Compare commits

...
29 Commits
Author SHA1 Message Date
Si Beaumont ff0f781cf7 extras: Add EC toolbox abstractions and OPRF(P-384, SHA-384) VOPRF API (#292)
## Motivation

We would like to provide support for the P384-SHA384 Verifiable Oblivious Pseudorandom Function (VOPRF) as defined in [RFC 9497: VOPRF Protocol](https://www.rfc-editor.org/rfc/rfc9497.html#name-voprf-protocol).

## Modifications

- Add protocols for some "ECToolbox" abstractions: `Group`, `GroupElement`, `GroupScalar`, `HashToGroup`.
- Add implementations backed by BoringSSL.
- Duplicate some utilities from Crypto into _CryptoExtras: hexadecimal bytes and I2OSP.
- Add VOPRF API, rooted at `enum P384._VOPRF`.

## Results

- New internal abstractions that make it easier to bring up new implementations.
- New API for `OPRF(P-384, SHA-384)` in VOPRF mode.

## Tests

- Test vectors for internal `HashToField` implementation.
- Test vectors for internal VOPRF implementation.
- Tests of the VOPRF public API.
2024-11-22 09:30:35 +00:00
Dimitri Bouniol a3b7196be9 Potentially Inconsistent Coordinate Slicing (#299)
* Replaced potentially unexpected accesses to coordinate slices with consistent ranges no matter if the raw representation is a root or a slice of data

* Updated uses of `.suffix(from:)` with `.dropFirst()` to prevent issues with potential non-0 start indexes
2024-11-21 17:11:58 +00:00
Rick Newton-RogersandGeorge Barnett b6a32eb851 add .spi.yml (#297)
Add .spi.yml for `Crypto` target so that DocC documentation is
generated.

Co-authored-by: George Barnett <gbarnett@apple.com>
2024-11-15 09:52:22 +00:00
Rick Newton-Rogers e14f4bb065 add .editorconfig file (#296) 2024-11-15 09:48:15 +00:00
Kyle 06dc63c6d8 Fix WASI platform build issue (#291) 2024-10-30 13:36:45 +00:00
Paul ToffoloniandCory Benfield 62958ed575 Fix Windows CI (#289)
Co-authored-by: Cory Benfield <lukasa@apple.com>
2024-10-29 14:29:05 +00:00
Franz Busch e9d46e7bf9 [CI] Enable Windows CI (#288) 2024-10-29 13:57:41 +00:00
finagolfin 8fa345c208 Add support for new Android overlay, swiftlang/swift#74758 (#286) 2024-10-21 15:30:06 +00:00
Adam Kopeć 7cb61135cb New API Proposal: Password based KDF (#98)
Motivation:

I've added a proposed implementation of PBKDF2, as per proposal in issue #59 to allow easy creation of SymmetricKey from relatively low entropy data sets, such as user provided passwords.

Importance:

This API change is attempting to simplify the creation of SymmetricKeys from user entered passwords.
It uses an ageing algorithm – PBKDF2, but it is still widely supported and popular. A newer algorithm – scrypt is added as a better, more secure alternative.

Modifications:

I've proposed a new API:  KDF namespace with Scrypt struct and an Insecure namespace with PBKDF2 struct with associated hash function and a single static method which takes a passphrase and salt, the size of the resulting key in bytes and optionally the number of rounds to use in the PBKDF2 algorithm.
In the case of scrypt algorithm there is no associated hash function due to the nature of the algorithm, the single static method takes passphrase and salt, the size of the resulting key in bytes and optionally: the number of rounds, block size and parallelism factor.
I implemented this functions using both CommonCrypto and BoringSSL where available.
The function names and parameters are heavily influenced by current HKDF implementation.

Additionally, I've added test cases to test the proposed API. I've decided to use test vectors described in RFC6070 for PBKDF2 and test vectors described in RFC7914 for scrypt.

Result:

This will add a new API to _CryptoKitExtras.
It follows the naming conventions from CryptoKit's HKDF implementation.
2024-10-18 17:16:05 +00:00
Cory Benfield 4518e5e680 Fix windows build and update BoringSSL to 2587c4974dbe9872451151c8e975f58567a1ce0d (#279)
* Update vendoring script

* Update BoringSSL to 2587c4974dbe9872451151c8e975f58567a1ce0d
2024-10-17 12:26:13 +00:00
Si Beaumont 4324b40570 wrapper: Add modular arithmetic functions to ArbitraryPrecisionInteger (#284) 2024-10-17 12:23:48 +00:00
Si Beaumont f98e46e42a wrapper: Move EC types and random bytes into CryptoBoringWrapper (#277)
* wrapper: Remove _boring suffix from files in CryptoBoringWrapper

* wrapper: Move EC types and random bytes into CryptoBoringWrapper

* tests: Remove unused conditional compilation conditions in wrapper tests
2024-10-17 10:03:07 +01:00
Franz Busch ae321d29be [CI] Adopt GitHub actions (#280) 2024-10-16 18:04:07 +01:00
Si Beaumont 21f7878f2b Refactor and adopt package access to remove symlinked files (#265) 2024-10-08 14:10:48 +01:00
Cory Benfield 735a29ccf7 Update BoringSSL to 76968bb3d53982560bcf08bcd0ba3e1865fe15cd (#271)
Also fixes the build issues on Windows, resolves #262
2024-10-07 10:19:12 +01:00
Gautier Delorme b639b5bd1f Support loading RSAPSS public keys with parameters (#268)
* Support loading RSAPSS public keys with parameters

Parameters are stripped and the key is treated as a regular public key.

* address review + reset algorithm as well

* run update_cmakelists.sh

* attempt to fix the cmake build

* include FetchContent

* early return for non-PSS keys

* simplify code
2024-10-03 18:33:02 +01:00
Si Beaumont ffca28be3c crypto: Equatable and Hashable conformance for CryptoKitError and CryptoKitASN1Error (#266) 2024-10-02 09:59:06 +01:00
Si Beaumont 35703579f6 Drop support for Swift 5.8 (#259) 2024-09-19 13:55:18 +01:00
George Barnett a24950b8aa Only include privacy manifest on Darwin (#257)
Motivation:

The privacy manifest is only required on Darwin. It's currently included
on all platforms which results in build warnings.

Modifications:

- Only include manifest on Darwin
- Exclude the privacy manifest from the build on other platforms

Result:

- Fewer warnings
- Resolves #256
2024-09-18 17:21:21 +01:00
Paul Toffoloni e8f2ddd49e extras: Add API to create RSA keys with random primes (#250)
This adds a _createFromNumbers API which creates an RSA key generating p and q primes based on n, e and d

Motivation:

There might be cases where private RSA keys have to be generated based on their raw elements (modulus n, exponent e and private exponent d) without p and q. These are random primes which are integral parts of the key but can be generated based on the former numbers, and are not always included when representing RSA keys, for example these parameters are not required in JWKs

Modifications:

This adds a static _createFromNumbers method which creates a new RSA key given its raw modulus, exponent and private exponent, generating the needed p and q primes without having to provide them as parameters. The algorithm to generate primes is based on a hybrid approach of the Miller Rabin primality test and the proof of Fact 1 in Dan Boneh's Twenty Years of Attacks on the RSA Cryptosystem paper. To write this algorithm, some additions had to be made to the ArbitraryPrecisionInteger

Result:

Private RSA keys can now be generated without providing p and q
2024-09-16 13:52:23 +01:00
Paul ToffoloniandCory Benfield 55201a8f83 extras: Add RSA primitives decoding API (#249)
Motivation:

At times we might need to get the raw elements of an RSA key, e.g when debugging.

Modifications:

Add a getKeyPrimitives method to the RSA (public and private) key structures that returns the elements.
To do this we leverage the built in BoringSSL methods (CCryptoBoringSSL_RSA_get0_n etc.) which return the raw elements of the key. In the Security version of the key we convert the key to a BoringSSL one and just call the BoringSSL method.

Once we get ASN.1 support in _CryptoExtras it might make sense switching the implementation of this method to decode the ASN.1 structure and returns the elements from there.

Result:

New APIs to fetch the key primitives.

---------

Co-authored-by: Cory Benfield <lukasa@apple.com>
2024-09-16 04:34:21 -07:00
Cory Benfield 81bee98e70 Update BoringSSL to 6a2ccdcc2ed1d37a43a2183658d2ae61fd5ce208 (#255) 2024-09-16 10:02:51 +01:00
Franz Busch ce204da6f6 Add release.yml (#253) 2024-09-06 13:57:39 +01:00
Si Beaumont 9f95b4d033 Add support for AES CFB mode in _CryptoExtras (#252) 2024-09-03 09:28:28 +01:00
Jake Petroules a53a7e8f85 Convert PrivacyInfo.xcprivacy symlinks to files (#251)
It is expected behavior of the swift-build build system to copy links as-is, so presenting these as symlinks results in broken symlinks being present in the produced resource bundles, which won't work properly.

On Windows, it even causes the build entirely to fail due to a related bug: https://github.com/apple/swift-foundation/issues/845

Convert these to flat files instead.
2024-08-13 07:53:29 +01:00
Si BeaumontandCory Benfield 8dafe0fdce extras: Provide initializers for RSA keys from RSA numbers as bytes (#247)
Co-authored-by: Cory Benfield <lukasa@apple.com>
2024-07-22 04:01:44 -07:00
Nicolas BachschmidtandCory Benfield e2dd5a547b Reduce the number of heap allocations when computing hash digests (#238)
* Remove array allocation when computing digest

Instead of creating a temporary array, the digest bytes are stored in a temporary stack-allocated buffer.

This reduces the number of heap allocations needed to hash a message from 7 to 6.

* Store digest context inline

This reduces the number of heap allocations needed to hash a message from 6 to 2.

Before:
  1. the DigestContext object is created
  2. the EVP_MD_CTX struct is allocated by EVP_MD_CTX_new
  3. md_data is allocated when initializing the context

After:
  1. the DigestContext object is created

In both cases, the number is doubled because COW happens during finalize().

* Use temporary variable to finalize context

This reduces the number of heap allocations needed to hash a message from 2 to 1.

---------

Co-authored-by: Cory Benfield <lukasa@apple.com>
2024-07-18 08:34:52 -07:00
Si Beaumont 46072478ca build: Revert to using @_implementationOnly for all BoringSSL modules (#246) 2024-07-03 14:37:23 +01:00
Si BeaumontandCory Benfield c370ad81a1 symbols (#245)
Co-authored-by: Cory Benfield <lukasa@apple.com>
2024-07-02 11:10:27 -07:00
618 changed files with 68825 additions and 40251 deletions

