2016-04-13 06:47:05

by Dan Carpenter

[permalink] [raw]
Subject: [patch] md/raid0: check for create_strip_zones() errors

My static checker complains that if create_strip_zones() fails then we
use "priv_conf" without initializing it. Fix this by checking for
failure.

Signed-off-by: Dan Carpenter <[email protected]>

diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
index 2ea12c6..1d80e3c 100644
--- a/drivers/md/raid0.c
+++ b/drivers/md/raid0.c
@@ -507,6 +507,7 @@ static void *raid0_takeover_raid45(struct mddev *mddev)
{
struct md_rdev *rdev;
struct r0conf *priv_conf;
+ int ret;

if (mddev->degraded != 1) {
printk(KERN_ERR "md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n",
@@ -534,13 +535,16 @@ static void *raid0_takeover_raid45(struct mddev *mddev)
/* make sure it will be not marked as dirty */
mddev->recovery_cp = MaxSector;

- create_strip_zones(mddev, &priv_conf);
+ ret = create_strip_zones(mddev, &priv_conf);
+ if (ret)
+ return ERR_PTR(ret);
return priv_conf;
}

static void *raid0_takeover_raid10(struct mddev *mddev)
{
struct r0conf *priv_conf;
+ int ret;

/* Check layout:
* - far_copies must be 1
@@ -575,7 +579,9 @@ static void *raid0_takeover_raid10(struct mddev *mddev)
/* make sure it will be not marked as dirty */
mddev->recovery_cp = MaxSector;

- create_strip_zones(mddev, &priv_conf);
+ ret = create_strip_zones(mddev, &priv_conf);
+ if (ret)
+ return ERR_PTR(ret);
return priv_conf;
}

@@ -583,6 +589,7 @@ static void *raid0_takeover_raid1(struct mddev *mddev)
{
struct r0conf *priv_conf;
int chunksect;
+ int ret;

/* Check layout:
* - (N - 1) mirror drives must be already faulty
@@ -617,7 +624,9 @@ static void *raid0_takeover_raid1(struct mddev *mddev)
/* make sure it will be not marked as dirty */
mddev->recovery_cp = MaxSector;

- create_strip_zones(mddev, &priv_conf);
+ ret = create_strip_zones(mddev, &priv_conf);
+ if (ret)
+ return ERR_PTR(ret);
return priv_conf;
}



2016-04-13 17:02:45

by Shaohua Li

[permalink] [raw]
Subject: Re: [patch] md/raid0: check for create_strip_zones() errors

On Wed, Apr 13, 2016 at 09:46:45AM +0300, Dan Carpenter wrote:
> My static checker complains that if create_strip_zones() fails then we
> use "priv_conf" without initializing it. Fix this by checking for
> failure.

It's more convenient setting '*private_conf = ERR_PTR(-ENOMEM);' at the
begining of create_strip_zones() when it returns -ENOMEM. create_strip_zones
already sets private_conf correctly in other cases.

Thanks,
Shaohua

2016-04-13 17:54:26

by Dan Carpenter

[permalink] [raw]
Subject: Re: [patch] md/raid0: check for create_strip_zones() errors

On Wed, Apr 13, 2016 at 10:02:40AM -0700, Shaohua Li wrote:
> On Wed, Apr 13, 2016 at 09:46:45AM +0300, Dan Carpenter wrote:
> > My static checker complains that if create_strip_zones() fails then we
> > use "priv_conf" without initializing it. Fix this by checking for
> > failure.
>
> It's more convenient setting '*private_conf = ERR_PTR(-ENOMEM);' at the
> begining of create_strip_zones() when it returns -ENOMEM. create_strip_zones
> already sets private_conf correctly in other cases.
>

Yeah. I'll send v2.

regards,
dan carpenter

2016-04-14 09:32:05

by Dan Carpenter

[permalink] [raw]
Subject: [patch v2] md/raid0: fix uninitialized variable bug

If this function fails the callers expect that *private_conf is set to
an ERR_PTR() but that isn't true for the first error path where we can't
allocate "conf". It leads to some uninitialized variable bugs.

Signed-off-by: Dan Carpenter <[email protected]>
---
v2: Shaohua suggested a different fix

diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
index 2ea12c6..f63dbb6 100644
--- a/drivers/md/raid0.c
+++ b/drivers/md/raid0.c
@@ -85,6 +85,7 @@ static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf)
struct r0conf *conf = kzalloc(sizeof(*conf), GFP_KERNEL);
unsigned short blksize = 512;

+ *private_conf = ERR_PTR(-ENOMEM);
if (!conf)
return -ENOMEM;
rdev_for_each(rdev1, mddev) {

2016-04-14 16:57:06

by Shaohua Li

[permalink] [raw]
Subject: Re: [patch v2] md/raid0: fix uninitialized variable bug

On Thu, Apr 14, 2016 at 12:31:49PM +0300, Dan Carpenter wrote:
> If this function fails the callers expect that *private_conf is set to
> an ERR_PTR() but that isn't true for the first error path where we can't
> allocate "conf". It leads to some uninitialized variable bugs.
>
> Signed-off-by: Dan Carpenter <[email protected]>

Applied, thanks!