[patch] Fix naming ethernet devices in udevstart

[patch] Fix naming ethernet devices in udevstart

[Tom Rini]

I noticed somewhat recently that my enet devices weren't being renamed
on boot anymore.  I don't quite know when this got broken (or rather, if
it was supposed to be working.  I swear it worked for me once..), but
the following seems to do it.  In udev_scan_class(), look for not just
%s/%s/dev (which everything with a dev node has), but %s/%s/dev* (both
of my enet devices, sis900 & 3c59x only have device) and if that
exists, pass this along to udev.  This does mean that other things
which udev doesn't know how to handle
(/sys/class/usb_host/usb%d/device) will get passed along, but I don't
know of a bitter solution.  Tested on my Debian/unstable machine.

Signed-off-by: Tom Rini <trini@kernel.crashing.org>

 udevstart.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletion(-)
--- 1.13/udevstart.c	2004-08-04 15:37:55 -07:00
+++ edited/udevstart.c	2004-09-06 20:06:56 -07:00
@@ -220,11 +220,12 @@
 						for (dent3 = readdir(dir3); dent3 != NULL; dent3 = readdir(dir3)) {
 							char filename[MAX_PATHLEN];
 
-							if (strcmp(dent3->d_name, "dev") = 0) {
+							if (strncmp(dent3->d_name, "dev", 3) = 0) {
 								snprintf(filename, MAX_PATHLEN, "/class/%s/%s",
 									 dent->d_name, dent2->d_name);
 								filename[MAX_PATHLEN-1] = '\0';
 								device_list_insert(filename, dent->d_name, &device_list);
+								break;
 							}
 						}
 						closedir(dir3);

-- 
Tom Rini
http://gate.crashing.org/~trini/


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idP47&alloc_id\x10808&op=click
_______________________________________________
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: [patch] Fix naming ethernet devices in udevstart

[Kay Sievers]

On Mon, 2004-09-06 at 20:18 -0700, Tom Rini wrote:
> I noticed somewhat recently that my enet devices weren't being renamed
> on boot anymore.  I don't quite know when this got broken (or rather, if
> it was supposed to be working.  I swear it worked for me once..), but
> the following seems to do it.

I think it never worked in the udevstart case. It worked only with the
hotplug-event-udev, I expect.

> In udev_scan_class(), look for not just
> %s/%s/dev (which everything with a dev node has), but %s/%s/dev* (both
> of my enet devices, sis900 & 3c59x only have device) and if that
> exists, pass this along to udev.

Yeah, network devices don't have a devnode and therefore no "dev", but
they are all in /sys/class/net/. We may just test if we are there
instead of the "device" match.
Btw: Greg, what's the reason not to use libsysfs for the directory walk?

> This does mean that other things
> which udev doesn't know how to handle
> (/sys/class/usb_host/usb%d/device) will get passed along, but I don't
> know of a bitter solution.  Tested on my Debian/unstable machine.
             ^
Yeah, nice typo :)

Best,
Kay



-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idP47&alloc_id\x10808&op=click
_______________________________________________
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: [patch] Fix naming ethernet devices in udevstart

[Kay Sievers]

[-- Attachment #1: Type: text/plain, Size: 1391 bytes --]

On Tue, Sep 07, 2004 at 12:46:43PM +0200, Kay Sievers wrote:
> On Mon, 2004-09-06 at 20:18 -0700, Tom Rini wrote:
> > I noticed somewhat recently that my enet devices weren't being renamed
> > on boot anymore.  I don't quite know when this got broken (or rather, if
> > it was supposed to be working.  I swear it worked for me once..), but
> > the following seems to do it.
> 
> I think it never worked in the udevstart case. It worked only with the
> hotplug-event-udev, I expect.
> 
> > In udev_scan_class(), look for not just
> > %s/%s/dev (which everything with a dev node has), but %s/%s/dev* (both
> > of my enet devices, sis900 & 3c59x only have device) and if that
> > exists, pass this along to udev.
> 
> Yeah, network devices don't have a devnode and therefore no "dev", but
> they are all in /sys/class/net/. We may just test if we are there
> instead of the "device" match.

How about something like this. It adds all the net devices without
looking at the attributes and keeps the remaining logic like it is.

