[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [Nbd] [PATCH 6/9] nbd: Restructure debugging prints



On Fri, 2015-02-13 at 12:24 +0100, Markus Pargmann wrote:
> On Fri, Feb 13, 2015 at 02:05:27AM -0800, Joe Perches wrote:
> > On Fri, 2015-02-13 at 10:58 +0100, Markus Pargmann wrote:
> > > On Thu, Feb 12, 2015 at 01:08:48PM -0800, Joe Perches wrote:
> > > > On Thu, 2015-02-12 at 21:57 +0100, Markus Pargmann wrote:
> > > > > dprintk has some name collisions with other frameworks and drivers. It
> > > > > is also not necessary to have these custom debug print filters. Dynamic
> > > > > debug offers the same amount of filtered debugging.
> > > > > 
> > > > > This patch replaces all dprintks with dev_dbg(). It also removes the
> > > > > ioctl dprintk which prints the ingoing ioctls which should be
> > > > > replaceable by strace or similar stuff.
> > > > 
> > > > Perhaps add
> > > > 
> > > > #define nbd_dbg(nbd, fmt, ...)					\
> > > > 	dev_dbg(disk_to_dev((nbd)->disk), "%s: " fmt,		\
> > > > 		nbd->disk->disk_name, ##__VA_ARGS__)
> > > 
> > > I am not really happy with those custom debug print macros. What do you
> > > think about an inline function 'nbd_to_dev' instead?
> > 
> > Wouldn't that change the output?
> > What would the output look like?
> 
> I meant a function that just translates struct nbd_device* to struct
> device*. That wouldn't change the output.
> > > > > diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
> > > > []
> > > > > +static void nbd_end_request(struct nbd_device *nbd, struct request *req)
> > > > >  {
> > > > >  	int error = req->errors ? -EIO : 0;
> > > > >  	struct request_queue *q = req->q;
> > > > >  	unsigned long flags;
> > > > >  
> > > > > -	dprintk(DBG_BLKDEV, "%s: request %p: %s\n", req->rq_disk->disk_name,
> > > > > -			req, error ? "failed" : "done");
> > > > > +	dev_dbg(disk_to_dev(nbd->disk), "%s: request %p: %s\n",
> > > > > +		req->rq_disk->disk_name, req, error ? "failed" : "done");
> > > > 
> > > > so this becomes
> > > > 
> > > > 	nbd_dbg(nbd, "request %p: %s\n",
> > > > 		req, error ? "failed" : "done");
> > > 
> > > so this would be:
> > >  	nbd_dbg(nbd_to_dev(nbd), "request %p: %s\n",
> > >  		req, error ? "failed" : "done");
> > 
> > I don't see much value in that style,
> > but I don't manage the code either.
> 
> Oh sorry, I meant to write:
> 	dev_dbg(nbd_to_dev(nbd), "request %p: %s\n",
> 		req, error ? "failed" : "done");
> 
> So the normal dev_dbg call is still there and the expression to get the
> device from a nbd_device struct is shorter.

Is nbd->disk->disk_name the same string as
disk_to_dev((nbd)->disk)->name?

What's the output of the conversion in this patch?

-	dprintk(DBG_BLKDEV, "%s: request %p: %s\n", req->rq_disk->disk_name,
-			req, error ? "failed" : "done"); 
+	dev_dbg(disk_to_dev(nbd->disk), "%s: request %p: %s\n",
+		req->rq_disk->disk_name, req, error ? "failed" : "done");

Should this conversion be as you wrote above

	dev_dbg(nbd_to_dev(nbd), "request %p: %s\n",
 		req, error ? "failed" : "done");

If so, then there's probably not much use in a
custom nbd_dbg macro.

There is some value though in classifying blocks of
debugging output akin to the use of netif_msg_<type>.

That is lost with these conversions.





Reply to: