Re: [PATCH] powerpc/powernv: Add a null pointer check to scom_debug_init_one

[Andrew Donnellan]

On Fri, 2023-12-08 at 16:59 +0800, Kunwu Chan wrote:
> kasprintf() returns a pointer to dynamically allocated memory
> which can be NULL upon failure.
> Add a null pointer check, and release 'ent' to avoid memory leaks.
> 
> Fixes: bfd2f0d49aef ("powerpc/powernv: Get rid of old scom_controller
> abstraction")

[+ robh]

This commit just reshuffled around some existing code. The commit that
appears to have added this is actually b7c670d673d1 ("powerpc: Convert
to using %pOF instead of full_name").

> Cc: Kunwu Chan <kunwu.chan@xxxxxxxxxxx>
> Signed-off-by: Kunwu Chan <chentao@xxxxxxxxxx>
> ---
>  arch/powerpc/platforms/powernv/opal-xscom.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/arch/powerpc/platforms/powernv/opal-xscom.c
> b/arch/powerpc/platforms/powernv/opal-xscom.c
> index 262cd6fac907..748c2b97fa53 100644
> --- a/arch/powerpc/platforms/powernv/opal-xscom.c
> +++ b/arch/powerpc/platforms/powernv/opal-xscom.c
> @@ -165,6 +165,11 @@ static int scom_debug_init_one(struct dentry
> *root, struct device_node *dn,
>  	ent->chip = chip;
>  	snprintf(ent->name, 16, "%08x", chip);
>  	ent->path.data = (void *)kasprintf(GFP_KERNEL, "%pOF", dn);
> +	if (!ent->path.data) {
> +		kfree(ent);
> +		return -ENOMEM;

The caller of this function (scom_debug_init()) uses a bitwise OR to
aggregate the return codes of multiple calls to scom_debug_init_one().
This doesn't really work for returning specific error codes, so I'd
just return -1 here (or change the way the return codes are handled on
the caller side).

> +	}
> +
>  	ent->path.size = strlen((char *)ent->path.data);
>  
>  	dir = debugfs_create_dir(ent->name, root);

-- 
Andrew Donnellan    OzLabs, ADL Canberra
ajd@xxxxxxxxxxxxx   IBM Australia Limited