On Wed, Apr 28, 2021 at 03:02:13PM +0200, Arusekk wrote: > The instructions at the entry point of the executable being inspected. > The flow of the tool: > - parse ELF headers of the binary to be inspected, > - locate its entry point position in the file, > - write short code at the location (this short code has used sendfile so far), > - execute the patched binary, > - parse the output and extract information about the relevant mappings. > This can be seen as equivalent to setting LD_TRACE_LOADED_OBJECTS, > but also works for static binaries, and is a bit safer. > > The problem was reported at: > https://github.com/Gallopsled/pwntools/issues/1871 Oh, this patches just the userspace binarz, ok. > > Linus did object to blindly switching over all instances. > > I know, I read that, but I thought that pointing a real use case, combined > with the new interface being used all throughout the other code, might be > convincing. > I would be happy with only changing the f_ops of /proc/.../maps, even if only > on MMU-enabled systems, but I thought that consistence would be better. > This is my first time contributing to Linux, so I am very sorry for any wrong > assumptions, and glad to learn more. Unless Linus changed his mind just patching the file you care about for now seems like the best idea.
