* [PATCH v2 0/3] overflow.h: Add arithmetic shift helper @ 2018-08-01 0:00 Kees Cook 2018-08-01 0:00 ` [PATCH v2 1/3] " Kees Cook ` (2 more replies) 0 siblings, 3 replies; 13+ messages in thread From: Kees Cook @ 2018-08-01 0:00 UTC (permalink / raw) To: Rasmus Villemoes Cc: Kees Cook, Jason Gunthorpe, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, linux-kernel This adds the shift overflow helper, selftests, and first usage. -Kees Jason Gunthorpe (1): overflow.h: Add arithmetic shift helper Kees Cook (1): test_overflow: Add shift overflow tests Leon Romanovsky (1): RDMA/mlx5: Fix shift overflow in mlx5_ib_create_wq drivers/infiniband/hw/mlx5/qp.c | 5 +- include/linux/overflow.h | 31 +++++++ lib/test_overflow.c | 140 +++++++++++++++++++++++++++++++- 3 files changed, 174 insertions(+), 2 deletions(-) -- 2.17.1 ^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH v2 1/3] overflow.h: Add arithmetic shift helper 2018-08-01 0:00 [PATCH v2 0/3] overflow.h: Add arithmetic shift helper Kees Cook @ 2018-08-01 0:00 ` Kees Cook 2018-08-01 1:59 ` Randy Dunlap ` (2 more replies) 2018-08-01 0:00 ` [PATCH v2 2/3] test_overflow: Add shift overflow tests Kees Cook 2018-08-01 0:00 ` [PATCH v2 3/3] RDMA/mlx5: Fix shift overflow in mlx5_ib_create_wq Kees Cook 2 siblings, 3 replies; 13+ messages in thread From: Kees Cook @ 2018-08-01 0:00 UTC (permalink / raw) To: Rasmus Villemoes Cc: Kees Cook, Jason Gunthorpe, Leon Romanovsky, Jason Gunthorpe, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, linux-kernel From: Jason Gunthorpe <jgg@mellanox.com> Add shift_overflow() helper to assist driver authors in ensuring that shift operations don't cause overflows or other odd conditions. Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> [kees: tweaked comments and commit log, dropped unneeded assignment] Signed-off-by: Kees Cook <keescook@chromium.org> --- include/linux/overflow.h | 31 +++++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) diff --git a/include/linux/overflow.h b/include/linux/overflow.h index 8712ff70995f..69fc366ce865 100644 --- a/include/linux/overflow.h +++ b/include/linux/overflow.h @@ -202,6 +202,37 @@ #endif /* COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW */ +/** check_shift_overflow() - Calculate a left-shifted value and check overflow + * + * @a: Value to be shifted + * @b: How many bits left to shift + * @d: Pointer to where to store the result + * + * Computes *@d = (@a << @s) + * + * Returns true if '*d' cannot hold the result or when 'a << s' doesn't + * make sense. Example conditions: + * - 'a << s' causes bits to be lost when stored in *d. + * - 's' is garbage (e.g. negative) or so large that the result of + * 'a << s' is guaranteed to be 0. + * - 'a' is negative. + * - 'a << s' sets the sign bit, if any, in '*d'. + * + * '*d' will hold the results of the attempted shift, but is not + * considered "safe for use" if false is returned. + */ +#define check_shift_overflow(a, s, d) ({ \ + typeof(a) _a = a; \ + typeof(s) _s = s; \ + typeof(d) _d = d; \ + u64 _a_full = _a; \ + unsigned int _to_shift = \ + _s >= 0 && _s < 8 * sizeof(*d) ? _s : 0; \ + *_d = (_a_full << _to_shift); \ + (_to_shift != _s || *_d < 0 || _a < 0 || \ + (*_d >> _to_shift) != _a); \ +}) + /** * array_size() - Calculate size of 2-dimensional array. * -- 2.17.1 ^ permalink raw reply related [flat|nested] 13+ messages in thread
* Re: [PATCH v2 1/3] overflow.h: Add arithmetic shift helper 2018-08-01 0:00 ` [PATCH v2 1/3] " Kees Cook @ 2018-08-01 1:59 ` Randy Dunlap 2018-08-01 2:15 ` Jason Gunthorpe 2018-08-01 7:57 ` Dan Carpenter 2 siblings, 0 replies; 13+ messages in thread From: Randy Dunlap @ 2018-08-01 1:59 UTC (permalink / raw) To: Kees Cook, Rasmus Villemoes Cc: Jason Gunthorpe, Leon Romanovsky, Jason Gunthorpe, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, linux-kernel On 07/31/2018 05:00 PM, Kees Cook wrote: > From: Jason Gunthorpe <jgg@mellanox.com> > > Add shift_overflow() helper to assist driver authors in ensuring that > shift operations don't cause overflows or other odd conditions. > > Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> > Signed-off-by: Leon Romanovsky <leonro@mellanox.com> > [kees: tweaked comments and commit log, dropped unneeded assignment] > Signed-off-by: Kees Cook <keescook@chromium.org> > --- > include/linux/overflow.h | 31 +++++++++++++++++++++++++++++++ > 1 file changed, 31 insertions(+) > > diff --git a/include/linux/overflow.h b/include/linux/overflow.h > index 8712ff70995f..69fc366ce865 100644 > --- a/include/linux/overflow.h > +++ b/include/linux/overflow.h > @@ -202,6 +202,37 @@ > > #endif /* COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW */ > > +/** check_shift_overflow() - Calculate a left-shifted value and check overflow > + * > + * @a: Value to be shifted > + * @b: How many bits left to shift > + * @d: Pointer to where to store the result > + * > + * Computes *@d = (@a << @s) @b) > + * > + * Returns true if '*d' cannot hold the result or when 'a << s' doesn't > + * make sense. Example conditions: > + * - 'a << s' causes bits to be lost when stored in *d. > + * - 's' is garbage (e.g. negative) or so large that the result of > + * 'a << s' is guaranteed to be 0. > + * - 'a' is negative. > + * - 'a << s' sets the sign bit, if any, in '*d'. > + * > + * '*d' will hold the results of the attempted shift, but is not > + * considered "safe for use" if false is returned. > + */ > +#define check_shift_overflow(a, s, d) ({ \ > + typeof(a) _a = a; \ > + typeof(s) _s = s; \ > + typeof(d) _d = d; \ > + u64 _a_full = _a; \ > + unsigned int _to_shift = \ > + _s >= 0 && _s < 8 * sizeof(*d) ? _s : 0; \ > + *_d = (_a_full << _to_shift); \ > + (_to_shift != _s || *_d < 0 || _a < 0 || \ > + (*_d >> _to_shift) != _a); \ > +}) > + > /** > * array_size() - Calculate size of 2-dimensional array. > * > -- ~Randy ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH v2 1/3] overflow.h: Add arithmetic shift helper 2018-08-01 0:00 ` [PATCH v2 1/3] " Kees Cook 2018-08-01 1:59 ` Randy Dunlap @ 2018-08-01 2:15 ` Jason Gunthorpe 2018-08-01 4:22 ` Kees Cook 2018-08-01 7:57 ` Dan Carpenter 2 siblings, 1 reply; 13+ messages in thread From: Jason Gunthorpe @ 2018-08-01 2:15 UTC (permalink / raw) To: Kees Cook Cc: Rasmus Villemoes, Leon Romanovsky, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, linux-kernel On Tue, Jul 31, 2018 at 05:00:37PM -0700, Kees Cook wrote: > From: Jason Gunthorpe <jgg@mellanox.com> > > Add shift_overflow() helper to assist driver authors in ensuring that > shift operations don't cause overflows or other odd conditions. > > Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> > Signed-off-by: Leon Romanovsky <leonro@mellanox.com> > [kees: tweaked comments and commit log, dropped unneeded assignment] > Signed-off-by: Kees Cook <keescook@chromium.org> > --- > include/linux/overflow.h | 31 +++++++++++++++++++++++++++++++ > 1 file changed, 31 insertions(+) > > diff --git a/include/linux/overflow.h b/include/linux/overflow.h > index 8712ff70995f..69fc366ce865 100644 > --- a/include/linux/overflow.h > +++ b/include/linux/overflow.h > @@ -202,6 +202,37 @@ > > #endif /* COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW */ > > +/** check_shift_overflow() - Calculate a left-shifted value and check overflow > + * > + * @a: Value to be shifted > + * @b: How many bits left to shift The above @b should be @s > +#define check_shift_overflow(a, s, d) ({ \ Should I run this series through the rdma tree? Thanks, Jason ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH v2 1/3] overflow.h: Add arithmetic shift helper 2018-08-01 2:15 ` Jason Gunthorpe @ 2018-08-01 4:22 ` Kees Cook 0 siblings, 0 replies; 13+ messages in thread From: Kees Cook @ 2018-08-01 4:22 UTC (permalink / raw) To: Jason Gunthorpe Cc: Rasmus Villemoes, Leon Romanovsky, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, LKML On Tue, Jul 31, 2018 at 7:15 PM, Jason Gunthorpe <jgg@ziepe.ca> wrote: > On Tue, Jul 31, 2018 at 05:00:37PM -0700, Kees Cook wrote: >> From: Jason Gunthorpe <jgg@mellanox.com> >> >> Add shift_overflow() helper to assist driver authors in ensuring that >> shift operations don't cause overflows or other odd conditions. >> >> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> >> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> >> [kees: tweaked comments and commit log, dropped unneeded assignment] >> Signed-off-by: Kees Cook <keescook@chromium.org> >> --- >> include/linux/overflow.h | 31 +++++++++++++++++++++++++++++++ >> 1 file changed, 31 insertions(+) >> >> diff --git a/include/linux/overflow.h b/include/linux/overflow.h >> index 8712ff70995f..69fc366ce865 100644 >> --- a/include/linux/overflow.h >> +++ b/include/linux/overflow.h >> @@ -202,6 +202,37 @@ >> >> #endif /* COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW */ >> >> +/** check_shift_overflow() - Calculate a left-shifted value and check overflow >> + * >> + * @a: Value to be shifted >> + * @b: How many bits left to shift > > The above @b should be @s Ooops, copy-paste-o. :) > >> +#define check_shift_overflow(a, s, d) ({ \ > > Should I run this series through the rdma tree? I'd like to get Rasmus's Ack, but otherwise, yes, that'd be fine. -Kees -- Kees Cook Pixel Security ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH v2 1/3] overflow.h: Add arithmetic shift helper 2018-08-01 0:00 ` [PATCH v2 1/3] " Kees Cook 2018-08-01 1:59 ` Randy Dunlap 2018-08-01 2:15 ` Jason Gunthorpe @ 2018-08-01 7:57 ` Dan Carpenter 2018-08-01 8:07 ` Dan Carpenter 2 siblings, 1 reply; 13+ messages in thread From: Dan Carpenter @ 2018-08-01 7:57 UTC (permalink / raw) To: Kees Cook Cc: Rasmus Villemoes, Jason Gunthorpe, Leon Romanovsky, Jason Gunthorpe, Leon Romanovsky, Bart Van Assche, Doug Ledford, linux-rdma, linux-kernel The idea is nice, but I don't like the API. The "_overflow" feels too specific because maybe we could check for other things in the future. Normally boolean macros should say they are boolean in the name and I would prefer if it returned zero on failure. if (!checked_shift(dest, mask, shift)) { if (!shift_ok(dest, mask, shift)) { if (!safe_shift(dest, mask, shift)) { regards, dan carpenter ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH v2 1/3] overflow.h: Add arithmetic shift helper 2018-08-01 7:57 ` Dan Carpenter @ 2018-08-01 8:07 ` Dan Carpenter 2018-08-01 15:38 ` Kees Cook 2018-08-01 15:48 ` Jason Gunthorpe 0 siblings, 2 replies; 13+ messages in thread From: Dan Carpenter @ 2018-08-01 8:07 UTC (permalink / raw) To: Kees Cook Cc: Rasmus Villemoes, Jason Gunthorpe, Leon Romanovsky, Jason