summaryrefslogtreecommitdiff
path: root/src/lib/libcrypto (follow)
Commit message (Collapse)AuthorAgeFilesLines
...
* Move EC_GROUP_check() to ec_lib.ctb2024-10-193-115/+57
| | | | EC_GROUP_check() is quite simple. It doesn't need to use its own file.
* Simplify EC_get_builtin_curves().tb2024-10-181-4/+5
| | | | | When determining the minimum of nitems and EC_CURVE_LIST_LENGTH we need neither an extra variable nor a ternary operator.
* Use better naming in ec_curve.ctb2024-10-181-33/+33
| | | | | | | Rename struct ec_list_element into struct ec_curve. Accordingly, curve_list becomes struct ec_curve ec_curve_list[]. Adjust internal API to match. suggested by jsing
* Enforce that EC Parameters correspond to a builtin curvetb2024-10-183-3/+227
| | | | | | | | | | | | | | | | | | | | | | | | | | | | EC parameters are very general. While there are some minimal sanity checks, for the parameters due to DoS risks found in the last decade, the elliptic curve code is poorly written and a target rich environment for NULL dereferences, busy loops, expensive computations and whatever other nastiness you can think of. It is not too hard to come up with parameters that reach very ugly code. While we have removed for the worst of it (the "fast" nist code and GF2m come to mind), the code very much resembles the Augean Stables. Unfortunately, curve parameters are still in use - even mandatory in some contexts - for example in machine-readable travel documents signed by ICAO country signing certification authorities (see ICAO Doc 9303). To avoid many of these DoS vectors, start enforcing that we know what the curve parameters are about, namely that they correspond to a builtin curve. This way we know that the parameters are at least as good as the standards we implement and checking this is cheap: Translate curve parameters into the ad hoc representation in the builtin curve code and check there's a match. That's very cheap since most curves are distinguished by cofactor and parameter length and we need to use an actual parameter comparison for at most half a dozen curves, usually only one or two. ok jsing
* Remove now unused x86cpuid.pl.jsing2024-10-181-153/+0
|
* Provide crypto_cpu_caps_init() for i386.jsing2024-10-183-10/+120
| | | | | | | This is the same CPU capabilities code that is now used for amd64. Like amd64 we now only populate OPENSSL_ia32cap_P with bits used by perlasm. Discussed with tb@
* Remove now unused x86_64cpuid.pl.jsing2024-10-181-147/+0
|
* Provide crypto_cpu_caps_init() for amd64.jsing2024-10-184-11/+126
| | | | | | | | | | | This is a CPU capability detection implementation in C, with minimal inline assembly (for cpuid and xgetbv). This replaces the assembly mess generated by x86_64cpuid.pl. Rather than populating OPENSSL_ia32cap_P directly with CPUID output, just set the bits that the remaining perlasm checks (namely AESNI, AVX, FXSR, INTEL, HT, MMX, PCLMUL, SSE, SSE2 and SSSE3). ok joshua@ tb@
* Inline last use of OPENSSL_load_builtin_modules()tb2024-10-184-77/+6
| | | | | | | This used to be a trivial wrapper of the ASN1_add_oid_module() horror. It's no longer exported, so it can go away. It moves from the terribly named file conf_mall.c to the equally terribly named file conf_sap.c. I have no idea what mall and sap are supposed to mean in this context.
* Move EC_GROUP_new_curve_GFp() into ec_lib.ctb2024-10-183-99/+23
| | | | Another single-function file goes away.
