Re: Problems using scsi_id with udevstart

[Kay Sievers <kay.sievers@vrfy.org>]

[-- Attachment #1: Type: text/plain, Size: 2422 bytes --]

On Wed, Oct 06, 2004 at 03:24:58PM -0700, Greg KH wrote:
> On Wed, Oct 06, 2004 at 11:39:17PM +0200, Kay Sievers wrote:
> > On Wed, Oct 06, 2004 at 03:22:03PM -0600, Andrew Patterson wrote:
> > > On Wed, 2004-10-06 at 21:13 +0200, Kay Sievers wrote:
> > > > On Wed, 2004-10-06 at 12:42 -0600, Andrew Patterson wrote:
> > > > > On Wed, 2004-10-06 at 09:59 -0700, Patrick Mansfield wrote:
> > > > > > On Wed, Oct 06, 2004 at 10:26:13AM -0600, Andrew Patterson wrote:
> > > > > > 
> > > > > > > Note: here is the error you get when scsi_id is run without any
> > > > > > > parameters:
> > > > > > > 
> > > > > > > # /sbin/scsi_id
> > > > > > > -s must be specified
> > > > > > 
> > > > > > For the udev PROGRAM rules with no arguments, it is effectively invoked
> > > > > > like this (well there are other environment values but scsi_id only cares
> > > > > > about DEVPATH):
> > > > > > 
> > > > > > DEVPATH=/block/sda /sbin/scsi_id block
> > > > > > 
> > > > > 
> > > > > This is what I thought was supposed to happen.  I printed out the
> > > > > arguments and the environment when scsi_id was invoked through
> > > > > udevstart.  No arguments were passed and UDEVPATH was always set
> > > > > to /class/net/lo for every path in /sys.  Note that DEVNAME does seem to
> > > > > be set correctly.
> > > > 
> > > > Yes, this is a bug. We need to export the DEVPATH for the callout if we
> > > > run as udevstart. I will fix that now.
> > > 
> > > So did you confirm that the subsystem is indeed being passed as an
> > > argument when running udevstart as well?  My testing shows no.  I think
> > > you need both to make this work.
> > 
> > Yes, when udevstart was running we didn't set the environment and the
> > subsystem argument for the callouts the dev.d/ scripts.
> > 
> > Here is a fix, that sets that with every udevstart iteration, corrects
> > argv[0] to be the basename() only not the whole path  and adds a test
> > for invoking callouts without arguments.
> 
> Applied, thanks.

Here is the correction for the dev.d/ scripts too. We should pass
the right argv[0] here too. A script may depend on the right value, as
udev does with udev/udevstart.

Here is the old version:
  [pid  4692] execve("/etc/dev.d/default/log.dev", ["./udev", "block"], [/* 41 vars */]) = 0

this the new one:
  [pid  9832] execve("/etc/dev.d/default/log.dev", ["/etc/dev.d/default/log.dev", "block"], [/* 41 vars */]) = 0

Thanks,
Kay

[-- Attachment #2: udev-dev.d-argv0-fix-01.patch --]
[-- Type: text/plain, Size: 963 bytes --]

===== dev_d.c 1.10 vs edited =====
--- 1.10/dev_d.c	2004-09-20 16:02:44 +02:00
+++ edited/dev_d.c	2004-10-07 00:57:31 +02:00
@@ -39,6 +39,7 @@ static int run_program(char *name)
 {
 	pid_t pid;
 	int fd;
+	char *argv[3];
 
 	dbg("running %s", name);
 
@@ -54,7 +55,12 @@ static int run_program(char *name)
 			dup2(fd, STDERR_FILENO);
 		}
 		close(fd);
-		execv(name, main_argv);
+
+		argv[0] = name;
+		argv[1] = main_argv[1];
+		argv[2] = NULL;
+
+		execv(name, argv);
 		dbg("exec of child failed");
 		exit(1);
 	case -1:
===== namedev.c 1.148 vs edited =====
--- 1.148/namedev.c	2004-10-06 23:27:35 +02:00
+++ edited/namedev.c	2004-10-07 00:51:40 +02:00
@@ -450,11 +450,6 @@ static int execute_program(const char *p
 		dbg("execute '%s' with subsystem '%s' argument", arg, argv[1]);
 	}
 
-	/* set basename() only */
-	pos = strrchr(argv[0], '/');
-	if (pos != NULL)
-		argv[0] = &pos[1];
-
 	retval = pipe(fds);
 	if (retval != 0) {
 		dbg("pipe failed");