udev-039-6 on Fedora Core 3

FYI,

I added several patches to udev-039 which will be on FC3.
I only describe the most patches here. The patches itsself are in the src.rpm and can be posted on request. The two relevant and important patches against 039 are attached.

ru
Harald

****

Patch1: udev-039-static.patch
makes inline getpwnam and getgrnam functions, which are used for the static version of udev, which is used in initramfs

Patch2: udev-025-volsbin.patch
moves volume_id from /usr/sbin to /sbin (such tools should not live in /usr, because /usr could be nfs-mounted

Patch3: udev-039-wait-error-log.patch
wait_for_sysfs patch

Patch4: udev-039-logging.patch
logging patch

Patch5: udev-039-wait-ppp-class.patch
wait_for_sysfs patch

Patch6: udev-039-wait-dm-block.patch
wait_for_sysfs patch

Patch7: udev-039-static2.patch
include wait_for_sysfs in udev for the static udev version used in initramfs

Patch8: udev-039-netif.patch
only call dev_d_send for network interfaces, if the name was changed

Patch9: udev-039-dummy.patch
wait_for_sysfs patch

Patch10: udev-039-volume_id-nonblock.patch
do not open(2) with O_NONBLOCK

Patch11: udev-039-norm.patch
add NOREMOVE keyword to the rules. this was necessary because removable ide devices were added/removed for every open(2)/close(2) added to the rules:
KERNEL="hd[a-z]*", NOREMOVE="1"

this could eventually be refined by checking the "removable" sysfs-attribute, but not for partitions..

Patch22: udev-032-symlink.patch
replace format directives in every symlink part

Patch24: udev-038-volmak.patch
Makefile patch for volume_id

@@ -898,14 +898,18 @@
 
 			if (dev->symlink[0] != '\0') {
 				char temp[NAME_SIZE];
-
+				char *p;
+				int len;
 				info("configured rule in '%s' at line %i applied, added symlink '%s'",
 				     dev->config_file, dev->config_line, dev->symlink);
-				strfieldcpy(temp, dev->symlink);
-				apply_format(udev, temp, sizeof(temp), class_dev, sysfs_device);
-				if (udev->symlink[0] != '\0')
-					strfieldcat(udev->symlink, " ");
-				strfieldcat(udev->symlink, temp);
+				foreach_strpart(dev->symlink, " ", p, len) {
+					strncpy(temp, p, len);
+					temp[len] = '\0';
+					apply_format(udev, temp, sizeof(temp), class_dev, sysfs_device);
+					if (udev->symlink[0] != '\0')
+						strfieldcat(udev->symlink, " ");
+					strfieldcat(udev->symlink, temp);
+				}
 			}
 
 			if (dev->name[0] != '\0') {

@@ -44,6 +44,7 @@
 #define FIELD_KERNEL		"KERNEL"
 #define FIELD_NAME		"NAME"
 #define FIELD_SYMLINK		"SYMLINK"
+#define FIELD_NOREMOVE		"NOREMOVE"
 #define FIELD_OWNER		"OWNER"
 #define FIELD_GROUP		"GROUP"
 #define FIELD_MODE		"MODE"

@@ -83,6 +84,7 @@
 	struct sysfs_pair sysfs_pair[MAX_SYSFS_PAIRS];
 	char owner[OWNER_SIZE];
 	char group[GROUP_SIZE];
+	char noremove;
 	unsigned int mode;
 	int partitions;
 	char config_file[NAME_SIZE];

@@ -113,6 +113,11 @@
 
 	strfieldcpy(filename, udev_root);
 	strfieldcat(filename, dev->name);
+	
+	if ( dev->noremove != 0) {
+		info("not removing device node '%s', because of rule", filename);
+		return 0;
+	}
 
 	info("removing device node '%s'", filename);
 	retval = secure_unlink(filename);