No files matched your search

-15
View File
@@ -1,15 +0,0 @@
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
<key>NSPrivacyTracking</key>
<false/>
<key>NSPrivacyAccessedAPITypes</key>
<array/>
<key>NSPrivacyCollectedDataTypes</key>
<array/>
<key>NSPrivacyTrackingDomains</key>
<array/>
</dict>
</plist>
+8
View File
@@ -0,0 +1,8 @@
root = true
[*]
indent_style = space
indent_size = 4
end_of_line = lf
insert_final_newline = true
trim_trailing_whitespace = true
+1 -1
View File
@@ -9,7 +9,7 @@ about: If something isn't working as expected
### New Issue Checklist
- [ ] I'm using the latest version of Swift Crypto (`main` branch)
- [ ] I read the [Contribution Guidelines](https://github.com/apple/swift-crypto/blob/master/CONTRIBUTING.md)
- [ ] I read the [Contribution Guidelines](https://github.com/apple/swift-crypto/blob/main/CONTRIBUTING.md)
- [ ] I searched for [existing GitHub issues](https://github.com/apple/swift-crypto/issues)
### Expected behavior
+2 -2
View File
@@ -6,8 +6,8 @@ about: Not sure how something works or how to implement some functionality? Ask
### Question Checklist
- [ ] I'm using the latest version of Swift Crypto (master branch)
- [ ] I read the [Contribution Guidelines](https://github.com/apple/swift-crypto/blob/master/CONTRIBUTING.md)
- [ ] I'm using the latest version of Swift Crypto (main branch)
- [ ] I read the [Contribution Guidelines](https://github.com/apple/swift-crypto/blob/main/CONTRIBUTING.md)
- [ ] I searched for [existing GitHub issues](https://github.com/apple/swift-crypto/issues)
### Question Subject
+2 -2
View File
@@ -8,8 +8,8 @@ about: If a recent release broke a feature (Please make sure you know the last k
### New Regression Checklist
- [ ] I'm using the latest version of Swift Crypto (master branch)
- [ ] I read the [Contribution Guidelines](https://github.com/apple/swift-crypto/blob/master/CONTRIBUTING.md)
- [ ] I'm using the latest version of Swift Crypto (main branch)
- [ ] I read the [Contribution Guidelines](https://github.com/apple/swift-crypto/blob/main/CONTRIBUTING.md)
- [ ] I searched for [existing GitHub issues](https://github.com/apple/swift-crypto/issues)
### Regression Information
+14
View File
@@ -0,0 +1,14 @@
changelog:
categories:
- title: SemVer Major
labels:
- ⚠️ semver/major
- title: SemVer Minor
labels:
- semver/minor
- title: SemVer Patch
labels:
- semver/patch
- title: Other Changes
labels:
- semver/none
+22
View File
@@ -0,0 +1,22 @@
name: Main
on:
push:
branches: [main]
schedule:
- cron: "0 8,20 * * *"
jobs:
unit-tests:
name: Unit tests
uses: apple/swift-nio/.github/workflows/unit_tests.yml@main
with:
linux_5_9_arguments_override: "--explicit-target-dependency-import-check error"
linux_5_10_arguments_override: "--explicit-target-dependency-import-check error"
linux_6_0_arguments_override: "--explicit-target-dependency-import-check error"
linux_nightly_6_0_arguments_override: "--explicit-target-dependency-import-check error"
linux_nightly_main_arguments_override: "--explicit-target-dependency-import-check error"
cxx-interop:
name: Cxx interop
uses: apple/swift-nio/.github/workflows/cxx_interop.yml@main
+41
View File
@@ -0,0 +1,41 @@
name: PR
on:
pull_request:
types: [opened, reopened, synchronize]
jobs:
soundness:
name: Soundness
uses: swiftlang/github-workflows/.github/workflows/soundness.yml@main
with:
license_header_check_project_name: "SwiftCrypto"
format_check_enabled: false
docs_check_enabled: false
cmake-lists:
name: Check cmake lists
uses: apple/swift-nio/.github/workflows/swift_matrix.yml@main
with:
name: "Check cmake lists"
matrix_linux_command: ./scripts/check-cmake-lists.sh
unit-tests:
name: Unit tests
uses: apple/swift-nio/.github/workflows/unit_tests.yml@main
with:
linux_5_9_arguments_override: "--explicit-target-dependency-import-check error"
linux_5_10_arguments_override: "--explicit-target-dependency-import-check error"
linux_6_0_arguments_override: "--explicit-target-dependency-import-check error"
linux_nightly_6_0_arguments_override: "--explicit-target-dependency-import-check error"
linux_nightly_main_arguments_override: "--explicit-target-dependency-import-check error"
windows_6_0_enabled: true
windows_nightly_6_0_enabled: true
windows_nightly_main_enabled: true
windows_6_0_arguments_override: "--explicit-target-dependency-import-check error"
windows_nightly_6_0_arguments_override: "--explicit-target-dependency-import-check error"
windows_nightly_main_arguments_override: "--explicit-target-dependency-import-check error"
cxx-interop:
name: Cxx interop
uses: apple/swift-nio/.github/workflows/cxx_interop.yml@main
+18
View File
@@ -0,0 +1,18 @@
name: PR label
on:
pull_request:
types: [labeled, unlabeled, opened, reopened, synchronize]
jobs:
semver-label-check:
name: Semantic Version label check
runs-on: ubuntu-latest
timeout-minutes: 1
steps:
- name: Checkout repository
uses: actions/checkout@v4
with:
persist-credentials: false
- name: Check for Semantic Version label
uses: apple/swift-nio/.github/actions/pull_request_semver_label_checker@main
+48
View File
@@ -0,0 +1,48 @@
.gitignore
**/.gitignore
.licenseignore
.unacceptablelanguageignore
.gitattributes
.git-blame-ignore-revs
.mailfilter
.mailmap
.spi.yml
.swift-format
.editorconfig
.github/*
*.md
*.txt
*.yml
*.yaml
*.json
Package.swift
**/Package.swift
Package@-*.swift
**/Package@-*.swift
Package.resolved
**/Package.resolved
Makefile
*.modulemap
**/*.modulemap
**/*.docc/*
*.xcprivacy
**/*.xcprivacy
*.symlink
**/*.symlink
Dockerfile
**/Dockerfile
Snippets/*
dev/git.commit.template
dev/update-benchmark-thresholds
*.crt
**/*.crt
*.pem
**/*.pem
*.der
**/*.der
.swiftformat
Sources/CCryptoBoringSSL/*
**/*.swift.gyb
scripts/*.patch
scripts/gyb
scripts/gyb.py
+4
View File
@@ -0,0 +1,4 @@
version: 1
builder:
configs:
- documentation_targets: [Crypto]
+1
View File
@@ -0,0 +1 @@
Sources/CCryptoBoringSSL/*
-5
View File
@@ -1,5 +0,0 @@
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<array/>
</plist>
+15 -1
View File
@@ -6,7 +6,7 @@
## Licensed under Apache License v2.0
##
## See LICENSE.txt for license information
## See CONTRIBUTORS.md for the list of SwiftCrypto project authors
## See CONTRIBUTORS.txt for the list of SwiftCrypto project authors
##
## SPDX-License-Identifier: Apache-2.0
##
@@ -31,6 +31,10 @@ endif()
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
set(CMAKE_Swift_MODULE_DIRECTORY ${CMAKE_BINARY_DIR}/swift)
set(SWIFT_CRYPTO_COMPILE_OPTIONS
-package-name com.apple.swift-crypto
)
option(BUILD_SHARED_LIBS "Build shared libraries by default" YES)
if(BUILD_SHARED_LIBS)
@@ -48,5 +52,15 @@ if(NOT CMAKE_SYSTEM_NAME STREQUAL Darwin)
find_package(Foundation CONFIG)
endif()
include(FetchContent)
find_package(SwiftASN1 CONFIG)
if(NOT SwiftASN1_FOUND)
message("-- Vending swift-asn1")
FetchContent_Declare(ASN1
GIT_REPOSITORY https://github.com/apple/swift-asn1
GIT_TAG 1.1.0)
FetchContent_MakeAvailable(ASN1)
endif()
add_subdirectory(Sources)
add_subdirectory(cmake/modules)
File renamed without changes.
+1 -1
View File
@@ -28,7 +28,7 @@ components that this product depends on.
This product contains test vectors from Google's wycheproof project.
* LICENSE (Apache License 2.0):
* https://github.com/google/wycheproof/blob/master/LICENSE
* https://github.com/C2SP/wycheproof/blob/31387e2cd596587c859c611027b6a44d2e2b65ff/LICENSE
* HOMEPAGE:
* https://github.com/google/wycheproof
+42 -26
View File
@@ -1,4 +1,4 @@
// swift-tools-version:5.8
// swift-tools-version:5.9
//===----------------------------------------------------------------------===//
//
// This source file is part of the SwiftCrypto open source project
@@ -7,7 +7,7 @@
// Licensed under Apache License v2.0
//
// See LICENSE.txt for license information
// See CONTRIBUTORS.md for the list of SwiftCrypto project authors
// See CONTRIBUTORS.txt for the list of SwiftCrypto project authors
//
// SPDX-License-Identifier: Apache-2.0
//
@@ -20,7 +20,7 @@
// Sources/CCryptoBoringSSL directory. The source repository is at
// https://boringssl.googlesource.com/boringssl.
//
// BoringSSL Commit: dbad745811195c00b729efd0ee0a09b7d9fce1d2
// BoringSSL Commit: 2587c4974dbe9872451151c8e975f58567a1ce0d
import PackageDescription
@@ -57,6 +57,18 @@ if development {
]
}
// This doesn't work when cross-compiling: the privacy manifest will be included in the Bundle and
// Foundation will be linked. This is, however, strictly better than unconditionally adding the
// resource.
#if canImport(Darwin)
let privacyManifestExclude: [String] = []
let privacyManifestResource: [PackageDescription.Resource] = [.copy("PrivacyInfo.xcprivacy")]
#else
// Exclude on other platforms to avoid build warnings.
let privacyManifestExclude: [String] = ["PrivacyInfo.xcprivacy"]
let privacyManifestResource: [PackageDescription.Resource] = []
#endif
let package = Package(
name: "swift-crypto",
platforms: [
@@ -72,11 +84,13 @@ let package = Package(
.library(name: "CCryptoBoringSSL", type: .static, targets: ["CCryptoBoringSSL"]),
MANGLE_END */
],
dependencies: [],
dependencies: [
.package(url: "https://github.com/apple/swift-asn1.git", from: "1.2.0")
],
targets: [
.target(
name: "CCryptoBoringSSL",
exclude: [
exclude: privacyManifestExclude + [
"hash.txt",
"include/boringssl_prefix_symbols_nasm.inc",
"CMakeLists.txt",
@@ -88,9 +102,7 @@ let package = Package(
"crypto/bio/socket_helper.c",
"crypto/bio/socket.c"
],
resources: [
.copy("PrivacyInfo.xcprivacy"),
],
resources: privacyManifestResource,
cSettings: [
// These defines come from BoringSSL's build system
.define("_HAS_EXCEPTIONS", to: "0", .when(platforms: [Platform.windows])),
@@ -107,26 +119,22 @@ let package = Package(
.target(
name: "CCryptoBoringSSLShims",
dependencies: ["CCryptoBoringSSL"],
exclude: [
exclude: privacyManifestExclude + [
"CMakeLists.txt"
],
resources: [
.copy("PrivacyInfo.xcprivacy"),
]
resources: privacyManifestResource
),
.target(
name: "Crypto",
dependencies: dependencies,
exclude: [
exclude: privacyManifestExclude + [
"CMakeLists.txt",
"AEADs/Nonces.swift.gyb",
"Digests/Digests.swift.gyb",
"Key Agreement/ECDH.swift.gyb",
"Signatures/ECDSA.swift.gyb",
],
resources: [
.copy("PrivacyInfo.xcprivacy"),
],
resources: privacyManifestResource,
swiftSettings: swiftSettings
),
.target(
@@ -135,14 +143,13 @@ let package = Package(
"CCryptoBoringSSL",
"CCryptoBoringSSLShims",
"CryptoBoringWrapper",
"Crypto"
"Crypto",
.product(name: "SwiftASN1", package: "swift-asn1")
],
exclude: [
exclude: privacyManifestExclude + [
"CMakeLists.txt",
],
resources: [
.copy("PrivacyInfo.xcprivacy"),
],
resources: privacyManifestResource,
swiftSettings: swiftSettings
),
.target(
@@ -151,12 +158,10 @@ let package = Package(
"CCryptoBoringSSL",
"CCryptoBoringSSLShims"
],
exclude: [
exclude: privacyManifestExclude + [
"CMakeLists.txt",
],
resources: [
.copy("PrivacyInfo.xcprivacy"),
]
resources: privacyManifestResource
),
.executableTarget(name: "crypto-shasum", dependencies: ["Crypto"]),
.testTarget(
@@ -167,7 +172,18 @@ let package = Package(
],
swiftSettings: swiftSettings
),
.testTarget(name: "_CryptoExtrasTests", dependencies: ["_CryptoExtras"]),
.testTarget(
name: "_CryptoExtrasTests",
dependencies: ["_CryptoExtras"],
resources: [
.copy("ECToolbox/H2CVectors/P256_XMD-SHA-256_SSWU_RO_.json"),
.copy("ECToolbox/H2CVectors/P384_XMD-SHA-384_SSWU_RO_.json"),
.copy("OPRFs/OPRFVectors/OPRFVectors-VOPRFDraft8.json"),
.copy("OPRFs/OPRFVectors/OPRFVectors-VOPRFDraft19.json"),
.copy("OPRFs/OPRFVectors/OPRFVectors-edgecases.json"),
],
swiftSettings: swiftSettings
),
.testTarget(name: "CryptoBoringWrapperTests", dependencies: ["CryptoBoringWrapper"]),
],
cxxLanguageStandard: .cxx11
+2 -1
View File
@@ -117,7 +117,8 @@ Swift Crypto | Minimum Swift Version
`2.2.0 ..< 2.4.2` | 5.5
`2.4.2 ..< 3.1.0` | 5.6
`3.1.0 ..< 3.3.0` | 5.7
`3.3.0 ...` | 5.8
`3.3.0 ..< 3.8.0` | 5.8
`3.8.0 ...` | 5.9
### Compatibility
+90 -148
View File
@@ -6,7 +6,7 @@
## Licensed under Apache License v2.0
##
## See LICENSE.txt for license information
## See CONTRIBUTORS.md for the list of SwiftCrypto project authors
## See CONTRIBUTORS.txt for the list of SwiftCrypto project authors
##
## SPDX-License-Identifier: Apache-2.0
##
@@ -91,6 +91,7 @@ add_library(CCryptoBoringSSL STATIC
"crypto/dh_extra/dh_asn1.c"
"crypto/dh_extra/params.c"
"crypto/digest_extra/digest_extra.c"
"crypto/dilithium/dilithium.c"
"crypto/dsa/dsa.c"
"crypto/dsa/dsa_asn1.c"
"crypto/ec_extra/ec_asn1.c"
@@ -100,10 +101,11 @@ add_library(CCryptoBoringSSL STATIC
"crypto/ecdsa_extra/ecdsa_asn1.c"
"crypto/engine/engine.c"
"crypto/err/err.c"
"crypto/err/err_data.c"
"crypto/evp/evp.c"
"crypto/evp/evp_asn1.c"
"crypto/evp/evp_ctx.c"
"crypto/evp/p_dh.c"
"crypto/evp/p_dh_asn1.c"
"crypto/evp/p_dsa_asn1.c"
"crypto/evp/p_ec.c"
"crypto/evp/p_ec_asn1.c"
@@ -119,89 +121,17 @@ add_library(CCryptoBoringSSL STATIC
"crypto/evp/scrypt.c"
"crypto/evp/sign.c"
"crypto/ex_data.c"
"crypto/fipsmodule/aes/aes.c"
"crypto/fipsmodule/aes/aes_nohw.c"
"crypto/fipsmodule/aes/key_wrap.c"
"crypto/fipsmodule/aes/mode_wrappers.c"
"crypto/fipsmodule/bn/add.c"
"crypto/fipsmodule/bn/asm/x86_64-gcc.c"
"crypto/fipsmodule/bn/bn.c"
"crypto/fipsmodule/bn/bytes.c"
"crypto/fipsmodule/bn/cmp.c"
"crypto/fipsmodule/bn/ctx.c"
"crypto/fipsmodule/bn/div.c"
"crypto/fipsmodule/bn/div_extra.c"
"crypto/fipsmodule/bn/exponentiation.c"
"crypto/fipsmodule/bn/gcd.c"
"crypto/fipsmodule/bn/gcd_extra.c"
"crypto/fipsmodule/bn/generic.c"
"crypto/fipsmodule/bn/jacobi.c"
"crypto/fipsmodule/bn/montgomery.c"
"crypto/fipsmodule/bn/montgomery_inv.c"
"crypto/fipsmodule/bn/mul.c"
"crypto/fipsmodule/bn/prime.c"
"crypto/fipsmodule/bn/random.c"
"crypto/fipsmodule/bn/rsaz_exp.c"
"crypto/fipsmodule/bn/shift.c"
"crypto/fipsmodule/bn/sqrt.c"
"crypto/fipsmodule/cipher/aead.c"
"crypto/fipsmodule/cipher/cipher.c"
"crypto/fipsmodule/cipher/e_aes.c"
"crypto/fipsmodule/cipher/e_aesccm.c"
"crypto/fipsmodule/cmac/cmac.c"
"crypto/fipsmodule/dh/check.c"
"crypto/fipsmodule/dh/dh.c"
"crypto/fipsmodule/digest/digest.c"
"crypto/fipsmodule/digest/digests.c"
"crypto/fipsmodule/digestsign/digestsign.c"
"crypto/fipsmodule/ec/ec.c"
"crypto/fipsmodule/ec/ec_key.c"
"crypto/fipsmodule/ec/ec_montgomery.c"
"crypto/fipsmodule/ec/felem.c"
"crypto/fipsmodule/ec/oct.c"
"crypto/fipsmodule/ec/p224-64.c"
"crypto/fipsmodule/ec/p256-nistz.c"
"crypto/fipsmodule/ec/p256.c"
"crypto/fipsmodule/ec/scalar.c"
"crypto/fipsmodule/ec/simple.c"
"crypto/fipsmodule/ec/simple_mul.c"
"crypto/fipsmodule/ec/util.c"
"crypto/fipsmodule/ec/wnaf.c"
"crypto/fipsmodule/ecdh/ecdh.c"
"crypto/fipsmodule/ecdsa/ecdsa.c"
"crypto/fipsmodule/bcm.c"
"crypto/fipsmodule/fips_shared_support.c"
"crypto/fipsmodule/hkdf/hkdf.c"
"crypto/fipsmodule/hmac/hmac.c"
"crypto/fipsmodule/md4/md4.c"
"crypto/fipsmodule/md5/md5.c"
"crypto/fipsmodule/modes/cbc.c"
"crypto/fipsmodule/modes/cfb.c"
"crypto/fipsmodule/modes/ctr.c"
"crypto/fipsmodule/modes/gcm.c"
"crypto/fipsmodule/modes/gcm_nohw.c"
"crypto/fipsmodule/modes/ofb.c"
"crypto/fipsmodule/modes/polyval.c"
"crypto/fipsmodule/rand/ctrdrbg.c"
"crypto/fipsmodule/rand/fork_detect.c"
"crypto/fipsmodule/rand/rand.c"
"crypto/fipsmodule/rand/urandom.c"
"crypto/fipsmodule/rsa/blinding.c"
"crypto/fipsmodule/rsa/padding.c"
"crypto/fipsmodule/rsa/rsa.c"
"crypto/fipsmodule/rsa/rsa_impl.c"
"crypto/fipsmodule/self_check/fips.c"
"crypto/fipsmodule/self_check/self_check.c"
"crypto/fipsmodule/service_indicator/service_indicator.c"
"crypto/fipsmodule/sha/sha1.c"
"crypto/fipsmodule/sha/sha256.c"
"crypto/fipsmodule/sha/sha512.c"
"crypto/fipsmodule/tls/kdf.c"
"crypto/hpke/hpke.c"
"crypto/hrss/hrss.c"
"crypto/keccak/keccak.c"
"crypto/kyber/kyber.c"
"crypto/lhash/lhash.c"
"crypto/md4/md4.c"
"crypto/md5/md5.c"
"crypto/mem.c"
"crypto/mldsa/mldsa.c"
"crypto/obj/obj.c"
"crypto/obj/obj_xref.c"
"crypto/pem/pem_all.c"
@@ -222,26 +152,37 @@ add_library(CCryptoBoringSSL STATIC
"crypto/poly1305/poly1305_vec.c"
"crypto/pool/pool.c"
"crypto/rand_extra/deterministic.c"
"crypto/rand_extra/fork_detect.c"
"crypto/rand_extra/forkunsafe.c"
"crypto/rand_extra/getentropy.c"
"crypto/rand_extra/ios.c"
"crypto/rand_extra/passive.c"
"crypto/rand_extra/rand_extra.c"
"crypto/rand_extra/trusty.c"
"crypto/rand_extra/urandom.c"
"crypto/rand_extra/windows.c"
"crypto/rc4/rc4.c"
"crypto/refcount.c"
"crypto/rsa_extra/rsa_asn1.c"
"crypto/rsa_extra/rsa_crypt.c"
"crypto/rsa_extra/rsa_extra.c"
"crypto/rsa_extra/rsa_print.c"
"crypto/sha/sha1.c"
"crypto/sha/sha256.c"
"crypto/sha/sha512.c"
"crypto/siphash/siphash.c"
"crypto/spx/address.c"
"crypto/spx/fors.c"
"crypto/spx/merkle.c"
"crypto/slhdsa/fors.c"
"crypto/slhdsa/merkle.c"
"crypto/slhdsa/slhdsa.c"
"crypto/slhdsa/thash.c"
"crypto/slhdsa/wots.c"
"crypto/spx/spx.c"
"crypto/spx/spx_address.c"
"crypto/spx/spx_fors.c"
"crypto/spx/spx_merkle.c"
"crypto/spx/spx_thash.c"
"crypto/spx/spx_util.c"
"crypto/spx/thash.c"
"crypto/spx/wots.c"
"crypto/spx/spx_wots.c"
"crypto/stack/stack.c"
"crypto/thread.c"
"crypto/thread_none.c"
@@ -319,80 +260,81 @@ add_library(CCryptoBoringSSL STATIC
"crypto/x509/x_spki.c"
"crypto/x509/x_val.c"
"crypto/x509/x_x509.c"
"crypto/x509/x_x509a.c")
"crypto/x509/x_x509a.c"
"gen/crypto/err_data.c")
if(CMAKE_SYSTEM_NAME STREQUAL Darwin AND CMAKE_SYSTEM_PROCESSOR MATCHES "amd64|x86_64")
target_sources(CCryptoBoringSSL PRIVATE
crypto/chacha/chacha-x86_64-mac.mac.x86_64.S
crypto/cipher_extra/aes128gcmsiv-x86_64-mac.mac.x86_64.S
crypto/cipher_extra/chacha20_poly1305_x86_64-mac.mac.x86_64.S
crypto/fipsmodule/aesni-gcm-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/aesni-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/ghash-ssse3-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/ghash-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/md5-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/p256-x86_64-asm-mac.mac.x86_64.S
crypto/fipsmodule/p256_beeu-x86_64-asm-mac.mac.x86_64.S
crypto/fipsmodule/rdrand-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/rsaz-avx2-mac.mac.x86_64.S
crypto/fipsmodule/sha1-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/sha256-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/sha512-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/vpaes-x86_64-mac.mac.x86_64.S
crypto/fipsmodule/x86_64-mont-mac.mac.x86_64.S
crypto/fipsmodule/x86_64-mont5-mac.mac.x86_64.S)
gen/bcm/aesni-gcm-x86_64-apple.S
gen/bcm/aesni-x86_64-apple.S
gen/bcm/ghash-ssse3-x86_64-apple.S
gen/bcm/ghash-x86_64-apple.S
gen/bcm/p256-x86_64-asm-apple.S
gen/bcm/p256_beeu-x86_64-asm-apple.S
gen/bcm/rdrand-x86_64-apple.S
gen/bcm/rsaz-avx2-apple.S
gen/bcm/sha1-x86_64-apple.S
gen/bcm/sha256-x86_64-apple.S
gen/bcm/sha512-x86_64-apple.S
gen/bcm/vpaes-x86_64-apple.S
gen/bcm/x86_64-mont-apple.S
gen/bcm/x86_64-mont5-apple.S
gen/crypto/aes128gcmsiv-x86_64-apple.S
gen/crypto/chacha-x86_64-apple.S
gen/crypto/chacha20_poly1305_x86_64-apple.S
gen/crypto/md5-x86_64-apple.S)
elseif(CMAKE_SYSTEM_NAME MATCHES "Linux|Android" AND CMAKE_SYSTEM_PROCESSOR MATCHES "amd64|x86_64")
target_sources(CCryptoBoringSSL PRIVATE
crypto/chacha/chacha-x86_64-linux.linux.x86_64.S
crypto/cipher_extra/aes128gcmsiv-x86_64-linux.linux.x86_64.S
crypto/cipher_extra/chacha20_poly1305_x86_64-linux.linux.x86_64.S
crypto/fipsmodule/aesni-gcm-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/aesni-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/ghash-ssse3-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/ghash-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/md5-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/p256-x86_64-asm-linux.linux.x86_64.S
crypto/fipsmodule/p256_beeu-x86_64-asm-linux.linux.x86_64.S
crypto/fipsmodule/rdrand-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/rsaz-avx2-linux.linux.x86_64.S
crypto/fipsmodule/sha1-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/sha256-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/sha512-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/vpaes-x86_64-linux.linux.x86_64.S
crypto/fipsmodule/x86_64-mont-linux.linux.x86_64.S
crypto/fipsmodule/x86_64-mont5-linux.linux.x86_64.S)
gen/bcm/aesni-gcm-x86_64-linux.S
gen/bcm/aesni-x86_64-linux.S
gen/bcm/ghash-ssse3-x86_64-linux.S
gen/bcm/ghash-x86_64-linux.S
gen/bcm/p256-x86_64-asm-linux.S
gen/bcm/p256_beeu-x86_64-asm-linux.S
gen/bcm/rdrand-x86_64-linux.S
gen/bcm/rsaz-avx2-linux.S
gen/bcm/sha1-x86_64-linux.S
gen/bcm/sha256-x86_64-linux.S
gen/bcm/sha512-x86_64-linux.S
gen/bcm/vpaes-x86_64-linux.S
gen/bcm/x86_64-mont-linux.S
gen/bcm/x86_64-mont5-linux.S
gen/crypto/aes128gcmsiv-x86_64-linux.S
gen/crypto/chacha-x86_64-linux.S
gen/crypto/chacha20_poly1305_x86_64-linux.S
gen/crypto/md5-x86_64-linux.S)
elseif(CMAKE_SYSTEM_NAME STREQUAL Darwin AND CMAKE_SYSTEM_PROCESSOR MATCHES "arm64|aarch64")
target_sources(CCryptoBoringSSL PRIVATE
crypto/chacha/chacha-armv8-ios.ios.aarch64.S
crypto/cipher_extra/chacha20_poly1305_armv8-ios.ios.aarch64.S
crypto/fipsmodule/aesv8-armv8-ios.ios.aarch64.S
