Skip to contentSkip to conversation

obsolete modparam change busted.

PATCH Dave Jones 2 participants 5 messages Aug 8, 2005 – Aug 16, 2005

lore
Share & tools
lore

Dave Jones

author
#0 lore raw

Circa 2.6.10, the module loader started barfing if modprobe.conf contained obsolete parameters.

However this change was broken, and if the modprobe.conf has trailing whitespace, modules fail to load with the following helpful message..

snd_intel8x0: Unknown parameter `'

This ends up screwing over people who upgrade from previously working configurations, so we've been backing out that change in Fedora for a while with the patch below.

It doesn't look like the right thing to do, but it has got things working again at least. Probably we should just check explicity for whitespace and ignore it somewhere else in the module loader. Rusty?

Dave

kernel/module.c +0 -2
diff -urNp --exclude-from=/home/davej/.exclude linux-1503/kernel/module.c linux-1700/kernel/module.c
@@ -1707,8 +1707,6 @@ static struct module *load_module(void _
 				 / sizeof(struct kernel_param),
 				 NULL);
 	}
-	if (err < 0)
-		goto arch_cleanup;
 
 	err = mod_sysfs_setup(mod, 
 			      (struct kernel_param *)
#1 lore raw
Quoted 3 lines · Dave Jones

However this change was broken, and if the modprobe.conf has trailing whitespace, modules fail to load with the following helpful message..

Hi Dave,

This fix should be preferable, I think.

Name: Ignore trailing whitespace on kernel parameters correctly

Dave Jones says:

... if the modprobe.conf has trailing whitespace, modules fail to load with the following helpful message..

snd_intel8x0: Unknown parameter `'

tmp/foo.c renamed from kernel/params.c +2 -5
@@ -1,0 +1,0 @@
@@ -135,11 +80,8 @@
 
 	DEBUGP("Parsing ARGS: %s\n", args);
 
-	while (*args) {
-		int ret;
-
-		args = next_arg(args, &param, &val);
-		ret = parse_one(param, val, params, num, unknown);
+	while (*(args = next_arg(args, &param, &val))) {
+		int ret = parse_one(param, val, params, num, unknown);
 		switch (ret) {
 		case -ENOENT:
 			printk(KERN_ERR "%s: Unknown parameter `%s'\n",
Signature · 1 line
A bad analogy is like a leaky screwdriver -- Richard Braakman
author
#2 lore raw
Quoted 11 lines · Rusty Russell
Quoted 3 lines · Dave Jones

However this change was broken, and if the modprobe.conf has trailing whitespace, modules fail to load with the following helpful message..

Hi Dave,

This fix should be preferable, I think.

Name: Ignore trailing whitespace on kernel parameters correctly

Hey Rusty,
This seems to have introduced a different regression :-)

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=165633

We're now munching the end of the boot command line it seems.

Dave

#3 lore raw
Quoted 1 line · Dave Jones

We're now munching the end of the boot command line it seems.

Wow, if we had testcases in the kernel source, I would not have to keep rewriting them (badly) every time I touched this code.

Throw away that stupid patch, apply this stupid patch.

Name: Ignore trailing whitespace on kernel parameters correctly: Fixed version

Dave Jones says:

... if the modprobe.conf has trailing whitespace, modules fail to load with the following helpful message..

snd_intel8x0: Unknown parameter `'

Previous version truncated last argument.

kernel/params.c +6 -2
Index: linux-2.6.13-rc6-git7-Module/kernel/params.c
===================================================================
@@ -80,8 +80,6 @@
 	int in_quote = 0, quoted = 0;
 	char *next;
 
-	/* Chew any extra spaces */
-	while (*args == ' ') args++;
 	if (*args == '"') {
 		args++;
 		in_quote = 1;
@@ -121,6 +119,9 @@
 		next = args + i + 1;
 	} else
 		next = args + i;
+
+	/* Chew up trailing spaces. */
+	while (*next == ' ') next++;
 	return next;
 }
 
@@ -134,6 +135,9 @@
 	char *param, *val;
 
 	DEBUGP("Parsing ARGS: %s\n", args);
+	
+	/* Chew leading spaces */
+	while (*args == ' ') args++;
 
 	while (*args) {
 		int ret;
Signature · 1 line
A bad analogy is like a leaky screwdriver -- Richard Braakman
author
#4 lore raw
Quoted 5 lines · Rusty Russell
Quoted 1 line · Dave Jones

We're now munching the end of the boot command line it seems.

Wow, if we had testcases in the kernel source, I would not have to keep rewriting them (badly) every time I touched this code.

Maybe someone should write a userspace test harness for kernel code. :-P

Quoted 1 line

Throw away that stupid patch, apply this stupid patch.

I'll throw it into the Fedora kernel tomorrow, and re-test. Thanks for chasing this so quickly.

Dave