[PATCH v2] cifs: call print_hex_dump instead of custom implementation

[Andy Shevchenko]

On Mon, 2014-05-12 at 12:23 -0400, Jeff Layton wrote:
> On Mon, 12 May 2014 19:15:09 +0300
> Andy Shevchenko <andriy.shevchenko at linux.intel.com> wrote:
> 
> > On Mon, 2014-05-12 at 11:42 -0400, Jeff Layton wrote:
> > > On Mon, 12 May 2014 18:15:20 +0300
> > > Andy Shevchenko <andriy.shevchenko at linux.intel.com> wrote:
> > > 
> > > > This patch converts custom dumper to use native print_hex_dump() instead. The
> > > > dump will have a given label and addresses per each line which differs it from
> > > > the original code.
> > > > 
> > > > Signed-off-by: Andy Shevchenko <andriy.shevchenko at linux.intel.com>
> > > > ---
> > > >  fs/cifs/cifs_debug.c | 21 ++-------------------
> > > >  1 file changed, 2 insertions(+), 19 deletions(-)
> > > > 
> > > > diff --git a/fs/cifs/cifs_debug.c b/fs/cifs/cifs_debug.c
> > > > index fa78b68..e7b87ce 100644
> > > > --- a/fs/cifs/cifs_debug.c
> > > > +++ b/fs/cifs/cifs_debug.c
> > > > @@ -34,27 +34,10 @@
> > > >  void
> > > >  cifs_dump_mem(char *label, void *data, int length)
> > > >  {
> > > > -	int i, j;
> > > > -	int *intptr = data;
> > > > -	char *charptr = data;
> > > > -	char buf[10], line[80];
> > > > -
> > > >  	printk(KERN_DEBUG "%s: dump of %d bytes of data at 0x%p\n",
> > > >  		label, length, data);
> > > > -	for (i = 0; i < length; i += 16) {
> > > > -		line[0] = 0;
> > > > -		for (j = 0; (j < 4) && (i + j * 4 < length); j++) {
> > > > -			sprintf(buf, " %08x", intptr[i / 4 + j]);
> > > > -			strcat(line, buf);
> > > > -		}
> > > > -		buf[0] = ' ';
> > > > -		buf[2] = 0;
> > > > -		for (j = 0; (j < 16) && (i + j < length); j++) {
> > > > -			buf[1] = isprint(charptr[i + j]) ? charptr[i + j] : '.';
> > > > -			strcat(line, buf);
> > > > -		}
> > > > -		printk(KERN_DEBUG "%s\n", line);
> > > > -	}
> > > > +	print_hex_dump(KERN_DEBUG, label, DUMP_PREFIX_ADDRESS, 16, 4,
> > > > +		       data, length, true);
> > > 
> > > I'm not sure we want "label" as the prefix_str as it'll go on every
> > > line. Maybe just use "" in place of "label" there?
> > 
> > It was in my initial v1. I thought it would be better to have.
> 
> Most of the labels are stuff like "Bad SMB:" and I'm not sure there's a
> lot of value in repeating that on every line. The existing code
> certainly doesn't and that sort of thing tends to clutter up logfiles.

Got it.

> 
> > What about addresses? Would you like to see it or offset is enough, or
> > drop them as well?
> > 
> 
> The addresses aren't terribly helpful in general, and printing them
> might be considered information leakage. I'd suggest going with just
> the offsets.

Yeah, but since printk at the top of function already prints virtual
address... Maybe you could fix it in separate patch if you want, I leave
as agreed — just offsets and no label.

> 
> > Moreover, I found another place where similar change could be done
> > (dump_smb), so, would it be better to do in separate patch?
> > 
> 
> Sure, that certainly wouldn't hurt. Your call on whether to do two
> patches or one.

I'm going to resend a whole patchseries as v3 with coverletter and
changelog, since I missed CC you in previous patch, and AB's
Reviewed-by. Also found rebase error which makes compilation fail.
And printk->pr_* change as new patch as well, will be 3 all together.

> 
> 
> > > 
> > > >  }
> > > >  
> > > >  #ifdef CONFIG_CIFS_DEBUG
> > > 
> > > Otherwise, this looks good.
> > > 
> > > Acked-by: Jeff Layton <jlayton at poochiereds.net>
> > 
> > 
> 


-- 
Andy Shevchenko <andriy.shevchenko at intel.com>
Intel Finland Oy