use kmap atomic for kernels past 2.6.37

[tsahee]

kmap_atomic is the recommended API, and it is much faster then kmap.

In my test, using kmap_atomic made memory dumps with LiME faster by an order of magnitude. I don't know what was the original reasoning for using kmap. I only put here a test for kernel version 2.6.37, but if you think we need to add some other compile or run time choice between kmap_atomic and kmap, I'd be happy to adjust.

Thank you!
Tsahi.

[kd8bny]

Cool Ill pull this in to test out !

[tsahee]

@kd8bny did you get a chance to look at it?

[kd8bny]

I have not and I really need to.
My apologies for falling behind on this, Im going to merge and test with the 5.10 fixes

[qodroi]

kmap_atomic caused my kernel to issue a bug - scheduling while atomic, with a message that preemption was disabled somewhere in the code. See what you can do about that, maybe revert to the original one? I'll see if I can find a solution without using the original kmap instead, due to performance issues, by your words.

Edit: See #110