One of the "legitimate" uses of strncpy() is copying a NUL-terminated
string into a fixed-size non-NUL-terminated character array. To avoid
the weaknesses and ambiguity of intent when using strncpy(), provide
replacement functions that explicitly distinguish between trailing
padding and not, and require the destination buffer size be discoverable
by the compiler.
For example:
struct obj {
int foo;
char small[4] __nonstring;
char big[8] __nonstring;
int bar;
};
struct obj p;
/* This will truncate to 4 chars with no trailing NUL */
strncpy(p.small, "hello", sizeof(p.small));
/* p.small contains 'h', 'e', 'l', 'l' */
/* This will NUL pad to 8 chars. */
strncpy(p.big, "hello", sizeof(p.big));
/* p.big contains 'h', 'e', 'l', 'l', 'o', '\0', '\0', '\0' */
When the "__nonstring" attributes are missing, the intent of the
programmer becomes ambiguous for whether the lack of a trailing NUL
in the p.small copy is a bug. Additionally, it's not clear whether
the trailing padding in the p.big copy is _needed_. Both cases
become unambiguous with:
strtomem(p.small, "hello");
strtomem_pad(p.big, "hello", 0);
See also https://github.com/KSPP/linux/issues/90
Expand the memcpy KUnit tests to include these functions.
Cc: Wolfram Sang <wsa+renesas@sang-engineering.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Kees Cook <keescook@chromium.org>
When building m68k:allmodconfig, recent versions of gcc generate the
following error if the length of UTS_RELEASE is less than 8 bytes.
In function 'memcpy_and_pad',
inlined from 'nvmet_execute_disc_identify' at
drivers/nvme/target/discovery.c:268:2: arch/m68k/include/asm/string.h:72:25: error:
'__builtin_memcpy' reading 8 bytes from a region of size 7
Discussions around the problem suggest that this only happens if an
architecture does not provide strlen(), if -ffreestanding is provided as
compiler option, and if CONFIG_FORTIFY_SOURCE=n. All of this is the case
for m68k. The exact reasons are unknown, but seem to be related to the
ability of the compiler to evaluate the return value of strlen() and
the resulting execution flow in memcpy_and_pad(). It would be possible
to work around the problem by using sizeof(UTS_RELEASE) instead of
strlen(UTS_RELEASE), but that would only postpone the problem until the
function is called in a similar way. Uninline memcpy_and_pad() instead
to solve the problem for good.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Andy Shevchenko <andriy.shevchenko@intel.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A common idiom in kernel code is to wipe the contents of a structure
starting from a given member. These open-coded cases are usually difficult
to read and very sensitive to struct layout changes. Like memset_after(),
introduce a new helper, memset_startat() that takes the target struct
instance, the byte to write, and the member name where zeroing should
start.
Note that this doesn't zero padding preceding the target member. For
those cases, memset_after() should be used on the preceding member.
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Francis Laniel <laniel_francis@privacyrequired.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: Daniel Axtens <dja@axtens.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
When commit a28a6e860c ("string.h: move fortified functions definitions
in a dedicated header.") moved the fortify-specific code, some helpers
were left behind. Move the remaining fortify-specific helpers into
fortify-string.h so they're together where they're used. This requires
that any FORTIFY helper function prototypes be conditionally built to
avoid "no prototype" warnings. Additionally removes unused helpers.
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Daniel Axtens <dja@axtens.net>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Acked-by: Francis Laniel <laniel_francis@privacyrequired.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Patch series "Fortify strscpy()", v7.
This patch implements a fortified version of strscpy() enabled by setting
CONFIG_FORTIFY_SOURCE=y. The new version ensures the following before
calling vanilla strscpy():
1. There is no read overflow because either size is smaller than src
length or we shrink size to src length by calling fortified strnlen().
2. There is no write overflow because we either failed during
compilation or at runtime by checking that size is smaller than dest
size. Note that, if src and dst size cannot be got, the patch defaults
to call vanilla strscpy().
The patches adds the following:
1. Implement the fortified version of strscpy().
2. Add a new LKDTM test to ensures the fortified version still returns
the same value as the vanilla one while panic'ing when there is a write
overflow.
