>> I'm going over the code in bs_rdwr_request: the data is read/written >> from/to the device using pread64/pwrite64. I understand that it treats >> the device as a file and just reads/writes from/to it with some offset. >> What I don't understand is if the data is copied between user & kernel >> space. Eventually, it will go from tgt (user) to scsi-ml (kernel) somehow. > > Yes, data is copied between user and kernel spaces, that is, we copy > data between an user-space buffer and page cache. We could improve > this area but there are still tons of things that I want to do before > that.
Can you estimate the performance penalty due to these copies? Do you have an idea how this could be done? Eli _______________________________________________ Stgt-devel mailing list [email protected] https://lists.berlios.de/mailman/listinfo/stgt-devel
