summaryrefslogtreecommitdiff
path: root/src/lib/libcrypto/bn (follow)
Commit message (Collapse)AuthorAgeFilesLines
...
* Flip roles of lowercase and uppercase A and B.tb2022-06-201-44/+44
| | | | | | | This matches Cohen's text better and makes the entire thing easier to read. suggested by jsing
* Clean up BN_kronecker()tb2022-06-201-73/+88
| | | | | | | | | | Instead of "Cohen's step N" explain in words what is being done. Things such as (A & B & 2) != 0 being equivalent to (-1)^((A-1)(B-1)/4) being negative are not entirely obvious... Remove the strange error dance and adjust variable names to what Cohen's book uses. Simplify various curly bits. ok jsing
* Fix some bizarre indentation and line breaks.tb2022-06-201-8/+7
|
* Fix prime recognition when doing trial divisionstb2022-06-181-2/+2
| | | | | | | | | If gcd(a, primes[i]) == 0 then a could still be a prime, namely in the case that a == primes[i], so check for that case as well. Problem noted by Martin Grenouilloux ok jsing
* Avoid strict aliasing violations in BN_nist_mod_*()jsing2022-05-071-86/+137
| | | | | | | | | | | | | | | | | | | | | | | The optimised code path switches from processing data via unsigned long to processing data via unsigned int, which requires type punning. This is currently attempted via a union (for one case), however this fails since a pointer to a union member is passed to another function (these unions were added to "fix strict-aliasing compiler warning" - it would seem the warnings stopped but the undefined behaviour remained). The second case does not use a union and simply casts from one type to another. Undefined behaviour is currently triggered when compiling with clang 14 using -03 and -fstrict-aliasing, while disabling assembly (in order to use this C code). The resulting binary produces incorrect results. Avoid strict aliasing violations by copying from an unsigned long array to an unsigned int array, then copying back the result. Any sensible compiler will omit the copies, while avoiding undefined behaviour that would result from unsafe type punning via pointer type casting. Thanks to Guido Vranken for reporting the issue and testing the fix. ok tb@
* Avoid use of uninitialized in BN_mod_exp_recp()tb2022-04-201-2/+3
| | | | | | | | | | If either of the two initial BN_CTX_get() fails, we will call BN_RECP_CTX_free() on the uninitialized recp, which won't end well, so hoist the BN_RECP_CTX_init() call a few lines up. From Pauli, OpenSSL ad249412 ok inoguchi jsing
* Fix infinite loop in BN_mod_sqrt()tb2022-03-151-14/+15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | A bug in the implementation of the Tonelli-Shanks algorithm can lead to an infinite loop. This loop can be hit in various ways, in particular on decompressing an elliptic curve public key via EC_POINT_oct2point() - to do this, one must solve y^2 = x^3 + ax + b for y, given x. If a certificate uses explicit encoding for elliptic curve parameters, this operation needs to be done during certificate verification, leading to a DoS. In particular, everything dealing with untrusted certificates is affected, notably TLS servers explicitly configured to request client certificates (httpd, smtpd, various VPN implementations, ...). Ordinary TLS servers do not consume untrusted certificates. The problem is that we cannot assume that x^3 + ax + b is actually a square on untrusted input and neither can we assume that the modulus p is a prime. Ensuring that p is a prime is too expensive (it would likely itself lead to a DoS). To avoid the infinite loop, fix the logic to be more resilient and explicitly limit the number of iterations that can be done. The bug is such that the infinite loop can also be hit for primes = 3 (mod 4) but fortunately that case is optimized earlier. It's also worth noting that there is a size bound on the field size enforced via OPENSSL_ECC_MAX_FIELD_BITS (= 661), which help mitigate further DoS vectors in presence of this fix. Reported by Tavis Ormandy and David Benjamin, Google Patch based on the fixes by David Benjamin and Tomas Mraz, OpenSSL ok beck inoguchi
* Avoid a NULL dereference in BN_mod_exp2_mont()tb2022-02-071-2/+2
| | | | | | | | | | This is a very rarely used function and the crash is hard to reach in practice. Instead of implementing BN_is_odd() badly by hand, just call the real thing. Reported by Guido Vranken ok beck jsing
* Check for zero modulus in BN_MONT_CTX_set().tb2022-02-071-1/+4
| | | | | | From OpenSSL 6a009812, prompted by a report by Guido Vranken ok beck jsing
* Add and fix check for BN functions return valueinoguchi2022-01-201-4/+5
| | | | ok jsing@ millert@ tb@
* Add check for BN functions return valueinoguchi2022-01-201-3/+5
| | | | | | | CID 21665 24835 comment from jsing@ and tb@ ok jsing@ millert@ tb@
* Move BN structs to bn_lcl.htb2022-01-142-50/+46
| | | | | | This makes all structs in bn.h opaque that are also opaque in OpenSSL. ok inoguchi jsing
* Unifdef LIBRESSL_OPAQUE_* and LIBRESSL_NEXT_APItb2022-01-141-65/+1
| | | | | This marks the start of major surgery in libcrypto. Do not attempt to build the tree for a while (~50 commits).
