Re: [PATCH] Bluetooth: Fix uuid output in debugfs

[Johan Hedberg <johan.hedberg@gmail.com>]

Hi Gustavo,

On Fri, Dec 21, 2012, Gustavo Padovan wrote:
> The uuid should be printed in the CPU endianness and not in little-endian.
> 
> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> ---
>  net/bluetooth/hci_sysfs.c | 25 ++++++++++++-------------
>  1 file changed, 12 insertions(+), 13 deletions(-)
> 
> diff --git a/net/bluetooth/hci_sysfs.c b/net/bluetooth/hci_sysfs.c
> index 55cceee..05b78c7 100644
> --- a/net/bluetooth/hci_sysfs.c
> +++ b/net/bluetooth/hci_sysfs.c
> @@ -461,19 +461,18 @@ static const struct file_operations blacklist_fops = {
>  
>  static void print_bt_uuid(struct seq_file *f, u8 *uuid)
>  {
> -	__be32 data0, data4;
> -	__be16 data1, data2, data3, data5;
> -
> -	memcpy(&data0, &uuid[0], 4);
> -	memcpy(&data1, &uuid[4], 2);
> -	memcpy(&data2, &uuid[6], 2);
> -	memcpy(&data3, &uuid[8], 2);
> -	memcpy(&data4, &uuid[10], 4);
> -	memcpy(&data5, &uuid[14], 2);
> -
> -	seq_printf(f, "%.8x-%.4x-%.4x-%.4x-%.8x%.4x\n",
> -		   ntohl(data0), ntohs(data1), ntohs(data2), ntohs(data3),
> -		   ntohl(data4), ntohs(data5));
> +	u32 data0, data5;
> +	u16 data1, data2, data3, data4;
> +
> +	data5 = le32_to_cpu(*(__le32 *)uuid);
> +	data4 = le16_to_cpu(*(__le16 *)(uuid + 4));
> +	data3 = le16_to_cpu(*(__le16 *)(uuid + 6));
> +	data2 = le16_to_cpu(*(__le16 *)(uuid + 8));
> +	data1 = le16_to_cpu(*(__le16 *)(uuid + 10));
> +	data0 = le32_to_cpu(*(__le32 *)(uuid + 12));

This looks prone to unaligned access violations if the "u8 *uuid"
pointer doesn't start off with a nicely aligned address. The use of the
unaligned getter macros would also look nicer since you wouldn't have to
do explicit type casting.

Johan