RE: [PATCH 1/3] spi/qspi: Add memory mapped read support.

["Gupta, Pekon" <pekon-l0cyMroinI0@public.gmane.org>]

> From:  Poddar, Sourav
> > On Tuesday 15 October 2013 06:16 PM, Mark Brown wrote:
> > On Tue, Oct 15, 2013 at 05:19:07PM +0530, Sourav Poddar wrote:
> >> On Tuesday 15 October 2013 04:46 PM, Mark Brown wrote:
> >>> Can you fix this by enabling the clock is enabled when you return the
> >>> buffer to the MTD layer and then disabling the clock when the buffer is
> >>> released?
> >> Sorry, I did not get you here. With memory mapped read, there is no
> >> buffer exchanged, everything takes place at the mtd layer only, what gets
> >> exchanged is just the memory mapped address.
> > The buffer is the memory mapped address - part of getting the address
> > should be preparing the hardware for it.
> >
> >>       if (spi->mode&&  SPI_RX_MMAP) {
> >>           printk("memory mapped mode set\n");
> >> -        flash->mmap_read = true;
> >> +        flash->mmap_read = spi->memory_map;
> > So this probably needs to be a function call to get the buffer (and a
> > corresponding one to free it).
> So, the flow can be something like this:
> 
> drivers/mtd/devices/m25p80.c
> get_flash_buf()
> {
>      lock();
> 
>       t[0] = GET_BUFFER;
>       t[1] = buf;
>       ......
> 
>       spi_sync();
> 
>      unlock();
> }
> 
Problem here..
spi_sync() is not blocking, that means it would just add a spi_message
to queue and return. And it depends on kthread_worker when it pumps
this spi_message to QPSI controller driver for actual configuration.
So this is actually a race-condition. You cannot use spi_sync() to configure.

(refer my comments in previous emails).
If you really want to configure QSPI controller just before memcpy(),
Then you need to somehow prevent spi kthead_worker from accessing.
This you can do by locking the spi_meesage queue/list at that time.


> mtd_read
> {
>      get_flash_buf();
> 
>      if (flash->buf) {
>          memcpy();
>         return 0;
>    }
> }
> 
> Not sure, if free buf is needed as devm_* variant is used to allocate that
> memory.
> 
> 
> }

With regards, pekon

------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from 
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60135031&iu=/4140/ostg.clktrk