Q: Probing for devices > buswidth???

Q: Probing for devices > buswidth???

[Eric W. Biederman]

I'm trying to bringin some mtd map drivers for some non cfi devices.

Trying to understand some anomolous results I discovered that
gen_probe_newchip is looking for:
16 bit devices on an 8 bit bus.
byte interleaved 16 bit devices on a 16 bit bus.
And some really bizare things for a 32 bit bus.

Is there any good reason for doing it that way?

If not I'm going to figure out how to kill that extra confusing code.

Eric

Re: Q: Probing for devices > buswidth???

[David Woodhouse]

ebiederman@lnxi.com said:
> Trying to understand some anomolous results I discovered that
> gen_probe_newchip is looking for: 16 bit devices on an 8 bit bus. byte
> interleaved 16 bit devices on a 16 bit bus. And some really bizare
> things for a 32 bit bus.

> Is there any good reason for doing it that way?

What it's probing for is a 16-bit device in 8-bit mode. This is _different_ 
to an 8-bit device :)


--
dwmw2

Re: Q: Probing for devices > buswidth???

[Eric W. Biederman]

David Woodhouse <dwmw2@infradead.org> writes:

> ebiederman@lnxi.com said:
> > Trying to understand some anomolous results I discovered that
> > gen_probe_newchip is looking for: 16 bit devices on an 8 bit bus. byte
> > interleaved 16 bit devices on a 16 bit bus. And some really bizare
> > things for a 32 bit bus.
> 
> > Is there any good reason for doing it that way?
> 
> What it's probing for is a 16-bit device in 8-bit mode. This is _different_ 
> to an 8-bit device :)

Just so I understand.  There are chips with a 16-bit interface that
you can put in a crippled mode such that they pretend use only an 8
bit bus.  But the id logic is the same as when the are on a 16-bit
bus?

I'm just trying to comprehend this weird situation.

Eric

Re: Q: Probing for devices > buswidth???

[David Woodhouse]

ebiederman@lnxi.com said:
>  Just so I understand.  There are chips with a 16-bit interface that
> you can put in a crippled mode such that they pretend use only an 8
> bit bus.  But the id logic is the same as when the are on a 16-bit
> bus?

Basically, yes. They are 16-bit devices, with the A0 line selecting which 
16-bit word you should look at - and all the 'magic' addresses must be 
placed on the address bus starting at A0. But they decided they wanted to 
be able to use them with an 8-bit bus, so they added an 'A-1' address line 
too.

> I'm just trying to comprehend this weird situation.

Do yourself a favour - don't. Just accept it. Or do me a favour and take 
responsibility for it from now on :)

--
dwmw2

Re: Q: Probing for devices > buswidth???

[Eric W. Biederman]

David Woodhouse <dwmw2@infradead.org> writes:

> ebiederman@lnxi.com said:
> >  Just so I understand.  There are chips with a 16-bit interface that
> > you can put in a crippled mode such that they pretend use only an 8
> > bit bus.  But the id logic is the same as when the are on a 16-bit
> > bus?
> 
> Basically, yes. They are 16-bit devices, with the A0 line selecting which 
> 16-bit word you should look at - and all the 'magic' addresses must be 
> placed on the address bus starting at A0. But they decided they wanted to 
> be able to use them with an 8-bit bus, so they added an 'A-1' address line 
> too.
> 
> > I'm just trying to comprehend this weird situation.
> 
> Do yourself a favour - don't. Just accept it. Or do me a favour and take 
> responsibility for it from now on :)

I'll probably take responsiblity for the legacy jedec side of things.
Or at the very least I'll be a very active tester/user.  So I want
enough understanding so that I don't look at someone else's code and
say that is broken, and be wrong.  I have fixed a whole bunch of small
bugs already.  Currently I have 4 map drivers completed and a 5th one
to write, and none of them use these nice new chips.  The worst case
is the driver for the boot rom on the eepro100 it doesn't even
implement block erase, just chip erase.  

Anyway my smorgase board currently looks like:
eepro100 boot rom.
BIOS on AMD766 southbrige.
BIOS on the l440gx motherboard.
BIOS on Compaq DS10
BIOS connected to the ICH2 (Intel firmware hub.) 
   [This one I should be getting started on in a couple of hours].

