Re: [PATCH 3/4] gcov: add gcov profiling infrastructure

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

On Wed, 03 Jun 2009 17:26:22 +0200
Peter Oberparleiter <oberpar@linux.vnet.ibm.com> wrote:

> > Right - the sscanf would make sense if kernel parameters could contain 
> > spaces (in that case it catches <number><blanks><garbage> input) which 
> > it can't so strtoul() would indeed make more sense. I'll prepare an 
> > updated patch and send it out later today.
> 
> See below for the updated patch that uses strtoul instead of sscanf.
> This patch replaces kernel-constructor-support.patch in the -mm tree:

umm, no it doesn't.  I get the below incremental patch, against
gcov-add-gcov-profiling-infrastructure.patch:

--- a/kernel/gcov/fs.c~gcov-add-gcov-profiling-infrastructure-update
+++ a/kernel/gcov/fs.c
@@ -70,15 +70,8 @@ static int gcov_persist = 1;
 
 static int __init gcov_persist_setup(char *str)
 {
-	int val;
-	char delim;
-
-	if (sscanf(str, "%d %c", &val, &delim) != 1) {
-		pr_warning("invalid gcov_persist parameter '%s'\n", str);
-		return 0;
-	}
-	pr_info("setting gcov_persist to %d\n", val);
-	gcov_persist = val;
+	gcov_persist = simple_strtoul(str, NULL, 0);
+	pr_info("setting gcov_persist to %d\n", gcov_persist);
 
 	return 1;
 }
_

arguably we should use strict_strtoul(), but the kernel is a lot less
fussy about boot parameters than it is with sysfs writes, etc.  If you
fat-finger your grub.conf, you lose and we don't tell you.