[PATCH] Xilinx UART Lite 2.6.18 driver

[Peter Korsgaard]

>>>>> "David" == David H Lynch <dhlii at dlasys.net> writes:

Hi,

 David>     The 8250 is certainly the oldest driver - but it is also
 David> by far the most used. It is also the most heavily maintained,

Yes, because there's so many more-or-less compatible variants of the
hardware. Most drivers don't see a lot of development once they are
past the initial development/testing..

 David> There are complexities to the 8250 that proved to be difficult
 David> and troubling when I based my UartLite driver on it, Those
 David> problems were because the 8250driver suports so many 8250
 David> variants with so many idiosyncracies.

Exactly. Try to compare it to one of the "modern" drivers like
atmel_serial.c - It's almost 3 times as big.

 >> Remove support? I don't remember anything about that.

 David>     I may have misread something there - I thought your
 David> initial driver allowed stting the number of Uarts, and I
 David> beleive your current one does not.  If I am wrong, then I
 David> think support for more Uarts needs to get added to the todo
 David> list.  That is not something important to me. But it probably
 David> will be for others.

There was a discussion about how many device nodes to register with
lanana.org and hence how many devices to support. I went with 4 which
seems to be pretty standard (E.G. virtex.c supports up to 4 8250s). If
more is needed a bigger range can be requested with lanana.org.

Again, let's see a real need before changing anything.

 David> Despite the fact hat you apear to run combined 16550/Uartlite
 David> implimentations, My expectation is that will be rare.  A small
 David> savings in FPGA spare is rarely going to justify more complex
 David> hardware and software.  Multiple port systems are mostly
 David> likely to be all one type or another.

Almost all our designs use both.

P.S.: Would it be possible to fix up your quoting?

-- 
Bye, Peter Korsgaard