I don't know if this is the proper place to follow up on your review,
then, but I did appreciate your advice, and would like to comment on a
few points:
It's conditional on MODULE_LICENSE itself; that's merely compatibility
cruft. (I know this will have to go eventually, but I'm in no hurry.)
For 8-bits-in-a-byte? Does anyone do that? (I count nearly 500
instances of this exact loop in the current kernel.)
There are many timing values that occur only once throughout the code;
I'm not sure that replacing them all with #defines would be more
readable than a good set of comments.
Even worse, in cases such as above, the 20 is purely arbitrary. So is
the udelay(100) in the loop. What matters is the product, ie. we will
give up after 2000us, while polling a few times in the interval.
I'll try to look around for inspiration and see if there are other
drivers doing similar things.
On x86-64, I assume? It's dead silent on i386, unfortunately. Thanks
for the heads up!
Does it matter, as long as it's not any less permissive? I mean:
$ git grep -l '(at your option) any later version' | wc -l
5854
Contacting all copyright holders would be a hassle, so I just want to
make sure there's a good reason for it.
Again, thanks for your generous help!
--
Staff meeting in the conference room in 3 minutes.
_______________________________________________
Kernel-mentors mailing list
Kernel-mentors@selenic.comhttp://selenic.com/mailman/listinfo/kernel-mentors