = open(UDEVD_LOCK, O_RDWR | O_CREAT, 0x640);
-	if (lock_file < 0)
-		return -1;
-
-	/* see if we can lock */
-	if (lockf(lock_file, F_TLOCK, 0) < 0) {
-		dbg("file is already locked, exit");
-		close(lock_file);
-		return -1;
-	}
-
-	snprintf(string, sizeof(string), "%d\n", getpid());
-	write(lock_file, string, strlen(string));
-
-	return 0;
-}
-
 int main(int argc, char *argv[])
 {
 	int ssock;
@@ -368,10 +343,6 @@
 	int retval;
 
 	init_logging("udevd");
-
-	/* only let one version of the daemon run at any one time */
-	if (one_and_only() != 0)
-		exit(0);
 
 	signal(SIGINT, sig_handler);
 	signal(SIGTERM, sig_handler);

Re: udev 16, two copies of udevd

[Greg KH]

On Fri, Feb 06, 2004 at 01:47:02AM +0100, Kay Sievers wrote:
> On Thu, Feb 05, 2004 at 04:32:16PM -0800, Greg KH wrote:
> > On Fri, Feb 06, 2004 at 12:49:46AM +0100, Kay Sievers wrote:
> > > 
> > > Jon, please try the current bk tree. We switched away from the socket file,
> > > that gets missing with the tmpfs mount.
> > > It should work now, cause we can't listen twice on the same socket in
> > > abstract namespace.
> > > 
> > > Greg, it seems we can remove the lock file code completely.
> > 
> > I agree, it should be safe to remove that now, which is a very good
> > thing.
> > 
> > Care to send a patch?  Or I can do it.
> 
> Here we go...

Applied, thanks.

greg k-h


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel