Subject: [PATCH] slab: Adds missing kmalloc() checks.


Adds two missing kmalloc() checks in kmem_cache_init(). Note that if the
allocation fails, there is nothing to do, so we panic(); and the __LINE__
is used because is good to know which allocation has failed.

Signed-off-by: Luiz Capitulino <[email protected]>

mm/slab.c | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/mm/slab.c b/mm/slab.c
index e5ec26e..3f23ad2 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -1115,6 +1115,8 @@ void __init kmem_cache_init(void)
void * ptr;

ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
+ if (!ptr)
+ panic("%s at %d: Could not allocate memory.\n", __FUNCTION__, __LINE__);

local_irq_disable();
BUG_ON(ac_data(&cache_cache) != &initarray_cache.cache);
@@ -1124,6 +1126,8 @@ void __init kmem_cache_init(void)
local_irq_enable();

ptr = kmalloc(sizeof(struct arraycache_init), GFP_KERNEL);
+ if (!ptr)
+ panic("%s at %d: Could not allocate memory.\n", __FUNCTION__, __LINE__);

local_irq_disable();
BUG_ON(ac_data(malloc_sizes[INDEX_AC].cs_cachep)


--
Luiz Fernando N. Capitulino


2006-01-06 15:25:05

by Arjan van de Ven

[permalink] [raw]
Subject: Re: [PATCH] slab: Adds missing kmalloc() checks.

On Fri, 2006-01-06 at 13:12 -0200, Luiz Fernando Capitulino wrote:
> Adds two missing kmalloc() checks in kmem_cache_init(). Note that if the
> allocation fails, there is nothing to do, so we panic();

ok so what good does this do? if you die this early.. you are in deeper
problems, and can't boot. while this makes the code bigger...


Subject: Re: [PATCH] slab: Adds missing kmalloc() checks.

On Fri, 06 Jan 2006 16:24:47 +0100
Arjan van de Ven <[email protected]> wrote:

| On Fri, 2006-01-06 at 13:12 -0200, Luiz Fernando Capitulino wrote:
| > Adds two missing kmalloc() checks in kmem_cache_init(). Note that if the
| > allocation fails, there is nothing to do, so we panic();
|
| ok so what good does this do? if you die this early.. you are in deeper
| problems, and can't boot. while this makes the code bigger...

Well, you'll get a panic with a message saying you have no memory to
boot, instead of a OOPS with a kernel NULL pointer derefecence, which
will make you look for a bug.

--
Luiz Fernando N. Capitulino

2006-01-07 02:13:26

by Pekka Enberg

[permalink] [raw]
Subject: Re: [PATCH] slab: Adds missing kmalloc() checks.

Hi,

On Fri, 2006-01-06 at 13:12 -0200, Luiz Fernando Capitulino wrote:
> | > Adds two missing kmalloc() checks in kmem_cache_init(). Note that if the
> | > allocation fails, there is nothing to do, so we panic();

On Fri, 06 Jan 2006 16:24:47 +0100
Arjan van de Ven <[email protected]> wrote:
> | ok so what good does this do? if you die this early.. you are in deeper
> | problems, and can't boot. while this makes the code bigger...

On Fri, 2006-01-06 at 13:30 -0200, Luiz Fernando Capitulino wrote:
> Well, you'll get a panic with a message saying you have no memory to
> boot, instead of a OOPS with a kernel NULL pointer derefecence, which
> will make you look for a bug.

The code is in init section so I don't think size is an issue. A plain
BUG_ON would be better though as it can be disabled by the embedded
folk.

Pekka

Subject: Re: [PATCH] slab: Adds missing kmalloc() checks.


Hi Pekka,

On Sat, 07 Jan 2006 04:12:50 +0200
Pekka Enberg <[email protected]> wrote:

| Hi,
|
| On Fri, 2006-01-06 at 13:12 -0200, Luiz Fernando Capitulino wrote:
| > | > Adds two missing kmalloc() checks in kmem_cache_init(). Note that if the
| > | > allocation fails, there is nothing to do, so we panic();
|
| On Fri, 06 Jan 2006 16:24:47 +0100
| Arjan van de Ven <[email protected]> wrote:
| > | ok so what good does this do? if you die this early.. you are in deeper
| > | problems, and can't boot. while this makes the code bigger...
|
| On Fri, 2006-01-06 at 13:30 -0200, Luiz Fernando Capitulino wrote:
| > Well, you'll get a panic with a message saying you have no memory to
| > boot, instead of a OOPS with a kernel NULL pointer derefecence, which
| > will make you look for a bug.
|
| The code is in init section so I don't think size is an issue. A plain
| BUG_ON would be better though as it can be disabled by the embedded
| folk.

Okay, but a quick look at the initialization functions called from
'init/main.c' shows that some of them does the same thing (test an
error condition, and panic() if necessary).

Should they be changed to call BUG_ON() instead?

PS: I'm asking becuase I also found several untested returns, I'll
send patches for they too.

--
Luiz Fernando N. Capitulino