3. Correct some typos in LKDTM related file.
I based my modifications on top of two patches from Daniel Axtens which
modify calls to __builtin_object_size, in fortified string functions, to
ensure the true size of char * are returned and not the surrounding
structure size.
About performance, I measured the slow down of fortified strscpy(), using
the vanilla one as baseline. The hardware I used is an Intel i3 2130 CPU
clocked at 3.4 GHz. I ran "Linux 5.10.0-rc4+ SMP PREEMPT" inside qemu
3.10 with 4 CPU cores. The following code, called through LKDTM, was used
as a benchmark:
#define TIMES 10000
char *src;
char dst[7];
int i;
ktime_t begin;
src = kstrdup("foobar", GFP_KERNEL);
if (src == NULL)
return;
begin = ktime_get();
for (i = 0; i < TIMES; i++)
strscpy(dst, src, strlen(src));
pr_info("%d fortified strscpy() tooks %lld", TIMES, ktime_get() - begin);
begin = ktime_get();
for (i = 0; i < TIMES; i++)
__real_strscpy(dst, src, strlen(src));
pr_info("%d vanilla strscpy() tooks %lld", TIMES, ktime_get() - begin);
kfree(src);
I called the above code 30 times to compute stats for each version (in ns,
round to int):
| version | mean | std | median | 95th |
| --------- | ------- | ------ | ------- | ------- |
| fortified | 245_069 | 54_657 | 216_230 | 331_122 |
| vanilla | 172_501 | 70_281 | 143_539 | 219_553 |
On average, fortified strscpy() is approximately 1.42 times slower than
vanilla strscpy(). For the 95th percentile, the fortified version is
about 1.50 times slower.
So, clearly the stats are not in favor of fortified strscpy(). But, the
fortified version loops the string twice (one in strnlen() and another in
vanilla strscpy()) while the vanilla one only loops once. This can
explain why fortified strscpy() is slower than the vanilla one.
This patch (of 5):
When the fortify feature was first introduced in commit 6974f0c455
("include/linux/string.h: add the option of fortified string.h
functions"), Daniel Micay observed:
* It should be possible to optionally use __builtin_object_size(x, 1) for
some functions (C strings) to detect intra-object overflows (like
glibc's _FORTIFY_SOURCE=2), but for now this takes the conservative
approach to avoid likely compatibility issues.
This is a case that often cannot be caught by KASAN. Consider:
struct foo {
char a[10];
char b[10];
}
void test() {
char *msg;
struct foo foo;
msg = kmalloc(16, GFP_KERNEL);
strcpy(msg, "Hello world!!");
// this copy overwrites foo.b
strcpy(foo.a, msg);
}
The questionable copy overflows foo.a and writes to foo.b as well. It
cannot be detected by KASAN. Currently it is also not detected by
fortify, because strcpy considers __builtin_object_size(x, 0), which
considers the size of the surrounding object (here, struct foo). However,
if we switch the string functions over to use __builtin_object_size(x, 1),
the compiler will measure the size of the closest surrounding subobject
(here, foo.a), rather than the size of the surrounding object as a whole.
See https://gcc.gnu.org/onlinedocs/gcc/Object-Size-Checking.html for more
info.
Only do this for string functions: we cannot use it on things like memcpy,
memmove, memcmp and memchr_inv due to code like this which purposefully
operates on multiple structure members: (arch/x86/kernel/traps.c)
/*
* regs->sp points to the failing IRET frame on the
* ESPFIX64 stack. Copy it to the entry stack. This fills
* in gpregs->ss through gpregs->ip.
*
*/
memmove(&gpregs->ip, (void *)regs->sp, 5*8);
This change passes an allyesconfig on powerpc and x86, and an x86 kernel
built with it survives running with syz-stress from syzkaller, so it seems
safe so far.
Link: https://lkml.kernel.org/r/20201122162451.27551-1-laniel_francis@privacyrequired.com
Link: https://lkml.kernel.org/r/20201122162451.27551-2-laniel_francis@privacyrequired.com
Signed-off-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Francis Laniel <laniel_francis@privacyrequired.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Daniel Micay <danielmicay@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In reaction to a proposal to introduce a memcpy_mcsafe_fast()
implementation Linus points out that memcpy_mcsafe() is poorly named
relative to communicating the scope of the interface. Specifically what
addresses are valid to pass as source, destination, and what faults /
exceptions are handled.
Of particular concern is that even though x86 might be able to handle
the semantics of copy_mc_to_user() with its common copy_user_generic()
implementation other archs likely need / want an explicit path for this
case:
On Fri, May 1, 2020 at 11:28 AM Linus Torvalds <torvalds@linux-foundation.org> wrote:
>
> On Thu, Apr 30, 2020 at 6:21 PM Dan Williams <dan.j.williams@intel.com> wrote:
> >
> > However now I see that copy_user_generic() works for the wrong reason.
> > It works because the exception on the source address due to poison
> > looks no different than a write fault on the user address to the
> > caller, it's still just a short copy. So it makes copy_to_user() work
> > for the wrong reason relative to the name.
>
> Right.
>
> And it won't work that way on other architectures. On x86, we have a
> generic function that can take faults on either side, and we use it
> for both cases (and for the "in_user" case too), but that's an
> artifact of the architecture oddity.
>
> In fact, it's probably wrong even on x86 - because it can hide bugs -
> but writing those things is painful enough that everybody prefers
> having just one function.
Replace a single top-level memcpy_mcsafe() with either
copy_mc_to_user(), or copy_mc_to_kernel().
Introduce an x86 copy_mc_fragile() name as the rename for the
low-level x86 implementation formerly named memcpy_mcsafe(). It is used
as the slow / careful backend that is supplanted by a fast
copy_mc_generic() in a follow-on patch.
One side-effect of this reorganization is that separating copy_mc_64.S
to its own file means that perf no longer needs to track dependencies
for its memcpy_64.S benchmarks.
[ bp: Massage a bit. ]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Cc: <stable@vger.kernel.org>
Link: http://lore.kernel.org/r/CAHk-=wjSqtXAqfUJxFtWNwmguFASTgB0dz1dT3V-78Quiezqbg@mail.gmail.com
Link: https://lkml.kernel.org/r/160195561680.2163339.11574962055305783722.stgit@dwillia2-desk3.amr.corp.intel.com
The memcmp KASAN self-test fails on a kernel with both KASAN and
FORTIFY_SOURCE.
When FORTIFY_SOURCE is on, a number of functions are replaced with
fortified versions, which attempt to check the sizes of the operands.
However, these functions often directly invoke __builtin_foo() once they
have performed the fortify check. Using __builtins may bypass KASAN
checks if the compiler decides to inline it's own implementation as
sequence of instructions, rather than emit a function call that goes out
to a KASAN-instrumented implementation.
Why is only memcmp affected?
============================
Of the string and string-like functions that kasan_test tests, only memcmp
is replaced by an inline sequence of instructions in my testing on x86
with gcc version 9.2.1 20191008 (Ubuntu 9.2.1-9ubuntu2).
I believe this is due to compiler heuristics. For example, if I annotate
kmalloc calls with the alloc_size annotation (and disable some fortify
compile-time checking!), the compiler will replace every memset except the
one in kmalloc_uaf_memset with inline instructions. (I have some WIP
patches to add this annotation.)
Does this affect other functions in string.h?
=============================================
Yes. Anything that uses __builtin_* rather than __real_* could be
affected. This looks like:
- strncpy
- strcat
- strlen
- strlcpy maybe, under some circumstances?
- strncat under some circumstances
- memset
- memcpy
- memmove
- memcmp (as noted)
- memchr
- strcpy
Whether a function call is emitted always depends on the compiler. Most
bugs should get caught by FORTIFY_SOURCE, but the missed memcmp test shows
that this is not always the case.
Isn't FORTIFY_SOURCE disabled with KASAN?
========================================-
The string headers on all arches supporting KASAN disable fortify with
kasan, but only when address sanitisation is _also_ disabled. For example
from x86:
#if defined(CONFIG_KASAN) && !defined(__SANITIZE_ADDRESS__)
/*
* For files that are not instrumented (e.g. mm/slub.c) we
* should use not instrumented version of mem* functions.
