Skip to contentSkip to conversation

udev - Shouldn't we blacklist some devices?

DISCUSSION Kay Sievers 2 participants 7 messages Jan 3, 2004 – Jan 4, 2004

lore
Share & tools
lore
#0 lore raw

While connecting my flash reader to the USB bus I get:

Jan  3 20:56:58 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:56:59 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:56:59 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Jan  3 20:56:59 pim udev[31627]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:0/dev'
Jan  3 20:56:59 pim udev[31650]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:1/dev'
Jan  3 20:56:59 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:57:00 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:57:00 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Jan  3 20:57:00 pim udev[31627]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:0/dev'
Jan  3 20:57:00 pim udev[31650]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:1/dev'
Jan  3 20:57:00 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:57:01 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:57:01 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'

Shouldn't we blacklist devices like /sys/class/scsi_device/*, /sys/class/scsi_host/* ?

thanks,
Kay

-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id78&alloc_id371&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#1 lore raw
Quoted 17 lines · Kay Sievers

While connecting my flash reader to the USB bus I get:

Jan  3 20:56:58 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:56:59 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:56:59 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Jan  3 20:56:59 pim udev[31627]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:0/dev'
Jan  3 20:56:59 pim udev[31650]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:1/dev'
Jan  3 20:56:59 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:57:00 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:57:00 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Jan  3 20:57:00 pim udev[31627]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:0/dev'
Jan  3 20:57:00 pim udev[31650]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:1/dev'
Jan  3 20:57:00 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:57:01 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:57:01 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Quoted 1 line

Shouldn't we blacklist devices like /sys/class/scsi_device/*, /sys/class/scsi_host/* ?

Yes we should, as we try to remove those device nodes when the devices get removed (look for the 'trying to remove host54' messages), which could be very dangerous.

thanks,

greg k-h

-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id78&alloc_id371&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#2 lore raw
In reply to Kay Sievers
Quoted 7 lines · Greg KH
Quoted 6 lines · Kay Sievers

While connecting my flash reader to the USB bus I get:

Jan  3 20:56:58 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:56:59 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:56:59 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Quoted 6 lines
Quoted 1 line

Shouldn't we blacklist devices like /sys/class/scsi_device/*, /sys/class/scsi_host/* ?

Yes we should, as we try to remove those device nodes when the devices get removed (look for the 'trying to remove host54' messages), which could be very dangerous.

So, here we go:

Blacklist devices that we don't want to process, like:
/sys/class/scsi_device/*, /sys/class/scsi_host/*,
cause these devices are handled by the block device events we receive and we don't try to remove non existing nodes.

thanks,
Kay

udev.c +17 -0
diff -Nru a/udev.c b/udev.c
@@ -139,12 +139,19 @@
 	return retval;
 }
 
+static char *dev_path_blacklist[] = {
+	"class/scsi_host",
+	"class/scsi_device",
+	""
+};
+
 static inline int udev_hotplug(int argc, char **argv)
 {
 	char *action;
 	char *devpath;
 	char *subsystem;
 	int retval = -EINVAL;
+	int i;
 
 	subsystem = argv[1];
 
@@ -160,6 +167,16 @@
 	    !strstr(devpath, "block")) {
 		dbg("not a block or class device");
 		goto exit;
+	}
+
+	/* skip blacklisted path */
+	i = 0;
+	while (dev_path_blacklist[i][0] != '\0') {
+		if (strstr(devpath, dev_path_blacklist[i])) {
+			dbg("device with path '%s' is blacklisted", devpath);
+			goto exit;
+		}
+		i++;
 	}
 
 	/* but we don't care about net class devices */
#3 lore raw
In reply to Kay Sievers
Quoted 15 lines · Kay Sievers
Quoted 7 lines · Greg KH
Quoted 6 lines · Kay Sievers

While connecting my flash reader to the USB bus I get:

Jan  3 20:56:58 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:56:59 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:56:59 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Quoted 6 lines
Quoted 1 line

Shouldn't we blacklist devices like /sys/class/scsi_device/*, /sys/class/scsi_host/* ?

Yes we should, as we try to remove those device nodes when the devices get removed (look for the 'trying to remove host54' messages), which could be very dangerous.

Quoted 6 lines

So, here we go:

Blacklist devices that we don't want to process, like:
/sys/class/scsi_device/*, /sys/class/scsi_host/*,
cause these devices are handled by the block device events we receive and we don't try to remove non existing nodes.