Eric

Re: Q: Probing for devices > buswidth???

[Kenneth Johansson]

"Eric W. Biederman" wrote:
> 
> David Woodhouse <dwmw2@infradead.org> writes:
> 
> > ebiederman@lnxi.com said:
> > > Trying to understand some anomolous results I discovered that
> > > gen_probe_newchip is looking for: 16 bit devices on an 8 bit bus. byte
> > > interleaved 16 bit devices on a 16 bit bus. And some really bizare
> > > things for a 32 bit bus.
> >
> > > Is there any good reason for doing it that way?
> >
> > What it's probing for is a 16-bit device in 8-bit mode. This is _different_
> > to an 8-bit device :)
> 
> Just so I understand.  There are chips with a 16-bit interface that
> you can put in a crippled mode such that they pretend use only an 8
> bit bus.  But the id logic is the same as when the are on a 16-bit
> bus?

Yes. I have a pcmcia linear flash device that is using Intel strataflash
28F640xxx. The chips is 16 bit but used as 8bit. To make things interesting
they also interleave the chip so odd bytes goes to one chip and even to the
other.

This makes the effective block size 256KB and with JFFS2 I need 5 blocks free
minimum so I wast 1.2 MB regardless of total flash bank size :( 


I get this rather confusing message from the probe.
Physically mapped flash: Found 2 x16 devices at 0x0 in 16-bit mode


What I have is: 2x16 bit chips on a 16 bit bus in 8 bit mode and interleaved.


-- 
Kenneth Johansson	
Ericsson Business Innovation AB   Tel: +46 8 404 71 83
Viderögatan 3                     Fax: +46 8 404 72 72
164 80 Stockholm                  kenneth.johansson@inn.ericsson.se

Re: Q: Probing for devices > buswidth???

[David Woodhouse]

kenneth.johansson@inn.ericsson.se said:
>  Yes. I have a pcmcia linear flash device that is using Intel
> strataflash 28F640xxx. The chips is 16 bit but used as 8bit. To make
> things interesting they also interleave the chip so odd bytes goes to
> one chip and even to the other.

You have map driver code for PCMCIA devices? Is this a generic PCMCIA 
driver or a platform-specific hack?

--
dwmw2

Re: Q: Probing for devices > buswidth???

[Kenneth Johansson]

David Woodhouse wrote:
> 
> kenneth.johansson@inn.ericsson.se said:
> >  Yes. I have a pcmcia linear flash device that is using Intel
> > strataflash 28F640xxx. The chips is 16 bit but used as 8bit. To make
> > things interesting they also interleave the chip so odd bytes goes to
> > one chip and even to the other.
> 
> You have map driver code for PCMCIA devices? Is this a generic PCMCIA
> driver or a platform-specific hack?

Well this is linear flash this means I don't have to know anything about
pcmcia it is directly connected to the bus. This gives max 64Meg memory.

On the other hand it also works as a proper pcmcia device and can be used in
real pcmcia/cardbus environment.

I made a big fuckup when I started to use them and could no get it to work so
I made a small driver for real pcmcia that i used in a x86 laptop just to se
that the chip worked. This driver is way to simple to be usefull as a generic
mapping driver for MTD. But a real driver is not to hard to do. The PCMCIA
subsystem is just a bit confusing and has it's own bulkmem system that needs
to be removed and converted to MTD. 

The fuckup was that every pcmcia linear flash decode the entire 64Meg memory
range. I had 16Meg flash and expected to get 4 views of the same memmory in
that 64Meg rage. This did not happen and it never hit me to try to use the
device in a memory region that was modulo 64Meg so I did only get strange
values out of it:(

I needed the memory to exist from -1 to -16Meg as the memory is used for
booting a PowerPC CPU and they boot from fffffffc.

-- 
Kenneth Johansson	
Ericsson Business Innovation AB   Tel: +46 8 404 71 83
Viderögatan 3                     Fax: +46 8 404 72 72
164 80 Stockholm                  kenneth.johansson@inn.ericsson.se