]> git.baikalelectronics.ru Git - kernel.git/commit
nfsd: escape high characters in binary data
authorJ. Bruce Fields <bfields@redhat.com>
Wed, 19 Jun 2019 16:30:13 +0000 (12:30 -0400)
committerJ. Bruce Fields <bfields@redhat.com>
Wed, 3 Jul 2019 21:52:50 +0000 (17:52 -0400)
commitcdf7ecd296ea8c53189b72ededf60adeca08e639
tree730c7927a9a945e91401dfb0a904f3ca94031519
parent1e864c460ca91df1ecc9fbfac5fe2b3d9aaee884
nfsd: escape high characters in binary data

I'm exposing some information about NFS clients in pseudofiles.  I
expect to eventually have simple tools to help read those pseudofiles.

But it's also helpful if the raw files are human-readable to the extent
possible.  It aids debugging and makes them usable on systems that don't
have the latest nfs-utils.

A minor challenge there is opaque client-generated protocol objects like
state owners and client identifiers.  Some clients generate those to
include handy information in plain ascii.  But they may also include
arbitrary byte sequences.

I think the simplest approach is to limit to isprint(c) && isascii(c)
and escape everything else.

That means you can just cat the file and get something that looks OK.
Also, I'm trying to keep these files legal YAML, which requires them to
UTF-8, and this is a simple way to guarantee that.

Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
fs/seq_file.c
include/linux/seq_file.h
include/linux/string_helpers.h
lib/string_helpers.c