* Merge EC_GROUP_new_curve_GFp() with ec_group_new_curve()tb2024-10-181-12/+5
| | | | | | The latter was used for EC_GROUP_new_curve_GF2m() and is now pointless. ok jsing
* Split ec_asn1_parameters2group() into digestible piecestb2024-10-171-96/+144
| | | | | | | | | | | | | | | | | This becomes a simple wrapper function that currently does three checks: 1. ensure the fieldID is for a prime field 2. check that the purported prime is of reasonable size, extract and set curve coefficients and point conversion form 3. extract and set generator, order, cofactor and seed. Sanity checks such as the Hasse bound are dealt with in the EC_GROUP API, so need not be repeated here. They will become redundant once we enforce that the parameters represent a builtin curve anyway. ok jsing
* Provide crypto_cpu_caps_init() as a CPU capability detection entry point.jsing2024-10-173-5/+18
| | | | | | | This can be overridden on a per-architecture basis. The default version calls OPENSSL_cpuid_setup(), which will be eventually replaced/removed. ok joshua@ tb@
* Remove run once code from OPENSSL_cpuid_setup().jsing2024-10-171-5/+1
| | | | | | | OPENSSL_cpuid_setup() is no longer exported and is now only ever run under pthread_once(). ok joshua@ tb@
* Unindent error check in EC_GROUP_set_generator()tb2024-10-151-5/+5
|
* Switch ec_asn1_group2parameters() to get0_{order,cofactor}()tb2024-10-151-11/+14
| | | | | | | | These are more ergonomic, result in more readable code, avoid a copy and we no longer ignore a possible memory allocation error due to API misdesign and bad code. ok jsing
* Provide EC_GROUP_get0_cofactor() for internal usetb2024-10-152-2/+9
| | | | | | While this is public API in OpenSSL, there are no plans to expose it. ok jsing
* Make NULL checks in ec_asn1_group2curve() explicittb2024-10-141-3/+5
|
* Fix field element encoding for elliptic curve coefficientstb2024-10-141-45/+36
| | | | | | | | | | | | SEC 1, section 2.3.5, is explicit that the encoding of an element of the field of definition for an elliptic curve needs to be a zero-padded octet string whose length matches the byte size of the field's degree. So use BN_bn2binpad() to fix this. Factor things into a simple helper to avoid copy-pasting. This gets rid of some of the most grotesque code in this file. ok jsing
* Drop an obvious comment and fix indent for setting the seedtb2024-10-141-4/+3
| | | | | | Also remove a pointless cast. ok jsing
* In ec_asn1_group2curve() rename ok to ret, per usualtb2024-10-141-4/+4
| | | | ok jsing
* Use a and b for the curve coefficientstb2024-10-111-10/+11
| | | | | | | No idea how anyone would think that tmp_1 and tmp_2 are better suited for this. ok jsing
* Clean up ec_asn1_group2fieldid()tb2024-10-111-32/+28
| | | | | | | | This drops some unnecessary freeing that was turned into a double free reachable via public API in OpenSSL 1.1. Other than that it unindents code and uses better variable names. ok jsing
* Fix a long-standing bug in ec_asn1_group2pkparameters()tb2024-10-111-2/+2
| | | | | | | Only check for the OPENSSL_EC_NAMED_CURVE being set to treat the curve parameters as named curve parameters. ok jsing
* Use defines for the CHOICE variants of ECPKPARAMETERStb2024-10-111-7/+10
| | | | ok jsing
* Remove NULL check in ec_asn1_pkparameters2group()tb2024-10-111-5/+1
| | | | | | The callers already ensure that params != NULL. ok jsing
* Use better variable names in ec_asn1_pkparameters2group()tb2024-10-111-10/+13
| | | | ok jsing
* First cleanup pass over ec_asn1_group2pkparameters()tb2024-10-111-27/+25
| | | | | | | Use better variable names and do things in a slightly more sensible order. This way the code becomes almost self-documenting. ok jsing
* err.c: zap stray spacetb2024-10-111-2/+2
|
* Apply some style(9) and tidy some comments.jsing2024-10-111-62/+73
|
* Remove pointless locking from err_build_SYS_str_reasons().jsing2024-10-111-20/+2
| | | | | | | | err_build_SYS_str_reasons() is only called during initialisation, under pthread_once(). As such, there is no need to grab CRYPTO_LOCK_ERR or check for reentrant calls. ok tb@
