Hi All,
I'm somewhat of a linux newbie who has been charged with writing a device driver for a custom ISA card. The card has a number of IO ports and some memory. I need to access the IO ports as fast as possible.
I've been working my way through Oreilly's linux device driver book, and looking around the web for info.
The original software is written in dos/borland c++.

Should I use the read and write file operations to do it? I"m wondering if they have internal buffering, or any other hidden gotchas that will slow access down.

Should ioctl be used?

are there any other methods that I don't know about?
Thanks in advance for any help you can give

-Edgewise

Define fast (sec, millisec, usec, nanosec)? Read and write may or may not be fast depending on how you look at it. There is no internal buffer unless you make one, as you will actually define the behavior of the command in your module. I mostly use ioctl as a method of defining commands that can be sent to the driver.

I'm looking at the usec range. I guess my confusion (or aprehension) comes from using the Read and Write system calls. The prototypes don't seem to lend themselves to performing discrete IO. I.E. Read takes a file pointer, a buffer pointer, a count, and an offset.

The time sensitive data is written to (or read from) byte ports. I want to be able to send the driver an address and a byte of data, and have it write it (or read the port) as soon as possible.


-Edgewise

That being the case you probably do want to implement an ioctl command, or maybe three, one to set the address, and one for reading a byte and one for writing a byte.