It also removes certain levels of indirection and much simplifies the
udevstart process. We surely don't need to open and close the udevdb
for every node while iterating over the list. (We are about 5% faster on
my box)

It's not well tested, so it would be nice if someone can have a look
at it, before a broken udevstart renders any system unbootable.

Thanks,
Kay

[-- Attachment #2: udevstart-cleanup-01.patch --]
[-- Type: text/plain, Size: 7535 bytes --]

===== udev.c 1.60 vs edited =====
--- 1.60/udev.c	2004-08-10 17:11:56 +02:00
+++ edited/udev.c	2004-09-07 17:48:42 +02:00
@@ -40,9 +40,6 @@
 char **main_argv;
 char **main_envp;
 
-/* local variables */
-static int is_udevstart;
-
 #ifdef LOG
 unsigned char logname[LOGNAME_SIZE];
 void log_message(int level, const char *format, ...)
@@ -79,36 +76,72 @@ static char *subsystem_blacklist[] = {
 	""
 };
 
-int __udev_hotplug(char *action, const char *devpath, const char *subsystem)
+int main(int argc, char *argv[], char *envp[])
 {
-	int retval = -EINVAL;
-	int i;
+	main_argv = argv;
+	main_envp = envp;
 	struct sigaction act;
-	const int nofake = 0;
+	char *action;
+	char *devpath = "";
+	char *subsystem = "";
+	int i;
+	int retval = -EINVAL;
+	enum {
+		ADD,
+		REMOVE,
+		UDEVSTART,
+	} act_type;
 
-	dbg("looking at '%s'", devpath);
+	dbg("version %s", UDEV_VERSION);
 
-	/* we only care about class devices and block stuff */
-	if (!strstr(devpath, "class") &&
-	    !strstr(devpath, "block")) {
-		dbg("not a block or class device");
-		goto exit;
-	}
+	/* initialize our configuration */
+	udev_init_config();
 
-	i = 0;
-	while (subsystem_blacklist[i][0] != '\0') {
-		if (strcmp(subsystem, subsystem_blacklist[i]) == 0) {
-			dbg("don't care about '%s' devices", subsystem);
+	if (strstr(argv[0], "udevstart")) {
+		act_type = UDEVSTART;
+	} else {
+		action = get_action();
+		if (!action) {
+			dbg("no action?");
+			goto exit;
+		}
+		if (strcmp(action, "add") == 0) {
+			act_type = ADD;
+		} else if (strcmp(action, "remove") == 0) {
+			act_type = REMOVE;
+		} else {
+			dbg("unknown action '%s'", action);
 			goto exit;
 		}
-		i++;
-	}
 
-	/* initialize udev database */
-	retval = udevdb_init(UDEVDB_DEFAULT);
-	if (retval != 0) {
-		dbg("unable to initialize database");
-		goto exit;
+		devpath = get_devpath();
+		if (!devpath) {
+			dbg("no devpath?");
+			goto exit;
+		}
+		dbg("looking at '%s'", devpath);
+
+		/* we only care about class devices and block stuff */
+		if (strstr(devpath, "class") && strstr(devpath, "block")) {
+			dbg("not a block or class device");
+			goto exit;
+		}
+
+		subsystem = get_subsystem(main_argv[1]);
+		if (!subsystem) {
+			dbg("no subsystem?");
+			goto exit;
+		}
+
+		/* skip blacklisted subsystems */
+		i = 0;
+		while (subsystem_blacklist[i][0] != '\0') {
+			if (strcmp(subsystem, subsystem_blacklist[i]) == 0) {
+				dbg("don't care about '%s' devices", subsystem);
+				goto exit;
+			}
+			i++;
+		}
 	}
 
 	/* set signal handlers */
@@ -118,76 +151,31 @@ int __udev_hotplug(char *action, const c
 	sigaction(SIGINT, &act, NULL);
 	sigaction(SIGTERM, &act, NULL);
 
-	if (strcmp(action, "add") == 0) {
-		/* Already done. */
-		if (!is_udevstart)
-			namedev_init();
-		retval = udev_add_device(devpath, subsystem, nofake);
-		goto action_done;
+	/* initialize udev database */
+	if (udevdb_init(UDEVDB_DEFAULT) != 0) {
+		dbg("unable to initialize database");
+		goto exit;
 	}
 
