The loop exits early iff there is a nonzero limb, so i==0 means that
all limbs are 0, whether the number of limbs is 0 or not.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Fix a bug introduced in "Fix multiplication producing a negative zero" that
caused the sign to be forced to +1 when A > 0, B < 0 and B's low-order limb
is 0.
Add a non-regression test. More generally, systematically test combinations
of leading zeros, trailing zeros and signs.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
In mbedtls_mpi_read_string, if the string is empty, return an empty bignum
rather than a bignum with one limb with the value 0.
Both representations are correct, so this is not, in principle, a
user-visible change. The change does leak however through
mbedtls_mpi_write_string in base 16 (but not in other bases), as it writes a
bignum with 0 limbs as "" but a bignum with the value 0 and at least one
limb as "00".
This change makes it possible to construct an empty bignum through
mbedtls_mpi_read_string, which is especially useful to construct test
cases (a common use of mbedtls_mpi_read_string, as most formats use in
production encode numbers in binary, to be read with mbedtls_mpi_read_binary
or mbedtls_mpi_read_binary_le).
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Fix mbedtls_mpi_mul_mpi() when one of the operands is zero and the
other is negative. The sign of the result must be 1, since some
library functions do not treat {-1, 0, NULL} or {-1, n, {0}} as
representing the value 0.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Shifting TA and TB before the loop is not necessary. If A != 0, it will be
done at the start of the loop iteration. If A == 0, then lz==0 and G is
correctly set to B after 0 loop iterations.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Fix a null pointer dereference in mbedtls_mpi_exp_mod(X, A, N, E, _RR) when
A is the value 0 represented with 0 limbs.
Make the code a little more robust against similar bugs.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Unrelated to RSA (only used in ECP), but while improving one
mbedtls_safe_cond_xxx function, let's improve the other as well.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
mbedtls_mpi_cf_bool_eq() is a verbatim copy of mbedtls_ssl_cf_bool_eq()
Deduplication will be part of a future task.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
Calling mbedtls_mpi_cmp_int reveals the number of leading zero limbs
to an adversary who is capable of very fine-grained timing
measurements. This is very little information, but could be practical
with secp521r1 (1/512 chance of the leading limb being 0) if the
adversary can measure the precise timing of a large number of
signature operations.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
The idiom "resize an mpi to a given size" appeared 4 times. Unify it
in a single function. Guarantee that the value is set to 0, which is
required by some of the callers and not a significant expense where
not required.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Since the internal function mpi_fill_random_internal() assumes that X
has the right size, there is no need to call grow().
To further simplify the function, set the sign outside, and zero out
the non-randomized part directly.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
In real life, min << N and the probability that mbedtls_mpi_random()
fails to find a suitable value after 30 iterations is less than one in
a billion. But at least for testing purposes, it's useful to not
outright reject "silly" small values of N, and for such values, 30
iterations is not enough to have a good probability of success.
Pick 250 iterations, which is enough for cases like (min=3, N=4), but
not for cases like (min=255, N=256).
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This comment is no longer in the specific context of generating a
random point on an elliptic curve.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
mbedtls_mpi_random() uses mbedtls_mpi_cmp_mpi_ct(), which requires its
two arguments to have the same storage size. This was not the case
when the upper bound passed to mbedtls_mpi_random() had leading zero
limbs.
Fix this by forcing the result MPI to the desired size. Since this is
not what mbedtls_mpi_fill_random() does, don't call it from
mbedtls_mpi_random(), but instead call a new auxiliary function.
Add tests to cover this and other conditions with varying sizes for
the two arguments.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Since mbedtls_mpi_random() is not specific to ECC code, move it from
the ECP module to the bignum module.
This increases the code size in builds without short Weierstrass
curves (including builds without ECC at all) that do not optimize out
unused functions.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Move the handling of the sign out of the base-specific loops. This
both simplifies the code, and corrects an edge case: the code in the
non-hexadecimal case depended on mbedtls_mpi_mul_int() preserving the
sign bit when multiplying a "negative zero" MPI by an integer, which
used to be the case but stopped with PR #2512.
Fix#4295. Thanks to Guido Vranken for analyzing the cause of the bug.
Credit to OSS-Fuzz.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Fix a buffer overflow in mbedtls_mpi_sub_abs() when calculating
|A| - |B| where |B| is larger than |A| and has more limbs (so the
function should return MBEDTLS_ERR_MPI_NEGATIVE_VALUE).
Fix#4042
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
T1 is set to a 2-limb value. The first operation that takes it as
input is mbedtls_mpi_mul_int, which makes it grow to 3 limbs. Later it
is shifted left, which causes it to grow again. Set its size to the
final size from the start. This saves two calls to calloc(), at the
expense of a slowdown in some operations involving T1 as input since
it now has more leading zeros.
Setting T1 to 3 limbs initially instead of 2 saves about 6% of the
calloc() calls in test_suite_ecp and does not incur a performance
penalty. Setting T1 to A->n + 2 limbs instead of 2 saves about 20% of
the calloc calls and does not cause a measurable performance
difference on my Linux PC.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Note a possible microoptimization in mbedtls_mpi_mul_hlp that I tried
in the hope of reducing the number of allocations, but turned out to
be counterproductive.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Rewrite mbedtls_mpi_mul_int to call mpi_mul_hlp directly rather than
create a temporary mpi object. This has the benefit of not performing
an allocation when the multiplication is in place (mpi operand aliased
with the result) and the result mpi is large enough.
This saves about 40% of the calloc() calls in test_suite_ecp. There is
no measurable performance difference on my Linux PC.
The cost is a few bytes in bignum.o.
When there is no aliasing, or when there is aliasing but the mpi
object needs to be enlarged, the performance difference is negligible.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
mbedtls_mpi_sub_abs systematically allocated a new mpi when the result
was aliased with the right operand (i.e. X = A - X). This aliasing
very commonly happens during ECP operations. Rewrite the function to
allocate only if the result might not fit otherwise.
This costs a few bytes of code size in bignum.o, and might make
mbedtls_mpi_sub_abs very very slightly slower when no reallocation is
done. However, there is a substantial performance gain in ECP
operations with Montgomery curves (10-20% on my PC).
test_suite_ecp drops from 1422794 to 1271506 calls to calloc().
This commit also fixes a bug whereby mbedtls_mpi_sub_abs would leak
memory when X == B (so TB was in use) and the result was negative.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