How about this patch instead (not tested)? It removes the explicit test for net class devices.

thanks,

greg k-h

=== udev.c 1.27 vs edited ==--- 1.27/udev.c Tue Dec 30 16:33:12 2003
+++ edited/udev.c Sat Jan 3 15:29:00 2004
@@ -139,12 +139,20 @@
return retval;
}

+static char *subsystem_blacklist[] = {
+ "net",
+ "scsi_host",
+ "scsi_device",
+ "",
+};
+
static inline int udev_hotplug(int argc, char **argv) {
char *action;
char *devpath;
char *subsystem;
int retval = -EINVAL;
+ int i;

subsystem = argv[1];

@@ -162,10 +170,14 @@
goto exit;
}

- /* but we don't care about net class devices */
- if (strcmp(subsystem, "net") = 0) {
- dbg("don't care about net devices");
- goto exit;
+ /* skip blacklisted subsystems */
+ i = 0;
+ while (subsystem_blacklist[i][0] != '\0') {
+ if (strcmp(subsystem, subsystem_blacklist[i])) {
+ dbg("don't care about '%s' devices", subsystem);
+ goto exit;
+ }
+ i++;
}

action = get_action();

-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id78&alloc_id371&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#4 lore raw
In reply to Kay Sievers
Quoted 18 lines · Greg KH
Quoted 17 lines · Kay Sievers
Quoted 7 lines · Greg KH
Quoted 6 lines · Kay Sievers

While connecting my flash reader to the USB bus I get:

Jan  3 20:56:58 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:56:59 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:56:59 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Quoted 6 lines
Quoted 1 line

Shouldn't we blacklist devices like /sys/class/scsi_device/*, /sys/class/scsi_host/* ?

Yes we should, as we try to remove those device nodes when the devices get removed (look for the 'trying to remove host54' messages), which could be very dangerous.

So, here we go:

Quoted 7 lines
Quoted 4 lines

Blacklist devices that we don't want to process, like:
/sys/class/scsi_device/*, /sys/class/scsi_host/*,
cause these devices are handled by the block device events we receive and we don't try to remove non existing nodes.

How about this patch instead (not tested)? It removes the explicit test for net class devices.

Nice, looks good.
I've tested it:

Jan  4 00:38:40 pim udev[4389]: udev_hotplug: looking at '/class/scsi_device/62:0:0:0'
Jan  4 00:38:40 pim udev[4389]: udev_hotplug: don't care about 'scsi_device' devices
Jan  4 00:38:41 pim udev[4562]: udev_hotplug: looking at '/class/scsi_host/host62'
Jan  4 00:38:41 pim udev[4562]: udev_hotplug: don't care about 'scsi_host' devices

Kay

-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id78&alloc_id371&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#5 lore raw
In reply to Kay Sievers
Quoted 27 lines · Kay Sievers
Quoted 18 lines · Greg KH
Quoted 17 lines · Kay Sievers
Quoted 7 lines · Greg KH
Quoted 6 lines · Kay Sievers

While connecting my flash reader to the USB bus I get:

Jan  3 20:56:58 pim udev[31673]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:2/dev'
Jan  3 20:56:59 pim udev[31695]: sleep_for_dev: looking for '/sys/class/scsi_device/54:0:0:3/dev'
Jan  3 20:56:59 pim udev[31514]: sleep_for_dev: looking for '/sys/class/scsi_host/host54/dev'
Quoted 6 lines
Quoted 1 line

Shouldn't we blacklist devices like /sys/class/scsi_device/*, /sys/class/scsi_host/* ?

Yes we should, as we try to remove those device nodes when the devices get removed (look for the 'trying to remove host54' messages), which could be very dangerous.

So, here we go:

Quoted 7 lines
Quoted 4 lines

Blacklist devices that we don't want to process, like:
/sys/class/scsi_device/*, /sys/class/scsi_host/*,
cause these devices are handled by the block device events we receive and we don't try to remove non existing nodes.

How about this patch instead (not tested)? It removes the explicit test for net class devices.

Quoted 8 lines

Nice, looks good.
I've tested it:

Jan  4 00:38:40 pim udev[4389]: udev_hotplug: looking at '/class/scsi_device/62:0:0:0'
Jan  4 00:38:40 pim udev[4389]: udev_hotplug: don't care about 'scsi_device' devices
Jan  4 00:38:41 pim udev[4562]: udev_hotplug: looking at '/class/scsi_host/host62'
Jan  4 00:38:41 pim udev[4562]: udev_hotplug: don't care about 'scsi_host' devices

Oh, my testing is good, but reading to the end of the debug is better :) It should be:
if (strcmp(subsystem, subsystem_blacklist[i]) = 0) {

otherwise we skip everything.

Kay

-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id78&alloc_id371&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#6 lore raw
In reply to Kay Sievers
Quoted 6 lines · Kay Sievers

Oh, my testing is good, but reading to the end of the debug is better :) It should be:
if (strcmp(subsystem, subsystem_blacklist[i]) = 0) {

otherwise we skip everything.

Heh, oops. Thanks, I've fixed that and pushed the change.

thanks,

greg k-h

-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id78&alloc_id371&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel