linux-kernel - [PATCH 0/5] uaccess: Updates to scoped_user

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <20251220215608.434614-1-david.laight.linux@gmail.com>
Date: Sat, 20 Dec 2025 21:56:03 +0000
From: david.laight.linux@...il.com
To: Thomas Gleixner <tglx@...utronix.de>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Peter Zijlstra <peterz@...radead.org>,
	Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
	Kees Cook <kees@...nel.org>,
	linux-kernel@...r.kernel.org,
	akpm@...ux-foundation.org,
	Al Viro <viro@...iv.linux.org.uk>
Cc: David Laight <david.laight.linux@...il.com>
Subject: [PATCH 0/5] uaccess: Updates to scoped_user_access()

From: David Laight <david.laight.linux@...il.com>

Converting kernel/signal.c to use scoped_user_access() had compilation
warnings because some of the pointers are 'pointer to const'.
This is fixed in patch 1.

Patches 2 and 3 factor out the 'autoterminating nested for loops'.
I'm sure there'll be a 'bikeshed' discussion about the names.

I think the two 'typeof(uptr)' could be 'auto', compiles for me
but IIRC there are issues with old compilers.

Patch 4 stops warnings from -Wshadow (enabled by W=2).
I did think about making the _diag_xxx conditional on a W=2 build,
but since the pre-processor just emits #pragma lines and they are
smaller that the for() loop it doesn't seem worth while.

Patch 5 is the change to signal.c that prompted patch 1.
The generated code looks fine, but I've not tested it.
Most of the changes are to 'compat' code - so are probably not
usually performance critical.
IIRC the non-compat code uses copy_to/from_user() for the structures
so is probably slower than the compat code's member by member copy.

David Laight (5):
  uaccess: Fix scoped_user_read_access() for 'pointer to const'
  compiler.h: Add generic support for 'autoterminating nested for()
    loops'
  uaccess.h: Use with() and and_with() in __scoped_user_access()
  uaccess: Disable -Wshadow in __scoped_user_access()
  signal: Use scoped_user_access() instead of __put/get_user()

 include/linux/compiler.h | 26 +++++++++++++++
 include/linux/uaccess.h  | 32 +++++++++---------
 kernel/signal.c          | 72 +++++++++++++++++++++++-----------------
 3 files changed, 84 insertions(+), 46 deletions(-)

-- 
2.39.5