crypto/fipsmodule/aesv8-gcm-armv8-ios.ios.aarch64.S
crypto/fipsmodule/armv8-mont-ios.ios.aarch64.S
crypto/fipsmodule/bn-armv8-ios.ios.aarch64.S
crypto/fipsmodule/ghash-neon-armv8-ios.ios.aarch64.S
crypto/fipsmodule/ghashv8-armv8-ios.ios.aarch64.S
crypto/fipsmodule/p256-armv8-asm-ios.ios.aarch64.S
crypto/fipsmodule/p256_beeu-armv8-asm-ios.ios.aarch64.S
crypto/fipsmodule/sha1-armv8-ios.ios.aarch64.S
crypto/fipsmodule/sha256-armv8-ios.ios.aarch64.S
crypto/fipsmodule/sha512-armv8-ios.ios.aarch64.S
crypto/fipsmodule/vpaes-armv8-ios.ios.aarch64.S)
gen/bcm/aesv8-armv8-apple.S
gen/bcm/aesv8-gcm-armv8-apple.S
gen/bcm/armv8-mont-apple.S
gen/bcm/bn-armv8-apple.S
gen/bcm/ghash-neon-armv8-apple.S
gen/bcm/ghashv8-armv8-apple.S
gen/bcm/p256-armv8-asm-apple.S
gen/bcm/p256_beeu-armv8-asm-apple.S
gen/bcm/sha1-armv8-apple.S
gen/bcm/sha256-armv8-apple.S
gen/bcm/sha512-armv8-apple.S
gen/bcm/vpaes-armv8-apple.S
gen/crypto/chacha-armv8-apple.S
gen/crypto/chacha20_poly1305_armv8-apple.S)
elseif(CMAKE_SYSTEM_NAME MATCHES "Linux|Android" AND CMAKE_SYSTEM_PROCESSOR MATCHES "arm64|aarch64")
target_sources(CCryptoBoringSSL PRIVATE
crypto/chacha/chacha-armv8-linux.linux.aarch64.S
crypto/cipher_extra/chacha20_poly1305_armv8-linux.linux.aarch64.S
crypto/fipsmodule/aesv8-armv8-linux.linux.aarch64.S
crypto/fipsmodule/aesv8-gcm-armv8-linux.linux.aarch64.S
crypto/fipsmodule/armv8-mont-linux.linux.aarch64.S
crypto/fipsmodule/bn-armv8-linux.linux.aarch64.S
crypto/fipsmodule/ghash-neon-armv8-linux.linux.aarch64.S
crypto/fipsmodule/ghashv8-armv8-linux.linux.aarch64.S
crypto/fipsmodule/p256-armv8-asm-linux.linux.aarch64.S
crypto/fipsmodule/p256_beeu-armv8-asm-linux.linux.aarch64.S
crypto/fipsmodule/sha1-armv8-linux.linux.aarch64.S
crypto/fipsmodule/sha256-armv8-linux.linux.aarch64.S
crypto/fipsmodule/sha512-armv8-linux.linux.aarch64.S
crypto/fipsmodule/vpaes-armv8-linux.linux.aarch64.S)
gen/bcm/aesv8-armv8-linux.S
gen/bcm/aesv8-gcm-armv8-linux.S
gen/bcm/armv8-mont-linux.S
gen/bcm/bn-armv8-linux.S
gen/bcm/ghash-neon-armv8-linux.S
gen/bcm/ghashv8-armv8-linux.S
gen/bcm/p256-armv8-asm-linux.S
gen/bcm/p256_beeu-armv8-asm-linux.S
gen/bcm/sha1-armv8-linux.S
gen/bcm/sha256-armv8-linux.S
gen/bcm/sha512-armv8-linux.S
gen/bcm/vpaes-armv8-linux.S
gen/crypto/chacha-armv8-linux.S
gen/crypto/chacha20_poly1305_armv8-linux.S)
endif()
target_include_directories(CCryptoBoringSSL PUBLIC
-1
View File
@@ -1 +0,0 @@
../../.PrivacyInfo.xcprivacy
+15
View File
@@ -0,0 +1,15 @@
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
<key>NSPrivacyTracking</key>
<false/>
<key>NSPrivacyAccessedAPITypes</key>
<array/>
<key>NSPrivacyCollectedDataTypes</key>
<array/>
<key>NSPrivacyTrackingDomains</key>
<array/>
</dict>
</plist>
@@ -307,6 +307,10 @@ static int base64_decode_quad(uint8_t *out, size_t *out_num_bytes,
(in[2] == '=') << 1 |
(in[3] == '=');
// In presence of padding, the lowest bits of v are unused. Canonical encoding
// (RFC 4648, section 3.5) requires that these bits all be set to zero. Common
// PEM parsers accept noncanonical base64, adding to the malleability of the
// format. This decoder follows OpenSSL's and Go's PEM parsers and accepts it.
switch (padding_pattern) {
case 0:
// The common case of no padding.
@@ -1,4 +1,4 @@
/* Copyright (c) 2020, Google Inc.
/* Copyright (c) 2024, Google Inc.
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
@@ -12,11 +12,19 @@
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
#ifndef OPENSSL_HEADER_CRYPTO_FORK_DETECT_H
#define OPENSSL_HEADER_CRYPTO_FORK_DETECT_H
#ifndef OPENSSL_HEADER_CRYPTO_BCM_SUPPORT_H
#define OPENSSL_HEADER_CRYPTO_BCM_SUPPORT_H
#include <CCryptoBoringSSL_base.h>
#include <stdio.h>
// Provided by libcrypto, called from BCM
#if defined(__cplusplus)
extern "C" {
#endif
#if defined(OPENSSL_LINUX)
// On linux we use MADVISE instead of pthread_atfork(), due
// to concerns about clone() being used for address space
@@ -29,15 +37,54 @@
// iOS doesn't normally allow fork in apps, but it's there.
#define OPENSSL_FORK_DETECTION
#define OPENSSL_FORK_DETECTION_PTHREAD_ATFORK
#elif defined(OPENSSL_WINDOWS) || defined(OPENSSL_TRUSTY)
#elif defined(OPENSSL_WINDOWS) || defined(OPENSSL_TRUSTY) || \
defined(__ZEPHYR__) || defined(CROS_EC)
// These platforms do not fork.
#define OPENSSL_DOES_NOT_FORK
#endif
#if defined(__cplusplus)
extern "C" {
#if defined(BORINGSSL_UNSAFE_DETERMINISTIC_MODE)
#define OPENSSL_RAND_DETERMINISTIC
#elif defined(OPENSSL_TRUSTY)
#define OPENSSL_RAND_TRUSTY
#elif defined(OPENSSL_WINDOWS)
#define OPENSSL_RAND_WINDOWS
#elif defined(OPENSSL_LINUX)
#define OPENSSL_RAND_URANDOM
#elif defined(OPENSSL_APPLE) && !defined(OPENSSL_MACOS)
// Unlike macOS, iOS and similar hide away getentropy().
#define OPENSSL_RAND_IOS
#else
// By default if you are integrating BoringSSL we expect you to
// provide getentropy from the <unistd.h> header file.
#define OPENSSL_RAND_GETENTROPY
#endif
// Provided by libcrypto, called from BCM
// CRYPTO_init_sysrand initializes long-lived resources needed to draw entropy
// from the operating system, if the operating system requires initialization.
void CRYPTO_init_sysrand(void);
// CRYPTO_sysrand fills |len| bytes at |buf| with entropy from the operating
// system.
void CRYPTO_sysrand(uint8_t *buf, size_t len);
// CRYPTO_sysrand_if_available fills |len| bytes at |buf| with entropy from the
// operating system, or early /dev/urandom data, and returns 1, _if_ the entropy
// pool is initialized or if getrandom() is not available and not in FIPS mode.
// Otherwise it will not block and will instead fill |buf| with all zeros and
// return 0.
int CRYPTO_sysrand_if_available(uint8_t *buf, size_t len);
// CRYPTO_sysrand_for_seed fills |len| bytes at |buf| with entropy from the
// operating system. It may draw from the |GRND_RANDOM| pool on Android,
// depending on the vendor's configuration.
void CRYPTO_sysrand_for_seed(uint8_t *buf, size_t len);
// RAND_need_entropy is called whenever the BCM module has stopped because it
// has run out of entropy.
void RAND_need_entropy(size_t bytes_needed);
// crypto_get_fork_generation returns the fork generation number for the current
// process, or zero if not supported on the platform. The fork generation number
@@ -60,8 +107,13 @@ OPENSSL_EXPORT uint64_t CRYPTO_get_fork_generation(void);
OPENSSL_EXPORT void CRYPTO_fork_detect_force_madv_wipeonfork_for_testing(
int on);
// CRYPTO_get_stderr returns stderr. This function exists to avoid BCM needing
// a data dependency on libc.
FILE *CRYPTO_get_stderr(void);
#if defined(__cplusplus)
} // extern C
#endif
#endif // OPENSSL_HEADER_CRYPTO_FORK_DETECT_H
#endif // OPENSSL_HEADER_CRYPTO_BCM_SUPPORT_H
+12 -12
View File
@@ -658,38 +658,38 @@ void BIO_meth_free(BIO_METHOD *method) {
}
int BIO_meth_set_create(BIO_METHOD *method,
int (*create)(BIO *)) {
method->create = create;
int (*create_func)(BIO *)) {
method->create = create_func;
return 1;
}
int BIO_meth_set_destroy(BIO_METHOD *method,
int (*destroy)(BIO *)) {
method->destroy = destroy;
int (*destroy_func)(BIO *)) {
method->destroy = destroy_func;
return 1;
}
int BIO_meth_set_write(BIO_METHOD *method,
int (*write)(BIO *, const char *, int)) {
method->bwrite = write;
int (*write_func)(BIO *, const char *, int)) {
method->bwrite = write_func;
return 1;
}
int BIO_meth_set_read(BIO_METHOD *method,
int (*read)(BIO *, char *, int)) {
method->bread = read;
int (*read_func)(BIO *, char *, int)) {
method->bread = read_func;
return 1;
}
int BIO_meth_set_gets(BIO_METHOD *method,
int (*gets)(BIO *, char *, int)) {
method->bgets = gets;
int (*gets_func)(BIO *, char *, int)) {
method->bgets = gets_func;
return 1;
}
int BIO_meth_set_ctrl(BIO_METHOD *method,
long (*ctrl)(BIO *, int, long, void *)) {
method->ctrl = ctrl;
long (*ctrl_func)(BIO *, int, long, void *)) {
method->ctrl = ctrl_func;
return 1;
}
@@ -487,7 +487,11 @@ static long conn_callback_ctrl(BIO *bio, int cmd, bio_info_cb fp) {
// convention.
OPENSSL_MSVC_PRAGMA(warning(push))
OPENSSL_MSVC_PRAGMA(warning(disable : 4191))
OPENSSL_CLANG_PRAGMA("clang diagnostic push")
OPENSSL_CLANG_PRAGMA("clang diagnostic ignored \"-Wunknown-warning-option\"")
OPENSSL_CLANG_PRAGMA("clang diagnostic ignored \"-Wcast-function-type\"")
data->info_callback = (int (*)(const struct bio_st *, int, int))fp;
OPENSSL_CLANG_PRAGMA("clang diagnostic pop")
OPENSSL_MSVC_PRAGMA(warning(pop))
break;
default:
@@ -507,11 +507,9 @@ int CBS_get_asn1_int64(CBS *cbs, int64_t *out) {
return 0;
}
uint8_t sign_extend[sizeof(int64_t)];
memset(sign_extend, is_negative ? 0xff : 0, sizeof(sign_extend));
for (size_t i = 0; i < len; i++) {
sign_extend[i] = data[len - i - 1];
}
memcpy(out, sign_extend, sizeof(sign_extend));
OPENSSL_memset(sign_extend, is_negative ? 0xff : 0, sizeof(sign_extend));
OPENSSL_memcpy(sign_extend + sizeof(int64_t) - len, data, len);
*out = CRYPTO_load_u64_be(sign_extend);
return 1;
}
@@ -18,11 +18,12 @@
static int is_valid_code_point(uint32_t v) {
// References in the following are to Unicode 9.0.0.
// References in the following are to Unicode 15.0.0.
if (// The Unicode space runs from zero to 0x10ffff (3.4 D9).
v > 0x10ffff ||
// Values 0x...fffe, 0x...ffff, and 0xfdd0-0xfdef are permanently reserved
// (3.4 D14)
// as noncharacters (3.4 D14). See also 23.7. As our APIs are intended for
// "open interchange", such as ASN.1, we reject them.
(v & 0xfffe) == 0xfffe ||
(v >= 0xfdd0 && v <= 0xfdef) ||
// Surrogate code points are invalid (3.2 C1).
File diff suppressed because it is too large. Load diff
@@ -1,973 +0,0 @@
#define BORINGSSL_PREFIX CCryptoBoringSSL
#if defined(__i386__) && defined(_WIN32)
; This file is generated from a similarly-named Perl script in the BoringSSL
; source tree. Do not edit by hand.
%ifdef BORINGSSL_PREFIX
%include "boringssl_prefix_symbols_nasm.inc"
%endif
%ifidn __OUTPUT_FORMAT__, win32
%ifidn __OUTPUT_FORMAT__,obj
section code use32 class=code align=64
%elifidn __OUTPUT_FORMAT__,win32
$@feat.00 equ 1
section .text code align=64
%else
section .text code
%endif
global _ChaCha20_ctr32_nohw
align 16
_ChaCha20_ctr32_nohw:
L$_ChaCha20_ctr32_nohw_begin:
push ebp
push ebx
push esi
push edi
mov esi,DWORD [32+esp]
mov edi,DWORD [36+esp]
sub esp,132
mov eax,DWORD [esi]
mov ebx,DWORD [4+esi]
mov ecx,DWORD [8+esi]
mov edx,DWORD [12+esi]
mov DWORD [80+esp],eax
mov DWORD [84+esp],ebx
mov DWORD [88+esp],ecx
mov DWORD [92+esp],edx
mov eax,DWORD [16+esi]
mov ebx,DWORD [20+esi]
mov ecx,DWORD [24+esi]
mov edx,DWORD [28+esi]
mov DWORD [96+esp],eax
mov DWORD [100+esp],ebx
mov DWORD [104+esp],ecx
mov DWORD [108+esp],edx
mov eax,DWORD [edi]
mov ebx,DWORD [4+edi]
mov ecx,DWORD [8+edi]
mov edx,DWORD [12+edi]
sub eax,1
mov DWORD [112+esp],eax
mov DWORD [116+esp],ebx
mov DWORD [120+esp],ecx
mov DWORD [124+esp],edx
jmp NEAR L$000entry
align 16
L$001outer_loop:
mov DWORD [156+esp],ebx
mov DWORD [152+esp],eax
mov DWORD [160+esp],ecx
L$000entry:
mov eax,1634760805
mov DWORD [4+esp],857760878
mov DWORD [8+esp],2036477234
mov DWORD [12+esp],1797285236
mov ebx,DWORD [84+esp]
mov ebp,DWORD [88+esp]
mov ecx,DWORD [104+esp]
mov esi,DWORD [108+esp]
mov edx,DWORD [116+esp]
mov edi,DWORD [120+esp]
mov DWORD [20+esp],ebx
mov DWORD [24+esp],ebp
mov DWORD [40+esp],ecx
mov DWORD [44+esp],esi
mov DWORD [52+esp],edx
mov DWORD [56+esp],edi
mov ebx,DWORD [92+esp]
mov edi,DWORD [124+esp]
mov edx,DWORD [112+esp]
mov ebp,DWORD [80+esp]
mov ecx,DWORD [96+esp]
mov esi,DWORD [100+esp]
add edx,1
mov DWORD [28+esp],ebx
mov DWORD [60+esp],edi
mov DWORD [112+esp],edx
mov ebx,10
jmp NEAR L$002loop
align 16
L$002loop:
add eax,ebp
mov DWORD [128+esp],ebx
mov ebx,ebp
xor edx,eax
rol edx,16
add ecx,edx
xor ebx,ecx
mov edi,DWORD [52+esp]
rol ebx,12
mov ebp,DWORD [20+esp]
add eax,ebx
xor edx,eax
mov DWORD [esp],eax
rol edx,8
mov eax,DWORD [4+esp]
add ecx,edx
mov DWORD [48+esp],edx
xor ebx,ecx
add eax,ebp
rol ebx,7
xor edi,eax
mov DWORD [32+esp],ecx
rol edi,16
mov DWORD [16+esp],ebx
add esi,edi
mov ecx,DWORD [40+esp]
xor ebp,esi
mov edx,DWORD [56+esp]
rol ebp,12
mov ebx,DWORD [24+esp]
add eax,ebp
xor edi,eax
mov DWORD [4+esp],eax
rol edi,8
mov eax,DWORD [8+esp]
add esi,edi
mov DWORD [52+esp],edi
xor ebp,esi
add eax,ebx
rol ebp,7
xor edx,eax
mov DWORD [36+esp],esi
rol edx,16
mov DWORD [20+esp],ebp
add ecx,edx
mov esi,DWORD [44+esp]
xor ebx,ecx
mov edi,DWORD [60+esp]
rol ebx,12
mov ebp,DWORD [28+esp]
add eax,ebx
xor edx,eax
mov DWORD [8+esp],eax
rol edx,8
mov eax,DWORD [12+esp]
add ecx,edx
mov DWORD [56+esp],edx
xor ebx,ecx
add eax,ebp
rol ebx,7
xor edi,eax
rol edi,16
mov DWORD [24+esp],ebx
add esi,edi
xor ebp,esi
rol ebp,12
mov ebx,DWORD [20+esp]
add eax,ebp
xor edi,eax
mov DWORD [12+esp],eax
rol edi,8
mov eax,DWORD [esp]
add esi,edi
mov edx,edi
xor ebp,esi
add eax,ebx
rol ebp,7
xor edx,eax
rol edx,16
mov DWORD [28+esp],ebp
add ecx,edx
xor ebx,ecx
mov edi,DWORD [48+esp]
rol ebx,12
mov ebp,DWORD [24+esp]
add eax,ebx
xor edx,eax
mov DWORD [esp],eax
rol edx,8
mov eax,DWORD [4+esp]
add ecx,edx
mov DWORD [60+esp],edx
xor ebx,ecx
add eax,ebp
rol ebx,7
xor edi,eax
mov DWORD [40+esp],ecx
rol edi,16
mov DWORD [20+esp],ebx
add esi,edi
mov ecx,DWORD [32+esp]
xor ebp,esi
mov edx,DWORD [52+esp]
rol ebp,12
mov ebx,DWORD [28+esp]
add eax,ebp
xor edi,eax
mov DWORD [4+esp],eax
rol edi,8
mov eax,DWORD [8+esp]
add esi,edi
mov DWORD [48+esp],edi
xor ebp,esi
add eax,ebx
rol ebp,7
xor edx,eax
mov DWORD [44+esp],esi
rol edx,16
mov DWORD [24+esp],ebp
add ecx,edx
mov esi,DWORD [36+esp]
xor ebx,ecx
mov edi,DWORD [56+esp]
rol ebx,12
mov ebp,DWORD [16+esp]
add eax,ebx
xor edx,eax
mov DWORD [8+esp],eax
rol edx,8
mov eax,DWORD [12+esp]
add ecx,edx
mov DWORD [52+esp],edx
xor ebx,ecx
add eax,ebp
rol ebx,7
xor edi,eax
rol edi,16
mov DWORD [28+esp],ebx
add esi,edi
xor ebp,esi
mov edx,DWORD [48+esp]
rol ebp,12
mov ebx,DWORD [128+esp]
add eax,ebp
xor edi,eax
mov DWORD [12+esp],eax
rol edi,8
mov eax,DWORD [esp]
add esi,edi
mov DWORD [56+esp],edi
xor ebp,esi
rol ebp,7
dec ebx
jnz NEAR L$002loop
mov ebx,DWORD [160+esp]
add eax,1634760805
add ebp,DWORD [80+esp]
add ecx,DWORD [96+esp]
add esi,DWORD [100+esp]
cmp ebx,64
jb NEAR L$003tail
mov ebx,DWORD [156+esp]
add edx,DWORD [112+esp]
add edi,DWORD [120+esp]
xor eax,DWORD [ebx]
xor ebp,DWORD [16+ebx]
mov DWORD [esp],eax
mov eax,DWORD [152+esp]
xor ecx,DWORD [32+ebx]
xor esi,DWORD [36+ebx]
xor edx,DWORD [48+ebx]
xor edi,DWORD [56+ebx]
mov DWORD [16+eax],ebp
mov DWORD [32+eax],ecx
mov DWORD [36+eax],esi
mov DWORD [48+eax],edx
mov DWORD [56+eax],edi
mov ebp,DWORD [4+esp]
mov ecx,DWORD [8+esp]
mov esi,DWORD [12+esp]
mov edx,DWORD [20+esp]
mov edi,DWORD [24+esp]
add ebp,857760878
add ecx,2036477234
add esi,1797285236
add edx,DWORD [84+esp]
add edi,DWORD [88+esp]
xor ebp,DWORD [4+ebx]
xor ecx,DWORD [8+ebx]
xor esi,DWORD [12+ebx]
xor edx,DWORD [20+ebx]
xor edi,DWORD [24+ebx]
mov DWORD [4+eax],ebp
mov DWORD [8+eax],ecx
mov DWORD [12+eax],esi
mov DWORD [20+eax],edx
mov DWORD [24+eax],edi
mov ebp,DWORD [28+esp]
mov ecx,DWORD [40+esp]
mov esi,DWORD [44+esp]
mov edx,DWORD [52+esp]
mov edi,DWORD [60+esp]
add ebp,DWORD [92+esp]
add ecx,DWORD [104+esp]
add esi,DWORD [108+esp]
add edx,DWORD [116+esp]
add edi,DWORD [124+esp]
xor ebp,DWORD [28+ebx]
xor ecx,DWORD [40+ebx]
xor esi,DWORD [44+ebx]
xor edx,DWORD [52+ebx]
xor edi,DWORD [60+ebx]
lea ebx,[64+ebx]
mov DWORD [28+eax],ebp
mov ebp,DWORD [esp]
mov DWORD [40+eax],ecx
mov ecx,DWORD [160+esp]
mov DWORD [44+eax],esi
mov DWORD [52+eax],edx
mov DWORD [60+eax],edi
mov DWORD [eax],ebp
lea eax,[64+eax]
sub ecx,64
jnz NEAR L$001outer_loop
jmp NEAR L$004done
L$003tail:
add edx,DWORD [112+esp]
add edi,DWORD [120+esp]
mov DWORD [esp],eax
mov DWORD [16+esp],ebp
mov DWORD [32+esp],ecx
mov DWORD [36+esp],esi
mov DWORD [48+esp],edx
mov DWORD [56+esp],edi
mov ebp,DWORD [4+esp]
mov ecx,DWORD [8+esp]
mov esi,DWORD [12+esp]
mov edx,DWORD [20+esp]
mov edi,DWORD [24+esp]
add ebp,857760878
add ecx,2036477234
add esi,1797285236
add edx,DWORD [84+esp]
add edi,DWORD [88+esp]
mov DWORD [4+esp],ebp
mov DWORD [8+esp],ecx
mov DWORD [12+esp],esi
mov DWORD [20+esp],edx
mov DWORD [24+esp],edi
mov ebp,DWORD [28+esp]
mov ecx,DWORD [40+esp]
mov esi,DWORD [44+esp]
mov edx,DWORD [52+esp]
mov edi,DWORD [60+esp]
add ebp,DWORD [92+esp]
add ecx,DWORD [104+esp]
add esi,DWORD [108+esp]
add edx,DWORD [116+esp]
add edi,DWORD [124+esp]
mov DWORD [28+esp],ebp
mov ebp,DWORD [156+esp]
mov DWORD [40+esp],ecx
mov ecx,DWORD [152+esp]
mov DWORD [44+esp],esi
xor esi,esi
mov DWORD [52+esp],edx
mov DWORD [60+esp],edi
xor eax,eax
xor edx,edx
L$005tail_loop:
mov al,BYTE [ebp*1+esi]
mov dl,BYTE [esi*1+esp]
lea esi,[1+esi]
xor al,dl
mov BYTE [esi*1+ecx-1],al
dec ebx
jnz NEAR L$005tail_loop
L$004done:
add esp,132
pop edi
pop esi
pop ebx
pop ebp
ret
global _ChaCha20_ctr32_ssse3
align 16
_ChaCha20_ctr32_ssse3:
L$_ChaCha20_ctr32_ssse3_begin:
push ebp
push ebx
push esi
push edi
call L$pic_point
L$pic_point:
pop eax
mov edi,DWORD [20+esp]
mov esi,DWORD [24+esp]
mov ecx,DWORD [28+esp]
mov edx,DWORD [32+esp]
mov ebx,DWORD [36+esp]
mov ebp,esp
sub esp,524
and esp,-64
mov DWORD [512+esp],ebp
lea eax,[(L$ssse3_data-L$pic_point)+eax]
movdqu xmm3,[ebx]
cmp ecx,256
jb NEAR L$0061x
mov DWORD [516+esp],edx
mov DWORD [520+esp],ebx
sub ecx,256
lea ebp,[384+esp]
movdqu xmm7,[edx]
pshufd xmm0,xmm3,0
pshufd xmm1,xmm3,85
pshufd xmm2,xmm3,170
pshufd xmm3,xmm3,255
paddd xmm0,[48+eax]
pshufd xmm4,xmm7,0
pshufd xmm5,xmm7,85
psubd xmm0,[64+eax]
pshufd xmm6,xmm7,170
pshufd xmm7,xmm7,255
movdqa [64+ebp],xmm0
movdqa [80+ebp],xmm1
movdqa [96+ebp],xmm2
movdqa [112+ebp],xmm3
movdqu xmm3,[16+edx]
movdqa [ebp-64],xmm4
movdqa [ebp-48],xmm5
movdqa [ebp-32],xmm6
movdqa [ebp-16],xmm7
movdqa xmm7,[32+eax]
lea ebx,[128+esp]
pshufd xmm0,xmm3,0
pshufd xmm1,xmm3,85
pshufd xmm2,xmm3,170
pshufd xmm3,xmm3,255
pshufd xmm4,xmm7,0
pshufd xmm5,xmm7,85
pshufd xmm6,xmm7,170
pshufd xmm7,xmm7,255
movdqa [ebp],xmm0
movdqa [16+ebp],xmm1
movdqa [32+ebp],xmm2
movdqa [48+ebp],xmm3
movdqa [ebp-128],xmm4
movdqa [ebp-112],xmm5
movdqa [ebp-96],xmm6
movdqa [ebp-80],xmm7
lea esi,[128+esi]
lea edi,[128+edi]
jmp NEAR L$007outer_loop
align 16
L$007outer_loop:
movdqa xmm1,[ebp-112]
movdqa xmm2,[ebp-96]
movdqa xmm3,[ebp-80]
movdqa xmm5,[ebp-48]
movdqa xmm6,[ebp-32]
movdqa xmm7,[ebp-16]
movdqa [ebx-112],xmm1
movdqa [ebx-96],xmm2
movdqa [ebx-80],xmm3
movdqa [ebx-48],xmm5
movdqa [ebx-32],xmm6
movdqa [ebx-16],xmm7
movdqa xmm2,[32+ebp]
movdqa xmm3,[48+ebp]
movdqa xmm4,[64+ebp]
movdqa xmm5,[80+ebp]
movdqa xmm6,[96+ebp]
movdqa xmm7,[112+ebp]
paddd xmm4,[64+eax]
movdqa [32+ebx],xmm2
movdqa [48+ebx],xmm3
movdqa [64+ebx],xmm4
movdqa [80+ebx],xmm5
movdqa [96+ebx],xmm6
movdqa [112+ebx],xmm7
movdqa [64+ebp],xmm4
movdqa xmm0,[ebp-128]
movdqa xmm6,xmm4
movdqa xmm3,[ebp-64]
movdqa xmm4,[ebp]
movdqa xmm5,[16+ebp]
mov edx,10
nop
align 16
L$008loop:
paddd xmm0,xmm3
movdqa xmm2,xmm3
pxor xmm6,xmm0
pshufb xmm6,[eax]
paddd xmm4,xmm6
pxor xmm2,xmm4
movdqa xmm3,[ebx-48]
movdqa xmm1,xmm2
pslld xmm2,12
psrld xmm1,20
por xmm2,xmm1
movdqa xmm1,[ebx-112]
paddd xmm0,xmm2
movdqa xmm7,[80+ebx]
pxor xmm6,xmm0
movdqa [ebx-128],xmm0
pshufb xmm6,[16+eax]
paddd xmm4,xmm6
movdqa [64+ebx],xmm6
pxor xmm2,xmm4
paddd xmm1,xmm3
movdqa xmm0,xmm2
pslld xmm2,7
psrld xmm0,25
pxor xmm7,xmm1
por xmm2,xmm0
movdqa [ebx],xmm4
pshufb xmm7,[eax]
movdqa [ebx-64],xmm2
paddd xmm5,xmm7
movdqa xmm4,[32+ebx]
pxor xmm3,xmm5
movdqa xmm2,[ebx-32]
movdqa xmm0,xmm3
pslld xmm3,12
psrld xmm0,20
por xmm3,xmm0
movdqa xmm0,[ebx-96]
paddd xmm1,xmm3
movdqa xmm6,[96+ebx]
pxor xmm7,xmm1
movdqa [ebx-112],xmm1
pshufb xmm7,[16+eax]
paddd xmm5,xmm7
movdqa [80+ebx],xmm7
pxor xmm3,xmm5
paddd xmm0,xmm2
movdqa xmm1,xmm3
pslld xmm3,7
psrld xmm1,25
pxor xmm6,xmm0
por xmm3,xmm1
movdqa [16+ebx],xmm5
pshufb xmm6,[eax]
movdqa [ebx-48],xmm3
paddd xmm4,xmm6
movdqa xmm5,[48+ebx]
pxor xmm2,xmm4
movdqa xmm3,[ebx-16]
movdqa xmm1,xmm2
pslld xmm2,12
psrld xmm1,20
por xmm2,xmm1
movdqa xmm1,[ebx-80]
paddd xmm0,xmm2
movdqa xmm7,[112+ebx]
pxor xmm6,xmm0
movdqa [ebx-96],xmm0
pshufb xmm6,[16+eax]
paddd xmm4,xmm6
movdqa [96+ebx],xmm6
pxor xmm2,xmm4
paddd xmm1,xmm3
movdqa xmm0,xmm2
pslld xmm2,7
psrld xmm0,25
pxor xmm7,xmm1
por xmm2,xmm0
pshufb xmm7,[eax]
movdqa [ebx-32],xmm2
paddd xmm5,xmm7
pxor xmm3,xmm5
movdqa xmm2,[ebx-48]
movdqa xmm0,xmm3
pslld xmm3,12
psrld xmm0,20
por xmm3,xmm0
movdqa xmm0,[ebx-128]
paddd xmm1,xmm3
pxor xmm7,xmm1
movdqa [ebx-80],xmm1
pshufb xmm7,[16+eax]
paddd xmm5,xmm7
movdqa xmm6,xmm7
pxor xmm3,xmm5
paddd xmm0,xmm2
movdqa xmm1,xmm3
pslld xmm3,7
psrld xmm1,25
pxor xmm6,xmm0
por xmm3,xmm1
pshufb xmm6,[eax]
movdqa [ebx-16],xmm3
paddd xmm4,xmm6
pxor xmm2,xmm4
movdqa xmm3,[ebx-32]
movdqa xmm1,xmm2
pslld xmm2,12
psrld xmm1,20
por xmm2,xmm1
movdqa xmm1,[ebx-112]
paddd xmm0,xmm2
movdqa xmm7,[64+ebx]
pxor xmm6,xmm0
movdqa [ebx-128],xmm0
pshufb xmm6,[16+eax]
paddd xmm4,xmm6
movdqa [112+ebx],xmm6
pxor xmm2,xmm4
paddd xmm1,xmm3
movdqa xmm0,xmm2
pslld xmm2,7
psrld xmm0,25
pxor xmm7,xmm1