* Pull BN_{new,init,clear,clear_free,free} up to the top of bn_lib.cjsing2021-12-271-58/+58
| | | | Discussed with tb@
* Consistently call BN_init() before BN_with_flags()tb2021-12-261-13/+27
| | | | | | | | | | | | | | | | BN_with_flags() preserves the BN_FLG_MALLOCED flag of the destination which results in a potential use of an uninitialized bit. In practice this doesn't matter since we don't free the cloned BIGNUMs anyway. As jsing points out, these are mostly pointless noise and should be garbage collected. I'll leave that for another rainy day. Coverity flagged one instance BN_gcd_no_branch(), the rest was found by the ever so helpful grep(1). CID 345122 ok jsing
* Annotate the structs to be moved to bn_lcl.h in the next bumptb2021-12-041-1/+5
| | | | ok inoguchi jsing
* Use BN_is_negative(p) instead of p->neg in one place.tb2021-12-041-2/+2
|
* Implement the BN_to_montgomery() macro as a functiontb2021-12-042-2/+13
| | | | ok inoguchi jsing
* Implement the BN_is_negative macro as a functiontb2021-12-042-2/+13
| | | | ok inoguchi jsing
* Provide function implementations for various BN_* macrostb2021-12-042-4/+54
| | | | | | | | BN_abs_is_word, BN_is_{zero,one,word,odd}, BN_one, BN_zero_ex are now implemented as functions for internal use. They will be exposed publicly to replace the macros reaching into BIGNUM in the next bump. ok inoguchi jsing
* Provide replacement functions for the BN_{get,set,with}_flags() macros.tb2021-12-042-2/+34
| | | | ok inoguchi jsing
* Provide replacement functions for the BN_GENCB_set{,_old}() macrostb2021-12-042-3/+33
| | | | | | | | The function implementations are necessary to make BIGNUM opaque. They will be used in libcrypto internally until they will replace the macro implementations with the next bump. ok inoguchi jsing
* typo in commenttb2021-11-181-2/+2
|
* Switch to <endian.h> from <machine/endian.h> for better portability.bcook2021-11-091-3/+2
| | | | ok tb@
* Expose BN_bn2{,le}binpad() and BN_lebin2bn() in <openssl/bn.h>tb2021-09-101-3/+1
| | | | ok beck inoguchi
* Expose BN_RAND_* in <openssl/bn.h>tb2021-09-101-3/+1
| | | | ok beck jsing
* Prepare to provide BN_RAND_* flags for BN_rand_range()tb2021-09-101-1/+12
| | | | ok beck jsing
* Prepare to provide BN_bn2{,le}binpad() and BN_lebin2bn()tb2021-09-082-9/+137
| | | | | | | | | As found by jsg and patrick, this is needed for newer uboot and will also be used in upcoming elliptic curve work. This is from OpenSSL 1.1.1l with minor style tweaks. ok beck inoguchi
* whitespacetb2021-08-312-7/+7
|
* Unindent a bit of code that performs a few too many checks totb2020-09-121-10/+8
| | | | figure out whether top > 0 or top == 0.
* Avoid an out-of-bounds access in BN_rand()tb2020-09-121-3/+8
| | | | | | | | | | If BN_rand() is called with top > 0 and bits == 1, it would allocate a buf[] of size 1 and set the top bit of buf[1]. Found in OpenSSL commit efee575ad464bfb60bf72dcb73f9b51768f4b1a1 while looking for something else. ok beck djm inoguchi
* Change generating and checking of primes so that the error rate ofschwarze2019-08-251-18/+73
| | | | | | | | | | | not being prime depends on the intended use based on the size of the input. For larger primes this will result in more rounds of Miller-Rabin. The maximal error rate for primes with more than 1080 bits is lowered to 2^-128. Patch from Kurt Roeckx <kurt@roeckx.be> and Annie Yousar via OpenSSL commit feac7a1c Jul 25 18:55:16 2018 +0200, still under a free license. OK tb@.