*/
#define memcpy(dst, src, len) __memcpy(dst, src, len)
#define memmove(dst, src, len) __memmove(dst, src, len)
#define memset(s, c, n) __memset(s, c, n)
#ifndef __NO_FORTIFY
#define __NO_FORTIFY /* FORTIFY_SOURCE uses __builtin_memcpy, etc. */
#endif
#endif
This comes from commit 6974f0c455 ("include/linux/string.h: add the
option of fortified string.h functions"), and doesn't work when KASAN is
enabled and the file is supposed to be sanitised - as with test_kasan.c
I'm pretty sure this is not wrong, but not as expansive it should be:
* we shouldn't use __builtin_memcpy etc in files where we don't have
instrumentation - it could devolve into a function call to memcpy,
which will be instrumented. Rather, we should use __memcpy which
by convention is not instrumented.
* we also shouldn't be using __builtin_memcpy when we have a KASAN
instrumented file, because it could be replaced with inline asm
that will not be instrumented.
What is correct behaviour?
==========================
Firstly, there is some overlap between fortification and KASAN: both
provide some level of _runtime_ checking. Only fortify provides
compile-time checking.
KASAN and fortify can pick up different things at runtime:
- Some fortify functions, notably the string functions, could easily be
modified to consider sub-object sizes (e.g. members within a struct),
and I have some WIP patches to do this. KASAN cannot detect these
because it cannot insert poision between members of a struct.
- KASAN can detect many over-reads/over-writes when the sizes of both
operands are unknown, which fortify cannot.
So there are a couple of options:
1) Flip the test: disable fortify in santised files and enable it in
unsanitised files. This at least stops us missing KASAN checking, but
we lose the fortify checking.
2) Make the fortify code always call out to real versions. Do this only
for KASAN, for fear of losing the inlining opportunities we get from
__builtin_*.
(We can't use kasan_check_{read,write}: because the fortify functions are
_extern inline_, you can't include _static_ inline functions without a
compiler warning. kasan_check_{read,write} are static inline so we can't
use them even when they would otherwise be suitable.)
Take approach 2 and call out to real versions when KASAN is enabled.
Use __underlying_foo to distinguish from __real_foo: __real_foo always
refers to the kernel's implementation of foo, __underlying_foo could be
either the kernel implementation or the __builtin_foo implementation.
This is sometimes enough to make the memcmp test succeed with
FORTIFY_SOURCE enabled. It is at least enough to get the function call
into the module. One more fix is needed to make it reliable: see the next
patch.
Fixes: 6974f0c455 ("include/linux/string.h: add the option of fortified string.h functions")
Signed-off-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: David Gow <davidgow@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Daniel Micay <danielmicay@gmail.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Link: http://lkml.kernel.org/r/20200423154503.5103-3-dja@axtens.net
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Patch series "lib: rework bitmap_parse", v5.
Similarl to the recently revisited bitmap_parselist(), bitmap_parse() is
ineffective and overcomplicated. This series reworks it, aligns its
interface with bitmap_parselist() and makes it simpler to use.
The series also adds a test for the function and fixes usage of it in
cpumask_parse() according to the new design - drops the calculating of
length of an input string.
bitmap_parse() takes the array of numbers to be put into the map in the BE
order which is reversed to the natural LE order for bitmaps. For example,
to construct bitmap containing a bit on the position 42, we have to put a
line '400,0'. Current implementation reads chunk one by one from the
beginning ('400' before '0') and makes bitmap shift after each successful
parse. It makes the complexity of the whole process as O(n^2). We can do
it in reverse direction ('0' before '400') and avoid shifting, but it
requires reverse parsing helpers.
This patch (of 7):
New function works like strchrnul() with a length limited string.
Link: http://lkml.kernel.org/r/20200102043031.30357-2-yury.norov@gmail.com
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Amritha Nambiar <amritha.nambiar@intel.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: "Tobin C . Harding" <tobin@kernel.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Cc: Vineet Gupta <vineet.gupta1@synopsys.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We have a function to copy strings safely and we have a function to copy
strings and zero the tail of the destination (if source string is
shorter than destination buffer) but we do not have a function to do
both at once. This means developers must write this themselves if they
desire this functionality. This is a chore, and also leaves us open to
off by one errors unnecessarily.
Add a function that calls strscpy() then memset()s the tail to zero if
the source string is shorter than the destination buffer.
