On Wed, Sep 14, 2016 at 10:37:13AM +0200, Miklos Szeredi wrote: > Getting the page for reading is pretty complicated. This functionality is > also duplicated between generic_file_read() generic_file_splice_read(). > > So first extract the common code from do_generic_file_read() into a > separate helper function that takes a filp, the page index, the offset into > the page and the byte count and returns the page ready for reading (or an > error). > > This makes do_generic_file_read() much easier to read as well.
__generic_file_splice_read() horrors are not going to survive - see the patchset posted on fsdevel. do_generic_file_read() getting easier to read is certainly a good thing, provided that we don't screw the code generation for what's a fairly hot path. IOW, that one really needs profiling.