por xmm2,xmm0
movdqa [32+ebx],xmm4
pshufb xmm7,[eax]
movdqa [ebx-48],xmm2
paddd xmm5,xmm7
movdqa xmm4,[ebx]
pxor xmm3,xmm5
movdqa xmm2,[ebx-16]
movdqa xmm0,xmm3
pslld xmm3,12
psrld xmm0,20
por xmm3,xmm0
movdqa xmm0,[ebx-96]
paddd xmm1,xmm3
movdqa xmm6,[80+ebx]
pxor xmm7,xmm1
movdqa [ebx-112],xmm1
pshufb xmm7,[16+eax]
paddd xmm5,xmm7
movdqa [64+ebx],xmm7
pxor xmm3,xmm5
paddd xmm0,xmm2
movdqa xmm1,xmm3
pslld xmm3,7
psrld xmm1,25
pxor xmm6,xmm0
por xmm3,xmm1
movdqa [48+ebx],xmm5
pshufb xmm6,[eax]
movdqa [ebx-32],xmm3
paddd xmm4,xmm6
movdqa xmm5,[16+ebx]
pxor xmm2,xmm4
movdqa xmm3,[ebx-64]
movdqa xmm1,xmm2
pslld xmm2,12
psrld xmm1,20
por xmm2,xmm1
movdqa xmm1,[ebx-80]
paddd xmm0,xmm2
movdqa xmm7,[96+ebx]
pxor xmm6,xmm0
movdqa [ebx-96],xmm0
pshufb xmm6,[16+eax]
paddd xmm4,xmm6
movdqa [80+ebx],xmm6
pxor xmm2,xmm4
paddd xmm1,xmm3
movdqa xmm0,xmm2
pslld xmm2,7
psrld xmm0,25
pxor xmm7,xmm1
por xmm2,xmm0
pshufb xmm7,[eax]
movdqa [ebx-16],xmm2
paddd xmm5,xmm7
pxor xmm3,xmm5
movdqa xmm0,xmm3
pslld xmm3,12
psrld xmm0,20
por xmm3,xmm0
movdqa xmm0,[ebx-128]
paddd xmm1,xmm3
movdqa xmm6,[64+ebx]
pxor xmm7,xmm1
movdqa [ebx-80],xmm1
pshufb xmm7,[16+eax]
paddd xmm5,xmm7
movdqa [96+ebx],xmm7
pxor xmm3,xmm5
movdqa xmm1,xmm3
pslld xmm3,7
psrld xmm1,25
por xmm3,xmm1
dec edx
jnz NEAR L$008loop
movdqa [ebx-64],xmm3
movdqa [ebx],xmm4
movdqa [16+ebx],xmm5
movdqa [64+ebx],xmm6
movdqa [96+ebx],xmm7
movdqa xmm1,[ebx-112]
movdqa xmm2,[ebx-96]
movdqa xmm3,[ebx-80]
paddd xmm0,[ebp-128]
paddd xmm1,[ebp-112]
paddd xmm2,[ebp-96]
paddd xmm3,[ebp-80]
movdqa xmm6,xmm0
punpckldq xmm0,xmm1
movdqa xmm7,xmm2
punpckldq xmm2,xmm3
punpckhdq xmm6,xmm1
punpckhdq xmm7,xmm3
movdqa xmm1,xmm0
punpcklqdq xmm0,xmm2
movdqa xmm3,xmm6
punpcklqdq xmm6,xmm7
punpckhqdq xmm1,xmm2
punpckhqdq xmm3,xmm7
movdqu xmm4,[esi-128]
movdqu xmm5,[esi-64]
movdqu xmm2,[esi]
movdqu xmm7,[64+esi]
lea esi,[16+esi]
pxor xmm4,xmm0
movdqa xmm0,[ebx-64]
pxor xmm5,xmm1
movdqa xmm1,[ebx-48]
pxor xmm6,xmm2
movdqa xmm2,[ebx-32]
pxor xmm7,xmm3
movdqa xmm3,[ebx-16]
movdqu [edi-128],xmm4
movdqu [edi-64],xmm5
movdqu [edi],xmm6
movdqu [64+edi],xmm7
lea edi,[16+edi]
paddd xmm0,[ebp-64]
paddd xmm1,[ebp-48]
paddd xmm2,[ebp-32]
paddd xmm3,[ebp-16]
movdqa xmm6,xmm0
punpckldq xmm0,xmm1
movdqa xmm7,xmm2
punpckldq xmm2,xmm3
punpckhdq xmm6,xmm1
punpckhdq xmm7,xmm3
movdqa xmm1,xmm0
punpcklqdq xmm0,xmm2
movdqa xmm3,xmm6
punpcklqdq xmm6,xmm7
punpckhqdq xmm1,xmm2
punpckhqdq xmm3,xmm7
movdqu xmm4,[esi-128]
movdqu xmm5,[esi-64]
movdqu xmm2,[esi]
movdqu xmm7,[64+esi]
lea esi,[16+esi]
pxor xmm4,xmm0
movdqa xmm0,[ebx]
pxor xmm5,xmm1
movdqa xmm1,[16+ebx]
pxor xmm6,xmm2
movdqa xmm2,[32+ebx]
pxor xmm7,xmm3
movdqa xmm3,[48+ebx]
movdqu [edi-128],xmm4
movdqu [edi-64],xmm5
movdqu [edi],xmm6
movdqu [64+edi],xmm7
lea edi,[16+edi]
paddd xmm0,[ebp]
paddd xmm1,[16+ebp]
paddd xmm2,[32+ebp]
paddd xmm3,[48+ebp]
movdqa xmm6,xmm0
punpckldq xmm0,xmm1
movdqa xmm7,xmm2
punpckldq xmm2,xmm3
punpckhdq xmm6,xmm1
punpckhdq xmm7,xmm3
movdqa xmm1,xmm0
punpcklqdq xmm0,xmm2
movdqa xmm3,xmm6
punpcklqdq xmm6,xmm7
punpckhqdq xmm1,xmm2
punpckhqdq xmm3,xmm7
movdqu xmm4,[esi-128]
movdqu xmm5,[esi-64]
movdqu xmm2,[esi]
movdqu xmm7,[64+esi]
lea esi,[16+esi]
pxor xmm4,xmm0
movdqa xmm0,[64+ebx]
pxor xmm5,xmm1
movdqa xmm1,[80+ebx]
pxor xmm6,xmm2
movdqa xmm2,[96+ebx]
pxor xmm7,xmm3
movdqa xmm3,[112+ebx]
movdqu [edi-128],xmm4
movdqu [edi-64],xmm5
movdqu [edi],xmm6
movdqu [64+edi],xmm7
lea edi,[16+edi]
paddd xmm0,[64+ebp]
paddd xmm1,[80+ebp]
paddd xmm2,[96+ebp]
paddd xmm3,[112+ebp]
movdqa xmm6,xmm0
punpckldq xmm0,xmm1
movdqa xmm7,xmm2
punpckldq xmm2,xmm3
punpckhdq xmm6,xmm1
punpckhdq xmm7,xmm3
movdqa xmm1,xmm0
punpcklqdq xmm0,xmm2
movdqa xmm3,xmm6
punpcklqdq xmm6,xmm7
punpckhqdq xmm1,xmm2
punpckhqdq xmm3,xmm7
movdqu xmm4,[esi-128]
movdqu xmm5,[esi-64]
movdqu xmm2,[esi]
movdqu xmm7,[64+esi]
lea esi,[208+esi]
pxor xmm4,xmm0
pxor xmm5,xmm1
pxor xmm6,xmm2
pxor xmm7,xmm3
movdqu [edi-128],xmm4
movdqu [edi-64],xmm5
movdqu [edi],xmm6
movdqu [64+edi],xmm7
lea edi,[208+edi]
sub ecx,256
jnc NEAR L$007outer_loop
add ecx,256
jz NEAR L$009done
mov ebx,DWORD [520+esp]
lea esi,[esi-128]
mov edx,DWORD [516+esp]
lea edi,[edi-128]
movd xmm2,DWORD [64+ebp]
movdqu xmm3,[ebx]
paddd xmm2,[96+eax]
pand xmm3,[112+eax]
por xmm3,xmm2
L$0061x:
movdqa xmm0,[32+eax]
movdqu xmm1,[edx]
movdqu xmm2,[16+edx]
movdqa xmm6,[eax]
movdqa xmm7,[16+eax]
mov DWORD [48+esp],ebp
movdqa [esp],xmm0
movdqa [16+esp],xmm1
movdqa [32+esp],xmm2
movdqa [48+esp],xmm3
mov edx,10
jmp NEAR L$010loop1x
align 16
L$011outer1x:
movdqa xmm3,[80+eax]
movdqa xmm0,[esp]
movdqa xmm1,[16+esp]
movdqa xmm2,[32+esp]
paddd xmm3,[48+esp]
mov edx,10
movdqa [48+esp],xmm3
jmp NEAR L$010loop1x
align 16
L$010loop1x:
paddd xmm0,xmm1
pxor xmm3,xmm0
db 102,15,56,0,222
paddd xmm2,xmm3
pxor xmm1,xmm2
movdqa xmm4,xmm1
psrld xmm1,20
pslld xmm4,12
por xmm1,xmm4
paddd xmm0,xmm1
pxor xmm3,xmm0
db 102,15,56,0,223
paddd xmm2,xmm3
pxor xmm1,xmm2
movdqa xmm4,xmm1
psrld xmm1,25
pslld xmm4,7
por xmm1,xmm4
pshufd xmm2,xmm2,78
pshufd xmm1,xmm1,57
pshufd xmm3,xmm3,147
nop
paddd xmm0,xmm1
pxor xmm3,xmm0
db 102,15,56,0,222
paddd xmm2,xmm3
pxor xmm1,xmm2
movdqa xmm4,xmm1
psrld xmm1,20
pslld xmm4,12
por xmm1,xmm4
paddd xmm0,xmm1
pxor xmm3,xmm0
db 102,15,56,0,223
paddd xmm2,xmm3
pxor xmm1,xmm2
movdqa xmm4,xmm1
psrld xmm1,25
pslld xmm4,7
por xmm1,xmm4
pshufd xmm2,xmm2,78
pshufd xmm1,xmm1,147
pshufd xmm3,xmm3,57
dec edx
jnz NEAR L$010loop1x
paddd xmm0,[esp]
paddd xmm1,[16+esp]
paddd xmm2,[32+esp]
paddd xmm3,[48+esp]
cmp ecx,64
jb NEAR L$012tail
movdqu xmm4,[esi]
movdqu xmm5,[16+esi]
pxor xmm0,xmm4
movdqu xmm4,[32+esi]
pxor xmm1,xmm5
movdqu xmm5,[48+esi]
pxor xmm2,xmm4
pxor xmm3,xmm5
lea esi,[64+esi]
movdqu [edi],xmm0
movdqu [16+edi],xmm1
movdqu [32+edi],xmm2
movdqu [48+edi],xmm3
lea edi,[64+edi]
sub ecx,64
jnz NEAR L$011outer1x
jmp NEAR L$009done
L$012tail:
movdqa [esp],xmm0
movdqa [16+esp],xmm1
movdqa [32+esp],xmm2
movdqa [48+esp],xmm3
xor eax,eax
xor edx,edx
xor ebp,ebp
L$013tail_loop:
mov al,BYTE [ebp*1+esp]
mov dl,BYTE [ebp*1+esi]
lea ebp,[1+ebp]
xor al,dl
mov BYTE [ebp*1+edi-1],al
dec ecx
jnz NEAR L$013tail_loop
L$009done:
mov esp,DWORD [512+esp]
pop edi
pop esi
pop ebx
pop ebp
ret
align 64
L$ssse3_data:
db 2,3,0,1,6,7,4,5,10,11,8,9,14,15,12,13
db 3,0,1,2,7,4,5,6,11,8,9,10,15,12,13,14
dd 1634760805,857760878,2036477234,1797285236
dd 0,1,2,3
dd 4,4,4,4
dd 1,0,0,0
dd 4,0,0,0
dd 0,-1,-1,-1
align 64
db 67,104,97,67,104,97,50,48,32,102,111,114,32,120,56,54
db 44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32
db 60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111
db 114,103,62,0
%else
; Work around https://bugzilla.nasm.us/show_bug.cgi?id=3392738
ret
%endif
#endif // defined(__i386__) && defined(_WIN32)
#if defined(__linux__) && defined(__ELF__)
.section .note.GNU-stack,"",%progbits
#endif
@@ -85,17 +85,14 @@ static int des_cbc_cipher(EVP_CIPHER_CTX *ctx, uint8_t *out, const uint8_t *in,
}
static const EVP_CIPHER evp_des_cbc = {
/* nid = */ NID_des_cbc,
/* block_size = */ 8,
/* key_len = */ 8,
/* iv_len = */ 8,
/* ctx_size = */ sizeof(EVP_DES_KEY),
/* flags = */ EVP_CIPH_CBC_MODE,
/* app_data = */ NULL,
/* init = */ des_init_key,
/* cipher = */ des_cbc_cipher,
/* cleanup = */ NULL,
/* ctrl = */ NULL,
.nid = NID_des_cbc,
.block_size = 8,
.key_len = 8,
.iv_len = 8,
.ctx_size = sizeof(EVP_DES_KEY),
.flags = EVP_CIPH_CBC_MODE,
.init = des_init_key,
.cipher = des_cbc_cipher,
};
const EVP_CIPHER *EVP_des_cbc(void) { return &evp_des_cbc; }
@@ -115,17 +112,14 @@ static int des_ecb_cipher(EVP_CIPHER_CTX *ctx, uint8_t *out, const uint8_t *in,
}
static const EVP_CIPHER evp_des_ecb = {
/* nid = */ NID_des_ecb,
/* block_size = */ 8,
/* key_len = */ 8,
/* iv_len = */ 0,
/* ctx_size = */ sizeof(EVP_DES_KEY),
/* flags = */ EVP_CIPH_ECB_MODE,
/* app_data = */ NULL,
/* init = */ des_init_key,
/* cipher = */ des_ecb_cipher,
/* cleanup = */ NULL,
/* ctrl = */ NULL,
.nid = NID_des_ecb,
.block_size = 8,
.key_len = 8,
.iv_len = 0,
.ctx_size = sizeof(EVP_DES_KEY),
.flags = EVP_CIPH_ECB_MODE,
.init = des_init_key,
.cipher = des_ecb_cipher,
};
const EVP_CIPHER *EVP_des_ecb(void) { return &evp_des_ecb; }
@@ -155,17 +149,14 @@ static int des_ede3_cbc_cipher(EVP_CIPHER_CTX *ctx, uint8_t *out,
}
static const EVP_CIPHER evp_des_ede3_cbc = {
/* nid = */ NID_des_ede3_cbc,
/* block_size = */ 8,
/* key_len = */ 24,
/* iv_len = */ 8,
/* ctx_size = */ sizeof(DES_EDE_KEY),
/* flags = */ EVP_CIPH_CBC_MODE,
/* app_data = */ NULL,
/* init = */ des_ede3_init_key,
/* cipher = */ des_ede3_cbc_cipher,
/* cleanup = */ NULL,
/* ctrl = */ NULL,
.nid = NID_des_ede3_cbc,
.block_size = 8,
.key_len = 24,
.iv_len = 8,
.ctx_size = sizeof(DES_EDE_KEY),
.flags = EVP_CIPH_CBC_MODE,
.init = des_ede3_init_key,
.cipher = des_ede3_cbc_cipher,
};
const EVP_CIPHER *EVP_des_ede3_cbc(void) { return &evp_des_ede3_cbc; }
@@ -181,17 +172,14 @@ static int des_ede_init_key(EVP_CIPHER_CTX *ctx, const uint8_t *key,
}
static const EVP_CIPHER evp_des_ede_cbc = {
/* nid = */ NID_des_ede_cbc,
/* block_size = */ 8,
/* key_len = */ 16,
/* iv_len = */ 8,
/* ctx_size = */ sizeof(DES_EDE_KEY),
/* flags = */ EVP_CIPH_CBC_MODE,
/* app_data = */ NULL,
/* init = */ des_ede_init_key,
/* cipher = */ des_ede3_cbc_cipher,
/* cleanup = */ NULL,
/* ctrl = */ NULL,
.nid = NID_des_ede_cbc,
.block_size = 8,
.key_len = 16,
.iv_len = 8,
.ctx_size = sizeof(DES_EDE_KEY),
.flags = EVP_CIPH_CBC_MODE,
.init = des_ede_init_key,
.cipher = des_ede3_cbc_cipher,
};
const EVP_CIPHER *EVP_des_ede_cbc(void) { return &evp_des_ede_cbc; }
@@ -212,33 +200,27 @@ static int des_ede_ecb_cipher(EVP_CIPHER_CTX *ctx, uint8_t *out,
}
static const EVP_CIPHER evp_des_ede = {
/* nid = */ NID_des_ede_ecb,
/* block_size = */ 8,
/* key_len = */ 16,
/* iv_len = */ 0,
/* ctx_size = */ sizeof(DES_EDE_KEY),
/* flags = */ EVP_CIPH_ECB_MODE,
/* app_data = */ NULL,
/* init = */ des_ede_init_key,
/* cipher = */ des_ede_ecb_cipher,
/* cleanup = */ NULL,
/* ctrl = */ NULL,
.nid = NID_des_ede_ecb,
.block_size = 8,
.key_len = 16,
.iv_len = 0,
.ctx_size = sizeof(DES_EDE_KEY),
.flags = EVP_CIPH_ECB_MODE,
.init = des_ede_init_key,
.cipher = des_ede_ecb_cipher,
};
const EVP_CIPHER *EVP_des_ede(void) { return &evp_des_ede; }
static const EVP_CIPHER evp_des_ede3 = {
/* nid = */ NID_des_ede3_ecb,
/* block_size = */ 8,
/* key_len = */ 24,
/* iv_len = */ 0,
/* ctx_size = */ sizeof(DES_EDE_KEY),
/* flags = */ EVP_CIPH_ECB_MODE,
/* app_data = */ NULL,
/* init = */ des_ede3_init_key,
/* cipher = */ des_ede_ecb_cipher,
/* cleanup = */ NULL,
/* ctrl = */ NULL,
.nid = NID_des_ede3_ecb,
.block_size = 8,
.key_len = 24,
.iv_len = 0,
.ctx_size = sizeof(DES_EDE_KEY),
.flags = EVP_CIPH_ECB_MODE,
.init = des_ede3_init_key,
.cipher = des_ede_ecb_cipher,
};
const EVP_CIPHER *EVP_des_ede3(void) { return &evp_des_ede3; }
@@ -78,9 +78,13 @@ static int null_cipher(EVP_CIPHER_CTX *ctx, uint8_t *out,
}
static const EVP_CIPHER n_cipher = {
NID_undef, 1 /* block size */, 0 /* key_len */, 0 /* iv_len */,
0 /* ctx_size */, 0 /* flags */, NULL /* app_data */, null_init_key,
null_cipher, NULL /* cleanup */, NULL /* ctrl */,
.nid = NID_undef,
.block_size = 1,
.key_len = 0,
.iv_len = 0,
.ctx_size = 0,
.init = null_init_key,
.cipher = null_cipher,
};
const EVP_CIPHER *EVP_enc_null(void) { return &n_cipher; }
@@ -427,37 +427,29 @@ static int rc2_ctrl(EVP_CIPHER_CTX *ctx, int type, int arg, void *ptr) {
}
static const EVP_CIPHER rc2_40_cbc = {
NID_rc2_40_cbc,
8 /* block size */,
5 /* 40 bit */,
8 /* iv len */,
sizeof(EVP_RC2_KEY),
EVP_CIPH_CBC_MODE | EVP_CIPH_VARIABLE_LENGTH | EVP_CIPH_CTRL_INIT,
NULL /* app_data */,
rc2_init_key,
rc2_cbc_cipher,
NULL,
rc2_ctrl,
.nid = NID_rc2_40_cbc,
.block_size = 8,
.key_len = 5 /* 40 bit */,
.iv_len = 8,
.ctx_size = sizeof(EVP_RC2_KEY),
.flags = EVP_CIPH_CBC_MODE | EVP_CIPH_VARIABLE_LENGTH | EVP_CIPH_CTRL_INIT,
.init = rc2_init_key,
.cipher = rc2_cbc_cipher,
.ctrl = rc2_ctrl,
};
const EVP_CIPHER *EVP_rc2_40_cbc(void) {
return &rc2_40_cbc;
}
const EVP_CIPHER *EVP_rc2_40_cbc(void) { return &rc2_40_cbc; }
static const EVP_CIPHER rc2_cbc = {
NID_rc2_cbc,
8 /* block size */,
16 /* 128 bit */,
8 /* iv len */,
sizeof(EVP_RC2_KEY),
EVP_CIPH_CBC_MODE | EVP_CIPH_VARIABLE_LENGTH | EVP_CIPH_CTRL_INIT,
NULL /* app_data */,
rc2_init_key,
rc2_cbc_cipher,
NULL,
rc2_ctrl,
.nid = NID_rc2_cbc,
.block_size = 8,
.key_len = 16 /* 128 bit */,
.iv_len = 8,
.ctx_size = sizeof(EVP_RC2_KEY),
.flags = EVP_CIPH_CBC_MODE | EVP_CIPH_VARIABLE_LENGTH | EVP_CIPH_CTRL_INIT,
.init = rc2_init_key,
.cipher = rc2_cbc_cipher,
.ctrl = rc2_ctrl,
};
const EVP_CIPHER *EVP_rc2_cbc(void) {
return &rc2_cbc;
}
const EVP_CIPHER *EVP_rc2_cbc(void) { return &rc2_cbc; }
@@ -81,9 +81,14 @@ static int rc4_cipher(EVP_CIPHER_CTX *ctx, uint8_t *out, const uint8_t *in,
}
static const EVP_CIPHER rc4 = {
NID_rc4, 1 /* block_size */, 16 /* key_size */,
0 /* iv_len */, sizeof(RC4_KEY), EVP_CIPH_VARIABLE_LENGTH,
NULL /* app_data */, rc4_init_key, rc4_cipher,
NULL /* cleanup */, NULL /* ctrl */, };
.nid = NID_rc4,
.block_size = 1,
.key_len = 16,
.iv_len = 0,
.ctx_size = sizeof(RC4_KEY),
.flags = EVP_CIPH_VARIABLE_LENGTH,
.init = rc4_init_key,
.cipher = rc4_cipher,
};
const EVP_CIPHER *EVP_rc4(void) { return &rc4; }
@@ -442,6 +442,7 @@ static int aead_tls_get_iv(const EVP_AEAD_CTX *ctx, const uint8_t **out_iv,
const AEAD_TLS_CTX *tls_ctx = (AEAD_TLS_CTX *)&ctx->state;
const size_t iv_len = EVP_CIPHER_CTX_iv_length(&tls_ctx->cipher_ctx);
if (iv_len <= 1) {
OPENSSL_PUT_ERROR(CIPHER, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
return 0;
}
@@ -192,22 +192,65 @@ OPENSSL_INLINE int chacha20_poly1305_asm_capable(void) {
// Additional input parameters are passed in |aead_data->in|. On exit, it will
// write calculated tag value to |aead_data->out.tag|, which the caller must
// check.
#if defined(OPENSSL_X86_64)
extern void chacha20_poly1305_open_nohw(
uint8_t *out_plaintext, const uint8_t *ciphertext, size_t plaintext_len,
const uint8_t *ad, size_t ad_len, union chacha20_poly1305_open_data *data);
extern void chacha20_poly1305_open_avx2(
uint8_t *out_plaintext, const uint8_t *ciphertext, size_t plaintext_len,
const uint8_t *ad, size_t ad_len, union chacha20_poly1305_open_data *data);
OPENSSL_INLINE void chacha20_poly1305_open(uint8_t *out_plaintext,
const uint8_t *ciphertext,
size_t plaintext_len, const uint8_t *ad,
size_t ad_len,
union chacha20_poly1305_open_data *data) {
if (CRYPTO_is_AVX2_capable() && CRYPTO_is_BMI2_capable()) {
chacha20_poly1305_open_avx2(out_plaintext, ciphertext, plaintext_len, ad,
ad_len, data);
} else {
chacha20_poly1305_open_nohw(out_plaintext, ciphertext, plaintext_len, ad,
ad_len, data);
}
}
#else
extern void chacha20_poly1305_open(uint8_t *out_plaintext,
const uint8_t *ciphertext,
size_t plaintext_len, const uint8_t *ad,
size_t ad_len,
union chacha20_poly1305_open_data *data);
#endif
// chacha20_poly1305_open is defined in chacha20_poly1305_*.pl. It encrypts
// |plaintext_len| bytes from |plaintext| and writes them to |out_ciphertext|.
// Additional input parameters are passed in |aead_data->in|. The calculated tag
// value is over the computed ciphertext concatenated with |extra_ciphertext|
// and written to |aead_data->out.tag|.
#if defined(OPENSSL_X86_64)
extern void chacha20_poly1305_seal_nohw(
uint8_t *out_ciphertext, const uint8_t *plaintext, size_t plaintext_len,
const uint8_t *ad, size_t ad_len, union chacha20_poly1305_seal_data *data);
extern void chacha20_poly1305_seal_avx2(
uint8_t *out_ciphertext, const uint8_t *plaintext, size_t plaintext_len,
const uint8_t *ad, size_t ad_len, union chacha20_poly1305_seal_data *data);
OPENSSL_INLINE void chacha20_poly1305_seal(
uint8_t *out_ciphertext, const uint8_t *plaintext, size_t plaintext_len,
const uint8_t *ad, size_t ad_len, union chacha20_poly1305_seal_data *data) {
if (CRYPTO_is_AVX2_capable() && CRYPTO_is_BMI2_capable()) {
chacha20_poly1305_seal_avx2(out_ciphertext, plaintext, plaintext_len, ad,
ad_len, data);
} else {
chacha20_poly1305_seal_nohw(out_ciphertext, plaintext, plaintext_len, ad,
ad_len, data);
}
}
#else
extern void chacha20_poly1305_seal(uint8_t *out_ciphertext,
const uint8_t *plaintext,
size_t plaintext_len, const uint8_t *ad,
size_t ad_len,
union chacha20_poly1305_seal_data *data);
#endif
#else
OPENSSL_INLINE int chacha20_poly1305_asm_capable(void) { return 0; }
@@ -121,8 +121,8 @@ void EVP_tls_cbc_copy_mac(uint8_t *out, size_t md_size, const uint8_t *in,
size_t mac_end = in_len;
size_t mac_start = mac_end - md_size;
assert(orig_len >= in_len);
assert(in_len >= md_size);
declassify_assert(orig_len >= in_len);
declassify_assert(in_len >= md_size);
assert(md_size <= EVP_MAX_MD_SIZE);
assert(md_size > 0);
+66 -41
View File
@@ -56,6 +56,7 @@
#include <CCryptoBoringSSL_conf.h>
#include <assert.h>
#include <string.h>
#include <ctype.h>
@@ -65,7 +66,6 @@
#include <CCryptoBoringSSL_lhash.h>
#include <CCryptoBoringSSL_mem.h>
#include "conf_def.h"
#include "internal.h"
#include "../internal.h"
@@ -183,6 +183,26 @@ err:
return NULL;
}
static int is_comment(char c) { return c == '#'; }
static int is_quote(char c) { return c == '"' || c == '\'' || c == '`'; }
static int is_esc(char c) { return c == '\\'; }
static int is_conf_ws(char c) {
// This differs from |OPENSSL_isspace| in that CONF does not accept '\v' and
// '\f' as whitespace.
return c == ' ' || c == '\t' || c == '\r' || c == '\n';
}
static int is_name_char(char c) {
// Alphanumeric characters, and a handful of symbols, may appear in value and
// section names without escaping.
return OPENSSL_isalnum(c) || c == '_' || c == '!' || c == '.' || c == '%' ||
c == '&' || c == '*' || c == '+' || c == ',' || c == '/' || c == ';' ||
c == '?' || c == '@' || c == '^' || c == '~' || c == '|' || c == '-';
}
static int str_copy(CONF *conf, char *section, char **pto, char *from) {
int q, to = 0, len = 0;
char v;
@@ -199,13 +219,13 @@ static int str_copy(CONF *conf, char *section, char **pto, char *from) {
}
for (;;) {
if (IS_QUOTE(conf, *from)) {
if (is_quote(*from)) {
q = *from;
from++;
while (!IS_EOF(conf, *from) && (*from != q)) {
if (IS_ESC(conf, *from)) {
while (*from != '\0' && *from != q) {
if (is_esc(*from)) {
from++;
if (IS_EOF(conf, *from)) {
if (*from == '\0') {
break;
}
}
@@ -214,10 +234,10 @@ static int str_copy(CONF *conf, char *section, char **pto, char *from) {
if (*from == q) {
from++;
}
} else if (IS_ESC(conf, *from)) {
} else if (is_esc(*from)) {
from++;
v = *(from++);
if (IS_EOF(conf, v)) {
if (v == '\0') {
break;
} else if (v == 'r') {
v = '\r';
@@ -229,11 +249,13 @@ static int str_copy(CONF *conf, char *section, char **pto, char *from) {
v = '\t';
}
buf->data[to++] = v;
} else if (IS_EOF(conf, *from)) {
} else if (*from == '\0') {
break;
} else if (*from == '$') {
// Historically, $foo would expand to a previously-parsed value. This
// feature has been removed as it was unused and is a DoS vector.
// feature has been removed as it was unused and is a DoS vector. If
// trying to embed '$' in a line, either escape it or wrap the value in
// quotes.
OPENSSL_PUT_ERROR(CONF, CONF_R_VARIABLE_EXPANSION_NOT_SUPPORTED);
goto err;
} else {
@@ -312,36 +334,39 @@ static int add_string(const CONF *conf, CONF_SECTION *section,
return 1;
}
static char *eat_ws(CONF *conf, char *p) {
while (IS_WS(conf, *p) && !IS_EOF(conf, *p)) {
static char *eat_ws(char *p) {
while (*p != '\0' && is_conf_ws(*p)) {
p++;
}
return p;
}
#define scan_esc(conf, p) (((IS_EOF((conf), (p)[1])) ? ((p) + 1) : ((p) + 2)))
static char *scan_esc(char *p) {
assert(p[0] == '\\');
return p[1] == '\0' ? p + 1 : p + 2;
}
static char *eat_alpha_numeric(CONF *conf, char *p) {
static char *eat_name(char *p) {
for (;;) {
if (IS_ESC(conf, *p)) {
p = scan_esc(conf, p);
if (is_esc(*p)) {
p = scan_esc(p);
continue;
}
if (!IS_ALPHA_NUMERIC_PUNCT(conf, *p)) {
if (!is_name_char(*p)) {
return p;
}
p++;
}
}
static char *scan_quote(CONF *conf, char *p) {
static char *scan_quote(char *p) {
int q = *p;
p++;
while (!IS_EOF(conf, *p) && *p != q) {
if (IS_ESC(conf, *p)) {
while (*p != '\0' && *p != q) {
if (is_esc(*p)) {
p++;
if (IS_EOF(conf, *p)) {
if (*p == '\0') {
return p;
}
}
@@ -353,28 +378,28 @@ static char *scan_quote(CONF *conf, char *p) {
return p;
}
static void clear_comments(CONF *conf, char *p) {
static void clear_comments(char *p) {
for (;;) {
if (!IS_WS(conf, *p)) {
if (!is_conf_ws(*p)) {
break;
}
p++;
}
for (;;) {
if (IS_COMMENT(conf, *p)) {
if (is_comment(*p)) {
*p = '\0';
return;
}
if (IS_QUOTE(conf, *p)) {
p = scan_quote(conf, p);
if (is_quote(*p)) {
p = scan_quote(p);
continue;
}
if (IS_ESC(conf, *p)) {
p = scan_esc(conf, p);
if (is_esc(*p)) {
p = scan_esc(p);
continue;
}
if (IS_EOF(conf, *p)) {
if (*p == '\0') {
return;
} else {
p++;
@@ -454,7 +479,7 @@ int NCONF_load_bio(CONF *conf, BIO *in, long *out_error_line) {
// If we have bytes and the last char '\\' and
// second last char is not '\\'
p = &(buff->data[bufnum - 1]);
