Re: [RFC, 2.6.26.2-rc1] fs/partition/check.c: fix return value warning

Re: [RFC, 2.6.26.2-rc1] fs/partition/check.c: fix return value warning

[Linus Torvalds]

On Sat, 2 Aug 2008, Oliver Pinter wrote:
>
> Subject: [PATCH] fs/partition/check.c: fix return value warning
> 
> [ Upstream commit d805dda412346225a50af2d399d958a4bc676c38 ]

Why are these being submitted for stable kernels?

Oliver - stable kernels should _not_ get trivial fixes that don't actually 
have any problems associated with them. 

Every patch in a stable kernel has a downside. It had better have a big 
upside too.  These don't.

Please don't just send patches to the stable team because they "look 
fine". They should be real bugs, with real consequences. And no, a 
compiler warning is not a real consequence.

		Linus

[RFC, 2.6.26.2-rc1] fs/partition/check.c: fix return value warning

[Oliver Pinter]

>From d805dda412346225a50af2d399d958a4bc676c38 Mon Sep 17 00:00:00 2001
From: Abdel Benamrouche <draconux@gmail.com>
Date: Fri, 25 Jul 2008 01:48:25 -0700
Subject: [PATCH] fs/partition/check.c: fix return value warning

[ Upstream commit d805dda412346225a50af2d399d958a4bc676c38 ]

fs/partitions/check.c:381: warning: ignoring return value of ___device_add___,
  declared with attribute warn_unused_result

[akpm@linux-foundation.org: multiple-return-statements-per-function are evil]
Signed-off-by: Abdel Benamrouche <draconux@gmail.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
CC: Oliver Pinter <oliver.pntr@gmail.com>

diff --git a/fs/partitions/check.c b/fs/partitions/check.c
index efef715..2e6413f 100644
--- a/fs/partitions/check.c
+++ b/fs/partitions/check.c
@@ -344,18 +344,18 @@ static ssize_t whole_disk_show(struct device *dev,
 static DEVICE_ATTR(whole_disk, S_IRUSR | S_IRGRP | S_IROTH,
 		   whole_disk_show, NULL);
 
-void add_partition(struct gendisk *disk, int part, sector_t start, sector_t len, int flags)
+int add_partition(struct gendisk *disk, int part, sector_t start, sector_t len, int flags)
 {
 	struct hd_struct *p;
 	int err;
 
 	p = kzalloc(sizeof(*p), GFP_KERNEL);
 	if (!p)
-		return;
+		return -ENOMEM;
 
 	if (!init_part_stats(p)) {
-		kfree(p);
-		return;
+		err = -ENOMEM;
+		goto out0;
 	}
 	p->start_sect = start;
 	p->nr_sects = len;
@@ -378,15 +378,31 @@ void add_partition(struct gendisk *disk, int part, sector_t start, sector_t len,
 
 	/* delay uevent until 'holders' subdir is created */
 	p->dev.uevent_suppress = 1;
-	device_add(&p->dev);
+	err = device_add(&p->dev);
+	if (err)
+		goto out1;
 	partition_sysfs_add_subdir(p);
 	p->dev.uevent_suppress = 0;
-	if (flags & ADDPART_FLAG_WHOLEDISK)
+	if (flags & ADDPART_FLAG_WHOLEDISK) {
 		err = device_create_file(&p->dev, &dev_attr_whole_disk);
+		if (err)
+			goto out2;
+	}
 
 	/* suppress uevent if the disk supresses it */
 	if (!disk->dev.uevent_suppress)
 		kobject_uevent(&p->dev.kobj, KOBJ_ADD);
+
+	return 0;
+
+out2:
+	device_del(&p->dev);
+out1:
+	put_device(&p->dev);
+	free_part_stats(p);
+out0:
+	kfree(p);
+	return err;
 }
 
 /* Not exported, helper to add_disk(). */
diff --git a/include/linux/genhd.h b/include/linux/genhd.h
index e878741..118216f 100644
--- a/include/linux/genhd.h
+++ b/include/linux/genhd.h
@@ -541,7 +541,7 @@ extern dev_t blk_lookup_devt(const char *name, int part);
 extern char *disk_name (struct gendisk *hd, int part, char *buf);
 
 extern int rescan_partitions(struct gendisk *disk, struct block_device *bdev);
-extern void add_partition(struct gendisk *, int, sector_t, sector_t, int);
+extern int __must_check add_partition(struct gendisk *, int, sector_t, sector_t, int);
 extern void delete_partition(struct gendisk *, int);
 extern void printk_all_partitions(void);
 

Re: [RFC, 2.6.26.2-rc1] fs/partition/check.c: fix return value warning

[Willy Tarreau]

On Sat, Aug 02, 2008 at 12:56:10PM -0700, Linus Torvalds wrote:
> Please don't just send patches to the stable team because they "look 
> fine". They should be real bugs, with real consequences. And no, a 
> compiler warning is not a real consequence.

BTW, before Oliver does the mistake, I'd suggest never forwarding
network patches to stable either as Davem already maintains his
own -stable queue ;-)

Willy

Re: [RFC, 2.6.26.2-rc1] fs/partition/check.c: fix return value warning

[Oliver Pinter]

Willy, this good to know ...

and I think, this patch fixed a small leak:

+
+out2:
+       device_del(&p->dev);
+out1:
+       put_device(&p->dev);
+       free_part_stats(p);
+out0:
+       kfree(p);
+       return err;


On 8/2/08, Willy Tarreau <w@1wt.eu> wrote:
> On Sat, Aug 02, 2008 at 12:56:10PM -0700, Linus Torvalds wrote:
>> Please don't just send patches to the stable team because they "look
>> fine". They should be real bugs, with real consequences. And no, a
>> compiler warning is not a real consequence.
>
> BTW, before Oliver does the mistake, I'd suggest never forwarding
> network patches to stable either as Davem already maintains his
> own -stable queue ;-)
>
> Willy
>
>


-- 
Thanks,
Oliver

Re: [RFC, 2.6.26.2-rc1] fs/partition/check.c: fix return value warning

[Stefan Richter]

Oliver Pinter wrote:
> Willy, this good to know ...
> 
> and I think, this patch fixed a small leak:
> 
> +
> +out2:
> +       device_del(&p->dev);
> +out1:
> +       put_device(&p->dev);
> +       free_part_stats(p);
> +out0:
> +       kfree(p);
> +       return err;
> 
> 
> On 8/2/08, Willy Tarreau <w@1wt.eu> wrote:
>> On Sat, Aug 02, 2008 at 12:56:10PM -0700, Linus Torvalds wrote:
>>> Please don't just send patches to the stable team because they "look
>>> fine". They should be real bugs, with real consequences. And no, a
>>> compiler warning is not a real consequence.

Oliver, check under which circumstances the leak happens, and which 
problems the leak causes for the system.

Newly added failure handling paths, which by nature are insufficiently 
runtime-tested, need to be well considered before they are taken over 
into -stable even if the code looks OK.  I.e. they need to address bugs 
which are known to cause problems of certain severity in practice.
-- 
Stefan Richter
-=====-==--- =--- ---==
http://arcgraph.de/sr/