Gunthorpe, Leon Romanovsky, Bart Van Assche, Doug Ledford, linux-rdma, linux-kernel On Wed, Aug 01, 2018 at 10:57:44AM +0300, Dan Carpenter wrote: > The idea is nice, but I don't like the API. The "_overflow" feels too > specific because maybe we could check for other things in the future. > Normally boolean macros should say they are boolean in the name and I > would prefer if it returned zero on failure. > > if (!checked_shift(dest, mask, shift)) { > if (!shift_ok(dest, mask, shift)) { > if (!safe_shift(dest, mask, shift)) { Huh... It turns out I put the argument order different as well. If we wanted to keep it returning 1 on failure then some other names are: if (shift_failed(dest, mask, shift)) { if (shift_error(dest, mask, shift)) { if (shift_overflow(dest, mask, shift)) { regards, dan carpenter ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH v2 1/3] overflow.h: Add arithmetic shift helper 2018-08-01 8:07 ` Dan Carpenter @ 2018-08-01 15:38 ` Kees Cook 2018-08-01 15:48 ` Jason Gunthorpe 1 sibling, 0 replies; 13+ messages in thread From: Kees Cook @ 2018-08-01 15:38 UTC (permalink / raw) To: Dan Carpenter Cc: Rasmus Villemoes, Jason Gunthorpe, Leon Romanovsky, Jason Gunthorpe, Leon Romanovsky, Bart Van Assche, Doug Ledford, linux-rdma, LKML On Wed, Aug 1, 2018 at 1:07 AM, Dan Carpenter <dan.carpenter@oracle.com> wrote: > On Wed, Aug 01, 2018 at 10:57:44AM +0300, Dan Carpenter wrote: >> The idea is nice, but I don't like the API. The "_overflow" feels too >> specific because maybe we could check for other things in the future. >> Normally boolean macros should say they are boolean in the name and I >> would prefer if it returned zero on failure. >> >> if (!checked_shift(dest, mask, shift)) { >> if (!shift_ok(dest, mask, shift)) { >> if (!safe_shift(dest, mask, shift)) { > > Huh... It turns out I put the argument order different as well. > > If we wanted to keep it returning 1 on failure then some other names > are: > > if (shift_failed(dest, mask, shift)) { > if (shift_error(dest, mask, shift)) { > if (shift_overflow(dest, mask, shift)) { This is following the existing check_{add,mul}_overflow() helpers, which are based on the gcc helpers. I'd like to keep things consistent. -Kees -- Kees Cook Pixel Security ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH v2 1/3] overflow.h: Add arithmetic shift helper 2018-08-01 8:07 ` Dan Carpenter 2018-08-01 15:38 ` Kees Cook @ 2018-08-01 15:48 ` Jason Gunthorpe 1 sibling, 0 replies; 13+ messages in thread From: Jason Gunthorpe @ 2018-08-01 15:48 UTC (permalink / raw) To: Dan Carpenter Cc: Kees Cook, Rasmus Villemoes, Leon Romanovsky, Leon Romanovsky, Bart Van Assche, Doug Ledford, linux-rdma, linux-kernel On Wed, Aug 01, 2018 at 11:07:24AM +0300, Dan Carpenter wrote: > On Wed, Aug 01, 2018 at 10:57:44AM +0300, Dan Carpenter wrote: > > The idea is nice, but I don't like the API. The "_overflow" feels too > > specific because maybe we could check for other things in the future. > > Normally boolean macros should say they are boolean in the name and I > > would prefer if it returned zero on failure. > > > > if (!checked_shift(dest, mask, shift)) { > > if (!shift_ok(dest, mask, shift)) { > > if (!safe_shift(dest, mask, shift)) { > > Huh... It turns out I put the argument order different as well. > > If we wanted to keep it returning 1 on failure then some other names > are: > > if (shift_failed(dest, mask, shift)) { > if (shift_error(dest, mask, shift)) { > if (shift_overflow(dest, mask, shift)) { I think this ship