On Tue, 2013-08-20 at 23:29 +0200, Christoph Hellwig wrote:On Tue, Aug 20, 2013 at 08:07:51PM +0000, Nicholas A. Bellinger wrote:
It's also currently lacking the necessary sychronization between I/O
submission of COMPARE_AND_WRITE verify instance and write instance
user data, which is still being worked on in order to avoid additional
overhead in the main I/O fast path.
I don't think merging such a non-conforming implementation makes any sense.
Yes, I don't intend to merge anything that's not fully functional.
The idea was to get review going on these pieces first. I'll be posting
an PATCH-v2 to complete the implementation over the next days.
Also for a complex command like this with all it's race potential I would
really like to see some test cases to go along with it.
Yes, Eric @ PureStorage has a sg_compare_write that I'm using to test
this. It's probably about time that this be included in upstream
sg3-utils too..