* make BN_CTX_end(NULL) a NOOP for compatibility with documented behaviourschwarze2019-08-201-1/+4
| | | | | | | | | in OpenSSL 1.1.1 even though in general, letting random functions accept NULL is not advisable because it can hide programming errors; "yes please" tb@ "unfortunately I suspect you're right" jsing@ "oh well" deraadt@
* Make BN_num_bits_word() constant time.tb2019-06-171-48/+18
| | | | | | | | | | | | | | | | | Previously, this function would leak the most significant word of its argument due to branching and memory access pattern. This patch is enough to fix the use of BN_num_bits() on RSA prime factors in the library. The diff is a simplified and more readable (but perhaps less efficient) version of https://github.com/openssl/openssl/commit/972c87df by Andy Polyakov and David Benjamin (pre license change). Consult that commit message for details. Subsequent fixes to follow in the near future. Issue pointed out by David Schrammel and Samuel Weiser as part of a larger report. tests & ok inoguchi, ok jsing
* Add range checks to varios ASN1_INTEGER functions to ensure thebeck2019-03-231-1/+3
| | | | | | sizes used remain a positive integer. Should address issue 13799 from oss-fuzz ok tb@ jsing@
* Fix BN_is_prime_* calls in libcrypto, the API returns -1 on error.tb2019-01-201-10/+20
| | | | | | | From BoringSSL's commit 53409ee3d7595ed37da472bc73b010cd2c8a5ffd by David Benjamin. ok djm, jsing
* Flip reversed test in bn_rand_interval().tb2018-11-061-2/+2
| | | | ok jsing
* Introduce bn_rand_interval() that allows specifying an interval [a, b)tb2018-11-052-2/+30
| | | | | | from which a a BIGNUM is chosen uniformly at random. ok beck jsing
* Use a size_t instead of an int for the byte count in BN_swap_ct().tb2018-07-232-8/+11
| | | | | | | | Since bignums use ints for the same purpose, this still uses an int internally after an overflow check. Suggested by and discussed with jsing. ok inoguchi, jsing
* Clean up our disgusting implementations of BN_{,u}{add,sub}(), followingtb2018-07-231-157/+67
| | | | | | | | | changes made in OpenSSL by Davide Galassi and others, so that one can actually follow what is going on. There is no performance impact from this change as the code still does essentially the same thing. There's a ton of work still to be done to make the BN code less terrible. ok jsing, kn
* Eliminate the weird condition in the BN_swap_ct() API that at most one bittb2018-07-131-3/+3
| | | | | | | | be set in condition. This makes the constant time bit-twiddling a bit trickier, but it's not too bad. Thanks to halex for an extensive rubber ducking session over a non-spicy spicy tabouleh falafel.. ok jsing, kn
* Sync commentkn2018-07-111-3/+5
| | | | | | Makes it a tad easier to read through and compare with BN_swap_ct(). OK tb
* Turn yesterday's optimistic ! in an XXX comment into a more cautious ?tb2018-07-111-2/+2
|
* Provide BN_swap_ct(), a constant time function that conditionally swapstb2018-07-102-2/+53
| | | | | | | | | | two bignums. It's saner and substantially less ugly than the existing public BN_constantime_swap() function and will be used in forthcoming work on constant time ECC code. From Billy Brumley and his team. Thanks! ok jsing
* fix odd whitespacetb2018-06-101-3/+3
|
* Use explicit_bzero() rather than memset() when clearing a BIGNUM.jsing2018-05-121-2/+2
| | | | ok bcook@ tb@
* Provide BN_GENCB_new(), BN_GENCB_free() and BN_GENCB_get_arg()jsing2018-02-202-2/+32
|
* Provide BN_get_rfc2409_prime_*() and BN_get_rfc3526_prime_*().jsing2018-02-202-2/+58
|
* use freezero() instead of memset/explicit_bzero + free. Substantiallyderaadt2017-05-024-28/+14
| | | | | | | | | | reduces conditional logic (-218, +82). MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH cache alignment calculation bn/bn_exp.c wasn'tt quite right. Two other tricky bits with ASN1_STRING_FLAG_NDEF and BN_FLG_STATIC_DATA where the condition cannot be collapsed completely. Passes regress. ok beck
* Send the function codes from the error functions to the bit bucket,beck2017-01-2917-118/+70
| | | | | | as was done earlier in libssl. Thanks inoguchi@ for noticing libssl had more reacharounds into this. ok jsing@ inoguchi@
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-03-27 23:13:57 +0000