[U-Boot] [PATCH] cmd_sf: add handler for +len arg for all commands

[Maxime Hadjinlian]

Hi all,

On Wed, Oct 8, 2014 at 10:35 AM, Chin Liang See <clsee at altera.com> wrote:
> On Sun, 2014-10-05 at 22:55 +0200, marex at denx.de wrote:
>> On Sunday, October 05, 2014 at 10:35:06 PM, Maxime Hadjinlian wrote:
>> > On Sun, Oct 5, 2014 at 10:28 PM, Marek Vasut <marex at denx.de> wrote:
>> > > On Sunday, October 05, 2014 at 09:51:27 PM, Maxime Hadjinlian wrote:
>> > >> On Sun, Oct 5, 2014 at 9:45 PM, Marek Vasut <marex at denx.de> wrote:
>> > >> > On Sunday, October 05, 2014 at 09:27:00 PM, Maxime Hadjinlian wrote:
>> > >> >> On Sun, Oct 5, 2014 at 9:19 PM,  <thomas.langer at lantiq.com> wrote:
>> > >> >> >> On Sunday, October 05, 2014 at 08:40:26 PM, Maxime Hadjinlian wrote:
>> > >> >> >> > Hi Thomas, all,
>> > >> >> >> >
>> > >> >> >> > On Sun, Oct 5, 2014 at 7:43 PM,  <thomas.langer at lantiq.com> wrote:
>> > >> >> >> > > Hello Maxime,
>> > >> >> >> > >
>> > >> >> >> > > can you explain the usecase?
>> > >> >> >> > > I think, only the erase is executed per sector, all other
>> > >> >> >> > > commands are working fine with a byte oriented length.
>> > >> >> >> >
>> > >> >> >> > I need to write a file that is downloaded through TFTP. So I can
>> > >> >> >> > get the filesize through the variable of the same name, but if
>> > >> >> >> > it's not rounded, the write command may fail.
>> > >> >> >> > I can save the filesize in another variable, but at next boot,
>> > >> >> >> > when I need to read this file, I can't read the file, since I
>> > >> >> >> > only know it's size in byte, I need to be able to round it
>> > >> >> >> > again.
>> > >> >> >>
>> > >> >> >> I wonder, do all SPI flashes need to do sector-aligned writes ?
>> > >> >> >
>> > >> >> > All the serial flashes I have seen so far do support reading and
>> > >> >> > writing with any length, independent from the erase size.
>> > >> >> > Otherwise the current implementation of env_sf.c would also not
>> > >> >> > work.
>> > >> >>
>> > >> >> Well the QSPI, I used, does not. Every read and write has to be
>> > >> >> aligned. That's why I needed that in the first place.
>> > >> >
>> > >> > Aligned how exactly?
>> > >>
>> > >> sector-aligned.
>> > >> If I write anything that is not sector aligned, the 'sf' command will
>> > >> fail.
>> > >
>> > > Is that really a property of the SPI flash or the SPI controller ? And if
>> > > the later, is it a property of a particular one or is that a common
>> > > thing ?
>> >
>> > I'll have to check but if I remember correctly, it's the SPI flash.
>> > It's a Micron Serial NOR flash (N25Q512A83G1240x).
>> > And the controller is the Cadence QSPI controller available on the SoC
>> > FPGA platform.
>> > Maybe what caused my issues was the driver after all ?
>>
>> +CC Altera guys, they might help too.
>
> Hmmm.... quite strange to write for sector aligned.
> Wonder the write related to filesystem access?
>
> Within Cadence QSPI controller, read and write access need to be done
> within a page. But that would be taken care by controller when the
> access beyond the next page. From software perspective, we just tell
> controller the numbers of bytes to be transfer (for indirect transfer
> setup).
My needs have evolved a bit, and I no longer require this function.
But I wasn't able to right a file tftp into RAM using 'sf update'
without this patch.
I could try to reproduce the errors and maybe dig in a little more.
>
> Chin Liang
>
>>
>> Best regards,
>> Marek Vasut
>
>