This is the mail archive of the
ecos-patches@sourceware.org
mailing list for the eCos project.
Re: USB serial package
On Mon, Jun 16, 2008 at 11:56:22AM -0400, Frank Pagliughi wrote:
> Andrew Lunn wrote:
>> Hi Frank
>> Which 2.6 kernel did you test this with? I just tried my slightly out
>> of date 2.6.26-rc5 and it does not compile. Looking at
>> http://lxr.linux.no/ there has been some reworking going on in this
>> area. It might compile with 2.6.25. It would be nice to document what
>> is know to work.
>>
>> Thanks
>> Andrew
>>
>>
> BTW, please let me know if you have any thoughts, ideas, or suggestions
> on the submission.
I got it running this evening. The eCos example works great. I'm now
working on documentation, writing some sgml. I will also look at the
kernel module. I used uses module parameters when loading the
usbserial module, but a specific kernel modules might be nicer in some
situation.
> I wasn't sure which way to go on a number of items,
> like:
> - Should it implement an API of its own or just expose the endpoints
> for the standard USB API.
I think your API is O.K. Another option would be to make it look like
a serial port, so that you could open it with open("/dev/ttyUSB")
etc. That adds more overhead, but would allow select() to be used
etc. But that could be layered over the top of what you provide.
> - How best to make it device/driver independent.
You have to configure the code with the endpoints somehow. What you
have done in CDL seems good to me. I did something very similar in my
own private code.
> - Did I leave out any configuration options that would make it more
> useful?
I don't think so, but i've not put it to serious use.
> I also couldn't figure out any way to add class-specific descriptors
> (which would make the ACM configuration usable with a Linux host). Any
> help here would be appreciated.
Sorry, no idea. I would have to take a look at my USB book to first
figure out the encoding.
> I ask because I have the start of several other classes that could be
> submitted (Printer, Mass Storage, & HID), and would undoubtedly be
> modeled after this one.
Such class drivers would be great. However do you need to get the
class specific descriptors working first?
Thanks for the code.
Andrew