-	if (strcmp(action, "remove") == 0) {
+	switch(act_type) {
+	case UDEVSTART:
+		dbg("udevstart");
+		namedev_init();
+		udev_sleep = 0;
+		retval = udev_start();
+		break;
+	case ADD:
+		dbg("udev add");
+		namedev_init();
+		retval = udev_add_device(devpath, subsystem, NOFAKE);
+		break;
+	case REMOVE:
+		dbg("udev remove");
 		retval = udev_remove_device(devpath, subsystem);
-		goto action_done;
 	}
 
-	dbg("unknown action '%s'", action);
-	retval = -EINVAL;
-
-action_done:
 	udevdb_exit();
 
 exit:
 	return retval;
-}
-
-static int udev_hotplug(void)
-{
-	char *action;
-	char *devpath;
-	char *subsystem;
-
-	action = get_action();
-	if (!action) {
-		dbg("no action?");
-		return -EINVAL;
-	}
-
-	devpath = get_devpath();
-	if (!devpath) {
-		dbg("no devpath?");
-		return -EINVAL;
-	}
-
-	/* skip blacklisted subsystems */
-	subsystem = get_subsystem(main_argv[1]);
-	if (!subsystem) {
-		dbg("no subsystem?");
-		return -EINVAL;
-	}
-
-	return __udev_hotplug(action, devpath, subsystem);
-}
-
-int main(int argc, char *argv[], char *envp[])
-{
-	main_argv = argv;
-	main_envp = envp;
-
-	if (strstr(argv[0], "udevstart")) {
-		/* Setup env variables. */
-		setenv("UDEV_NO_SLEEP", "1", 1);
-		is_udevstart = 1;
-	}
-
-	/* initialize our configuration */
-	udev_init_config();
-
-	dbg("version %s", UDEV_VERSION);
-
-	if (is_udevstart) {
-		namedev_init();
-		return udev_start();
-	} else
-		return udev_hotplug();
 }
===== udev.h 1.58 vs edited =====
--- 1.58/udev.h	2004-08-05 00:43:58 +02:00
+++ edited/udev.h	2004-09-07 13:16:50 +02:00
@@ -37,6 +37,9 @@
 #define DEVPATH_SIZE			255
 #define SUBSYSTEM_SIZE			30
 
+#define FAKE				1
+#define NOFAKE				0
+
 /* length of public data */
 #define UDEVICE_LEN (offsetof(struct udevice, bus_id))
 
===== udevstart.c 1.14 vs edited =====
--- 1.14/udevstart.c	2004-08-10 17:11:37 +02:00
+++ edited/udevstart.c	2004-09-07 18:34:14 +02:00
@@ -70,18 +70,10 @@ static int device_list_insert(char *path
 	strfieldcpy(new_device->path, path);
 	strfieldcpy(new_device->subsys, subsystem);
 	list_add_tail(&new_device->list, &loop_device->list);
+	dbg("add '%s' from subsys '%s'", new_device->path, new_device->subsys);
 	return 0;
 }
 
-static void udev_exec(const char *path, const char* subsystem)
-{
-	/* Now call __udev_hotplug(). */
-	if (__udev_hotplug("add", path, subsystem)) {
-		dbg("Calling of udev_hotplug failed");
-		exit(1);
-	}
-}
-
 /* list of devices that we should run last due to any one of a number of reasons */
 static char *last_list[] = {
 	"/block/dm",	/* on here because dm wants to have the block devices around before it */
@@ -106,14 +98,14 @@ static void exec_list(struct list_head *
 		if (found)
 			continue;
 
-		udev_exec(loop_device->path, loop_device->subsys);
+		udev_add_device(loop_device->path, loop_device->subsys, NOFAKE);
 		list_del(&loop_device->list);
 		free(loop_device);
 	}
 
 	/* handle the rest of the devices left over, if any */
 	list_for_each_entry_safe(loop_device, tmp_device, device_list, list) {
-		udev_exec(loop_device->path, loop_device->subsys);
+		udev_add_device(loop_device->path, loop_device->subsys, NOFAKE);
 		list_del(&loop_device->list);
 		free(loop_device);
 	}
@@ -209,19 +201,27 @@ static void udev_scan_class(void)
 					    (strcmp(dent2->d_name, "..") == 0))
 						continue;
 
