FT> I'm talking about removing dmu_request->deps (dependent request FT> feature). Seems that you are talking about DMU_FLAG_COPY_FIRST FT> feature (it's ok for me). Ah, I see. Sorry for the confusion :) FT> My point is that the dependent request feature can be done in user FT> space easily. If a user-space program gets a request trying to FT> perform I/Os in the same block that an outgoing request with FT> DMU_FLAG_COPY_FIRST are performing I/Os too, it waits for the I/O FT> completion and then submit the new request to kernel space. Yes, I think this is a good idea. I will put it on my list of things to do next week. FT> I think that AIO is essential for the metadata issue even with FT> DMU_FLAG_COPY_FIRST and dependent request features, though the FT> DMU_FLAG_COPY_FIRST feature is important for writing a newly FT> allocated cow block. Yes, I fully intend to use AIO for metadata flushing eventually. -- Dan Smith IBM Linux Technology Center Open Hypervisor Team email: danms us ibm com
Attachment:
pgp0dlMuPC3SJ.pgp
Description: PGP signature