has sailed, the convention for these tests is already established in overflow.h. ie: check_add_overflow check_sub_overflow check_mul_overflow Jason ^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH v2 2/3] test_overflow: Add shift overflow tests 2018-08-01 0:00 [PATCH v2 0/3] overflow.h: Add arithmetic shift helper Kees Cook 2018-08-01 0:00 ` [PATCH v2 1/3] " Kees Cook @ 2018-08-01 0:00 ` Kees Cook 2018-08-01 2:13 ` Jason Gunthorpe 2018-08-01 0:00 ` [PATCH v2 3/3] RDMA/mlx5: Fix shift overflow in mlx5_ib_create_wq Kees Cook 2 siblings, 1 reply; 13+ messages in thread From: Kees Cook @ 2018-08-01 0:00 UTC (permalink / raw) To: Rasmus Villemoes Cc: Kees Cook, Jason Gunthorpe, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, linux-kernel This adds overflow tests for the new check_shift_overflow() helper to validate overflow, signedness glitches, storage glitches, etc. Co-developed-by: Rasmus Villemoes <linux@rasmusvillemoes.dk> Signed-off-by: Kees Cook <keescook@chromium.org> --- lib/test_overflow.c | 140 +++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 139 insertions(+), 1 deletion(-) diff --git a/lib/test_overflow.c b/lib/test_overflow.c index 2278fe05a1b0..face66e8aa37 100644 --- a/lib/test_overflow.c +++ b/lib/test_overflow.c @@ -252,7 +252,8 @@ static int __init test_ ## t ## _overflow(void) { \ int err = 0; \ unsigned i; \ \ - pr_info("%-3s: %zu tests\n", #t, ARRAY_SIZE(t ## _tests)); \ + pr_info("%-3s: %zu arithmetic tests\n", #t, \ + ARRAY_SIZE(t ## _tests)); \ for (i = 0; i < ARRAY_SIZE(t ## _tests); ++i) \ err |= do_test_ ## t(&t ## _tests[i]); \ return err; \ @@ -287,6 +288,142 @@ static int __init test_overflow_calculation(void) return err; } +static int __init test_overflow_shift(void) +{ + int err = 0; + +/* Args are: value, shift, type, expected result, overflow expected */ +#define TEST_ONE_SHIFT(a, s, t, expect, of) ({ \ + int __failed = 0; \ + typeof(a) __a = (a); \ + typeof(s) __s = (s); \ + t __e = (expect); \ + t __d; \ + bool __of = check_shift_overflow(__a, __s, &__d); \ + if (__of != of) { \ + pr_warn("expected (%s)(%s << %s) to%s overflow\n", \ + #t, #a, #s, of ? "" : " not"); \ + __failed = 1; \ + } else if (!__of && __d != __e) { \ + pr_warn("expected (%s)(%s << %s) == %s\n", \ + #t, #a, #s, #expect); \ + if ((t)-1 < 0) \ + pr_warn("got %lld\n", (s64)__d); \ + else \ + pr_warn("got %llu\n", (u64)__d); \ + __failed = 1; \ + } \ + if (!__failed) \ + pr_info("ok: (%s)(%s << %s) == %s\n", #t, #a, #s, \ + of ? "overflow" : #expect); \ + __failed; \ +}) + + /* Sane shifts */ + err |= TEST_ONE_SHIFT(1, 4, u8, 1 << 4, false); + err |= TEST_ONE_SHIFT(1, 7, u8, 1 << 7, false); + err |= TEST_ONE_SHIFT(1, 10, u16, 1 << 10, false); + err |= TEST_ONE_SHIFT(1, 15, u16, 1 << 15, false); + err |= TEST_ONE_SHIFT(1, 30, int, 1 << 30, false); + err |= TEST_ONE_SHIFT(1, 30, s32, 1 << 30, false); + err |= TEST_ONE_SHIFT(1, 20, unsigned int, 1U << 20, false); + err |= TEST_ONE_SHIFT(1, 31, unsigned int, 1U << 31, false); + err |= TEST_ONE_SHIFT(1, 20, u32, 1U << 20, false); + err |= TEST_ONE_SHIFT(1, 31, u32, 1U << 31, false); + err |= TEST_ONE_SHIFT(1, 40, u64, 1ULL << 40, false); + err |= TEST_ONE_SHIFT(1, 63, u64, 1ULL << 63, false); + + /* Sane shift: start and end with 0, without a too-wide shift. */ + err |= TEST_ONE_SHIFT(0, 7, u8, 0, false); + err |= TEST_ONE_SHIFT(0, 15, u16, 0, false); + err |= TEST_ONE_SHIFT(0, 31, unsigned int, 0, false); + err |= TEST_ONE_SHIFT(0, 31, u32, 0, false); + err |= TEST_ONE_SHIFT(0, 63, u64, 0, false); + + /* Sane