Thanx, that works. I had more crude patch - forcing AMD in to 1640 mode
and disabling linux serial chip autodetect:)
Oleg.
At 10:48 AM 5/12/99 -0700, Andy Waddell wrote:
>The on-board serial port implemented in the SC400 silicon has a bug which
>makes it behave slightly different than a 16550A. I have patched serial.c
>and it works fine, but I doubt that my patch is general pupose enough for
>the mainline source tree. The actual bug has to do with a one clock cycle
>delay in the assertion of LSR I think. Hope this helps.
>
>
> diff serial.c /usr/src/pristine-2.2.5/drivers/char/serial.c
>69,70d68
><
><
>613c611
>< printk("status = %x...", status);
>---
>> printk("status = %x...", status);
>621d618
><
>653c650
>< int status, iir;
>---
>> int status;
>676d672
>< iir = serial_in(info, UART_IIR);
>679c675
>< printk("status = %x iir=%x...\n", status,iid);
>---
>> printk("status = %x...", status);
>681,683d676
>< status = serial_inp(info, UART_LSR);
><
>< //#endif
>687,688c680
><
>< if ((status & UART_LSR_THRE) || (iir & UART_IIR_THRI))
>---
>> if (status & UART_LSR_THRE)
>2933d2924
><
>2935d2925
><
>[awaddell@freebird char]$
>
>-----Original Message-----
>From: Oleg Perelet [mailto:[EMAIL PROTECTED]]
>Sent: Wednesday, May 12, 1999 10:26 AM
>To: Chris Wyszkowski; 'Oleg Perelet'
>Cc: [EMAIL PROTECTED]
>Subject: RE: Running directly from ROM?
>
>
>Chris, I wrote pretty dumm AMD 22/23 register setup (why everybody is
>scared of 256 extra registers to load:) and protected mode switcher.
>If you have 4MB(or 8 or 16...) linear memory PCMCIA card and tools
>to dump image on it, I can send you boot image. The only thing that
>I'm not doing right now - DRAM type detection, mainly because our
>(home made) board has DRAM soldered on. AMD uses different DRAM
>SIMMS on their boards, but there is AMD asm code to do this detect.
>
>
>Also does anybody knows what is the problem with on chip AMD serial? If
>I use SIO com1 everything is fine, if I switch on on chip serial
>(1st serial connector) serial is pretty screwy (I have the same
>problems with qnx - so it's not about linux)
>
>
>
>At 12:50 PM 5/12/99 -0400, Chris Wyszkowski wrote:
>>Oleg,
>>
>>I have SC410 HW prototyped along the lines you are discussing. I am now
>>looking to see what I need in terms of a loader (I don't want a PC stype
>BIOS.)
>> Can you tell me what your solution was? I was thinking a MILO style
>loader
>>would be best, but it may be overkill.
>>
>>I have been watching this mailing list for some time and this seems to be a
>
>>recurring barrier, anyone have some definitive ROMable loaders available
>for
>>the AMD Elan family that allow boot from flash?
>>
>>Chris Wyszkowski
>>
>>On Wednesday, May 12, 1999 12:39 PM, Oleg Perelet [SMTP:[EMAIL PROTECTED]]
>>wrote:
>>> You can look at AMD Elan CPU's very need and flexible architecture,
>>> with 4MB Flash/ROM, 8MB of DRAM and simple network chip you can
>>> build pretty inexpensive system (<$400 in low quantities).
>>> And there is no problem running linux in there. (I run 2.2.3).
>>>
>>>
>>> Oleg.
>>>
>>>
>>> At 12:23 PM 5/12/99 +1000, you wrote:
>>> >Hi Folks,
>>> >
>>> >I'm currently investigating options for a minimal-cost x386 based
>diskless
>>> >embedded controller running a stripped-down Linux kernel for a
>reasonably
>>> >simple embedded TCP/IP application. I'm very interested in opinions or
>>> >experience anyone could offer.
>>> >
>>> >My choice of x386 is to allow easy hardware & software prototyping on
>>> >desktop
>>> >PC's and off-the-shelf embedded biscuit-PC-like boards. Ultimately, I'd
>like
>>> >to
>>> >build a tiny BIOS-less system with traditional (i.e. non-IDE)
>flash/eprom
>>> >memory and minimal RAM. Since the Linux kernel has a larger footprint
>than
>>> >traditional embedded OS's, it would be nice to run it directly from ROM
>with
>>> >a ROM-based root filesystem. I'd prefer not to simply copy the kernel
>into
>>> >RAM
>>> >or use an "initrd" RAM-disk for root filesystem stuff which is all
>>> >essentially
>>> >read-only. Demonstrating that Linux can run embedded with very little
>RAM is
>>> >one of the goals here, even at the expense of slightly more ROM.
>>> >
>>> >My guess is that much of the specifics here can be handled by a
>ROM-based
>>> >bootloader which sets up the x386 MMU to effectively "load" the kernel,
>and
>>> >from then on everything is much like running in a desktop box, albeit a
>>> >severely stripped one. Does anyone know of an existing bootloader that
>can
>>> >perform this sort of magic, or any other pointers to existing solutions?
>>> >
>>> >Thank you,
>>> >Graham
>>> >
>>
>