if (IS_ESC(conf, p[0]) && ((bufnum <= 1) || !IS_ESC(conf, p[-1]))) {
if (is_esc(p[0]) && ((bufnum <= 1) || !is_esc(p[-1]))) {
bufnum--;
again = 1;
}
@@ -465,20 +490,20 @@ int NCONF_load_bio(CONF *conf, BIO *in, long *out_error_line) {
bufnum = 0;
buf = buff->data;
clear_comments(conf, buf);
s = eat_ws(conf, buf);
if (IS_EOF(conf, *s)) {
clear_comments(buf);
s = eat_ws(buf);
if (*s == '\0') {
continue; // blank line
}
if (*s == '[') {
char *ss;
s++;
start = eat_ws(conf, s);
start = eat_ws(s);
ss = start;
again:
end = eat_alpha_numeric(conf, ss);
p = eat_ws(conf, end);
end = eat_name(ss);
p = eat_ws(end);
if (*p != ']') {
if (*p != '\0' && ss != p) {
ss = p;
@@ -502,27 +527,27 @@ int NCONF_load_bio(CONF *conf, BIO *in, long *out_error_line) {
} else {
pname = s;
psection = NULL;
end = eat_alpha_numeric(conf, s);
end = eat_name(s);
if ((end[0] == ':') && (end[1] == ':')) {
*end = '\0';
end += 2;
psection = pname;
pname = end;
end = eat_alpha_numeric(conf, end);
end = eat_name(end);
}
p = eat_ws(conf, end);
p = eat_ws(end);
if (*p != '=') {
OPENSSL_PUT_ERROR(CONF, CONF_R_MISSING_EQUAL_SIGN);
goto err;
}
*end = '\0';
p++;
start = eat_ws(conf, p);
while (!IS_EOF(conf, *p)) {
start = eat_ws(p);
while (*p != '\0') {
p++;
}
p--;
while ((p != start) && (IS_WS(conf, *p))) {
while (p != start && is_conf_ws(*p)) {
p--;
}
p++;
@@ -1,122 +0,0 @@
/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
* All rights reserved.
*
* This package is an SSL implementation written
* by Eric Young (eay@cryptsoft.com).
* The implementation was written so as to conform with Netscapes SSL.
*
* This library is free for commercial and non-commercial use as long as
* the following conditions are aheared to. The following conditions
* apply to all code found in this distribution, be it the RC4, RSA,
* lhash, DES, etc., code; not just the SSL code. The SSL documentation
* included with this distribution is covered by the same copyright terms
* except that the holder is Tim Hudson (tjh@cryptsoft.com).
*
* Copyright remains Eric Young's, and as such any Copyright notices in
* the code are not to be removed.
* If this package is used in a product, Eric Young should be given attribution
* as the author of the parts of the library used.
* This can be in the form of a textual message at program startup or
* in documentation (online or textual) provided with the package.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* "This product includes cryptographic software written by
* Eric Young (eay@cryptsoft.com)"
* The word 'cryptographic' can be left out if the rouines from the library
* being used are not cryptographic related :-).
* 4. If you include any Windows specific code (or a derivative thereof) from
* the apps directory (application code) you must include an acknowledgement:
* "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
*
* THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* The licence and distribution terms for any publically available version or
* derivative of this code cannot be changed. i.e. this code cannot simply be
* copied and put under another distribution licence
* [including the GNU Public Licence.]
*/
// This file was historically generated by keysets.pl in OpenSSL.
//
// TODO(davidben): Replace it with something more readable.
#define CONF_NUMBER 1
#define CONF_UPPER 2
#define CONF_LOWER 4
#define CONF_UNDER 256
#define CONF_PUNCTUATION 512
#define CONF_WS 16
#define CONF_ESC 32
#define CONF_QUOTE 64
#define CONF_COMMENT 128
#define CONF_EOF 8
#define CONF_HIGHBIT 4096
#define CONF_ALPHA (CONF_UPPER|CONF_LOWER)
#define CONF_ALPHA_NUMERIC (CONF_ALPHA|CONF_NUMBER|CONF_UNDER)
#define CONF_ALPHA_NUMERIC_PUNCT (CONF_ALPHA|CONF_NUMBER|CONF_UNDER| \
CONF_PUNCTUATION)
#define KEYTYPES(c) CONF_type_default
#define IS_COMMENT(c,a) (KEYTYPES(c)[(a)&0xff]&CONF_COMMENT)
#define IS_EOF(c,a) (KEYTYPES(c)[(a)&0xff]&CONF_EOF)
#define IS_ESC(c,a) (KEYTYPES(c)[(a)&0xff]&CONF_ESC)
#define IS_NUMBER(c,a) (KEYTYPES(c)[(a)&0xff]&CONF_NUMBER)
#define IS_WS(c,a) (KEYTYPES(c)[(a)&0xff]&CONF_WS)
#define IS_ALPHA_NUMERIC(c,a) (KEYTYPES(c)[(a)&0xff]&CONF_ALPHA_NUMERIC)
#define IS_ALPHA_NUMERIC_PUNCT(c,a) \
(KEYTYPES(c)[(a)&0xff]&CONF_ALPHA_NUMERIC_PUNCT)
#define IS_QUOTE(c,a) (KEYTYPES(c)[(a)&0xff]&CONF_QUOTE)
static const unsigned short CONF_type_default[256]={
0x0008,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0010,0x0010,0x0000,0x0000,0x0010,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,0x0000,
0x0010,0x0200,0x0040,0x0080,0x0000,0x0200,0x0200,0x0040,
0x0000,0x0000,0x0200,0x0200,0x0200,0x0200,0x0200,0x0200,
0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,0x0001,
0x0001,0x0001,0x0000,0x0200,0x0000,0x0000,0x0000,0x0200,
0x0200,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,
0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,
0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,0x0002,
0x0002,0x0002,0x0002,0x0000,0x0020,0x0000,0x0200,0x0100,
0x0040,0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,
0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,
0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,0x0004,
0x0004,0x0004,0x0004,0x0000,0x0200,0x0000,0x0200,0x0000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,0x1000,
};
@@ -15,7 +15,7 @@
#include "internal.h"
#if defined(OPENSSL_AARCH64) && defined(OPENSSL_APPLE) && \
!defined(OPENSSL_STATIC_ARMCAP)
!defined(OPENSSL_STATIC_ARMCAP) && !defined(OPENSSL_NO_ASM)
#include <sys/sysctl.h>
#include <sys/types.h>
@@ -15,7 +15,7 @@
#include "internal.h"
#if defined(OPENSSL_AARCH64) && defined(OPENSSL_FUCHSIA) && \
!defined(OPENSSL_STATIC_ARMCAP)
!defined(OPENSSL_STATIC_ARMCAP) && !defined(OPENSSL_NO_ASM)
#include <zircon/features.h>
#include <zircon/syscalls.h>
@@ -15,7 +15,7 @@
#include "internal.h"
#if defined(OPENSSL_AARCH64) && defined(OPENSSL_LINUX) && \
!defined(OPENSSL_STATIC_ARMCAP)
!defined(OPENSSL_STATIC_ARMCAP) && !defined(OPENSSL_NO_ASM)
#include <sys/auxv.h>
@@ -15,11 +15,11 @@
#include <CCryptoBoringSSL_cpu.h>
#if defined(OPENSSL_AARCH64) && defined(OPENSSL_OPENBSD) && \
!defined(OPENSSL_STATIC_ARMCAP)
!defined(OPENSSL_STATIC_ARMCAP) && !defined(OPENSSL_NO_ASM)
#include <sys/sysctl.h>
#include <machine/cpu.h>
#include <machine/armreg.h>
#include <machine/cpu.h>
#include <sys/sysctl.h>
#include <CCryptoBoringSSL_arm_arch.h>
@@ -27,7 +27,7 @@
void OPENSSL_cpuid_setup(void) {
int isar0_mib[] = { CTL_MACHDEP, CPU_ID_AA64ISAR0 };
int isar0_mib[] = {CTL_MACHDEP, CPU_ID_AA64ISAR0};
uint64_t cpu_id = 0;
size_t len = sizeof(cpu_id);
@@ -18,7 +18,8 @@
// expects userspace to simply read them. It traps the reads and fills in CPU
// capabilities.
#if defined(OPENSSL_AARCH64) && !defined(OPENSSL_STATIC_ARMCAP) && \
(defined(ANDROID_BAREMETAL) || defined(OPENSSL_FREEBSD))
(defined(ANDROID_BAREMETAL) || defined(OPENSSL_FREEBSD)) && \
!defined(OPENSSL_NO_ASM)
#include <CCryptoBoringSSL_arm_arch.h>
@@ -16,7 +16,7 @@
#include "internal.h"
#if defined(OPENSSL_AARCH64) && defined(OPENSSL_WINDOWS) && \
!defined(OPENSSL_STATIC_ARMCAP)
!defined(OPENSSL_STATIC_ARMCAP) && !defined(OPENSSL_NO_ASM)
#include <windows.h>
@@ -143,6 +143,9 @@ void OPENSSL_cpuid_setup(void) {
int CRYPTO_has_broken_NEON(void) { return 0; }
int CRYPTO_needs_hwcap2_workaround(void) { return g_needs_hwcap2_workaround; }
int CRYPTO_needs_hwcap2_workaround(void) {
OPENSSL_init_cpuid();
return g_needs_hwcap2_workaround;
}
#endif // OPENSSL_ARM && OPENSSL_LINUX && !OPENSSL_STATIC_ARMCAP
+71 -37
View File
@@ -173,20 +173,21 @@ void OPENSSL_cpuid_setup(void) {
OPENSSL_cpuid(&eax, &ebx, &ecx, &edx, 1);
const uint32_t base_family = (eax >> 8) & 15;
const uint32_t base_model = (eax >> 4) & 15;
uint32_t family = base_family;
uint32_t model = base_model;
if (base_family == 15) {
const uint32_t ext_family = (eax >> 20) & 255;
family += ext_family;
}
if (base_family == 6 || base_family == 15) {
const uint32_t ext_model = (eax >> 16) & 15;
model |= ext_model << 4;
}
if (is_amd) {
// See https://www.amd.com/system/files/TechDocs/25481.pdf, page 10.
const uint32_t base_family = (eax >> 8) & 15;
const uint32_t base_model = (eax >> 4) & 15;
uint32_t family = base_family;
uint32_t model = base_model;
if (base_family == 0xf) {
const uint32_t ext_family = (eax >> 20) & 255;
family += ext_family;
const uint32_t ext_model = (eax >> 16) & 15;
model |= ext_model << 4;
}
if (family < 0x17 || (family == 0x17 && 0x70 <= model && model <= 0x7f)) {
// Disable RDRAND on AMD families before 0x17 (Zen) due to reported
// failures after suspend.
@@ -208,15 +209,6 @@ void OPENSSL_cpuid_setup(void) {
// Reserved bit #30 is repurposed to signal an Intel CPU.
if (is_intel) {
edx |= (1u << 30);
// Clear the XSAVE bit on Knights Landing to mimic Silvermont. This enables
// some Silvermont-specific codepaths which perform better. See OpenSSL
// commit 64d92d74985ebb3d0be58a9718f9e080a14a8e7f and
// |CRYPTO_cpu_perf_is_like_silvermont|.
if ((eax & 0x0fff0ff0) == 0x00050670 /* Knights Landing */ ||
(eax & 0x0fff0ff0) == 0x00080650 /* Knights Mill (per SDE) */) {
ecx &= ~(1u << 26);
}
} else {
edx &= ~(1u << 30);
}
@@ -236,25 +228,67 @@ void OPENSSL_cpuid_setup(void) {
ecx &= ~(1u << 28); // AVX
ecx &= ~(1u << 12); // FMA
ecx &= ~(1u << 11); // AMD XOP
// Clear AVX2 and AVX512* bits.
//
// TODO(davidben): Should bits 17 and 26-28 also be cleared? Upstream
// doesn't clear those. See the comments in
// |CRYPTO_hardware_supports_XSAVE|.
extended_features[0] &=
~((1u << 5) | (1u << 16) | (1u << 21) | (1u << 30) | (1u << 31));
extended_features[0] &= ~(1u << 5); // AVX2
extended_features[1] &= ~(1u << 9); // VAES
extended_features[1] &= ~(1u << 10); // VPCLMULQDQ
}
// See Intel manual, volume 1, section 15.2.
// See Intel manual, volume 1, sections 15.2 ("Detection of AVX-512 Foundation
// Instructions") through 15.4 ("Detection of Intel AVX-512 Instruction Groups
// Operating at 256 and 128-bit Vector Lengths").
if ((xcr0 & 0xe6) != 0xe6) {
// Clear AVX512F. Note we don't touch other AVX512 extensions because they
// can be used with YMM.
extended_features[0] &= ~(1u << 16);
// Without XCR0.111xx11x, no AVX512 feature can be used. This includes ZMM
// registers, masking, SIMD registers 16-31 (even if accessed as YMM or
// XMM), and EVEX-coded instructions (even on YMM or XMM). Even if only
// XCR0.ZMM_Hi256 is missing, it isn't valid to use AVX512 features on
// shorter vectors, since AVX512 ties everything to the availability of
// 512-bit vectors. See the above-mentioned sections of the Intel manual,
// which say that *all* these XCR0 bits must be checked even when just using
// 128-bit or 256-bit vectors, and also volume 2a section 2.7.11 ("#UD
// Equations for EVEX") which says that all EVEX-coded instructions raise an
// undefined-instruction exception if any of these XCR0 bits is zero.
//
// AVX10 fixes this by reorganizing the features that used to be part of
// "AVX512" and allowing them to be used independently of 512-bit support.
// TODO: add AVX10 detection.
extended_features[0] &= ~(1u << 16); // AVX512F
extended_features[0] &= ~(1u << 17); // AVX512DQ
extended_features[0] &= ~(1u << 21); // AVX512IFMA
extended_features[0] &= ~(1u << 26); // AVX512PF
extended_features[0] &= ~(1u << 27); // AVX512ER
extended_features[0] &= ~(1u << 28); // AVX512CD
extended_features[0] &= ~(1u << 30); // AVX512BW
extended_features[0] &= ~(1u << 31); // AVX512VL
extended_features[1] &= ~(1u << 1); // AVX512VBMI
extended_features[1] &= ~(1u << 6); // AVX512VBMI2
extended_features[1] &= ~(1u << 11); // AVX512VNNI
extended_features[1] &= ~(1u << 12); // AVX512BITALG
extended_features[1] &= ~(1u << 14); // AVX512VPOPCNTDQ
}
// Disable ADX instructions on Knights Landing. See OpenSSL commit
// 64d92d74985ebb3d0be58a9718f9e080a14a8e7f.
if ((ecx & (1u << 26)) == 0) {
extended_features[0] &= ~(1u << 19);
// Repurpose the bit for the removed MPX feature to indicate when using zmm
// registers should be avoided even when they are supported. (When set, AVX512
// features can still be used, but only using ymm or xmm registers.) Skylake
// suffered from severe downclocking when zmm registers were used, which
// affected unrelated code running on the system, making zmm registers not too
// useful outside of benchmarks. The situation improved significantly by Ice
// Lake, but a small amount of downclocking remained. (See
// https://lore.kernel.org/linux-crypto/e8ce1146-3952-6977-1d0e-a22758e58914@intel.com/)
// We take a conservative approach of not allowing zmm registers until after
// Ice Lake and Tiger Lake, i.e. until Sapphire Rapids on the server side.
//
// AMD CPUs, which support AVX512 starting with Zen 4, have not been reported
// to have any downclocking problem when zmm registers are used.
if (is_intel && family == 6 &&
(model == 85 || // Skylake, Cascade Lake, Cooper Lake (server)
model == 106 || // Ice Lake (server)
model == 108 || // Ice Lake (micro server)
model == 125 || // Ice Lake (client)
model == 126 || // Ice Lake (mobile)
model == 140 || // Tiger Lake (mobile)
model == 141)) { // Tiger Lake (client)
extended_features[0] |= 1u << 14;
} else {
extended_features[0] &= ~(1u << 14);
}
OPENSSL_ia32cap_P[0] = edx;
+11 -62
View File
@@ -15,32 +15,16 @@
#include <CCryptoBoringSSL_crypto.h>
#include <assert.h>
#include <stdio.h>
#include "fipsmodule/rand/fork_detect.h"
#include "fipsmodule/rand/internal.h"
#include "bcm_support.h"
#include "internal.h"
static_assert(sizeof(ossl_ssize_t) == sizeof(size_t),
"ossl_ssize_t should be the same size as size_t");
#if !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_STATIC_ARMCAP) && \
(defined(OPENSSL_X86) || defined(OPENSSL_X86_64) || \
defined(OPENSSL_ARM) || defined(OPENSSL_AARCH64))
// x86, x86_64, and the ARMs need to record the result of a cpuid/getauxval call
// for the asm to work correctly, unless compiled without asm code.
#define NEED_CPUID
#else
// Otherwise, don't emit a static initialiser.
#if !defined(BORINGSSL_NO_STATIC_INITIALIZER)
#define BORINGSSL_NO_STATIC_INITIALIZER
#endif
#endif // !NO_ASM && !STATIC_ARMCAP && (X86 || X86_64 || ARM || AARCH64)
// Our assembly does not use the GOT to reference symbols, which means
// references to visible symbols will often require a TEXTREL. This is
@@ -79,7 +63,7 @@ HIDDEN uint8_t BORINGSSL_function_hit[7] = {0};
HIDDEN uint32_t OPENSSL_ia32cap_P[4] = {0};
uint32_t OPENSSL_get_ia32cap(int idx) {
CRYPTO_library_init();
OPENSSL_init_cpuid();
return OPENSSL_ia32cap_P[idx];
}
@@ -121,60 +105,24 @@ HIDDEN uint32_t OPENSSL_armcap_P =
HIDDEN uint32_t OPENSSL_armcap_P = 0;
uint32_t *OPENSSL_get_armcap_pointer_for_test(void) {
CRYPTO_library_init();
OPENSSL_init_cpuid();
return &OPENSSL_armcap_P;
}
#endif
uint32_t OPENSSL_get_armcap(void) {
CRYPTO_library_init();
OPENSSL_init_cpuid();
return OPENSSL_armcap_P;
}
#endif
#if defined(BORINGSSL_FIPS)
// In FIPS mode, the power-on self-test function calls |CRYPTO_library_init|
// because we have to ensure that CPUID detection occurs first.
#define BORINGSSL_NO_STATIC_INITIALIZER
#endif
#if defined(OPENSSL_WINDOWS) && !defined(BORINGSSL_NO_STATIC_INITIALIZER)
#define OPENSSL_CDECL __cdecl
#else
#define OPENSSL_CDECL
#endif
#if defined(BORINGSSL_NO_STATIC_INITIALIZER)
static CRYPTO_once_t once = CRYPTO_ONCE_INIT;
#elif defined(_MSC_VER)
#pragma section(".CRT$XCU", read)
static void __cdecl do_library_init(void);
__declspec(allocate(".CRT$XCU")) void(*library_init_constructor)(void) =
do_library_init;
#else
static void do_library_init(void) __attribute__ ((constructor));
#endif
// do_library_init is the actual initialization function. If
// BORINGSSL_NO_STATIC_INITIALIZER isn't defined, this is set as a static
// initializer. Otherwise, it is called by CRYPTO_library_init.
static void OPENSSL_CDECL do_library_init(void) {
// WARNING: this function may only configure the capability variables. See the
// note above about the linker bug.
#if defined(NEED_CPUID)
OPENSSL_cpuid_setup();
static CRYPTO_once_t once = CRYPTO_ONCE_INIT;
void OPENSSL_init_cpuid(void) { CRYPTO_once(&once, OPENSSL_cpuid_setup); }
#endif
}
void CRYPTO_library_init(void) {
// TODO(davidben): It would be tidier if this build knob could be replaced
// with an internal lazy-init mechanism that would handle things correctly
// in-library. https://crbug.com/542879
#if defined(BORINGSSL_NO_STATIC_INITIALIZER)
CRYPTO_once(&once, do_library_init);
#endif
}
void CRYPTO_library_init(void) {}
int CRYPTO_is_confidential_build(void) {
#if defined(BORINGSSL_CONFIDENTIAL)
@@ -194,7 +142,7 @@ int CRYPTO_has_asm(void) {
void CRYPTO_pre_sandbox_init(void) {
// Read from /proc/cpuinfo if needed.
CRYPTO_library_init();
OPENSSL_init_cpuid();
// Open /dev/urandom if needed.
CRYPTO_init_sysrand();
// Set up MADV_WIPEONFORK state if needed.
@@ -235,8 +183,9 @@ int ENGINE_register_all_complete(void) { return 1; }
void OPENSSL_load_builtin_modules(void) {}
int OPENSSL_init_crypto(uint64_t opts, const OPENSSL_INIT_SETTINGS *settings) {
CRYPTO_library_init();
return 1;
}
void OPENSSL_cleanup(void) {}
FILE *CRYPTO_get_stderr(void) { return stderr; }
@@ -81,7 +81,7 @@ typedef uint64_t fe_limb_t;
#define assert_fe(f) \
do { \
for (unsigned _assert_fe_i = 0; _assert_fe_i < 5; _assert_fe_i++) { \
assert(f[_assert_fe_i] <= UINT64_C(0x8cccccccccccc)); \
declassify_assert(f[_assert_fe_i] <= UINT64_C(0x8cccccccccccc)); \
} \
} while (0)
@@ -98,7 +98,7 @@ typedef uint64_t fe_limb_t;
#define assert_fe_loose(f) \
do { \
for (unsigned _assert_fe_i = 0; _assert_fe_i < 5; _assert_fe_i++) { \
assert(f[_assert_fe_i] <= UINT64_C(0x1a666666666664)); \
declassify_assert(f[_assert_fe_i] <= UINT64_C(0x1a666666666664)); \
} \
} while (0)
@@ -120,8 +120,8 @@ typedef uint32_t fe_limb_t;
#define assert_fe(f) \
do { \
for (unsigned _assert_fe_i = 0; _assert_fe_i < 10; _assert_fe_i++) { \
assert(f[_assert_fe_i] <= \
((_assert_fe_i & 1) ? 0x2333333u : 0x4666666u)); \
declassify_assert(f[_assert_fe_i] <= \
((_assert_fe_i & 1) ? 0x2333333u : 0x4666666u)); \
} \
} while (0)
@@ -138,8 +138,8 @@ typedef uint32_t fe_limb_t;
#define assert_fe_loose(f) \
do { \
for (unsigned _assert_fe_i = 0; _assert_fe_i < 10; _assert_fe_i++) { \
assert(f[_assert_fe_i] <= \
((_assert_fe_i & 1) ? 0x6999999u : 0xd333332u)); \
declassify_assert(f[_assert_fe_i] <= \
((_assert_fe_i & 1) ? 0x6999999u : 0xd333332u)); \
} \
} while (0)
@@ -150,7 +150,7 @@ static_assert(sizeof(fe) == sizeof(fe_limb_t) * FE_NUM_LIMBS,
static void fe_frombytes_strict(fe *h, const uint8_t s[32]) {
// |fiat_25519_from_bytes| requires the top-most bit be clear.
assert((s[31] & 0x80) == 0);
declassify_assert((s[31] & 0x80) == 0);
fiat_25519_from_bytes(h->v, s);
assert_fe(h->v);
}
@@ -61,6 +61,8 @@
#include <CCryptoBoringSSL_blake2.h>
#include <CCryptoBoringSSL_bytestring.h>
#include <CCryptoBoringSSL_obj.h>
#include <CCryptoBoringSSL_md4.h>
#include <CCryptoBoringSSL_md5.h>
#include <CCryptoBoringSSL_nid.h>
#include "../asn1/internal.h"
@@ -220,6 +222,7 @@ int EVP_marshal_digest_algorithm(CBB *cbb, const EVP_MD *md) {
return 0;
}
// TODO(crbug.com/boringssl/710): Is this correct? See RFC 4055, section 2.1.
if (!CBB_add_asn1(&algorithm, &null, CBS_ASN1_NULL) ||
!CBB_flush(cbb)) {
return 0;
@@ -263,3 +266,90 @@ static const EVP_MD evp_md_blake2b256 = {
};
const EVP_MD *EVP_blake2b256(void) { return &evp_md_blake2b256; }
static void md4_init(EVP_MD_CTX *ctx) {
BSSL_CHECK(MD4_Init(ctx->md_data));
}
static void md4_update(EVP_MD_CTX *ctx, const void *data, size_t count) {
BSSL_CHECK(MD4_Update(ctx->md_data, data, count));
}
static void md4_final(EVP_MD_CTX *ctx, uint8_t *out) {
BSSL_CHECK(MD4_Final(out, ctx->md_data));
}
static const EVP_MD evp_md_md4 = {
NID_md4,
MD4_DIGEST_LENGTH,
0,
md4_init,
md4_update,
md4_final,
64,
sizeof(MD4_CTX),
};
const EVP_MD *EVP_md4(void) { return &evp_md_md4; }
static void md5_init(EVP_MD_CTX *ctx) {
BSSL_CHECK(MD5_Init(ctx->md_data));
}
static void md5_update(EVP_MD_CTX *ctx, const void *data, size_t count) {
BSSL_CHECK(MD5_Update(ctx->md_data, data, count));
}
static void md5_final(EVP_MD_CTX *ctx, uint8_t *out) {
BSSL_CHECK(MD5_Final(out, ctx->md_data));
}
static const EVP_MD evp_md_md5 = {
NID_md5,
MD5_DIGEST_LENGTH,
0,
md5_init,
md5_update,
md5_final,
64,
sizeof(MD5_CTX),
};
const EVP_MD *EVP_md5(void) { return &evp_md_md5; }
typedef struct {
MD5_CTX md5;
SHA_CTX sha1;
} MD5_SHA1_CTX;
static void md5_sha1_init(EVP_MD_CTX *md_ctx) {
MD5_SHA1_CTX *ctx = md_ctx->md_data;
BSSL_CHECK(MD5_Init(&ctx->md5) && SHA1_Init(&ctx->sha1));
}
static void md5_sha1_update(EVP_MD_CTX *md_ctx, const void *data,
size_t count) {
MD5_SHA1_CTX *ctx = md_ctx->md_data;
BSSL_CHECK(MD5_Update(&ctx->md5, data, count) &&
SHA1_Update(&ctx->sha1, data, count));
}
static void md5_sha1_final(EVP_MD_CTX *md_ctx, uint8_t *out) {
MD5_SHA1_CTX *ctx = md_ctx->md_data;
BSSL_CHECK(MD5_Final(out, &ctx->md5) &&