shift: start and end with 0, without reaching signed bit. */ + err |= TEST_ONE_SHIFT(0, 6, s8, 0, false); + err |= TEST_ONE_SHIFT(0, 14, s16, 0, false); + err |= TEST_ONE_SHIFT(0, 30, int, 0, false); + err |= TEST_ONE_SHIFT(0, 30, s32, 0, false); + err |= TEST_ONE_SHIFT(0, 62, s64, 0, false); + + /* Overflow: shifted the bit off the end. */ + err |= TEST_ONE_SHIFT(1, 8, u8, 0, true); + err |= TEST_ONE_SHIFT(1, 16, u16, 0, true); + err |= TEST_ONE_SHIFT(1, 32, unsigned int, 0, true); + err |= TEST_ONE_SHIFT(1, 32, u32, 0, true); + err |= TEST_ONE_SHIFT(1, 64, u64, 0, true); + + /* Overflow: high bit falls off. */ + /* 10010110 */ + err |= TEST_ONE_SHIFT(150, 1, u8, 0, true); + /* 1000100010010110 */ + err |= TEST_ONE_SHIFT(34966, 1, u16, 0, true); + /* 10000100000010001000100010010110 */ + err |= TEST_ONE_SHIFT(2215151766U, 1, u32, 0, true); + err |= TEST_ONE_SHIFT(2215151766U, 1, unsigned int, 0, true); + /* 1000001000010000010000000100000010000100000010001000100010010110 */ + err |= TEST_ONE_SHIFT(9372061470395238550ULL, 1, u64, 0, true); + + /* Nonsense: negative initial value. */ + err |= TEST_ONE_SHIFT(-1, 0, s8, 0, true); + err |= TEST_ONE_SHIFT(-5, 0, s16, 0, true); + err |= TEST_ONE_SHIFT(-10, 0, int, 0, true); + err |= TEST_ONE_SHIFT(-100, 0, s32, 0, true); + err |= TEST_ONE_SHIFT(-10000, 0, s64, 0, true); + + /* Nonsense: negative shift values. */ + err |= TEST_ONE_SHIFT(0, -5, s8, 0, true); + err |= TEST_ONE_SHIFT(0, -10, s16, 0, true); + err |= TEST_ONE_SHIFT(0, -15, int, 0, true); + err |= TEST_ONE_SHIFT(0, -20, s32, 0, true); + err |= TEST_ONE_SHIFT(0, -30, s64, 0, true); + + /* Overflow: shifted at or beyond entire type's bit width. */ + err |= TEST_ONE_SHIFT(0, 8, u8, 0, true); + err |= TEST_ONE_SHIFT(0, 9, u8, 0, true); + err |= TEST_ONE_SHIFT(0, 8, s8, 0, true); + err |= TEST_ONE_SHIFT(0, 9, s8, 0, true); + err |= TEST_ONE_SHIFT(0, 16, u16, 0, true); + err |= TEST_ONE_SHIFT(0, 17, u16, 0, true); + err |= TEST_ONE_SHIFT(0, 16, s16, 0, true); + err |= TEST_ONE_SHIFT(0, 17, s16, 0, true); + err |= TEST_ONE_SHIFT(0, 32, u32, 0, true); + err |= TEST_ONE_SHIFT(0, 33, u32, 0, true); + err |= TEST_ONE_SHIFT(0, 32, int, 0, true); + err |= TEST_ONE_SHIFT(0, 33, int, 0, true); + err |= TEST_ONE_SHIFT(0, 32, s32, 0, true); + err |= TEST_ONE_SHIFT(0, 33, s32, 0, true); + err |= TEST_ONE_SHIFT(0, 64, u64, 0, true); + err |= TEST_ONE_SHIFT(0, 65, u64, 0, true); + err |= TEST_ONE_SHIFT(0, 64, s64, 0, true); + err |= TEST_ONE_SHIFT(0, 65, s64, 0, true); + + /* Overflow: shifted into signed bit. */ + err |= TEST_ONE_SHIFT(1, 7, s8, 0, true); + err |= TEST_ONE_SHIFT(1, 15, s16, 0, true); + err |= TEST_ONE_SHIFT(1, 31, int, 0, true); + err |= TEST_ONE_SHIFT(1, 31, s32, 0, true); + err |= TEST_ONE_SHIFT(1, 63, s64, 0, true); + + /* + * Corner case: for unsigned types, we fail when we've shifted + * through the entire width of bits. For signed types, we might + * want to match this behavior, but that would mean noticing if + * we shift through all but the signed bit, and this is not + * currently detected (but we'll notice an overflow into the + * signed bit). So, for now, we will test this condition but + * mark it as not expected to overflow. + */ + err |= TEST_ONE_SHIFT(0, 7, s8, 0, false); + err |= TEST_ONE_SHIFT(0, 15, s16, 0, false); + err |= TEST_ONE_SHIFT(0, 31, int, 0, false); + err |= TEST_ONE_SHIFT(0, 31, s32, 0, false); + err |= TEST_ONE_SHIFT(0, 63, s64, 0, false); + + return err; +} + /* * Deal with the various forms of allocator arguments. See comments above * the DEFINE_TEST_ALLOC() instances for mapping of the "bits". @@ -397,6 +534,7 @@ static int __init test_module_init(void) int err = 0; err |= test_overflow_calculation(); + err |= test_overflow_shift(); err |= test_overflow_allocation(); if (err) { -- 2.17.1 ^ permalink raw reply related [flat|nested] 13+ messages in thread
