Re: [PATCH 2/4] dmi: Add a DMI firmware node and handling

[Jean Delvare <jdelvare@suse.de>]

Hi Corey,

I won't comment on the IPMI side of this as this isn't my area. However
I have a comment on the DMI part:

Le Friday 29 January 2016 à 16:43 -0600, minyard@acm.org a écrit :
> From: Corey Minyard <cminyard@mvista.com>
> 
> This is so that an IPMI platform device can be created from a
> DMI firmware entry.
> 
> Signed-off-by: Corey Minyard <cminyard@mvista.com>
> Cc: Jean Delvare <jdelvare@suse.de>
> Cc: Andy Lutomirski <luto@kernel.org>
> ---
>  drivers/firmware/dmi_scan.c | 34 ++++++++++++++++++++++++----------
>  include/linux/dmi.h         | 24 ++++++++++++++++++++++++
>  include/linux/fwnode.h      |  1 +
>  3 files changed, 49 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/firmware/dmi_scan.c b/drivers/firmware/dmi_scan.c
> index da471b2..13d9bca 100644
> --- a/drivers/firmware/dmi_scan.c
> +++ b/drivers/firmware/dmi_scan.c
> @@ -41,6 +41,16 @@ static struct dmi_memdev_info {
>  } *dmi_memdev;
>  static int dmi_memdev_nr;
>  
> +static void *dmi_zalloc(unsigned len)
> +{
> +	void *ret = dmi_alloc(len);
> +
> +	if (ret)
> +		memset(ret, 0, len);
> +
> +	return ret;
> +}
> +
>  static const char * __init dmi_string_nosave(const struct dmi_header *dm, u8 s)
>  {
>  	const u8 *bp = ((u8 *) dm) + dm->length;
> @@ -242,6 +252,12 @@ static void __init dmi_save_type(const struct dmi_header *dm, int slot,
> (...)
> @@ -250,15 +266,14 @@ static void __init dmi_save_one_device(int type, const char *name)
>  	if (dmi_find_device(type, name, NULL))
>  		return;
>  
> -	dev = dmi_alloc(sizeof(*dev) + strlen(name) + 1);
> +	dev = dmi_zalloc(sizeof(*dev) + strlen(name) + 1);
>  	if (!dev)
>  		return;
>  
>  	dev->type = type;
>  	strcpy((char *)(dev + 1), name);
>  	dev->name = (char *)(dev + 1);
> -	dev->device_data = NULL;

This change seems rather unrelated, and I'm not sure what purpose it
serves. On ia64 and arm64 it is clearly redundant as dmi_alloc calls
kzalloc directly. On x86_64, extend_brk is called instead (don't ask me
why, I have no clue) but looking at the code I see that it does
memset(ret, 0, size) as well so memory is also zeroed there. Which makes
dmi_alloc the same as dmi_zalloc on all 3 architectures.

So please revert this change. This will make your patch easier to
review, too.

-- 
Jean Delvare
SUSE L3 Support