Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
From:   "Fabio M. De Francesco" <fmdefrancesco@gmail.com>
To:     Ira Weiny <ira.weiny@intel.com>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Benjamin Philip <benjamin.philip495@gmail.com>,
        Bart Van Assche <bvanassche@acm.org>,
        "Martin K. Petersen" <martin.petersen@oracle.com>,
        Charlie Sands <sandsch@northvilleschools.net>,
        Mitali Borkar <mitaliborkar810@gmail.com>,
        Colin Ian King <colin.king@intel.com>,
        linux-staging@lists.linux.dev, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] staging: rts5208: Convert kmap() to kmap_local_page()
Date:   Mon, 28 Mar 2022 19:21:26 +0200
Message-ID: <1816944.CQOukoFCf9@leap>
In-Reply-To: <YkHXQRmExRFioEBN@iweiny-desk3>
References: <20220328112440.17756-1-fmdefrancesco@gmail.com> <YkHXQRmExRFioEBN@iweiny-desk3>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="us-ascii"
Precedence: bulk

On luned? 28 marzo 2022 17:41:53 CEST Ira Weiny wrote:
> On Mon, Mar 28, 2022 at 01:24:40PM +0200, Fabio M. De Francesco wrote:
> > The use of kmap() is being deprecated and kmap_local_page() is faster.
> > Use kmap_local_page() in place of kmap().
> 
> Thanks for the patch!  

Thanks for your review!

> I have just a couple of comments.
> 
> kmap_local_page() is not necessarily faster than kmap() but it is more correct
> in this case.  You should mention why.

Sure, my mistake. Thomas G. was talking about kmap_atomic() when he wrote that
it is "faster". It does not apply to kmap_local_page(). 

What could justify the use of kmap_local_page() is that "The mapping is per 
thread, CPU local and not globally visible.". Therefore this code is the right 
context where to use kmap_local_page() in place of kmap().

At this moment I think that I might change my commit message and write something
like the above.

However, I'll research more information during the next days. In the meantime 
I'm also going to take a look at the differences in implementation.

> Also to help with kmap_local_page() there are a number of helpers implemented
> in highmem.h for things like memcpy, memmove, etc.
> 
> Check out memcpy_page() for this use case.

Aren't memcpy_to_page() and memcpy_from_page() better suited for the two 
different branches of the "if" statement?

Thank you,

Fabio M. De Francesco

> 
> Thank you!
> Ira
> 
> > 
> > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
> > ---
> >  drivers/staging/rts5208/rtsx_transport.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/staging/rts5208/rtsx_transport.c b/drivers/staging/rts5208/rtsx_transport.c
> > index 805dc18fac0a..de690d7ee5e3 100644
> > --- a/drivers/staging/rts5208/rtsx_transport.c
> > +++ b/drivers/staging/rts5208/rtsx_transport.c
> > @@ -92,13 +92,13 @@ unsigned int rtsx_stor_access_xfer_buf(unsigned char *buffer,
> >  			while (sglen > 0) {
> >  				unsigned int plen = min(sglen, (unsigned int)
> >  						PAGE_SIZE - poff);
> > -				unsigned char *ptr = kmap(page);
> > +				unsigned char *ptr = kmap_local_page(page);
> >  
> >  				if (dir == TO_XFER_BUF)
> >  					memcpy(ptr + poff, buffer + cnt, plen);
> >  				else
> >  					memcpy(buffer + cnt, ptr + poff, plen);
> > -				kunmap(page);
> > +				kunmap_local(ptr);
> >  
> >  				/* Start at the beginning of the next page */
> >  				poff = 0;
> > -- 
> > 2.34.1
> > 
>