* Re: [PATCH v2 2/3] test_overflow: Add shift overflow tests 2018-08-01 0:00 ` [PATCH v2 2/3] test_overflow: Add shift overflow tests Kees Cook @ 2018-08-01 2:13 ` Jason Gunthorpe 2018-08-01 5:51 ` Kees Cook 0 siblings, 1 reply; 13+ messages in thread From: Jason Gunthorpe @ 2018-08-01 2:13 UTC (permalink / raw) To: Kees Cook Cc: Rasmus Villemoes, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, linux-kernel On Tue, Jul 31, 2018 at 05:00:38PM -0700, Kees Cook wrote: > + /* Overflow: high bit falls off. */ > + /* 10010110 */ > + err |= TEST_ONE_SHIFT(150, 1, u8, 0, true); > + /* 1000100010010110 */ > + err |= TEST_ONE_SHIFT(34966, 1, u16, 0, true); > + /* 10000100000010001000100010010110 */ > + err |= TEST_ONE_SHIFT(2215151766U, 1, u32, 0, true); > + err |= TEST_ONE_SHIFT(2215151766U, 1, unsigned int, 0, true); > + /* 1000001000010000010000000100000010000100000010001000100010010110 */ > + err |= TEST_ONE_SHIFT(9372061470395238550ULL, 1, u64, 0, true); This same idea should be repeated with signed outputs and check both overflow past the end (<<2) and overflow into the signed bit (<<1) /* Overflow, high bit falls into the sign bit or off the end */ /* 01001011 */ err |= TEST_ONE_SHIFT(75, 1, s8, 0, true); err |= TEST_ONE_SHIFT(75, 2, s8, 0, true); And also general type mismatch overflow: err |= TEST_ONE_SHIFT(0x100, 0, u8, 0, true); err |= TEST_ONE_SHIFT(0xFF, 0, s8, 0, true); Thanks, Jason ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH v2 2/3] test_overflow: Add shift overflow tests 2018-08-01 2:13 ` Jason Gunthorpe @ 2018-08-01 5:51 ` Kees Cook 0 siblings, 0 replies; 13+ messages in thread From: Kees Cook @ 2018-08-01 5:51 UTC (permalink / raw) To: Jason Gunthorpe Cc: Rasmus Villemoes, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, LKML On Tue, Jul 31, 2018 at 7:13 PM, Jason Gunthorpe <jgg@ziepe.ca> wrote: > On Tue, Jul 31, 2018 at 05:00:38PM -0700, Kees Cook wrote: >> + /* Overflow: high bit falls off. */ >> + /* 10010110 */ >> + err |= TEST_ONE_SHIFT(150, 1, u8, 0, true); >> + /* 1000100010010110 */ >> + err |= TEST_ONE_SHIFT(34966, 1, u16, 0, true); >> + /* 10000100000010001000100010010110 */ >> + err |= TEST_ONE_SHIFT(2215151766U, 1, u32, 0, true); >> + err |= TEST_ONE_SHIFT(2215151766U, 1, unsigned int, 0, true); >> + /* 1000001000010000010000000100000010000100000010001000100010010110 */ >> + err |= TEST_ONE_SHIFT(9372061470395238550ULL, 1, u64, 0, true); > > This same idea should be repeated with signed outputs and check both > overflow past the end (<<2) and overflow into the signed bit (<<1) > > /* Overflow, high bit falls into the sign bit or off the end */ > /* 01001011 */ > err |= TEST_ONE_SHIFT(75, 1, s8, 0, true); > err |= TEST_ONE_SHIFT(75, 2, s8, 0, true); > > And also general type mismatch overflow: > > err |= TEST_ONE_SHIFT(0x100, 0, u8, 0, true); > err |= TEST_ONE_SHIFT(0xFF, 0, s8, 0, true); Ah yes, thanks. I've added these for the next version now. -Kees -- Kees Cook Pixel Security ^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH v2 3/3] RDMA/mlx5: Fix shift overflow in mlx5_ib_create_wq 2018-08-01 0:00 [PATCH v2 0/3] overflow.h: Add arithmetic shift helper Kees Cook 2018-08-01 0:00 ` [PATCH v2 1/3] " Kees