Re: [PATCH] usb: musb: add printf attribute to log function

From: Joe Perches <joe@perches.com>
To: Greg KH <gregkh@linuxfoundation.org>, trix@redhat.com
Cc: b-liu@ti.com, linux-usb@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] usb: musb: add printf attribute to log function
Date: Tue, 22 Dec 2020 01:52:48 -0800	[thread overview]
Message-ID: <6ea843411793073040bb8d518fca84f5b66b86aa.camel@perches.com> (raw)
In-Reply-To: <X+Gzvo3vPBFGKAy+@kroah.com>

On Tue, 2020-12-22 at 09:52 +0100, Greg KH wrote:
> On Mon, Dec 21, 2020 at 08:25:47AM -0800, trix@redhat.com wrote:
> > From: Tom Rix <trix@redhat.com>
> > 
> > Attributing the function allows the compiler to more thoroughly
> > check the use of the function with -Wformat and similar flags.
> > 
> > Signed-off-by: Tom Rix <trix@redhat.com>
> > ---
> >  drivers/usb/musb/musb_debug.h | 1 +
> >  1 file changed, 1 insertion(+)
> > 
> > diff --git a/drivers/usb/musb/musb_debug.h b/drivers/usb/musb/musb_debug.h
> > index e5b3506c7b3f..dfc0d02695fa 100644
> > --- a/drivers/usb/musb/musb_debug.h
> > +++ b/drivers/usb/musb/musb_debug.h
> > @@ -17,6 +17,7 @@
> >  #define INFO(fmt, args...) yprintk(KERN_INFO, fmt, ## args)
> >  #define ERR(fmt, args...) yprintk(KERN_ERR, fmt, ## args)
> >  
> > 
> > +__printf(2, 3)
> >  void musb_dbg(struct musb *musb, const char *fmt, ...);
> 
> While I understand the need for this, did this find any problems?
> If not, then it's not worth adding,

I have to disagree with that Greg.  While the driver isn't in active
development, a trivial mod to make it less likely a defect is introduced
by any additional code is still a useful addition.

> the driver-specific debugging macros
> should be removed entirely and just use dev_dbg() and friends instead.

Read the suggested change I posted in reply.

btw: the musb_dbg function is actually a trace function and not a
dmesg/logging mechanism.

drivers/usb/musb/musb_trace.c:void musb_dbg(struct musb *musb, const char *fmt, ...)
drivers/usb/musb/musb_trace.c-{
drivers/usb/musb/musb_trace.c-  struct va_format vaf;
drivers/usb/musb/musb_trace.c-  va_list args;
drivers/usb/musb/musb_trace.c-
drivers/usb/musb/musb_trace.c-  va_start(args, fmt);
drivers/usb/musb/musb_trace.c-  vaf.fmt = fmt;
drivers/usb/musb/musb_trace.c-  vaf.va = &args;
drivers/usb/musb/musb_trace.c-
drivers/usb/musb/musb_trace.c-  trace_musb_log(musb, &vaf);
drivers/usb/musb/musb_trace.c-
drivers/usb/musb/musb_trace.c-  va_end(args);
drivers/usb/musb/musb_trace.c-}

drivers/usb/musb/musb_trace.h:TRACE_EVENT(musb_log,
drivers/usb/musb/musb_trace.h-  TP_PROTO(struct musb *musb, struct va_format *vaf),
drivers/usb/musb/musb_trace.h-  TP_ARGS(musb, vaf),
drivers/usb/musb/musb_trace.h-  TP_STRUCT__entry(
drivers/usb/musb/musb_trace.h-          __string(name, dev_name(musb->controller))
drivers/usb/musb/musb_trace.h-          __dynamic_array(char, msg, MUSB_MSG_MAX)
drivers/usb/musb/musb_trace.h-  ),
drivers/usb/musb/musb_trace.h-  TP_fast_assign(
drivers/usb/musb/musb_trace.h-          __assign_str(name, dev_name(musb->controller));
drivers/usb/musb/musb_trace.h-          vsnprintf(__get_str(msg), MUSB_MSG_MAX, vaf->fmt, *vaf->va);
drivers/usb/musb/musb_trace.h-  ),
drivers/usb/musb/musb_trace.h-  TP_printk("%s: %s", __get_str(name), __get_str(msg))
drivers/usb/musb/musb_trace.h-);

Is that trace mechanism useful though?  I think it's somewhat odd.