Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Fri, 15 Jul 2016 14:44:14 -0700
From: Kees Cook <>
Cc: Kees Cook <>,
	Balbir Singh <>,
	Daniel Micay <>,
	Josh Poimboeuf <>,
	Rik van Riel <>,
	Casey Schaufler <>,
	PaX Team <>,
	Brad Spengler <>,
	Russell King <>,
	Catalin Marinas <>,
	Will Deacon <>,
	Ard Biesheuvel <>,
	Benjamin Herrenschmidt <>,
	Michael Ellerman <>,
	Tony Luck <>,
	Fenghua Yu <>,
	"David S. Miller" <>,,
	Christoph Lameter <>,
	Pekka Enberg <>,
	David Rientjes <>,
	Joonsoo Kim <>,
	Andrew Morton <>,
	Andy Lutomirski <>,
	Borislav Petkov <>,
	Mathias Krause <>,
	Jan Kara <>,
	Vitaly Wool <>,
	Andrea Arcangeli <>,
	Dmitry Vyukov <>,
	Laura Abbott <>,,,,,,,
Subject: [PATCH v3 00/11] mm: Hardened usercopy


[I'm going to carry this series in my kspp -next tree now, though I'd
really love to have some explicit Acked-bys or Reviewed-bys. If you've
looked through it or tested it, please consider it. :) (I added Valdis
and mpe's Tested-bys where they seemed correct, thank you!)]

This is a start of the mainline port of PAX_USERCOPY[1]. After I started
writing tests (now in lkdtm in -next) for Casey's earlier port[2], I kept
tweaking things further and further until I ended up with a whole new
patch series. To that end, I took Rik and other people's feedback along
with other changes and clean-ups.

Based on my understanding, PAX_USERCOPY was designed to catch a
few classes of flaws (mainly bad bounds checking) around the use of
copy_to_user()/copy_from_user(). These changes don't touch get_user() and
put_user(), since these operate on constant sized lengths, and tend to be
much less vulnerable. There are effectively three distinct protections in
the whole series, each of which I've given a separate CONFIG, though this
patch set is only the first of the three intended protections. (Generally
speaking, PAX_USERCOPY covers what I'm calling CONFIG_HARDENED_USERCOPY

This series, which adds CONFIG_HARDENED_USERCOPY, checks that objects
being copied to/from userspace meet certain criteria:
- if address is a heap object, the size must not exceed the object's
  allocated size. (This will catch all kinds of heap overflow flaws.)
- if address range is in the current process stack, it must be within the
  current stack frame (if such checking is possible) or at least entirely
  within the current process's stack. (This could catch large lengths that
  would have extended beyond the current process stack, or overflows if
  their length extends back into the original stack.)
- if the address range is part of kernel data, rodata, or bss, allow it.
- if address range is page-allocated, that it doesn't span multiple
- if address is within the kernel text, reject it.
- everything else is accepted

The patches in the series are:
- Support for arch-specific stack frame checking (which will likely be
  replaced in the future by Josh's more comprehensive unwinder):
        1- mm: Implement stack frame object validation
- The core copy_to/from_user() checks, without the slab object checks:
        2- mm: Hardened usercopy
- Per-arch enablement of the protection:
        3- x86/uaccess: Enable hardened usercopy
        4- ARM: uaccess: Enable hardened usercopy
        5- arm64/uaccess: Enable hardened usercopy
        6- ia64/uaccess: Enable hardened usercopy
        7- powerpc/uaccess: Enable hardened usercopy
        8- sparc/uaccess: Enable hardened usercopy
        9- s390/uaccess: Enable hardened usercopy
- The heap allocator implementation of object size checking:
       10- mm: SLAB hardened usercopy support
       11- mm: SLUB hardened usercopy support

Some notes:

- This is expected to apply on top of -next which contains fixes for the
  position of _etext on both arm and arm64, though it has minor conflicts
  with KASAN that are trivial to fix up. Living in -next are also tests
  for this protection in lkdtm, prefixed with USERCOPY_.

- I couldn't detect a measurable performance change with these features
  enabled. Kernel build times were unchanged, hackbench was unchanged,
  etc. I think we could flip this to "on by default" at some point, but
  for now, I'm leaving it off until I can get some more definitive
  measurements. I would love if someone with greater familiarity with
  perf could give this a spin and report results.

- The SLOB support extracted from grsecurity seems entirely broken. I
  have no idea what's going on there, I spent my time testing SLAB and
  SLUB. Having someone else look at SLOB would be nice, but this series
  doesn't depend on it.

Additional features that would be nice, but aren't blocking this series:

- Needs more architecture support for stack frame checking (only x86 now,
  but it seems Josh will have a good solution for this soon).



[1] "grsecurity - test kernel patch"

- switch to using BUG for better Oops integration
- when checking page allocations, check each for Reserved
- use enums for the stack check return for readability

- added s390 support
- handle slub red zone
- disallow writes to rodata area
- stack frame walker now CONFIG-controlled arch-specific helper

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.