Re: [PATCH 1/2] firewire: firewire is a big-endian bus

[Stefan Richter <stefanr@s5r6.in-berlin.de>]

On Apr 20 Joe Perches wrote:
> On Tue, 2015-04-21 at 02:36 +0200, Laurent Vivier wrote:
> > So, dump config_rom data as big-endian values.
> > 
> > The value given by /sys/bus/firewire/devices/fw0 were correctly
> > given on a big-endian host (like powermac) not on a little-endian host
> > (like PC), for instance:
> []
> > diff --git a/drivers/firewire/core-device.c b/drivers/firewire/core-device.c
> []
> > @@ -399,14 +399,14 @@ static ssize_t config_rom_show(struct device *dev,
> >  			       struct device_attribute *attr, char *buf)
> >  {
> >  	struct fw_device *device = fw_device(dev);
> > -	size_t length;
> > +	size_t i;
> >  
> >  	down_read(&fw_device_rwsem);
> > -	length = device->config_rom_length * 4;
> > -	memcpy(buf, device->config_rom, length);
> > +	for (i = 0; i < device->config_rom_length; i++)
> > +		((u32 *)buf)[i] = be32_to_cpu(device->config_rom[i]);
> 
> Is buf guaranteed to be appropriately aligned on a u32?

Good catch.  Though as far as I know, the buffer is page-sized and thereby
certainly page-aligned.  This could be looked up in the sysfs core code.

BTW, theoretically speaking it should be cpu_to_be32() here, and the buffer
pointer should be annotated as a big endian pointer.  (But as I noted in
the previous reply, the patch will not be applied either way.)
-- 
Stefan Richter
-=====-===== -=-- =-=-=
http://arcgraph.de/sr/