[PATCH] Fix a memory leak in pdaudiocf

[PATCH] Fix a memory leak in pdaudiocf

[Jesper Juhl]

There's a potential memory leak in 
sound/pcmcia/pdaudiocf/pdaudiocf.c::pdacf_config()

If we leave via one of the *failed: labels we may leak 'parse', so add a 
kfree(parse) to the end of the function and also make sure to set 'parse' 
to NULL after the kfree() call a little further up so we don't do a 
double-free of the pointer if we hit one of the *failed: labels after the 
first kfree().

Since I don't have the hardware I can't test the patch beyond making sure 
it compiles cleanly, but I feel pretty confident that it is correct.

Please consider for inclusion.


Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
---

 sound/pcmcia/pdaudiocf/pdaudiocf.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletion(-)

--- linux-2.6.17-rc4-git2-orig/sound/pcmcia/pdaudiocf/pdaudiocf.c	2006-05-13 21:28:55.000000000 +0200
+++ linux-2.6.17-rc4-git2/sound/pcmcia/pdaudiocf/pdaudiocf.c	2006-05-13 23:27:46.000000000 +0200
@@ -226,7 +226,7 @@ static int pdacf_config(struct pcmcia_de
 
 	snd_printdd(KERN_DEBUG "pdacf_config called\n");
 	parse = kmalloc(sizeof(*parse), GFP_KERNEL);
-	if (! parse) {
+	if (!parse) {
 		snd_printk(KERN_ERR "pdacf_config: cannot allocate\n");
 		return -ENOMEM;
 	}
@@ -242,6 +242,7 @@ static int pdacf_config(struct pcmcia_de
 	link->conf.ConfigBase = parse->config.base;
 	link->conf.ConfigIndex = 0x5;
 	kfree(parse);
+	parse = NULL;
 
 	CS_CHECK(RequestIO, pcmcia_request_io(link, &link->io));
 	CS_CHECK(RequestIRQ, pcmcia_request_irq(link, &link->irq));
@@ -257,6 +258,7 @@ cs_failed:
 	cs_error(link, last_fn, last_ret);
 failed:
 	pcmcia_disable_device(link);
+	kfree(parse);
 	return -ENODEV;
 }
 





PS. Please keep me on Cc when replying since I'm not subscribed to alsa-devel.

Re: [Alsa-devel] [PATCH] Fix a memory leak in pdaudiocf

[Takashi Iwai]

At Sat, 13 May 2006 23:44:25 +0200,
Jesper Juhl wrote:
> 
> There's a potential memory leak in 
> sound/pcmcia/pdaudiocf/pdaudiocf.c::pdacf_config()
> 
> If we leave via one of the *failed: labels we may leak 'parse', so add a 
> kfree(parse) to the end of the function and also make sure to set 'parse' 
> to NULL after the kfree() call a little further up so we don't do a 
> double-free of the pointer if we hit one of the *failed: labels after the 
> first kfree().
> 
> Since I don't have the hardware I can't test the patch beyond making sure 
> it compiles cleanly, but I feel pretty confident that it is correct.
> 
> Please consider for inclusion.

This bug was already fixed in a different way on ALSA HG repo.

thanks,

Takashi