@@ -59,11 +59,11 @@
 	dbg_parse("name='%s', symlink='%s', bus='%s', place='%s', id='%s', "
 		  "sysfs_file[0]='%s', sysfs_value[0]='%s', "
 		  "kernel='%s', program='%s', result='%s'"
-		  "owner='%s', group='%s', mode=%#o",
+		  "owner='%s', group='%s', mode=%#o remove=%d",
 		  dev->name, dev->symlink, dev->bus, dev->place, dev->id,
 		  dev->sysfs_pair[0].file, dev->sysfs_pair[0].value,
 		  dev->kernel, dev->program, dev->result,
-		  dev->owner, dev->group, dev->mode);
+		  dev->owner, dev->group, dev->mode, dev->noremove);
 }
 
 void dump_config_dev_list(void)

@@ -265,6 +265,12 @@
 				continue;
 			}
 
+			if (strcasecmp(temp2, FIELD_NOREMOVE) == 0) {
+				dev.noremove = strtol(temp3, NULL, 10);
+				valid = 1;
+				continue;
+			}
+
 			if (strcasecmp(temp2, FIELD_RESULT) == 0) {
 				strfieldcpy(dev.result, temp3);
 				valid = 1;

@@ -738,12 +738,15 @@
 	list_for_each_entry(dev, &config_device_list, node) {
 		dbg("process rule");
 		if (match_rule(dev, class_dev, udev, sysfs_device) == 0) {
-			if (dev->name[0] == '\0' && dev->symlink[0] == '\0') {
+			if (dev->name[0] == '\0' && dev->symlink[0] == '\0' && dev->noremove == 0) {
 				info("configured rule in '%s' at line %i applied, '%s' is ignored",
 				     dev->config_file, dev->config_line, udev->kernel_name);
 				return -1;
 			}
+			if (dev->noremove != 0) {
+				udev->noremove = dev->noremove;
+			}
 
 			if (dev->symlink[0] != '\0') {
 				char temp[NAME_SIZE];
 

@@ -52,6 +52,7 @@
 	char owner[OWNER_SIZE];
 	char group[GROUP_SIZE];
 	char type;
+	char noremove;
 	int major;
 	int minor;
 	unsigned int mode;	/* not mode_t due to conflicting definitions in different libcs */

@@ -116,6 +116,7 @@
 	printf("F: %s\n", dev->config_file);
 	printf("L: %i\n", dev->config_line);
 	printf("U: %li\n", dev->config_uptime);
+	printf("R: %d\n", dev->noremove);
 	printf("\n");
 	return 0;
 }

Cool, somebody sharing...
I've looked at all patches in the SRPM. Some comments on it:

Why don't you compile it with klibc? Everything should work. The klibc-version is also be able to parse /etc/passwd to keep the same permissions file in initrd.

Right, should go in our tree.

Already in current version!

Already in current version!

Already in current version!

Already in current version!

Huh, #include of an .c file :)
The current version has a udev_sysfs.c file for that. I'm on converting everything to a be more like a toolbox. Something like this should only need a few lines of code now and you only need to open the device once.

Already in current version!

Already in current version!

Already in current version!

I see the need for that, yes. But I would like to have a match of the ACTION string instead. This would make it more flexible. Something like:

KERNEL="abc", ACTION="add", NAME="xyz"

Not needed, I think. Should work out of the box as the format replacement will not stop at the spaces in the symlink string. If you have problems with that, then we have a bug and please tell us.

Should go in our tree. What was the problem here, On my box, this is working.

Harald,
I will take care, that:
udev-038-volmak.patch,
udev-025-volsbin.patch
are getting upstream.

Please give me some feedback on:
udev-039-norm.patch,
udev-032-symlink.patch,
udev-039-static.patch

Thanks,
Kay

-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal Use IT products in your business? Tell us what you think of them. Give us Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Because maintaining another libc for all our architecures is a PITA...

what do you mean?

ok, I will remove that one...

-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal Use IT products in your business? Tell us what you think of them. Give us Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel