On Tue, Apr 27, 2021 at 08:34:15PM +0200, Arkadiusz Kozdra (Arusekk) wrote: > Since seq_read_iter looks mature enough to be used for all procfs files, > re-allow applications to perform zero-copy data forwarding from them. > According to the sendfile(2) man-page, it is still enough for the file > being read to support mmap-like operations, and the proc files support > memory mapping, so returning -EINVAL was an inconsistency. Linus did object to blindly switching over all instances. > Some executable-inspecting tools rely on patching entry point > instructions with minimal machine code that uses sendfile to read > /proc/self/maps to stdout. The sendfile call allows them to do it > faster and without excessive allocations. Patching what entry point?