SHA1_Final(out + MD5_DIGEST_LENGTH, &ctx->sha1));
}
const EVP_MD evp_md_md5_sha1 = {
NID_md5_sha1,
MD5_DIGEST_LENGTH + SHA_DIGEST_LENGTH,
0,
md5_sha1_init,
md5_sha1_update,
md5_sha1_final,
64,
sizeof(MD5_SHA1_CTX),
};
const EVP_MD *EVP_md5_sha1(void) { return &evp_md_md5_sha1; }
File diff suppressed because it is too large. Load diff
@@ -0,0 +1,58 @@
/* Copyright (c) 2023, Google LLC
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
* SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
#ifndef OPENSSL_HEADER_CRYPTO_DILITHIUM_INTERNAL_H
#define OPENSSL_HEADER_CRYPTO_DILITHIUM_INTERNAL_H
#include <CCryptoBoringSSL_base.h>
#include <experimental/CCryptoBoringSSL_dilithium.h>
#if defined(__cplusplus)
extern "C" {
#endif
// DILITHIUM_GENERATE_KEY_ENTROPY is the number of bytes of uniformly random
// entropy necessary to generate a key pair.
#define DILITHIUM_GENERATE_KEY_ENTROPY 32
// DILITHIUM_SIGNATURE_RANDOMIZER_BYTES is the number of bytes of uniformly
// random entropy necessary to generate a signature in randomized mode.
#define DILITHIUM_SIGNATURE_RANDOMIZER_BYTES 32
// DILITHIUM_generate_key_external_entropy generates a public/private key pair
// using the given seed, writes the encoded public key to
// |out_encoded_public_key| and sets |out_private_key| to the private key,
// returning 1 on success and 0 on failure. Returns 1 on success and 0 on
// failure.
OPENSSL_EXPORT int DILITHIUM_generate_key_external_entropy(
uint8_t out_encoded_public_key[DILITHIUM_PUBLIC_KEY_BYTES],
struct DILITHIUM_private_key *out_private_key,
const uint8_t entropy[DILITHIUM_GENERATE_KEY_ENTROPY]);
// DILITHIUM_sign_deterministic generates a signature for the message |msg| of
// length |msg_len| using |private_key| following the deterministic algorithm,
// and writes the encoded signature to |out_encoded_signature|. Returns 1 on
// success and 0 on failure.
OPENSSL_EXPORT int DILITHIUM_sign_deterministic(
uint8_t out_encoded_signature[DILITHIUM_SIGNATURE_BYTES],
const struct DILITHIUM_private_key *private_key, const uint8_t *msg,
size_t msg_len);
#if defined(__cplusplus)
} // extern C
#endif
#endif // OPENSSL_HEADER_CRYPTO_DILITHIUM_INTERNAL_H
+36 -15
View File
@@ -208,6 +208,11 @@ int DSA_set0_pqg(DSA *dsa, BIGNUM *p, BIGNUM *q, BIGNUM *g) {
int DSA_generate_parameters_ex(DSA *dsa, unsigned bits, const uint8_t *seed_in,
size_t seed_len, int *out_counter,
unsigned long *out_h, BN_GENCB *cb) {
if (bits > OPENSSL_DSA_MAX_MODULUS_BITS) {
OPENSSL_PUT_ERROR(DSA, DSA_R_INVALID_PARAMETERS);
return 0;
}
int ok = 0;
unsigned char seed[SHA256_DIGEST_LENGTH];
unsigned char md[SHA256_DIGEST_LENGTH];
@@ -274,6 +279,8 @@ int DSA_generate_parameters_ex(DSA *dsa, unsigned bits, const uint8_t *seed_in,
if (!RAND_bytes(seed, qsize)) {
goto err;
}
// DSA parameters are public.
CONSTTIME_DECLASSIFY(seed, qsize);
} else {
// If we come back through, use random seed next time.
seed_in = NULL;
@@ -477,11 +484,13 @@ DSA *DSAparams_dup(const DSA *dsa) {
}
int DSA_generate_key(DSA *dsa) {
int ok = 0;
BN_CTX *ctx = NULL;
BIGNUM *pub_key = NULL, *priv_key = NULL;
if (!dsa_check_key(dsa)) {
return 0;
}
ctx = BN_CTX_new();
int ok = 0;
BIGNUM *pub_key = NULL, *priv_key = NULL;
BN_CTX *ctx = BN_CTX_new();
if (ctx == NULL) {
goto err;
}
@@ -513,6 +522,9 @@ int DSA_generate_key(DSA *dsa) {
goto err;
}
// The public key is computed from the private key, but is public.
bn_declassify(pub_key);
dsa->priv_key = priv_key;
dsa->pub_key = pub_key;
ok = 1;
@@ -649,6 +661,10 @@ redo:
goto err;
}
// The signature is computed from the private key, but is public.
bn_declassify(r);
bn_declassify(s);
// Redo if r or s is zero as required by FIPS 186-3: this is
// very unlikely.
if (BN_is_zero(r) || BN_is_zero(s)) {
@@ -681,7 +697,7 @@ err:
return ret;
}
int DSA_do_verify(const uint8_t *digest, size_t digest_len, DSA_SIG *sig,
int DSA_do_verify(const uint8_t *digest, size_t digest_len, const DSA_SIG *sig,
const DSA *dsa) {
int valid;
if (!DSA_do_check_signature(&valid, digest, digest_len, sig, dsa)) {
@@ -691,7 +707,8 @@ int DSA_do_verify(const uint8_t *digest, size_t digest_len, DSA_SIG *sig,
}
int DSA_do_check_signature(int *out_valid, const uint8_t *digest,
size_t digest_len, DSA_SIG *sig, const DSA *dsa) {
size_t digest_len, const DSA_SIG *sig,
const DSA *dsa) {
*out_valid = 0;
if (!dsa_check_key(dsa)) {
return 0;
@@ -899,15 +916,19 @@ static int dsa_sign_setup(const DSA *dsa, BN_CTX *ctx, BIGNUM **out_kinv,
ctx) ||
// Compute r = (g^k mod p) mod q
!BN_mod_exp_mont_consttime(r, dsa->g, &k, dsa->p, ctx,
dsa->method_mont_p) ||
// Note |BN_mod| below is not constant-time and may leak information about
// |r|. |dsa->p| may be significantly larger than |dsa->q|, so this is not
// easily performed in constant-time with Montgomery reduction.
//
// However, |r| at this point is g^k (mod p). It is almost the value of
// |r| revealed in the signature anyway (g^k (mod p) (mod q)), going from
// it to |k| would require computing a discrete log.
!BN_mod(r, r, dsa->q, ctx) ||
dsa->method_mont_p)) {
OPENSSL_PUT_ERROR(DSA, ERR_R_BN_LIB);
goto err;
}
// Note |BN_mod| below is not constant-time and may leak information about
// |r|. |dsa->p| may be significantly larger than |dsa->q|, so this is not
// easily performed in constant-time with Montgomery reduction.
//
// However, |r| at this point is g^k (mod p). It is almost the value of |r|
// revealed in the signature anyway (g^k (mod p) (mod q)), going from it to
// |k| would require computing a discrete log.
bn_declassify(r);
if (!BN_mod(r, r, dsa->q, ctx) ||
// Compute part of 's = inv(k) (m + xr) mod q' using Fermat's Little
// Theorem.
!bn_mod_inverse_prime(kinv, &k, dsa->q, ctx, dsa->method_mont_q)) {
@@ -65,8 +65,6 @@
#include "../bytestring/internal.h"
#define OPENSSL_DSA_MAX_MODULUS_BITS 10000
// This function is in dsa_asn1.c rather than dsa.c because it is reachable from
// |EVP_PKEY| parsers. This makes it easier for the static linker to drop most
// of the DSA implementation.
@@ -119,8 +117,9 @@ int dsa_check_key(const DSA *dsa) {
if (dsa->priv_key != NULL) {
// The private key is a non-zero element of the scalar field, determined by
// |q|.
if (BN_is_negative(dsa->priv_key) || BN_is_zero(dsa->priv_key) ||
BN_cmp(dsa->priv_key, dsa->q) >= 0) {
if (BN_is_negative(dsa->priv_key) ||
constant_time_declassify_int(BN_is_zero(dsa->priv_key)) ||
constant_time_declassify_int(BN_cmp(dsa->priv_key, dsa->q) >= 0)) {
OPENSSL_PUT_ERROR(DSA, DSA_R_INVALID_PARAMETERS);
return 0;
}
@@ -478,6 +478,41 @@ int i2d_ECPrivateKey(const EC_KEY *key, uint8_t **outp) {
return CBB_finish_i2d(&cbb, outp);
}
EC_GROUP *d2i_ECPKParameters(EC_GROUP **out, const uint8_t **inp, long len) {
if (len < 0) {
return NULL;
}
CBS cbs;
CBS_init(&cbs, *inp, (size_t)len);
EC_GROUP *ret = EC_KEY_parse_parameters(&cbs);
if (ret == NULL) {
return NULL;
}
if (out != NULL) {
EC_GROUP_free(*out);
*out = ret;
}
*inp = CBS_data(&cbs);
return ret;
}
int i2d_ECPKParameters(const EC_GROUP *group, uint8_t **outp) {
if (group == NULL) {
OPENSSL_PUT_ERROR(EC, ERR_R_PASSED_NULL_PARAMETER);
return -1;
}
CBB cbb;
if (!CBB_init(&cbb, 0) || //
!EC_KEY_marshal_curve_name(&cbb, group)) {
CBB_cleanup(&cbb);
return -1;
}
return CBB_finish_i2d(&cbb, outp);
}
EC_KEY *d2i_ECParameters(EC_KEY **out_key, const uint8_t **inp, long len) {
if (len < 0) {
return NULL;
@@ -62,34 +62,87 @@
#include <CCryptoBoringSSL_mem.h>
#include "../bytestring/internal.h"
#include "../fipsmodule/ec/internal.h"
#include "../fipsmodule/ecdsa/internal.h"
#include "../internal.h"
static ECDSA_SIG *ecdsa_sig_from_fixed(const EC_KEY *key, const uint8_t *in,
size_t len) {
const EC_GROUP *group = EC_KEY_get0_group(key);
if (group == NULL) {
OPENSSL_PUT_ERROR(ECDSA, ERR_R_PASSED_NULL_PARAMETER);
return NULL;
}
size_t scalar_len = BN_num_bytes(EC_GROUP_get0_order(group));
if (len != 2 * scalar_len) {
OPENSSL_PUT_ERROR(ECDSA, ECDSA_R_BAD_SIGNATURE);
return NULL;
}
ECDSA_SIG *ret = ECDSA_SIG_new();
if (ret == NULL ||
!BN_bin2bn(in, scalar_len, ret->r) ||
!BN_bin2bn(in + scalar_len, scalar_len, ret->s)) {
ECDSA_SIG_free(ret);
return NULL;
}
return ret;
}
static int ecdsa_sig_to_fixed(const EC_KEY *key, uint8_t *out, size_t *out_len,
size_t max_out, const ECDSA_SIG *sig) {
const EC_GROUP *group = EC_KEY_get0_group(key);
if (group == NULL) {
OPENSSL_PUT_ERROR(ECDSA, ERR_R_PASSED_NULL_PARAMETER);
return 0;
}
size_t scalar_len = BN_num_bytes(EC_GROUP_get0_order(group));
if (max_out < 2 * scalar_len) {
OPENSSL_PUT_ERROR(EC, EC_R_BUFFER_TOO_SMALL);
return 0;
}
if (BN_is_negative(sig->r) ||
!BN_bn2bin_padded(out, scalar_len, sig->r) ||
BN_is_negative(sig->s) ||
!BN_bn2bin_padded(out + scalar_len, scalar_len, sig->s)) {
OPENSSL_PUT_ERROR(ECDSA, ECDSA_R_BAD_SIGNATURE);
return 0;
}
*out_len = 2 * scalar_len;
return 1;
}
int ECDSA_sign(int type, const uint8_t *digest, size_t digest_len, uint8_t *sig,
unsigned int *sig_len, const EC_KEY *eckey) {
unsigned int *out_sig_len, const EC_KEY *eckey) {
if (eckey->ecdsa_meth && eckey->ecdsa_meth->sign) {
return eckey->ecdsa_meth->sign(digest, digest_len, sig, sig_len,
return eckey->ecdsa_meth->sign(digest, digest_len, sig, out_sig_len,
(EC_KEY*) eckey /* cast away const */);
}
int ret = 0;
ECDSA_SIG *s = ECDSA_do_sign(digest, digest_len, eckey);
if (s == NULL) {
*sig_len = 0;
goto err;
*out_sig_len = 0;
uint8_t fixed[ECDSA_MAX_FIXED_LEN];
size_t fixed_len;
if (!ecdsa_sign_fixed(digest, digest_len, fixed, &fixed_len, sizeof(fixed),
eckey)) {
return 0;
}
// TODO(davidben): We can actually do better and go straight from the DER
// format to the fixed-width format without a malloc.
ECDSA_SIG *s = ecdsa_sig_from_fixed(eckey, fixed, fixed_len);
if (s == NULL) {
return 0;
}
int ret = 0;
CBB cbb;
CBB_init_fixed(&cbb, sig, ECDSA_size(eckey));
size_t len;
if (!ECDSA_SIG_marshal(&cbb, s) ||
!CBB_finish(&cbb, NULL, &len)) {
OPENSSL_PUT_ERROR(ECDSA, ECDSA_R_ENCODE_ERROR);
*sig_len = 0;
goto err;
}
*sig_len = (unsigned)len;
*out_sig_len = (unsigned)len;
ret = 1;
err:
@@ -99,12 +152,13 @@ err:
int ECDSA_verify(int type, const uint8_t *digest, size_t digest_len,
const uint8_t *sig, size_t sig_len, const EC_KEY *eckey) {
ECDSA_SIG *s;
// Decode the ECDSA signature.
//
// TODO(davidben): We can actually do better and go straight from the DER
// format to the fixed-width format without a malloc.
int ret = 0;
uint8_t *der = NULL;
// Decode the ECDSA signature.
s = ECDSA_SIG_from_bytes(sig, sig_len);
ECDSA_SIG *s = ECDSA_SIG_from_bytes(sig, sig_len);
if (s == NULL) {
goto err;
}
@@ -118,7 +172,10 @@ int ECDSA_verify(int type, const uint8_t *digest, size_t digest_len,
goto err;
}
ret = ECDSA_do_verify(digest, digest_len, s, eckey);
uint8_t fixed[ECDSA_MAX_FIXED_LEN];
size_t fixed_len;
ret = ecdsa_sig_to_fixed(eckey, fixed, &fixed_len, sizeof(fixed), s) &&
ecdsa_verify_fixed(digest, digest_len, fixed, fixed_len, eckey);
err:
OPENSSL_free(der);
@@ -147,6 +204,95 @@ size_t ECDSA_size(const EC_KEY *key) {
return ECDSA_SIG_max_len(group_order_size);
}
ECDSA_SIG *ECDSA_SIG_new(void) {
ECDSA_SIG *sig = OPENSSL_malloc(sizeof(ECDSA_SIG));
if (sig == NULL) {
return NULL;
}
sig->r = BN_new();
sig->s = BN_new();
if (sig->r == NULL || sig->s == NULL) {
ECDSA_SIG_free(sig);
return NULL;
}
return sig;
}
void ECDSA_SIG_free(ECDSA_SIG *sig) {
if (sig == NULL) {
return;
}
BN_free(sig->r);
BN_free(sig->s);
OPENSSL_free(sig);
}
const BIGNUM *ECDSA_SIG_get0_r(const ECDSA_SIG *sig) {
return sig->r;
}
const BIGNUM *ECDSA_SIG_get0_s(const ECDSA_SIG *sig) {
return sig->s;
}
void ECDSA_SIG_get0(const ECDSA_SIG *sig, const BIGNUM **out_r,
const BIGNUM **out_s) {
if (out_r != NULL) {
*out_r = sig->r;
}
if (out_s != NULL) {
*out_s = sig->s;
}
}
int ECDSA_SIG_set0(ECDSA_SIG *sig, BIGNUM *r, BIGNUM *s) {
if (r == NULL || s == NULL) {
return 0;
}
BN_free(sig->r);
BN_free(sig->s);
sig->r = r;
sig->s = s;
return 1;
}
int ECDSA_do_verify(const uint8_t *digest, size_t digest_len,
const ECDSA_SIG *sig, const EC_KEY *eckey) {
uint8_t fixed[ECDSA_MAX_FIXED_LEN];
size_t fixed_len;
return ecdsa_sig_to_fixed(eckey, fixed, &fixed_len, sizeof(fixed), sig) &&
ecdsa_verify_fixed(digest, digest_len, fixed, fixed_len, eckey);
}
// This function is only exported for testing and is not called in production
// code.
ECDSA_SIG *ECDSA_sign_with_nonce_and_leak_private_key_for_testing(
const uint8_t *digest, size_t digest_len, const EC_KEY *eckey,
const uint8_t *nonce, size_t nonce_len) {
uint8_t sig[ECDSA_MAX_FIXED_LEN];
size_t sig_len;
if (!ecdsa_sign_fixed_with_nonce_for_known_answer_test(
digest, digest_len, sig, &sig_len, sizeof(sig), eckey, nonce,
nonce_len)) {
return NULL;
}
return ecdsa_sig_from_fixed(eckey, sig, sig_len);
}
ECDSA_SIG *ECDSA_do_sign(const uint8_t *digest, size_t digest_len,
const EC_KEY *eckey) {
uint8_t sig[ECDSA_MAX_FIXED_LEN];
size_t sig_len;
if (!ecdsa_sign_fixed(digest, digest_len, sig, &sig_len, sizeof(sig),
eckey)) {
return NULL;
}
return ecdsa_sig_from_fixed(eckey, sig, sig_len);
}
ECDSA_SIG *ECDSA_SIG_parse(CBS *cbs) {
ECDSA_SIG *ret = ECDSA_SIG_new();
if (ret == NULL) {
+13 -10
View File
@@ -164,6 +164,17 @@ extern const uint32_t kOpenSSLReasonValues[];
extern const size_t kOpenSSLReasonValuesLen;
extern const char kOpenSSLReasonStringData[];
static char *strdup_libc_malloc(const char *str) {
// |strdup| is not in C until C23, so MSVC triggers deprecation warnings, and
// glibc and musl gate it on a feature macro. Reimplementing it is easier.
size_t len = strlen(str);
char *ret = malloc(len + 1);
if (ret != NULL) {
memcpy(ret, str, len + 1);
}
return ret;
}
// err_clear clears the given queued error.
static void err_clear(struct err_error_st *error) {
free(error->data);
@@ -174,13 +185,9 @@ static void err_copy(struct err_error_st *dst, const struct err_error_st *src) {
err_clear(dst);
dst->file = src->file;
if (src->data != NULL) {
// Disable deprecated functions on msvc so it doesn't complain about strdup.
OPENSSL_MSVC_PRAGMA(warning(push))
OPENSSL_MSVC_PRAGMA(warning(disable : 4996))
// We can't use OPENSSL_strdup because we don't want to call OPENSSL_malloc,
// which can affect the error stack.
dst->data = strdup(src->data);
OPENSSL_MSVC_PRAGMA(warning(pop))
dst->data = strdup_libc_malloc(src->data);
}
dst->packed = src->packed;
dst->line = src->line;
@@ -767,13 +774,9 @@ void ERR_set_error_data(char *data, int flags) {
assert(0);
return;
}
// Disable deprecated functions on msvc so it doesn't complain about strdup.
OPENSSL_MSVC_PRAGMA(warning(push))
OPENSSL_MSVC_PRAGMA(warning(disable : 4996))
// We can not use OPENSSL_strdup because we don't want to call OPENSSL_malloc,
// which can affect the error stack.
char *copy = strdup(data);
OPENSSL_MSVC_PRAGMA(warning(pop))
char *copy = strdup_libc_malloc(data);
if (copy != NULL) {
err_set_error_data(copy);
}
+41 -120
View File
@@ -59,12 +59,9 @@
#include <assert.h>
#include <string.h>
#include <CCryptoBoringSSL_dsa.h>
#include <CCryptoBoringSSL_ec.h>
#include <CCryptoBoringSSL_err.h>
#include <CCryptoBoringSSL_mem.h>
#include <CCryptoBoringSSL_nid.h>
#include <CCryptoBoringSSL_rsa.h>
#include <CCryptoBoringSSL_thread.h>
#include "internal.h"
@@ -149,9 +146,7 @@ int EVP_PKEY_cmp(const EVP_PKEY *a, const EVP_PKEY *b) {
int EVP_PKEY_copy_parameters(EVP_PKEY *to, const EVP_PKEY *from) {
if (to->type == EVP_PKEY_NONE) {
if (!EVP_PKEY_set_type(to, from->type)) {
return 0;
}
evp_pkey_set_method(to, from->ameth);
} else if (to->type != from->type) {
OPENSSL_PUT_ERROR(EVP, EVP_R_DIFFERENT_KEY_TYPES);
return 0;
@@ -225,117 +220,21 @@ static const EVP_PKEY_ASN1_METHOD *evp_pkey_asn1_find(int nid) {
}
}
static void evp_pkey_set_method(EVP_PKEY *pkey,
const EVP_PKEY_ASN1_METHOD *method) {
void evp_pkey_set_method(EVP_PKEY *pkey, const EVP_PKEY_ASN1_METHOD *method) {
free_it(pkey);
pkey->ameth = method;
pkey->type = pkey->ameth->pkey_id;
}
int EVP_PKEY_type(int nid) {
const EVP_PKEY_ASN1_METHOD *meth = evp_pkey_asn1_find(nid);
if (meth == NULL) {
return NID_undef;
}
return meth->pkey_id;
// In OpenSSL, this was used to map between type aliases. BoringSSL supports
// no type aliases, so this function is just the identity.
return nid;
}
int EVP_PKEY_set1_RSA(EVP_PKEY *pkey, RSA *key) {
if (EVP_PKEY_assign_RSA(pkey, key)) {
RSA_up_ref(key);
return 1;
}
return 0;
}
int EVP_PKEY_assign_RSA(EVP_PKEY *pkey, RSA *key) {
evp_pkey_set_method(pkey, &rsa_asn1_meth);
pkey->pkey = key;
return key != NULL;
}
RSA *EVP_PKEY_get0_RSA(const EVP_PKEY *pkey) {
if (pkey->type != EVP_PKEY_RSA) {
OPENSSL_PUT_ERROR(EVP, EVP_R_EXPECTING_AN_RSA_KEY);
return NULL;
}
return pkey->pkey;
}
RSA *EVP_PKEY_get1_RSA(const EVP_PKEY *pkey) {
RSA *rsa = EVP_PKEY_get0_RSA(pkey);
if (rsa != NULL) {
RSA_up_ref(rsa);
}
return rsa;
}
int EVP_PKEY_set1_DSA(EVP_PKEY *pkey, DSA *key) {
if (EVP_PKEY_assign_DSA(pkey, key)) {
DSA_up_ref(key);
return 1;
}
return 0;
}
int EVP_PKEY_assign_DSA(EVP_PKEY *pkey, DSA *key) {
evp_pkey_set_method(pkey, &dsa_asn1_meth);
pkey->pkey = key;
return key != NULL;
}
DSA *EVP_PKEY_get0_DSA(const EVP_PKEY *pkey) {
if (pkey->type != EVP_PKEY_DSA) {
OPENSSL_PUT_ERROR(EVP, EVP_R_EXPECTING_A_DSA_KEY);
return NULL;
}
return pkey->pkey;
}
DSA *EVP_PKEY_get1_DSA(const EVP_PKEY *pkey) {
DSA *dsa = EVP_PKEY_get0_DSA(pkey);
if (dsa != NULL) {
DSA_up_ref(dsa);
}
return dsa;
}
int EVP_PKEY_set1_EC_KEY(EVP_PKEY *pkey, EC_KEY *key) {
if (EVP_PKEY_assign_EC_KEY(pkey, key)) {
EC_KEY_up_ref(key);
return 1;
}
return 0;
}
int EVP_PKEY_assign_EC_KEY(EVP_PKEY *pkey, EC_KEY *key) {
evp_pkey_set_method(pkey, &ec_asn1_meth);
pkey->pkey = key;
return key != NULL;
}
EC_KEY *EVP_PKEY_get0_EC_KEY(const EVP_PKEY *pkey) {
if (pkey->type != EVP_PKEY_EC) {
OPENSSL_PUT_ERROR(EVP, EVP_R_EXPECTING_AN_EC_KEY_KEY);
return NULL;
}
return pkey->pkey;
}
EC_KEY *EVP_PKEY_get1_EC_KEY(const EVP_PKEY *pkey) {
EC_KEY *ec_key = EVP_PKEY_get0_EC_KEY(pkey);
if (ec_key != NULL) {
EC_KEY_up_ref(ec_key);
}
return ec_key;
}
DH *EVP_PKEY_get0_DH(const EVP_PKEY *pkey) { return NULL; }
DH *EVP_PKEY_get1_DH(const EVP_PKEY *pkey) { return NULL; }
int EVP_PKEY_assign(EVP_PKEY *pkey, int type, void *key) {
// This function can only be used to assign RSA, DSA, and EC keys. Other key
// types have internal representations which are not exposed through the
// This function can only be used to assign RSA, DSA, EC, and DH keys. Other
// key types have internal representations which are not exposed through the
// public API.
switch (type) {
case EVP_PKEY_RSA:
@@ -344,6 +243,8 @@ int EVP_PKEY_assign(EVP_PKEY *pkey, int type, void *key) {
return EVP_PKEY_assign_DSA(pkey, key);
case EVP_PKEY_EC:
return EVP_PKEY_assign_EC_KEY(pkey, key);
case EVP_PKEY_DH:
return EVP_PKEY_assign_DH(pkey, key);
}
OPENSSL_PUT_ERROR(EVP, EVP_R_UNSUPPORTED_ALGORITHM);
@@ -375,16 +276,26 @@ int EVP_PKEY_set_type(EVP_PKEY *pkey, int type) {
EVP_PKEY *EVP_PKEY_new_raw_private_key(int type, ENGINE *unused,
const uint8_t *in, size_t len) {
EVP_PKEY *ret = EVP_PKEY_new();
if (ret == NULL ||
!EVP_PKEY_set_type(ret, type)) {
goto err;
// To avoid pulling in all key types, look for specifically the key types that
// support |set_priv_raw|.
const EVP_PKEY_ASN1_METHOD *method;
switch (type) {
case EVP_PKEY_X25519:
method = &x25519_asn1_meth;
break;
case EVP_PKEY_ED25519:
method = &ed25519_asn1_meth;
break;
default:
OPENSSL_PUT_ERROR(EVP, EVP_R_UNSUPPORTED_ALGORITHM);
return 0;
}
if (ret->ameth->set_priv_raw == NULL) {
OPENSSL_PUT_ERROR(EVP, EVP_R_OPERATION_NOT_SUPPORTED_FOR_THIS_KEYTYPE);
EVP_PKEY *ret = EVP_PKEY_new();
if (ret == NULL) {
goto err;
}
evp_pkey_set_method(ret, method);
if (!ret->ameth->set_priv_raw(ret, in, len)) {
goto err;
@@ -399,16 +310,26 @@ err:
EVP_PKEY *EVP_PKEY_new_raw_public_key(int type, ENGINE *unused,
const uint8_t *in, size_t len) {
EVP_PKEY *ret = EVP_PKEY_new();
if (ret == NULL ||
!EVP_PKEY_set_type(ret, type)) {
goto err;
// To avoid pulling in all key types, look for specifically the key types that
// support |set_pub_raw|.
const EVP_PKEY_ASN1_METHOD *method;
switch (type) {
case EVP_PKEY_X25519:
method = &x25519_asn1_meth;
break;
case EVP_PKEY_ED25519:
method = &ed25519_asn1_meth;
break;
default:
OPENSSL_PUT_ERROR(EVP, EVP_R_UNSUPPORTED_ALGORITHM);
return 0;
}
if (ret->ameth->set_pub_raw == NULL) {
OPENSSL_PUT_ERROR(EVP, EVP_R_OPERATION_NOT_SUPPORTED_FOR_THIS_KEYTYPE);
EVP_PKEY *ret = EVP_PKEY_new();
if (ret == NULL) {
goto err;
}
evp_pkey_set_method(ret, method);
if (!ret->ameth->set_pub_raw(ret, in, len)) {
goto err;
+13 -13
View File
@@ -69,6 +69,7 @@
#include "../internal.h"
// We intentionally omit |dh_asn1_meth| from this list. It is not serializable.
static const EVP_PKEY_ASN1_METHOD *const kASN1Methods[] = {
&rsa_asn1_meth,
&ec_asn1_meth,
@@ -77,28 +78,26 @@ static const EVP_PKEY_ASN1_METHOD *const kASN1Methods[] = {
&x25519_asn1_meth,
};
static int parse_key_type(CBS *cbs, int *out_type) {
static const EVP_PKEY_ASN1_METHOD *parse_key_type(CBS *cbs) {