+					/* pass the net class as it is */
+					if (strcmp(dent->d_name, "net") == 0) {
+						snprintf(dirname2, MAX_PATHLEN, "/class/net/%s", dent2->d_name);
+						device_list_insert(dirname2, "net", &device_list);
+						continue;
+					}
+
 					snprintf(dirname2, MAX_PATHLEN, "%s/%s", dirname, dent2->d_name);
 					dirname2[MAX_PATHLEN-1] = '\0';
-
 					dir3 = opendir(dirname2);
 					if (dir3 != NULL) {
 						for (dent3 = readdir(dir3); dent3 != NULL; dent3 = readdir(dir3)) {
 							char filename[MAX_PATHLEN];
 
+							/* pass devices with a "dev" file */
 							if (strcmp(dent3->d_name, "dev") == 0) {
 								snprintf(filename, MAX_PATHLEN, "/class/%s/%s",
 									 dent->d_name, dent2->d_name);
 								filename[MAX_PATHLEN-1] = '\0';
 								device_list_insert(filename, dent->d_name, &device_list);
+								break;
 							}
 						}
 						closedir(dir3);
===== udevtest.c 1.9 vs edited =====
--- 1.9/udevtest.c	2004-04-01 23:46:07 +02:00
+++ edited/udevtest.c	2004-09-07 13:17:19 +02:00
@@ -102,7 +102,7 @@ int main(int argc, char *argv[], char *e
 		subsystem = argv[2];
 
 	/* simulate node creation with fake flag */
-	udev_add_device(devpath, subsystem, fake);
+	udev_add_device(devpath, subsystem, FAKE);
 
 exit:
 	return 0;

Re: [patch] Fix naming ethernet devices in udevstart

[Tom Rini]

On Tue, Sep 07, 2004 at 06:56:49PM +0200, Kay Sievers wrote:

> On Tue, Sep 07, 2004 at 12:46:43PM +0200, Kay Sievers wrote:
> > On Mon, 2004-09-06 at 20:18 -0700, Tom Rini wrote:
> > > I noticed somewhat recently that my enet devices weren't being renamed
> > > on boot anymore.  I don't quite know when this got broken (or rather, if
> > > it was supposed to be working.  I swear it worked for me once..), but
> > > the following seems to do it.
> > 
> > I think it never worked in the udevstart case. It worked only with the
> > hotplug-event-udev, I expect.
> > 
> > > In udev_scan_class(), look for not just
> > > %s/%s/dev (which everything with a dev node has), but %s/%s/dev* (both
> > > of my enet devices, sis900 & 3c59x only have device) and if that
> > > exists, pass this along to udev.
> > 
> > Yeah, network devices don't have a devnode and therefore no "dev", but
> > they are all in /sys/class/net/. We may just test if we are there
> > instead of the "device" match.
> 
> How about something like this. It adds all the net devices without
> looking at the attributes and keeps the remaining logic like it is.

I'll give it a shot on my ppc_8xx board and if that's OK, I'll figure
out how to give it a shot on gentoo and then my workstation
(debian/unstable).

-- 
Tom Rini
http://gate.crashing.org/~trini/


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idP47&alloc_id\x10808&op=click
_______________________________________________
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: [patch] Fix naming ethernet devices in udevstart

[Tom Rini]

On Tue, Sep 07, 2004 at 10:23:50AM -0700, Tom Rini wrote:
> On Tue, Sep 07, 2004 at 06:56:49PM +0200, Kay Sievers wrote:
> > On Tue, Sep 07, 2004 at 12:46:43PM +0200, Kay Sievers wrote:
> > > On Mon, 2004-09-06 at 20:18 -0700, Tom Rini wrote:
> > > > I noticed somewhat recently that my enet devices weren't being renamed
> > > > on boot anymore.  I don't quite know when this got broken (or rather, if
> > > > it was supposed to be working.  I swear it worked for me once..), but
> > > > the following seems to do it.
> > > 
> > > I think it never worked in the udevstart case. It worked only with the
> > > hotplug-event-udev, I expect.
> > > 
> > > > In udev_scan_class(), look for not just
> > > > %s/%s/dev (which everything with a dev node has), but %s/%s/dev* (both
> > > > of my enet devices, sis900 & 3c59x only have device) and if that
> > > > exists, pass this along to udev.
> > > 
> > > Yeah, network devices don't have a devnode and therefore no "dev", but
> > > they are all in /sys/class/net/. We may just test if we are there
> > > instead of the "device" match.
> > 
> > How about something like this. It adds all the net devices without
> > looking at the attributes and keeps the remaining logic like it is.
> 
> I'll give it a shot on my ppc_8xx board and if that's OK, I'll figure
> out how to give it a shot on gentoo and then my workstation
> (debian/unstable).

