Re: [PATCH 2/4] remoteproc: Split firmware name allocation from rproc_alloc()

[Bjorn Andersson <bjorn.andersson@linaro.org>]

On Tue 14 Apr 12:44 PDT 2020, Mathieu Poirier wrote:

> Hey Bjorn,
> 
> On Mon, Apr 13, 2020 at 05:55:06PM -0700, Bjorn Andersson wrote:
> > On Mon 13 Apr 13:56 PDT 2020, Alex Elder wrote:
> > 
> > > On 4/13/20 2:33 PM, Mathieu Poirier wrote:
> > > > Make the firmware name allocation a function on its own in order to
> > > > introduce more flexibility to function rproc_alloc().
> > > > 
> > > > Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
> > > 
> > > I didn't look at the larger context (MCU series); I'm only looking
> > > at this (and the others in this series) in isolation.  I like
> > > that you're encapsulating this stuff into functions but doing so
> > > doesn't really add any flexibility.
> > > 
> > > Two small suggestions for you to consider but they're truly
> > > more about style so it's entirely up to you.  Outside of that
> > > this looks straightforward to me, and the result of the series
> > > is an improvement.
> > > 
> > > I'll let you comment on my suggestions before offering my
> > > "reviewed-by" indication.
> > > 
> > > 					-Alex
> > > 
> > > > ---
> > > >  drivers/remoteproc/remoteproc_core.c | 66 ++++++++++++++++------------
> > > >  1 file changed, 39 insertions(+), 27 deletions(-)
> > > > 
> > > > diff --git a/drivers/remoteproc/remoteproc_core.c b/drivers/remoteproc/remoteproc_core.c
> > > > index 80056513ae71..4dee63f319ba 100644
> > > > --- a/drivers/remoteproc/remoteproc_core.c
> > > > +++ b/drivers/remoteproc/remoteproc_core.c
> > > > @@ -1979,6 +1979,33 @@ static const struct device_type rproc_type = {
> > > >  	.release	= rproc_type_release,
> > > >  };
> > > >  
> > > > +static int rproc_alloc_firmware(struct rproc *rproc,
> > > > +				const char *name, const char *firmware)
> > > > +{
> > > > +	char *p, *template = "rproc-%s-fw";
> > > > +	int name_len;
> > > 
> > > Not a big deal (and maybe it's not consistent with other nearby
> > > style) but template and name_len could be defined inside the
> > > "if (!firmware)" block.
> > > 
> > 
> > I prefer variables declared in the beginning of the function, so I'm
> > happy with this.
> > 
> > > > +	if (!firmware) {
> > > > +		/*
> > > > +		 * If the caller didn't pass in a firmware name then
> > > > +		 * construct a default name.
> > > > +		 */
> > > > +		name_len = strlen(name) + strlen(template) - 2 + 1;
> > > > +		p = kmalloc(name_len, GFP_KERNEL);
> > > 
> > > 
> > > I don't know if it would be an improvement, but you could
> > > check for a null p value below for both cases.  I.e.:
> > > 
> > > 		if (p)
> > > 			snprintf(p, ...);
> > > 
> > 
> > Moving the common NULL check and return out seems nice, but given that
> > we then have to have this positive conditional I think the end result is
> > more complex.
> > 
> > That said, if we're not just doing a verbatim copy from rproc_alloc() I
> > think we should make this function:
> > 
> > 	if (!firmware)
> > 		p = kasprintf(GFP_KERNEL, "rproc-%s-fw", name);
> > 	else
> > 		p = kstrdup_const(firmware, GFP_KERNEL);
> 
> If you really want to use kstrdup_const() the return value has to be casted to a
> "char *".  Variable 'p' can't be declared const "char *" because rproc->firmware is not
> a "const".  Simply put somewhere the "const" will need to be dropped or casted out.
> 

The firmware parameter to rproc_alloc() is const char * and there's a
couple of places where a really const string is passed, so by using
kstrdup_const() we don't end up duplicating const data on the heap.

And afaict we can make both p and rproc->firmware const char * to allow
this, or am I missing something?

Regards,
Bjorn