CBS oid;
if (!CBS_get_asn1(cbs, &oid, CBS_ASN1_OBJECT)) {
return 0;
return NULL;
}
for (unsigned i = 0; i < OPENSSL_ARRAY_SIZE(kASN1Methods); i++) {
const EVP_PKEY_ASN1_METHOD *method = kASN1Methods[i];
if (CBS_len(&oid) == method->oid_len &&
OPENSSL_memcmp(CBS_data(&oid), method->oid, method->oid_len) == 0) {
*out_type = method->pkey_id;
return 1;
return method;
}
}
return 0;
return NULL;
}
EVP_PKEY *EVP_parse_public_key(CBS *cbs) {
// Parse the SubjectPublicKeyInfo.
CBS spki, algorithm, key;
int type;
uint8_t padding;
if (!CBS_get_asn1(cbs, &spki, CBS_ASN1_SEQUENCE) ||
!CBS_get_asn1(&spki, &algorithm, CBS_ASN1_SEQUENCE) ||
@@ -107,7 +106,8 @@ EVP_PKEY *EVP_parse_public_key(CBS *cbs) {
OPENSSL_PUT_ERROR(EVP, EVP_R_DECODE_ERROR);
return NULL;
}
if (!parse_key_type(&algorithm, &type)) {
const EVP_PKEY_ASN1_METHOD *method = parse_key_type(&algorithm);
if (method == NULL) {
OPENSSL_PUT_ERROR(EVP, EVP_R_UNSUPPORTED_ALGORITHM);
return NULL;
}
@@ -121,10 +121,10 @@ EVP_PKEY *EVP_parse_public_key(CBS *cbs) {
// Set up an |EVP_PKEY| of the appropriate type.
EVP_PKEY *ret = EVP_PKEY_new();
if (ret == NULL ||
!EVP_PKEY_set_type(ret, type)) {
if (ret == NULL) {
goto err;
}
evp_pkey_set_method(ret, method);
// Call into the type-specific SPKI decoding function.
if (ret->ameth->pub_decode == NULL) {
@@ -155,7 +155,6 @@ EVP_PKEY *EVP_parse_private_key(CBS *cbs) {
// Parse the PrivateKeyInfo.
CBS pkcs8, algorithm, key;
uint64_t version;
int type;
if (!CBS_get_asn1(cbs, &pkcs8, CBS_ASN1_SEQUENCE) ||
!CBS_get_asn1_uint64(&pkcs8, &version) ||
version != 0 ||
@@ -164,7 +163,8 @@ EVP_PKEY *EVP_parse_private_key(CBS *cbs) {
OPENSSL_PUT_ERROR(EVP, EVP_R_DECODE_ERROR);
return NULL;
}
if (!parse_key_type(&algorithm, &type)) {
const EVP_PKEY_ASN1_METHOD *method = parse_key_type(&algorithm);
if (method == NULL) {
OPENSSL_PUT_ERROR(EVP, EVP_R_UNSUPPORTED_ALGORITHM);
return NULL;
}
@@ -173,10 +173,10 @@ EVP_PKEY *EVP_parse_private_key(CBS *cbs) {
// Set up an |EVP_PKEY| of the appropriate type.
EVP_PKEY *ret = EVP_PKEY_new();
if (ret == NULL ||
!EVP_PKEY_set_type(ret, type)) {
if (ret == NULL) {
goto err;
}
evp_pkey_set_method(ret, method);
// Call into the type-specific PrivateKeyInfo decoding function.
if (ret->ameth->priv_decode == NULL) {
@@ -213,6 +213,7 @@ OPENSSL_EXPORT int EVP_PKEY_CTX_ctrl(EVP_PKEY_CTX *ctx, int keytype, int optype,
#define EVP_PKEY_CTRL_HKDF_KEY (EVP_PKEY_ALG_CTRL + 16)
#define EVP_PKEY_CTRL_HKDF_SALT (EVP_PKEY_ALG_CTRL + 17)
#define EVP_PKEY_CTRL_HKDF_INFO (EVP_PKEY_ALG_CTRL + 18)
#define EVP_PKEY_CTRL_DH_PAD (EVP_PKEY_ALG_CTRL + 19)
struct evp_pkey_ctx_st {
// Method associated with this operation
@@ -288,12 +289,18 @@ extern const EVP_PKEY_ASN1_METHOD ec_asn1_meth;
extern const EVP_PKEY_ASN1_METHOD rsa_asn1_meth;
extern const EVP_PKEY_ASN1_METHOD ed25519_asn1_meth;
extern const EVP_PKEY_ASN1_METHOD x25519_asn1_meth;
extern const EVP_PKEY_ASN1_METHOD dh_asn1_meth;
extern const EVP_PKEY_METHOD rsa_pkey_meth;
extern const EVP_PKEY_METHOD ec_pkey_meth;
extern const EVP_PKEY_METHOD ed25519_pkey_meth;
extern const EVP_PKEY_METHOD x25519_pkey_meth;
extern const EVP_PKEY_METHOD hkdf_pkey_meth;
extern const EVP_PKEY_METHOD dh_pkey_meth;
// evp_pkey_set_method behaves like |EVP_PKEY_set_type|, but takes a pointer to
// a method table. This avoids depending on every |EVP_PKEY_ASN1_METHOD|.
void evp_pkey_set_method(EVP_PKEY *pkey, const EVP_PKEY_ASN1_METHOD *method);
#if defined(__cplusplus)
+137
View File
@@ -0,0 +1,137 @@
/*
* Copyright 2006-2019 The OpenSSL Project Authors. All Rights Reserved.
*
* Licensed under the OpenSSL license (the "License"). You may not use
* this file except in compliance with the License. You can obtain a copy
* in the file LICENSE in the source distribution or at
* https://www.openssl.org/source/license.html
*/
#include <CCryptoBoringSSL_evp.h>
#include <assert.h>
#include <CCryptoBoringSSL_dh.h>
#include <CCryptoBoringSSL_err.h>
#include <CCryptoBoringSSL_mem.h>
#include "internal.h"
typedef struct dh_pkey_ctx_st {
int pad;
} DH_PKEY_CTX;
static int pkey_dh_init(EVP_PKEY_CTX *ctx) {
DH_PKEY_CTX *dctx = OPENSSL_zalloc(sizeof(DH_PKEY_CTX));
if (dctx == NULL) {
return 0;
}
ctx->data = dctx;
return 1;
}
static int pkey_dh_copy(EVP_PKEY_CTX *dst, EVP_PKEY_CTX *src) {
if (!pkey_dh_init(dst)) {
return 0;
}
const DH_PKEY_CTX *sctx = src->data;
DH_PKEY_CTX *dctx = dst->data;
dctx->pad = sctx->pad;
return 1;
}
static void pkey_dh_cleanup(EVP_PKEY_CTX *ctx) {
OPENSSL_free(ctx->data);
ctx->data = NULL;
}
static int pkey_dh_keygen(EVP_PKEY_CTX *ctx, EVP_PKEY *pkey) {
DH *dh = DH_new();
if (dh == NULL || !EVP_PKEY_assign_DH(pkey, dh)) {
DH_free(dh);
return 0;
}
if (ctx->pkey != NULL && !EVP_PKEY_copy_parameters(pkey, ctx->pkey)) {
return 0;
}
return DH_generate_key(dh);
}
static int pkey_dh_derive(EVP_PKEY_CTX *ctx, uint8_t *out, size_t *out_len) {
DH_PKEY_CTX *dctx = ctx->data;
if (ctx->pkey == NULL || ctx->peerkey == NULL) {
OPENSSL_PUT_ERROR(EVP, EVP_R_KEYS_NOT_SET);
return 0;
}
DH *our_key = ctx->pkey->pkey;
DH *peer_key = ctx->peerkey->pkey;
if (our_key == NULL || peer_key == NULL) {
OPENSSL_PUT_ERROR(EVP, EVP_R_KEYS_NOT_SET);
return 0;
}
const BIGNUM *pub_key = DH_get0_pub_key(peer_key);
if (pub_key == NULL) {
OPENSSL_PUT_ERROR(EVP, EVP_R_KEYS_NOT_SET);
return 0;
}
if (out == NULL) {
*out_len = DH_size(our_key);
return 1;
}
if (*out_len < (size_t)DH_size(our_key)) {
OPENSSL_PUT_ERROR(EVP, EVP_R_BUFFER_TOO_SMALL);
return 0;
}
int ret = dctx->pad ? DH_compute_key_padded(out, pub_key, our_key)
: DH_compute_key(out, pub_key, our_key);
if (ret < 0) {
return 0;
}
assert(ret <= DH_size(our_key));
*out_len = (size_t)ret;
return 1;
}
static int pkey_dh_ctrl(EVP_PKEY_CTX *ctx, int type, int p1, void *p2) {
DH_PKEY_CTX *dctx = ctx->data;
switch (type) {
case EVP_PKEY_CTRL_PEER_KEY:
// |EVP_PKEY_derive_set_peer| requires the key implement this command,
// even if it is a no-op.
return 1;
case EVP_PKEY_CTRL_DH_PAD:
dctx->pad = p1;
return 1;
default:
OPENSSL_PUT_ERROR(EVP, EVP_R_COMMAND_NOT_SUPPORTED);
return 0;
}
}
const EVP_PKEY_METHOD dh_pkey_meth = {
.pkey_id = EVP_PKEY_DH,
.init = pkey_dh_init,
.copy = pkey_dh_copy,
.cleanup = pkey_dh_cleanup,
.keygen = pkey_dh_keygen,
.derive = pkey_dh_derive,
.ctrl = pkey_dh_ctrl,
};
int EVP_PKEY_CTX_set_dh_pad(EVP_PKEY_CTX *ctx, int pad) {
return EVP_PKEY_CTX_ctrl(ctx, EVP_PKEY_DH, EVP_PKEY_OP_DERIVE,
EVP_PKEY_CTRL_DH_PAD, pad, NULL);
}
@@ -0,0 +1,120 @@
/*
* Copyright 2006-2021 The OpenSSL Project Authors. All Rights Reserved.
*
* Licensed under the OpenSSL license (the "License"). You may not use
* this file except in compliance with the License. You can obtain a copy
* in the file LICENSE in the source distribution or at
* https://www.openssl.org/source/license.html
*/
#include <CCryptoBoringSSL_evp.h>
#include <CCryptoBoringSSL_bn.h>
#include <CCryptoBoringSSL_dh.h>
#include <CCryptoBoringSSL_err.h>
#include "internal.h"
#include "../internal.h"
static void dh_free(EVP_PKEY *pkey) {
DH_free(pkey->pkey);
pkey->pkey = NULL;
}
static int dh_size(const EVP_PKEY *pkey) { return DH_size(pkey->pkey); }
static int dh_bits(const EVP_PKEY *pkey) { return DH_bits(pkey->pkey); }
static int dh_param_missing(const EVP_PKEY *pkey) {
const DH *dh = pkey->pkey;
return dh == NULL || DH_get0_p(dh) == NULL || DH_get0_g(dh) == NULL;
}
static int dh_param_copy(EVP_PKEY *to, const EVP_PKEY *from) {
if (dh_param_missing(from)) {
OPENSSL_PUT_ERROR(EVP, EVP_R_MISSING_PARAMETERS);
return 0;
}
const DH *dh = from->pkey;
const BIGNUM *q_old = DH_get0_q(dh);
BIGNUM *p = BN_dup(DH_get0_p(dh));
BIGNUM *q = q_old == NULL ? NULL : BN_dup(q_old);
BIGNUM *g = BN_dup(DH_get0_g(dh));
if (p == NULL || (q_old != NULL && q == NULL) || g == NULL ||
!DH_set0_pqg(to->pkey, p, q, g)) {
BN_free(p);
BN_free(q);
BN_free(g);
return 0;
}
// |DH_set0_pqg| took ownership of |p|, |q|, and |g|.
return 1;
}
static int dh_param_cmp(const EVP_PKEY *a, const EVP_PKEY *b) {
if (dh_param_missing(a) || dh_param_missing(b)) {
return -2;
}
// Matching OpenSSL, only compare p and g for PKCS#3-style Diffie-Hellman.
// OpenSSL only checks q in X9.42-style Diffie-Hellman ("DHX").
const DH *a_dh = a->pkey;
const DH *b_dh = b->pkey;
return BN_cmp(DH_get0_p(a_dh), DH_get0_p(b_dh)) == 0 &&
BN_cmp(DH_get0_g(a_dh), DH_get0_g(b_dh)) == 0;
}
static int dh_pub_cmp(const EVP_PKEY *a, const EVP_PKEY *b) {
if (dh_param_cmp(a, b) <= 0) {
return 0;
}
const DH *a_dh = a->pkey;
const DH *b_dh = b->pkey;
return BN_cmp(DH_get0_pub_key(a_dh), DH_get0_pub_key(b_dh)) == 0;
}
const EVP_PKEY_ASN1_METHOD dh_asn1_meth = {
.pkey_id = EVP_PKEY_DH,
.pkey_method = &dh_pkey_meth,
.pub_cmp = dh_pub_cmp,
.pkey_size = dh_size,
.pkey_bits = dh_bits,
.param_missing = dh_param_missing,
.param_copy = dh_param_copy,
.param_cmp = dh_param_cmp,
.pkey_free = dh_free,
};
int EVP_PKEY_set1_DH(EVP_PKEY *pkey, DH *key) {
if (EVP_PKEY_assign_DH(pkey, key)) {
DH_up_ref(key);
return 1;
}
return 0;
}
int EVP_PKEY_assign_DH(EVP_PKEY *pkey, DH *key) {
evp_pkey_set_method(pkey, &dh_asn1_meth);
pkey->pkey = key;
return key != NULL;
}
DH *EVP_PKEY_get0_DH(const EVP_PKEY *pkey) {
if (pkey->type != EVP_PKEY_DH) {
OPENSSL_PUT_ERROR(EVP, EVP_R_EXPECTING_A_DH_KEY);
return NULL;
}
return pkey->pkey;
}
DH *EVP_PKEY_get1_DH(const EVP_PKEY *pkey) {
DH *dh = EVP_PKEY_get0_DH(pkey);
if (dh != NULL) {
DH_up_ref(dh);
}
return dh;
}
@@ -306,3 +306,33 @@ int EVP_PKEY_CTX_set_dsa_paramgen_q_bits(EVP_PKEY_CTX *ctx, int qbits) {
OPENSSL_PUT_ERROR(EVP, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
return 0;
}
int EVP_PKEY_set1_DSA(EVP_PKEY *pkey, DSA *key) {
if (EVP_PKEY_assign_DSA(pkey, key)) {
DSA_up_ref(key);
return 1;
}
return 0;
}
int EVP_PKEY_assign_DSA(EVP_PKEY *pkey, DSA *key) {
evp_pkey_set_method(pkey, &dsa_asn1_meth);
pkey->pkey = key;
return key != NULL;
}
DSA *EVP_PKEY_get0_DSA(const EVP_PKEY *pkey) {
if (pkey->type != EVP_PKEY_DSA) {
OPENSSL_PUT_ERROR(EVP, EVP_R_EXPECTING_A_DSA_KEY);
return NULL;
}
return pkey->pkey;
}
DSA *EVP_PKEY_get1_DSA(const EVP_PKEY *pkey) {
DSA *dsa = EVP_PKEY_get0_DSA(pkey);
if (dsa != NULL) {
DSA_up_ref(dsa);
}
return dsa;
}
+3 -4
View File
@@ -90,15 +90,14 @@ static int pkey_ec_init(EVP_PKEY_CTX *ctx) {
}
static int pkey_ec_copy(EVP_PKEY_CTX *dst, EVP_PKEY_CTX *src) {
EC_PKEY_CTX *dctx, *sctx;
if (!pkey_ec_init(dst)) {
return 0;
}
sctx = src->data;
dctx = dst->data;
const EC_PKEY_CTX *sctx = src->data;
EC_PKEY_CTX *dctx = dst->data;
dctx->md = sctx->md;
dctx->gen_group = sctx->gen_group;
return 1;
}
@@ -300,3 +300,33 @@ const EVP_PKEY_ASN1_METHOD ec_asn1_meth = {
int_ec_free,
};
int EVP_PKEY_set1_EC_KEY(EVP_PKEY *pkey, EC_KEY *key) {
if (EVP_PKEY_assign_EC_KEY(pkey, key)) {
EC_KEY_up_ref(key);
return 1;
}
return 0;
}
int EVP_PKEY_assign_EC_KEY(EVP_PKEY *pkey, EC_KEY *key) {
evp_pkey_set_method(pkey, &ec_asn1_meth);
pkey->pkey = key;
return key != NULL;
}
EC_KEY *EVP_PKEY_get0_EC_KEY(const EVP_PKEY *pkey) {
if (pkey->type != EVP_PKEY_EC) {
OPENSSL_PUT_ERROR(EVP, EVP_R_EXPECTING_AN_EC_KEY_KEY);
return NULL;
}
return pkey->pkey;
}
EC_KEY *EVP_PKEY_get1_EC_KEY(const EVP_PKEY *pkey) {
EC_KEY *ec_key = EVP_PKEY_get0_EC_KEY(pkey);
if (ec_key != NULL) {
EC_KEY_up_ref(ec_key);
}
return ec_key;
}
@@ -30,10 +30,7 @@ static int pkey_ed25519_keygen(EVP_PKEY_CTX *ctx, EVP_PKEY *pkey) {
return 0;
}
if (!EVP_PKEY_set_type(pkey, EVP_PKEY_ED25519)) {
OPENSSL_free(key);
return 0;
}
evp_pkey_set_method(pkey, &ed25519_asn1_meth);
uint8_t pubkey_unused[32];
ED25519_keypair(pubkey_unused, key->key);
@@ -209,3 +209,33 @@ const EVP_PKEY_ASN1_METHOD rsa_asn1_meth = {
int_rsa_free,
};
int EVP_PKEY_set1_RSA(EVP_PKEY *pkey, RSA *key) {
if (EVP_PKEY_assign_RSA(pkey, key)) {
RSA_up_ref(key);
return 1;
}
return 0;
}
int EVP_PKEY_assign_RSA(EVP_PKEY *pkey, RSA *key) {
evp_pkey_set_method(pkey, &rsa_asn1_meth);
pkey->pkey = key;
return key != NULL;
}
RSA *EVP_PKEY_get0_RSA(const EVP_PKEY *pkey) {
if (pkey->type != EVP_PKEY_RSA) {
OPENSSL_PUT_ERROR(EVP, EVP_R_EXPECTING_AN_RSA_KEY);
return NULL;
}
return pkey->pkey;
}
RSA *EVP_PKEY_get1_RSA(const EVP_PKEY *pkey) {
RSA *rsa = EVP_PKEY_get0_RSA(pkey);
if (rsa != NULL) {
RSA_up_ref(rsa);
}
return rsa;
}
@@ -30,10 +30,7 @@ static int pkey_x25519_keygen(EVP_PKEY_CTX *ctx, EVP_PKEY *pkey) {
return 0;
}
if (!EVP_PKEY_set_type(pkey, EVP_PKEY_X25519)) {
OPENSSL_free(key);
return 0;
}
evp_pkey_set_method(pkey, &x25519_asn1_meth);
X25519_keypair(key->pub, key->priv);
key->has_private = 1;
@@ -104,3 +104,24 @@ int AES_set_decrypt_key(const uint8_t *key, unsigned bits, AES_KEY *aeskey) {
return aes_nohw_set_decrypt_key(key, bits, aeskey);
}
}
#if defined(HWAES) && (defined(OPENSSL_X86) || defined(OPENSSL_X86_64))
// On x86 and x86_64, |aes_hw_set_decrypt_key|, we implement
// |aes_hw_encrypt_key_to_decrypt_key| in assembly and rely on C code to combine
// the operations.
int aes_hw_set_decrypt_key(const uint8_t *user_key, int bits, AES_KEY *key) {
int ret = aes_hw_set_encrypt_key(user_key, bits, key);
if (ret == 0) {
aes_hw_encrypt_key_to_decrypt_key(key);
}
return ret;
}
int aes_hw_set_encrypt_key(const uint8_t *user_key, int bits, AES_KEY *key) {
if (aes_hw_set_encrypt_key_alt_preferred()) {
return aes_hw_set_encrypt_key_alt(user_key, bits, key);
} else {
return aes_hw_set_encrypt_key_base(user_key, bits, key);
}
}
#endif
@@ -17,6 +17,8 @@
#include <stdlib.h>
#include <CCryptoBoringSSL_aes.h>
#include "../../internal.h"
#if defined(__cplusplus)
@@ -66,17 +68,41 @@ OPENSSL_INLINE int vpaes_capable(void) { return CRYPTO_is_NEON_capable(); }
#if defined(HWAES)
int aes_hw_set_encrypt_key(const uint8_t *user_key, const int bits,
AES_KEY *key);
int aes_hw_set_decrypt_key(const uint8_t *user_key, const int bits,
AES_KEY *key);
int aes_hw_set_encrypt_key(const uint8_t *user_key, int bits, AES_KEY *key);
int aes_hw_set_decrypt_key(const uint8_t *user_key, int bits, AES_KEY *key);
void aes_hw_encrypt(const uint8_t *in, uint8_t *out, const AES_KEY *key);
void aes_hw_decrypt(const uint8_t *in, uint8_t *out, const AES_KEY *key);
void aes_hw_cbc_encrypt(const uint8_t *in, uint8_t *out, size_t length,
const AES_KEY *key, uint8_t *ivec, const int enc);
const AES_KEY *key, uint8_t *ivec, int enc);
void aes_hw_ctr32_encrypt_blocks(const uint8_t *in, uint8_t *out, size_t len,
const AES_KEY *key, const uint8_t ivec[16]);
#if defined(OPENSSL_X86) || defined(OPENSSL_X86_64)
// On x86 and x86_64, |aes_hw_set_decrypt_key| is implemented in terms of
// |aes_hw_set_encrypt_key| and a conversion function.
void aes_hw_encrypt_key_to_decrypt_key(AES_KEY *key);
// There are two variants of this function, one which uses aeskeygenassist
// ("base") and one which uses aesenclast + pshufb ("alt"). aesenclast is
// overall faster but is slower on some older processors. It doesn't use AVX,
// but AVX is used as a proxy to detecting this. See
// https://groups.google.com/g/mailing.openssl.dev/c/OuFXwW4NfO8/m/7d2ZXVjkxVkJ
//
// TODO(davidben): It is unclear if the aeskeygenassist version is still
// worthwhile. However, the aesenclast version requires SSSE3. SSSE3 long
// predates AES-NI, but it's not clear if AES-NI implies SSSE3. In OpenSSL, the
// CCM AES-NI assembly seems to assume it does.
OPENSSL_INLINE int aes_hw_set_encrypt_key_alt_capable(void) {
return hwaes_capable() && CRYPTO_is_SSSE3_capable();
}
OPENSSL_INLINE int aes_hw_set_encrypt_key_alt_preferred(void) {
return hwaes_capable() && CRYPTO_is_AVX_capable();
}
int aes_hw_set_encrypt_key_base(const uint8_t *user_key, int bits,
AES_KEY *key);
int aes_hw_set_encrypt_key_alt(const uint8_t *user_key, int bits, AES_KEY *key);
#endif // OPENSSL_X86 || OPENSSL_X86_64
#else
// If HWAES isn't defined then we provide dummy functions for each of the hwaes
@@ -120,7 +146,7 @@ OPENSSL_INLINE void aes_hw_ctr32_encrypt_blocks(const uint8_t *in, uint8_t *out,
#if defined(HWAES_ECB)
void aes_hw_ecb_encrypt(const uint8_t *in, uint8_t *out, size_t length,
const AES_KEY *key, const int enc);
const AES_KEY *key, int enc);
#endif // HWAES_ECB
@@ -218,7 +244,7 @@ void aes_nohw_ctr32_encrypt_blocks(const uint8_t *in, uint8_t *out,
size_t blocks, const AES_KEY *key,
const uint8_t ivec[16]);
void aes_nohw_cbc_encrypt(const uint8_t *in, uint8_t *out, size_t len,
const AES_KEY *key, uint8_t *ivec, const int enc);
const AES_KEY *key, uint8_t *ivec, int enc);
#if defined(__cplusplus)
File diff suppressed because it is too large. Load diff
@@ -1,808 +0,0 @@
#define BORINGSSL_PREFIX CCryptoBoringSSL
#if defined(__arm__) && defined(__APPLE__)
// This file is generated from a similarly-named Perl script in the BoringSSL
// source tree. Do not edit by hand.
#include <CCryptoBoringSSL_asm_base.h>
#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_ARM) && defined(__APPLE__)
#include <CCryptoBoringSSL_arm_arch.h>
#if __ARM_MAX_ARCH__>=7
.text
.code 32
#undef __thumb2__
.align 5
Lrcon:
.long 0x01,0x01,0x01,0x01
.long 0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d @ rotate-n-splat
.long 0x1b,0x1b,0x1b,0x1b
.text
.globl _aes_hw_set_encrypt_key
.private_extern _aes_hw_set_encrypt_key
#ifdef __thumb2__
.thumb_func _aes_hw_set_encrypt_key
#endif
.align 5
_aes_hw_set_encrypt_key:
Lenc_key:
mov r3,#-1
cmp r0,#0
beq Lenc_key_abort
cmp r2,#0
beq Lenc_key_abort
mov r3,#-2
cmp r1,#128
blt Lenc_key_abort
cmp r1,#256
bgt Lenc_key_abort
tst r1,#0x3f
bne Lenc_key_abort
adr r3,Lrcon
cmp r1,#192
veor q0,q0,q0
vld1.8 {q3},[r0]!
mov r1,#8 @ reuse r1
vld1.32 {q1,q2},[r3]!
blt Loop128
beq L192
b L256
.align 4
Loop128:
vtbl.8 d20,{q3},d4
vtbl.8 d21,{q3},d5
vext.8 q9,q0,q3,#12
vst1.32 {q3},[r2]!
.byte 0x00,0x43,0xf0,0xf3 @ aese q10,q0
subs r1,r1,#1
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q10,q10,q1
veor q3,q3,q9
vshl.u8 q1,q1,#1
veor q3,q3,q10
bne Loop128
vld1.32 {q1},[r3]
vtbl.8 d20,{q3},d4
vtbl.8 d21,{q3},d5
vext.8 q9,q0,q3,#12
vst1.32 {q3},[r2]!
.byte 0x00,0x43,0xf0,0xf3 @ aese q10,q0
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q10,q10,q1
veor q3,q3,q9
vshl.u8 q1,q1,#1
veor q3,q3,q10
vtbl.8 d20,{q3},d4
vtbl.8 d21,{q3},d5
vext.8 q9,q0,q3,#12
vst1.32 {q3},[r2]!
.byte 0x00,0x43,0xf0,0xf3 @ aese q10,q0
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q10,q10,q1
veor q3,q3,q9
veor q3,q3,q10
vst1.32 {q3},[r2]
add r2,r2,#0x50
mov r12,#10
b Ldone
.align 4
L192:
vld1.8 {d16},[r0]!
vmov.i8 q10,#8 @ borrow q10
vst1.32 {q3},[r2]!
vsub.i8 q2,q2,q10 @ adjust the mask
Loop192:
vtbl.8 d20,{q8},d4
vtbl.8 d21,{q8},d5
vext.8 q9,q0,q3,#12
vst1.32 {d16},[r2]!
.byte 0x00,0x43,0xf0,0xf3 @ aese q10,q0
subs r1,r1,#1
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q3,q3,q9
vdup.32 q9,d7[1]
veor q9,q9,q8
veor q10,q10,q1
vext.8 q8,q0,q8,#12
vshl.u8 q1,q1,#1
veor q8,q8,q9
veor q3,q3,q10
veor q8,q8,q10
vst1.32 {q3},[r2]!
bne Loop192
mov r12,#12
add r2,r2,#0x20
b Ldone
.align 4
L256:
vld1.8 {q8},[r0]
mov r1,#7
mov r12,#14
vst1.32 {q3},[r2]!
Loop256:
vtbl.8 d20,{q8},d4
vtbl.8 d21,{q8},d5
vext.8 q9,q0,q3,#12
vst1.32 {q8},[r2]!
.byte 0x00,0x43,0xf0,0xf3 @ aese q10,q0
subs r1,r1,#1
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q3,q3,q9
vext.8 q9,q0,q9,#12
veor q10,q10,q1
veor q3,q3,q9
vshl.u8 q1,q1,#1
veor q3,q3,q10
vst1.32 {q3},[r2]!
beq Ldone
vdup.32 q10,d7[1]
vext.8 q9,q0,q8,#12
.byte 0x00,0x43,0xf0,0xf3 @ aese q10,q0
veor q8,q8,q9
vext.8 q9,q0,q9,#12
veor q8,q8,q9
vext.8 q9,q0,q9,#12
veor q8,q8,q9
veor q8,q8,q10
b Loop256
Ldone:
str r12,[r2]
mov r3,#0
Lenc_key_abort:
mov r0,r3 @ return value
bx lr
.globl _aes_hw_set_decrypt_key
.private_extern _aes_hw_set_decrypt_key
#ifdef __thumb2__
.thumb_func _aes_hw_set_decrypt_key
#endif
.align 5
_aes_hw_set_decrypt_key:
stmdb sp!,{r4,lr}
bl Lenc_key
cmp r0,#0
bne Ldec_key_abort
sub r2,r2,#240 @ restore original r2
mov r4,#-16
add r0,r2,r12,lsl#4 @ end of key schedule
vld1.32 {q0},[r2]
vld1.32 {q1},[r0]
vst1.32 {q0},[r0],r4
vst1.32 {q1},[r2]!
Loop_imc:
vld1.32 {q0},[r2]
vld1.32 {q1},[r0]
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
vst1.32 {q0},[r0],r4
vst1.32 {q1},[r2]!
cmp r0,r2
bhi Loop_imc
vld1.32 {q0},[r2]
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
vst1.32 {q0},[r0]
eor r0,r0,r0 @ return value
Ldec_key_abort:
ldmia sp!,{r4,pc}
.globl _aes_hw_encrypt
.private_extern _aes_hw_encrypt
#ifdef __thumb2__
.thumb_func _aes_hw_encrypt
#endif
.align 5
_aes_hw_encrypt:
AARCH64_VALID_CALL_TARGET
ldr r3,[r2,#240]
vld1.32 {q0},[r2]!
vld1.8 {q2},[r0]
sub r3,r3,#2
vld1.32 {q1},[r2]!
Loop_enc:
.byte 0x00,0x43,0xb0,0xf3 @ aese q2,q0
.byte 0x84,0x43,0xb0,0xf3 @ aesmc q2,q2
vld1.32 {q0},[r2]!
subs r3,r3,#2
.byte 0x02,0x43,0xb0,0xf3 @ aese q2,q1
.byte 0x84,0x43,0xb0,0xf3 @ aesmc q2,q2
vld1.32 {q1},[r2]!
bgt Loop_enc
.byte 0x00,0x43,0xb0,0xf3 @ aese q2,q0
.byte 0x84,0x43,0xb0,0xf3 @ aesmc q2,q2
vld1.32 {q0},[r2]
.byte 0x02,0x43,0xb0,0xf3 @ aese q2,q1
veor q2,q2,q0
vst1.8 {q2},[r1]
bx lr
.globl _aes_hw_decrypt
.private_extern _aes_hw_decrypt
#ifdef __thumb2__