Tested on all 3 and worked fine.

-- 
Tom Rini
http://gate.crashing.org/~trini/


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idP47&alloc_id\x10808&op=click
_______________________________________________
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: [patch] Fix naming ethernet devices in udevstart

[Greg KH]

On Tue, Sep 07, 2004 at 12:46:43PM +0200, Kay Sievers wrote:
> On Mon, 2004-09-06 at 20:18 -0700, Tom Rini wrote:
> > I noticed somewhat recently that my enet devices weren't being renamed
> > on boot anymore.  I don't quite know when this got broken (or rather, if
> > it was supposed to be working.  I swear it worked for me once..), but
> > the following seems to do it.
> 
> I think it never worked in the udevstart case. It worked only with the
> hotplug-event-udev, I expect.
> 
> > In udev_scan_class(), look for not just
> > %s/%s/dev (which everything with a dev node has), but %s/%s/dev* (both
> > of my enet devices, sis900 & 3c59x only have device) and if that
> > exists, pass this along to udev.
> 
> Yeah, network devices don't have a devnode and therefore no "dev", but
> they are all in /sys/class/net/. We may just test if we are there
> instead of the "device" match.
> Btw: Greg, what's the reason not to use libsysfs for the directory walk?

No reason.  Originally I wanted udevstart to be as small as possible, so
I didn't use anything else.  But now that it's part of udev, that
restriction is gone.

Patches gladly accepted :)

thanks,

greg k-h


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
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: [patch] Fix naming ethernet devices in udevstart

[Greg KH]

On Tue, Sep 07, 2004 at 06:56:49PM +0200, Kay Sievers wrote:
> On Tue, Sep 07, 2004 at 12:46:43PM +0200, Kay Sievers wrote:
> > On Mon, 2004-09-06 at 20:18 -0700, Tom Rini wrote:
> > > I noticed somewhat recently that my enet devices weren't being renamed
> > > on boot anymore.  I don't quite know when this got broken (or rather, if
> > > it was supposed to be working.  I swear it worked for me once..), but
> > > the following seems to do it.
> > 
> > I think it never worked in the udevstart case. It worked only with the
> > hotplug-event-udev, I expect.
> > 
> > > In udev_scan_class(), look for not just
> > > %s/%s/dev (which everything with a dev node has), but %s/%s/dev* (both
> > > of my enet devices, sis900 & 3c59x only have device) and if that
> > > exists, pass this along to udev.
> > 
> > Yeah, network devices don't have a devnode and therefore no "dev", but
> > they are all in /sys/class/net/. We may just test if we are there
> > instead of the "device" match.
> 
> How about something like this. It adds all the net devices without
> looking at the attributes and keeps the remaining logic like it is.
> 
> It also removes certain levels of indirection and much simplifies the
> udevstart process. We surely don't need to open and close the udevdb
> for every node while iterating over the list. (We are about 5% faster on
> my box)
> 
> It's not well tested, so it would be nice if someone can have a look
> at it, before a broken udevstart renders any system unbootable.

Applied, thanks.

greg k-h


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
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: [patch] Fix naming ethernet devices in udevstart

[Kevin P. Fleming]

Greg KH wrote:

> Applied, thanks.

Greg,

Can you post a message to this list when you release the udev-031 
tarball? I'm anxious to have this feature, as I too was wanting my net 
devices renamed on boot.


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
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: [patch] Fix naming ethernet devices in udevstart

[Greg KH]

On Fri, Sep 10, 2004 at 02:25:31PM -0700, Kevin P. Fleming wrote:
> Greg KH wrote:
> 
> >Applied, thanks.
> 
> Greg,
> 
> Can you post a message to this list when you release the udev-031 
> tarball?

Just did :)


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
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