Cook 2018-08-01 0:00 ` [PATCH v2 2/3] test_overflow: Add shift overflow tests Kees Cook @ 2018-08-01 0:00 ` Kees Cook 2 siblings, 0 replies; 13+ messages in thread From: Kees Cook @ 2018-08-01 0:00 UTC (permalink / raw) To: Rasmus Villemoes Cc: Kees Cook, Leon Romanovsky, stable, syzkaller, Jason Gunthorpe, Leon Romanovsky, Bart Van Assche, Doug Ledford, Dan Carpenter, linux-rdma, linux-kernel From: Leon Romanovsky <leonro@mellanox.com> [ 61.182439] UBSAN: Undefined behaviour in drivers/infiniband/hw/mlx5/qp.c:5366:34 [ 61.183673] shift exponent 4294967288 is too large for 32-bit type 'unsigned int' [ 61.185530] CPU: 0 PID: 639 Comm: qp Not tainted 4.18.0-rc1-00037-g4aa1d69a9c60-dirty #96 [ 61.186981] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-2.fc27 04/01/2014 [ 61.188315] Call Trace: [ 61.188661] dump_stack+0xc7/0x13b [ 61.190427] ubsan_epilogue+0x9/0x49 [ 61.190899] __ubsan_handle_shift_out_of_bounds+0x1ea/0x22f [ 61.197040] mlx5_ib_create_wq+0x1c99/0x1d50 [ 61.206632] ib_uverbs_ex_create_wq+0x499/0x820 [ 61.213892] ib_uverbs_write+0x77e/0xae0 [ 61.248018] vfs_write+0x121/0x3b0 [ 61.249831] ksys_write+0xa1/0x120 [ 61.254024] do_syscall_64+0x7c/0x2a0 [ 61.256178] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 61.259211] RIP: 0033:0x7f54bab70e99 [ 61.262125] Code: 00 f3 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 [ 61.268678] RSP: 002b:00007ffe1541c318 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 [ 61.271076] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f54bab70e99 [ 61.273795] RDX: 0000000000000070 RSI: 0000000020000240 RDI: 0000000000000003 [ 61.276982] RBP: 00007ffe1541c330 R08: 00000000200078e0 R09: 0000000000000002 [ 61.280035] R10: 0000000000000000 R11: 0000000000000246 R12: 00000000004005c0 [ 61.283279] R13: 00007ffe1541c420 R14: 0000000000000000 R15: 0000000000000000 Cc: <stable@vger.kernel.org> # 4.7 Fixes: 79b20a6c3014 ("IB/mlx5: Add receive Work Queue verbs") Cc: syzkaller <syzkaller@googlegroups.com> Reported-by: Noa Osherovich <noaos@mellanox.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Kees Cook <keescook@chromium.org> --- drivers/infiniband/hw/mlx5/qp.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index a4f1f638509f..7fc156db643d 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -5365,7 +5365,10 @@ static int set_user_rq_size(struct mlx5_ib_dev *dev, rwq->wqe_count = ucmd->rq_wqe_count; rwq->wqe_shift = ucmd->rq_wqe_shift; - rwq->buf_size = (rwq->wqe_count << rwq->wqe_shift); + if (check_shift_overflow(rwq->wqe_count, rwq->wqe_shift, + &rwq->buf_size)) + return -EINVAL; + rwq->log_rq_stride = rwq->wqe_shift; rwq->log_rq_size = ilog2(rwq->wqe_count); return 0; -- 2.17.1 ^ permalink raw reply related [flat|nested] 13+ messages in thread
end of thread, other threads:[~2018-08-01 15:48 UTC | newest] Thread overview: 13+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2018-08-01 0:00 [PATCH v2 0/3] overflow.h: Add arithmetic shift helper Kees Cook 2018-08-01 0:00 ` [PATCH v2 1/3] " Kees Cook 2018-08-01 1:59 ` Randy Dunlap 2018-08-01 2:15 ` Jason Gunthorpe 2018-08-01 4:22 ` Kees Cook 2018-08-01 7:57 ` Dan Carpenter 2018-08-01 8:07 ` Dan Carpenter 2018-08-01 15:38 ` Kees Cook 2018-08-01 15:48 ` Jason Gunthorpe 2018-08-01 0:00 ` [PATCH v2 2/3] test_overflow: Add shift overflow tests Kees Cook 2018-08-01 2:13 ` Jason Gunthorpe 2018-08-01 5:51 ` Kees Cook 2018-08-01 0:00 ` [PATCH v2 3/3] RDMA/mlx5: Fix shift overflow in mlx5_ib_create_wq Kees Cook
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).