* Convert err_clear_data() and err_clear() from macros to functions.jsing2024-10-111-23/+25
| | | | ok tb@
* Remove unused typedef for ERR_FNS.jsing2024-10-111-3/+1
|
* Add err_ prefix to build_SYS_str_reasons().jsing2024-10-111-3/+3
|
* Make ERR_str_{libraries,reasons,functs}[] const.jsing2024-10-111-12/+18
| | | | | | | | Provide err_load_const_strings(), which takes a const ERR_STRING_DATA * and does not perform a library error value fixup. Make ERR_str_*[] tables const. ok tb@
* Add ERR_LIB_SYS to the error values in ERR_str_functs[].jsing2024-10-111-12/+12
| | | | | | | | This is currently added via err_load_strings(), which means ERR_str_functs[] cannot be made const. Adding ERR_LIB_SYS means the fixup becomes unnecessary. ok tb@
* Remove params argument from ec_asn1_group2parameters()tb2024-10-111-13/+8
| | | | | | | Its only caller passes NULL, so we can simplify the entry point and the exit of this function a bit. ok jsing
* Simplify ec_asn1_group2pkparameters()tb2024-10-111-15/+8
| | | | | | | The parameters argument is always NULL, so we can simplify this helper accordingly. ok jsing
* Spell NULL in a simpler fashiontb2024-10-111-3/+5
| | | | | | | | priv_key->parameters is always NULL at this point, since its corresponding entry in the ASN.1 template has ASN1_TFLG_OPTIONAL set, so there is no point in pretending to pass it to ec_asn1_group2pkparameters(). ok jsing
* Clean up i2d_ECPKParameters()tb2024-10-111-10/+13
| | | | | | | | Use better variable names and turn it into single-exit. This changes the behavior slightly in that an error is pushed onto the stack also for i2d_ECPKPARAMETERS() return values < 0. ok jsing
* Remove unused CONF_imodule struct members and accessorstb2024-10-102-69/+2
| | | | ok beck jsing
* Remove the unused field_mod_func from EC_GROUPtb2024-10-031-4/+1
| | | | | | | | This was only used by the NIST method. For all other group methods it's an uninitialized pointer (as EC_GROUP_new() still uses the malloc + set all members to 0 idiom). ok jsing
* Make EC{,PK}PARAMETERS_it statictb2024-10-031-3/+3
| | | | They aren't used outside of this file.
* Fix ASN1_INTEGER_to_BN() misusetb2024-10-031-16/+16
| | | | | | | Same issue/leak as for BN_to_ASN1_INTEGER(). Stop reusing the elliptic curve parameters a and b for order and cofacter. It's confusing. ok jsing
* Switch field_bits to be an inttb2024-10-031-3/+3
| | | | ok jsing
* Fix BN_to_ASN1_INTEGER() misusetb2024-10-031-7/+6
| | | | | | | You can either let this API reuse an existing ASN1_INTEGER or you can let it allocate a new one. If you try to do both at the same time, you'll leak. ok jsing
* Provide OPENSSL_INIT_NO_ATEXIT nooptb2024-10-031-1/+2
| | | | | | | | | | | The brilliant idea of installing a fragile non-idempotent cleanup atexit handler as a library has bitten many people over time. This gets particularly exciting when you can't control who dlopens the lib first (don't we all love Python bindings) or if you are in a threaded context. Fake OpenSSL clones chose not to do this but now get to carry a noop flag since people start opting out of this madness (there's a good old tradition at work here). ok beck joshua jsing millert miod
* X509V3_EXT_get_nid.3: indicate what nid meanstb2024-10-031-3/+3
|
* Reorder functions.jsing2024-10-021-297/+286
| | | | | Reorder functions so that things are somewhat more logical, moving internal functions towards the top (and removing now unnecessary prototypes).
* Remove err_fns and associated machinery.jsing2024-10-021-142/+65
| | | | | | | | | | | | Like all good OpenSSL code, errors was built to be completely extensible. Thankfully, the ERR_{get,set}_implementation() functions were removed in r1.127 of err.c, which means that the extensibility can no longer be used. Take the first of many steps to clean up this code - remove err_fns and associated machinery, calling functions directly. Rename so that we have an 'err_' prefix rather than 'int_' (or nothing). ok joshua@ tb@
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-08-16 12:32:00 +0000