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Tobin C. Harding <tobin@kernel.org>
Signed-off-by: Shuah Khan <shuah@kernel.org>
A discussion came up in the trace triggers thread about converting a
bunch of:
strncmp(str, "const", sizeof("const") - 1)
use cases into a helper macro. It started with:
strncmp(str, const, sizeof(const) - 1)
But then Joe Perches mentioned that if a const is not used, the
sizeof() will be the size of a pointer, which can be bad. And that
gcc will optimize strlen("const") into "sizeof("const") - 1".
Thinking about this more, a quick grep in the kernel tree found several
(thousands!) of cases that use this construct. A quick grep also
revealed that there's probably several bugs in that use case. Some are
that people forgot the "- 1" (which I found) and others could be that
the constant for the sizeof is different than the constant (although, I
haven't found any of those, but I also didn't look hard).
I figured the best thing to do is to create a helper macro and place it
into include/linux/string.h. And go around and fix all the open coded
versions of it later.
Note, gcc appears to optimize this when we make it into an always_inline
static function, which removes a lot of issues that a macro produces.
Link: http://lkml.kernel.org/r/e3e754f2bd18e56eaa8baf79bee619316ebf4cfc.1545161087.git.tom.zanussi@linux.intel.com
Link: http://lkml.kernel.org/r/20181219211615.2298e781@gandalf.local.home
Link: http://lkml.kernel.org/r/CAHk-=wg_sR-UEC1ggmkZpypOUYanL5CMX4R7ceuaV4QMf5jBtg@mail.gmail.com
Cc: Tom Zanussi <zanussi@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Suggestions-by: Linus Torvalds <torvalds@linux-foundation.org>
Suggestions-by: Joe Perches <joe@perches.com>
Suggestions-by: Andreas Schwab <schwab@linux-m68k.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
This adds a helper to paste 2 pointer arrays together, useful for merging
various types of attribute arrays. There are a few places in the kernel
tree where this is open coded, and I just added one more in the STM class.
The naming is inspired by memset_p() and memcat(), and partial credit for
it goes to Andy Shevchenko.
This patch adds the function wrapped in a type-enforcing macro and a test
module.
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Tested-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Pull misc vfs updates from Al Viro:
"All kinds of misc stuff, without any unifying topic, from various
people.
Neil's d_anon patch, several bugfixes, introduction of kvmalloc
analogue of kmemdup_user(), extending bitfield.h to deal with
fixed-endians, assorted cleanups all over the place..."
* 'work.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (28 commits)
alpha: osf_sys.c: use timespec64 where appropriate
alpha: osf_sys.c: fix put_tv32 regression
jffs2: Fix use-after-free bug in jffs2_iget()'s error handling path
dcache: delete unused d_hash_mask
dcache: subtract d_hash_shift from 32 in advance
fs/buffer.c: fold init_buffer() into init_page_buffers()
fs: fold __inode_permission() into inode_permission()
fs: add RWF_APPEND
sctp: use vmemdup_user() rather than badly open-coding memdup_user()
snd_ctl_elem_init_enum_names(): switch to vmemdup_user()
replace_user_tlv(): switch to vmemdup_user()
new primitive: vmemdup_user()
memdup_user(): switch to GFP_USER
eventfd: fold eventfd_ctx_get() into eventfd_ctx_fileget()
eventfd: fold eventfd_ctx_read() into eventfd_read()
eventfd: convert to use anon_inode_getfd()
nfs4file: get rid of pointless include of btrfs.h
uvc_v4l2: clean copyin/copyout up
vme_user: don't use __copy_..._user()
usx2y: don't bother with memdup_user() for 16-byte structure
...
Pull cgroup updates from Tejun Heo:
"Nothing too interesting. Documentation updates and trivial changes;
however, this pull request does containt he previusly discussed
dropping of __must_check from strscpy()"
* 'for-4.16' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
Documentation: Fix 'file_mapped' -> 'mapped_file'
string: drop __must_check from strscpy() and restore strscpy() usages in cgroup
cgroup, docs: document the root cgroup behavior of cpu and io controllers
cgroup-v2.txt: fix typos
cgroup: Update documentation reference
Documentation/cgroup-v1: fix outdated programming details
cgroup, docs: document cgroup v2 device controller
