Re: linux-next: Tree for July 29

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 30 Jul 2008 09:06:50 +0200 Bernhard Walle <bwalle@suse.de> wrote:

> * Greg KH <greg@kroah.com> [2008-07-29 21:48]:
> > > Isn't this the opposite end of the same problem for which Bernhard
> > > has been repeatedly trying to find a taker for his patch:
> > > 
> > > http://article.gmane.org/gmane.linux.kernel.kexec/1882
> > 
> > Yes.  It's not the kobject patch at fault here, it's the use of kobjects
> > so early in the boot process.  That needs to be fixed.

It was a bit optimistic to stick an unconditional GFP_KERNEL allocation
into the previously-atomic kobject_init().

It's only 128 bytes, so why can't we fix both problems thusly?

--- a/lib/kobject.c~a
+++ a/lib/kobject.c
@@ -38,12 +38,10 @@ static int ptr_in_range(void *ptr, void 
 
 static void verify_dynamic_kobject_allocation(struct kobject *kobj)
 {
-	char *namebuf;
+	char namebuf[KSYM_NAME_LEN];
 	const char *ret;
 
-	namebuf = kzalloc(KSYM_NAME_LEN, GFP_KERNEL);
-	ret = kallsyms_lookup((unsigned long)kobj, NULL, NULL, NULL,
-			namebuf);
+	ret = kallsyms_lookup((unsigned long)kobj, NULL, NULL, NULL, namebuf);
 	/*
 	 * This is the X86_32-only part of this function.
 	 * This is here because it is valid to have a kobject
@@ -63,7 +61,7 @@ static void verify_dynamic_kobject_alloc
 	/* dump_stack(); */
 	pr_debug("---- end silly warning ----\n");
 out:
-	kfree(namebuf);
+	return;
 }
 #else
 static void verify_dynamic_kobject_allocation(struct kobject *kobj) { }
_


> Yes, but if somebody could tell me why nobody takes the patch, I would
> be happy. Then I would be able to improve the patch. :)

Copy me on the patch.  Then I merge it and people know there will be no
hiding from it.