.thumb_func _aes_hw_decrypt
#endif
.align 5
_aes_hw_decrypt:
AARCH64_VALID_CALL_TARGET
ldr r3,[r2,#240]
vld1.32 {q0},[r2]!
vld1.8 {q2},[r0]
sub r3,r3,#2
vld1.32 {q1},[r2]!
Loop_dec:
.byte 0x40,0x43,0xb0,0xf3 @ aesd q2,q0
.byte 0xc4,0x43,0xb0,0xf3 @ aesimc q2,q2
vld1.32 {q0},[r2]!
subs r3,r3,#2
.byte 0x42,0x43,0xb0,0xf3 @ aesd q2,q1
.byte 0xc4,0x43,0xb0,0xf3 @ aesimc q2,q2
vld1.32 {q1},[r2]!
bgt Loop_dec
.byte 0x40,0x43,0xb0,0xf3 @ aesd q2,q0
.byte 0xc4,0x43,0xb0,0xf3 @ aesimc q2,q2
vld1.32 {q0},[r2]
.byte 0x42,0x43,0xb0,0xf3 @ aesd q2,q1
veor q2,q2,q0
vst1.8 {q2},[r1]
bx lr
.globl _aes_hw_cbc_encrypt
.private_extern _aes_hw_cbc_encrypt
#ifdef __thumb2__
.thumb_func _aes_hw_cbc_encrypt
#endif
.align 5
_aes_hw_cbc_encrypt:
mov ip,sp
stmdb sp!,{r4,r5,r6,r7,r8,lr}
vstmdb sp!,{d8,d9,d10,d11,d12,d13,d14,d15} @ ABI specification says so
ldmia ip,{r4,r5} @ load remaining args
subs r2,r2,#16
mov r8,#16
blo Lcbc_abort
moveq r8,#0
cmp r5,#0 @ en- or decrypting?
ldr r5,[r3,#240]
and r2,r2,#-16
vld1.8 {q6},[r4]
vld1.8 {q0},[r0],r8
vld1.32 {q8,q9},[r3] @ load key schedule...
sub r5,r5,#6
add r7,r3,r5,lsl#4 @ pointer to last 7 round keys
sub r5,r5,#2
vld1.32 {q10,q11},[r7]!
vld1.32 {q12,q13},[r7]!
vld1.32 {q14,q15},[r7]!
vld1.32 {q7},[r7]
add r7,r3,#32
mov r6,r5
beq Lcbc_dec
cmp r5,#2
veor q0,q0,q6
veor q5,q8,q7
beq Lcbc_enc128
vld1.32 {q2,q3},[r7]
add r7,r3,#16
add r6,r3,#16*4
add r12,r3,#16*5
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
add r14,r3,#16*6
add r3,r3,#16*7
b Lenter_cbc_enc
.align 4
Loop_cbc_enc:
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vst1.8 {q6},[r1]!
Lenter_cbc_enc:
.byte 0x22,0x03,0xb0,0xf3 @ aese q0,q9
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x04,0x03,0xb0,0xf3 @ aese q0,q2
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vld1.32 {q8},[r6]
cmp r5,#4
.byte 0x06,0x03,0xb0,0xf3 @ aese q0,q3
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vld1.32 {q9},[r12]
beq Lcbc_enc192
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vld1.32 {q8},[r14]
.byte 0x22,0x03,0xb0,0xf3 @ aese q0,q9
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vld1.32 {q9},[r3]
nop
Lcbc_enc192:
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
subs r2,r2,#16
.byte 0x22,0x03,0xb0,0xf3 @ aese q0,q9
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
moveq r8,#0
.byte 0x24,0x03,0xb0,0xf3 @ aese q0,q10
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x26,0x03,0xb0,0xf3 @ aese q0,q11
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vld1.8 {q8},[r0],r8
.byte 0x28,0x03,0xb0,0xf3 @ aese q0,q12
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
veor q8,q8,q5
.byte 0x2a,0x03,0xb0,0xf3 @ aese q0,q13
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vld1.32 {q9},[r7] @ re-pre-load rndkey[1]
.byte 0x2c,0x03,0xb0,0xf3 @ aese q0,q14
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x2e,0x03,0xb0,0xf3 @ aese q0,q15
veor q6,q0,q7
bhs Loop_cbc_enc
vst1.8 {q6},[r1]!
b Lcbc_done
.align 5
Lcbc_enc128:
vld1.32 {q2,q3},[r7]
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
b Lenter_cbc_enc128
Loop_cbc_enc128:
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vst1.8 {q6},[r1]!
Lenter_cbc_enc128:
.byte 0x22,0x03,0xb0,0xf3 @ aese q0,q9
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
subs r2,r2,#16
.byte 0x04,0x03,0xb0,0xf3 @ aese q0,q2
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
moveq r8,#0
.byte 0x06,0x03,0xb0,0xf3 @ aese q0,q3
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x24,0x03,0xb0,0xf3 @ aese q0,q10
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x26,0x03,0xb0,0xf3 @ aese q0,q11
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
vld1.8 {q8},[r0],r8
.byte 0x28,0x03,0xb0,0xf3 @ aese q0,q12
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x2a,0x03,0xb0,0xf3 @ aese q0,q13
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x2c,0x03,0xb0,0xf3 @ aese q0,q14
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
veor q8,q8,q5
.byte 0x2e,0x03,0xb0,0xf3 @ aese q0,q15
veor q6,q0,q7
bhs Loop_cbc_enc128
vst1.8 {q6},[r1]!
b Lcbc_done
.align 5
Lcbc_dec:
vld1.8 {q10},[r0]!
subs r2,r2,#32 @ bias
add r6,r5,#2
vorr q3,q0,q0
vorr q1,q0,q0
vorr q11,q10,q10
blo Lcbc_dec_tail
vorr q1,q10,q10
vld1.8 {q10},[r0]!
vorr q2,q0,q0
vorr q3,q1,q1
vorr q11,q10,q10
Loop3x_cbc_dec:
.byte 0x60,0x03,0xb0,0xf3 @ aesd q0,q8
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
.byte 0x60,0x23,0xb0,0xf3 @ aesd q1,q8
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x60,0x43,0xf0,0xf3 @ aesd q10,q8
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
vld1.32 {q8},[r7]!
subs r6,r6,#2
.byte 0x62,0x03,0xb0,0xf3 @ aesd q0,q9
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
.byte 0x62,0x23,0xb0,0xf3 @ aesd q1,q9
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x62,0x43,0xf0,0xf3 @ aesd q10,q9
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
vld1.32 {q9},[r7]!
bgt Loop3x_cbc_dec
.byte 0x60,0x03,0xb0,0xf3 @ aesd q0,q8
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
.byte 0x60,0x23,0xb0,0xf3 @ aesd q1,q8
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x60,0x43,0xf0,0xf3 @ aesd q10,q8
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
veor q4,q6,q7
subs r2,r2,#0x30
veor q5,q2,q7
movlo r6,r2 @ r6, r6, is zero at this point
.byte 0x62,0x03,0xb0,0xf3 @ aesd q0,q9
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
.byte 0x62,0x23,0xb0,0xf3 @ aesd q1,q9
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x62,0x43,0xf0,0xf3 @ aesd q10,q9
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
veor q9,q3,q7
add r0,r0,r6 @ r0 is adjusted in such way that
@ at exit from the loop q1-q10
@ are loaded with last "words"
vorr q6,q11,q11
mov r7,r3
.byte 0x68,0x03,0xb0,0xf3 @ aesd q0,q12
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
.byte 0x68,0x23,0xb0,0xf3 @ aesd q1,q12
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x68,0x43,0xf0,0xf3 @ aesd q10,q12
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
vld1.8 {q2},[r0]!
.byte 0x6a,0x03,0xb0,0xf3 @ aesd q0,q13
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
.byte 0x6a,0x23,0xb0,0xf3 @ aesd q1,q13
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x6a,0x43,0xf0,0xf3 @ aesd q10,q13
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
vld1.8 {q3},[r0]!
.byte 0x6c,0x03,0xb0,0xf3 @ aesd q0,q14
.byte 0xc0,0x03,0xb0,0xf3 @ aesimc q0,q0
.byte 0x6c,0x23,0xb0,0xf3 @ aesd q1,q14
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x6c,0x43,0xf0,0xf3 @ aesd q10,q14
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
vld1.8 {q11},[r0]!
.byte 0x6e,0x03,0xb0,0xf3 @ aesd q0,q15
.byte 0x6e,0x23,0xb0,0xf3 @ aesd q1,q15
.byte 0x6e,0x43,0xf0,0xf3 @ aesd q10,q15
vld1.32 {q8},[r7]! @ re-pre-load rndkey[0]
add r6,r5,#2
veor q4,q4,q0
veor q5,q5,q1
veor q10,q10,q9
vld1.32 {q9},[r7]! @ re-pre-load rndkey[1]
vst1.8 {q4},[r1]!
vorr q0,q2,q2
vst1.8 {q5},[r1]!
vorr q1,q3,q3
vst1.8 {q10},[r1]!
vorr q10,q11,q11
bhs Loop3x_cbc_dec
cmn r2,#0x30
beq Lcbc_done
nop
Lcbc_dec_tail:
.byte 0x60,0x23,0xb0,0xf3 @ aesd q1,q8
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x60,0x43,0xf0,0xf3 @ aesd q10,q8
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
vld1.32 {q8},[r7]!
subs r6,r6,#2
.byte 0x62,0x23,0xb0,0xf3 @ aesd q1,q9
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x62,0x43,0xf0,0xf3 @ aesd q10,q9
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
vld1.32 {q9},[r7]!
bgt Lcbc_dec_tail
.byte 0x60,0x23,0xb0,0xf3 @ aesd q1,q8
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x60,0x43,0xf0,0xf3 @ aesd q10,q8
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
.byte 0x62,0x23,0xb0,0xf3 @ aesd q1,q9
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x62,0x43,0xf0,0xf3 @ aesd q10,q9
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
.byte 0x68,0x23,0xb0,0xf3 @ aesd q1,q12
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x68,0x43,0xf0,0xf3 @ aesd q10,q12
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
cmn r2,#0x20
.byte 0x6a,0x23,0xb0,0xf3 @ aesd q1,q13
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x6a,0x43,0xf0,0xf3 @ aesd q10,q13
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
veor q5,q6,q7
.byte 0x6c,0x23,0xb0,0xf3 @ aesd q1,q14
.byte 0xc2,0x23,0xb0,0xf3 @ aesimc q1,q1
.byte 0x6c,0x43,0xf0,0xf3 @ aesd q10,q14
.byte 0xe4,0x43,0xf0,0xf3 @ aesimc q10,q10
veor q9,q3,q7
.byte 0x6e,0x23,0xb0,0xf3 @ aesd q1,q15
.byte 0x6e,0x43,0xf0,0xf3 @ aesd q10,q15
beq Lcbc_dec_one
veor q5,q5,q1
veor q9,q9,q10
vorr q6,q11,q11
vst1.8 {q5},[r1]!
vst1.8 {q9},[r1]!
b Lcbc_done
Lcbc_dec_one:
veor q5,q5,q10
vorr q6,q11,q11
vst1.8 {q5},[r1]!
Lcbc_done:
vst1.8 {q6},[r4]
Lcbc_abort:
vldmia sp!,{d8,d9,d10,d11,d12,d13,d14,d15}
ldmia sp!,{r4,r5,r6,r7,r8,pc}
.globl _aes_hw_ctr32_encrypt_blocks
.private_extern _aes_hw_ctr32_encrypt_blocks
#ifdef __thumb2__
.thumb_func _aes_hw_ctr32_encrypt_blocks
#endif
.align 5
_aes_hw_ctr32_encrypt_blocks:
mov ip,sp
stmdb sp!,{r4,r5,r6,r7,r8,r9,r10,lr}
vstmdb sp!,{d8,d9,d10,d11,d12,d13,d14,d15} @ ABI specification says so
ldr r4, [ip] @ load remaining arg
ldr r5,[r3,#240]
ldr r8, [r4, #12]
vld1.32 {q0},[r4]
vld1.32 {q8,q9},[r3] @ load key schedule...
sub r5,r5,#4
mov r12,#16
cmp r2,#2
add r7,r3,r5,lsl#4 @ pointer to last 5 round keys
sub r5,r5,#2
vld1.32 {q12,q13},[r7]!
vld1.32 {q14,q15},[r7]!
vld1.32 {q7},[r7]
add r7,r3,#32
mov r6,r5
movlo r12,#0
@ ARM Cortex-A57 and Cortex-A72 cores running in 32-bit mode are
@ affected by silicon errata #1742098 [0] and #1655431 [1],
@ respectively, where the second instruction of an aese/aesmc
@ instruction pair may execute twice if an interrupt is taken right
@ after the first instruction consumes an input register of which a
@ single 32-bit lane has been updated the last time it was modified.
@
@ This function uses a counter in one 32-bit lane. The
@ could write to q1 and q10 directly, but that trips this bugs.
@ We write to q6 and copy to the final register as a workaround.
@
@ [0] ARM-EPM-049219 v23 Cortex-A57 MPCore Software Developers Errata Notice
@ [1] ARM-EPM-012079 v11.0 Cortex-A72 MPCore Software Developers Errata Notice
#ifndef __ARMEB__
rev r8, r8
#endif
add r10, r8, #1
vorr q6,q0,q0
rev r10, r10
vmov.32 d13[1],r10
add r8, r8, #2
vorr q1,q6,q6
bls Lctr32_tail
rev r12, r8
vmov.32 d13[1],r12
sub r2,r2,#3 @ bias
vorr q10,q6,q6
b Loop3x_ctr32
.align 4
Loop3x_ctr32:
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x20,0x23,0xb0,0xf3 @ aese q1,q8
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
.byte 0x20,0x43,0xf0,0xf3 @ aese q10,q8
.byte 0xa4,0x43,0xf0,0xf3 @ aesmc q10,q10
vld1.32 {q8},[r7]!
subs r6,r6,#2
.byte 0x22,0x03,0xb0,0xf3 @ aese q0,q9
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x22,0x23,0xb0,0xf3 @ aese q1,q9
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
.byte 0x22,0x43,0xf0,0xf3 @ aese q10,q9
.byte 0xa4,0x43,0xf0,0xf3 @ aesmc q10,q10
vld1.32 {q9},[r7]!
bgt Loop3x_ctr32
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x83,0xb0,0xf3 @ aesmc q4,q0
.byte 0x20,0x23,0xb0,0xf3 @ aese q1,q8
.byte 0x82,0xa3,0xb0,0xf3 @ aesmc q5,q1
vld1.8 {q2},[r0]!
add r9,r8,#1
.byte 0x20,0x43,0xf0,0xf3 @ aese q10,q8
.byte 0xa4,0x43,0xf0,0xf3 @ aesmc q10,q10
vld1.8 {q3},[r0]!
rev r9,r9
.byte 0x22,0x83,0xb0,0xf3 @ aese q4,q9
.byte 0x88,0x83,0xb0,0xf3 @ aesmc q4,q4
.byte 0x22,0xa3,0xb0,0xf3 @ aese q5,q9
.byte 0x8a,0xa3,0xb0,0xf3 @ aesmc q5,q5
vld1.8 {q11},[r0]!
mov r7,r3
.byte 0x22,0x43,0xf0,0xf3 @ aese q10,q9
.byte 0xa4,0x23,0xf0,0xf3 @ aesmc q9,q10
.byte 0x28,0x83,0xb0,0xf3 @ aese q4,q12
.byte 0x88,0x83,0xb0,0xf3 @ aesmc q4,q4
.byte 0x28,0xa3,0xb0,0xf3 @ aese q5,q12
.byte 0x8a,0xa3,0xb0,0xf3 @ aesmc q5,q5
veor q2,q2,q7
add r10,r8,#2
.byte 0x28,0x23,0xf0,0xf3 @ aese q9,q12
.byte 0xa2,0x23,0xf0,0xf3 @ aesmc q9,q9
veor q3,q3,q7
add r8,r8,#3
.byte 0x2a,0x83,0xb0,0xf3 @ aese q4,q13
.byte 0x88,0x83,0xb0,0xf3 @ aesmc q4,q4
.byte 0x2a,0xa3,0xb0,0xf3 @ aese q5,q13
.byte 0x8a,0xa3,0xb0,0xf3 @ aesmc q5,q5
@ Note the logic to update q0, q1, and q1 is written to work
@ around a bug in ARM Cortex-A57 and Cortex-A72 cores running in
@ 32-bit mode. See the comment above.
veor q11,q11,q7
vmov.32 d13[1], r9
.byte 0x2a,0x23,0xf0,0xf3 @ aese q9,q13
.byte 0xa2,0x23,0xf0,0xf3 @ aesmc q9,q9
vorr q0,q6,q6
rev r10,r10
.byte 0x2c,0x83,0xb0,0xf3 @ aese q4,q14
.byte 0x88,0x83,0xb0,0xf3 @ aesmc q4,q4
vmov.32 d13[1], r10
rev r12,r8
.byte 0x2c,0xa3,0xb0,0xf3 @ aese q5,q14
.byte 0x8a,0xa3,0xb0,0xf3 @ aesmc q5,q5
vorr q1,q6,q6
vmov.32 d13[1], r12
.byte 0x2c,0x23,0xf0,0xf3 @ aese q9,q14
.byte 0xa2,0x23,0xf0,0xf3 @ aesmc q9,q9
vorr q10,q6,q6
subs r2,r2,#3
.byte 0x2e,0x83,0xb0,0xf3 @ aese q4,q15
.byte 0x2e,0xa3,0xb0,0xf3 @ aese q5,q15
.byte 0x2e,0x23,0xf0,0xf3 @ aese q9,q15
veor q2,q2,q4
vld1.32 {q8},[r7]! @ re-pre-load rndkey[0]
vst1.8 {q2},[r1]!
veor q3,q3,q5
mov r6,r5
vst1.8 {q3},[r1]!
veor q11,q11,q9
vld1.32 {q9},[r7]! @ re-pre-load rndkey[1]
vst1.8 {q11},[r1]!
bhs Loop3x_ctr32
adds r2,r2,#3
beq Lctr32_done
cmp r2,#1
mov r12,#16
moveq r12,#0
Lctr32_tail:
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x20,0x23,0xb0,0xf3 @ aese q1,q8
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
vld1.32 {q8},[r7]!
subs r6,r6,#2
.byte 0x22,0x03,0xb0,0xf3 @ aese q0,q9
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x22,0x23,0xb0,0xf3 @ aese q1,q9
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
vld1.32 {q9},[r7]!
bgt Lctr32_tail
.byte 0x20,0x03,0xb0,0xf3 @ aese q0,q8
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x20,0x23,0xb0,0xf3 @ aese q1,q8
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
.byte 0x22,0x03,0xb0,0xf3 @ aese q0,q9
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x22,0x23,0xb0,0xf3 @ aese q1,q9
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
vld1.8 {q2},[r0],r12
.byte 0x28,0x03,0xb0,0xf3 @ aese q0,q12
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x28,0x23,0xb0,0xf3 @ aese q1,q12
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
vld1.8 {q3},[r0]
.byte 0x2a,0x03,0xb0,0xf3 @ aese q0,q13
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x2a,0x23,0xb0,0xf3 @ aese q1,q13
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
veor q2,q2,q7
.byte 0x2c,0x03,0xb0,0xf3 @ aese q0,q14
.byte 0x80,0x03,0xb0,0xf3 @ aesmc q0,q0
.byte 0x2c,0x23,0xb0,0xf3 @ aese q1,q14
.byte 0x82,0x23,0xb0,0xf3 @ aesmc q1,q1
veor q3,q3,q7
.byte 0x2e,0x03,0xb0,0xf3 @ aese q0,q15
.byte 0x2e,0x23,0xb0,0xf3 @ aese q1,q15
cmp r2,#1
veor q2,q2,q0
veor q3,q3,q1
vst1.8 {q2},[r1]!
beq Lctr32_done
vst1.8 {q3},[r1]
Lctr32_done:
vldmia sp!,{d8,d9,d10,d11,d12,d13,d14,d15}
ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,pc}
#endif
#endif // !OPENSSL_NO_ASM && defined(OPENSSL_ARM) && defined(__APPLE__)
#endif // defined(__arm__) && defined(__APPLE__)
#if defined(__linux__) && defined(__ELF__)
.section .note.GNU-stack,"",%progbits
#endif
@@ -1,950 +0,0 @@
#define BORINGSSL_PREFIX CCryptoBoringSSL
#if defined(__arm__) && defined(__APPLE__)
// This file is generated from a similarly-named Perl script in the BoringSSL
// source tree. Do not edit by hand.
#include <CCryptoBoringSSL_asm_base.h>
#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_ARM) && defined(__APPLE__)
#include <CCryptoBoringSSL_arm_arch.h>
@ Silence ARMv8 deprecated IT instruction warnings. This file is used by both
@ ARMv7 and ARMv8 processors and does not use ARMv8 instructions.
.text
#if defined(__thumb2__)
.syntax unified
.thumb
#else
.code 32
#endif
.globl _bn_mul_mont_nohw
.private_extern _bn_mul_mont_nohw
#ifdef __thumb2__
.thumb_func _bn_mul_mont_nohw
#endif
.align 5
_bn_mul_mont_nohw:
ldr ip,[sp,#4] @ load num
stmdb sp!,{r0,r2} @ sp points at argument block
cmp ip,#2
mov r0,ip @ load num
#ifdef __thumb2__
ittt lt
#endif
movlt r0,#0
addlt sp,sp,#2*4
blt Labrt
stmdb sp!,{r4,r5,r6,r7,r8,r9,r10,r11,r12,lr} @ save 10 registers
mov r0,r0,lsl#2 @ rescale r0 for byte count
sub sp,sp,r0 @ alloca(4*num)
sub sp,sp,#4 @ +extra dword
sub r0,r0,#4 @ "num=num-1"
add r4,r2,r0 @ &bp[num-1]
add r0,sp,r0 @ r0 to point at &tp[num-1]
ldr r8,[r0,#14*4] @ &n0
ldr r2,[r2] @ bp[0]
ldr r5,[r1],#4 @ ap[0],ap++
ldr r6,[r3],#4 @ np[0],np++
ldr r8,[r8] @ *n0
str r4,[r0,#15*4] @ save &bp[num]
umull r10,r11,r5,r2 @ ap[0]*bp[0]
str r8,[r0,#14*4] @ save n0 value
mul r8,r10,r8 @ "tp[0]"*n0
mov r12,#0
umlal r10,r12,r6,r8 @ np[0]*n0+"t[0]"
mov r4,sp
L1st:
ldr r5,[r1],#4 @ ap[j],ap++
mov r10,r11
ldr r6,[r3],#4 @ np[j],np++
mov r11,#0
umlal r10,r11,r5,r2 @ ap[j]*bp[0]
mov r14,#0
umlal r12,r14,r6,r8 @ np[j]*n0
adds r12,r12,r10
str r12,[r4],#4 @ tp[j-1]=,tp++
adc r12,r14,#0
cmp r4,r0
bne L1st
adds r12,r12,r11
ldr r4,[r0,#13*4] @ restore bp
mov r14,#0
ldr r8,[r0,#14*4] @ restore n0
adc r14,r14,#0
str r12,[r0] @ tp[num-1]=
mov r7,sp
str r14,[r0,#4] @ tp[num]=
Louter:
sub r7,r0,r7 @ "original" r0-1 value
sub r1,r1,r7 @ "rewind" ap to &ap[1]
ldr r2,[r4,#4]! @ *(++bp)
sub r3,r3,r7 @ "rewind" np to &np[1]
ldr r5,[r1,#-4] @ ap[0]
ldr r10,[sp] @ tp[0]
ldr r6,[r3,#-4] @ np[0]
ldr r7,[sp,#4] @ tp[1]
mov r11,#0
umlal r10,r11,r5,r2 @ ap[0]*bp[i]+tp[0]
str r4,[r0,#13*4] @ save bp
mul r8,r10,r8
mov r12,#0
umlal r10,r12,r6,r8 @ np[0]*n0+"tp[0]"
mov r4,sp
Linner:
ldr r5,[r1],#4 @ ap[j],ap++
adds r10,r11,r7 @ +=tp[j]
ldr r6,[r3],#4 @ np[j],np++
mov r11,#0
umlal r10,r11,r5,r2 @ ap[j]*bp[i]
mov r14,#0
umlal r12,r14,r6,r8 @ np[j]*n0
adc r11,r11,#0
ldr r7,[r4,#8] @ tp[j+1]
adds r12,r12,r10
str r12,[r4],#4 @ tp[j-1]=,tp++
adc r12,r14,#0
cmp r4,r0
bne Linner
adds r12,r12,r11
mov r14,#0
ldr r4,[r0,#13*4] @ restore bp
adc r14,r14,#0
ldr r8,[r0,#14*4] @ restore n0
adds r12,r12,r7
ldr r7,[r0,#15*4] @ restore &bp[num]
adc r14,r14,#0
str r12,[r0] @ tp[num-1]=
str r14,[r0,#4] @ tp[num]=
cmp r4,r7
#ifdef __thumb2__
itt ne
#endif
movne r7,sp
bne Louter
ldr r2,[r0,#12*4] @ pull rp
mov r5,sp
add r0,r0,#4 @ r0 to point at &tp[num]
sub r5,r0,r5 @ "original" num value
mov r4,sp @ "rewind" r4
mov r1,r4 @ "borrow" r1
sub r3,r3,r5 @ "rewind" r3 to &np[0]
subs r7,r7,r7 @ "clear" carry flag
Lsub: ldr r7,[r4],#4
ldr r6,[r3],#4
sbcs r7,r7,r6 @ tp[j]-np[j]
str r7,[r2],#4 @ rp[j]=
teq r4,r0 @ preserve carry
bne Lsub
sbcs r14,r14,#0 @ upmost carry
mov r4,sp @ "rewind" r4
sub r2,r2,r5 @ "rewind" r2
Lcopy: ldr r7,[r4] @ conditional copy
ldr r5,[r2]
str sp,[r4],#4 @ zap tp
#ifdef __thumb2__
it cc
#endif
movcc r5,r7
str r5,[r2],#4
teq r4,r0 @ preserve carry
bne Lcopy
mov sp,r0
add sp,sp,#4 @ skip over tp[num+1]
ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11,r12,lr} @ restore registers
add sp,sp,#2*4 @ skip over {r0,r2}
mov r0,#1
Labrt:
#if __ARM_ARCH>=5
bx lr @ bx lr
#else
tst lr,#1
moveq pc,lr @ be binary compatible with V4, yet
.word 0xe12fff1e @ interoperable with Thumb ISA:-)
#endif
#if __ARM_MAX_ARCH__>=7
.globl _bn_mul8x_mont_neon
.private_extern _bn_mul8x_mont_neon
#ifdef __thumb2__
.thumb_func _bn_mul8x_mont_neon
#endif
.align 5
_bn_mul8x_mont_neon:
mov ip,sp
stmdb sp!,{r4,r5,r6,r7,r8,r9,r10,r11}
vstmdb sp!,{d8,d9,d10,d11,d12,d13,d14,d15} @ ABI specification says so
ldmia ip,{r4,r5} @ load rest of parameter block
mov ip,sp
cmp r5,#8
bhi LNEON_8n
@ special case for r5==8, everything is in register bank...
vld1.32 {d28[0]}, [r2,:32]!
veor d8,d8,d8
sub r7,sp,r5,lsl#4
vld1.32 {d0,d1,d2,d3}, [r1]! @ can't specify :32 :-(
and r7,r7,#-64
vld1.32 {d30[0]}, [r4,:32]
mov sp,r7 @ alloca
vzip.16 d28,d8
vmull.u32 q6,d28,d0[0]
vmull.u32 q7,d28,d0[1]
vmull.u32 q8,d28,d1[0]
vshl.i64 d29,d13,#16
vmull.u32 q9,d28,d1[1]
vadd.u64 d29,d29,d12
veor d8,d8,d8
vmul.u32 d29,d29,d30
vmull.u32 q10,d28,d2[0]
vld1.32 {d4,d5,d6,d7}, [r3]!
vmull.u32 q11,d28,d2[1]
vmull.u32 q12,d28,d3[0]
vzip.16 d29,d8
vmull.u32 q13,d28,d3[1]
vmlal.u32 q6,d29,d4[0]
sub r9,r5,#1
vmlal.u32 q7,d29,d4[1]
vmlal.u32 q8,d29,d5[0]
vmlal.u32 q9,d29,d5[1]
vmlal.u32 q10,d29,d6[0]
vmov q5,q6
vmlal.u32 q11,d29,d6[1]
vmov q6,q7
vmlal.u32 q12,d29,d7[0]
vmov q7,q8
vmlal.u32 q13,d29,d7[1]
vmov q8,q9
vmov q9,q10
vshr.u64 d10,d10,#16
vmov q10,q11
vmov q11,q12
vadd.u64 d10,d10,d11
vmov q12,q13
veor q13,q13
vshr.u64 d10,d10,#16
b LNEON_outer8
.align 4
LNEON_outer8:
vld1.32 {d28[0]}, [r2,:32]!
veor d8,d8,d8
vzip.16 d28,d8
vadd.u64 d12,d12,d10
vmlal.u32 q6,d28,d0[0]
vmlal.u32 q7,d28,d0[1]
vmlal.u32 q8,d28,d1[0]
vshl.i64 d29,d13,#16
vmlal.u32 q9,d28,d1[1]
vadd.u64 d29,d29,d12
veor d8,d8,d8
subs r9,r9,#1
vmul.u32 d29,d29,d30
vmlal.u32 q10,d28,d2[0]
vmlal.u32 q11,d28,d2[1]
vmlal.u32 q12,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q13,d28,d3[1]
vmlal.u32 q6,d29,d4[0]
vmlal.u32 q7,d29,d4[1]
vmlal.u32 q8,d29,d5[0]
vmlal.u32 q9,d29,d5[1]
vmlal.u32 q10,d29,d6[0]
vmov q5,q6
vmlal.u32 q11,d29,d6[1]
vmov q6,q7
vmlal.u32 q12,d29,d7[0]
vmov q7,q8
vmlal.u32 q13,d29,d7[1]
vmov q8,q9
vmov q9,q10
vshr.u64 d10,d10,#16
vmov q10,q11
vmov q11,q12
vadd.u64 d10,d10,d11
vmov q12,q13
veor q13,q13
vshr.u64 d10,d10,#16
bne LNEON_outer8
vadd.u64 d12,d12,d10
mov r7,sp
vshr.u64 d10,d12,#16
mov r8,r5
vadd.u64 d13,d13,d10
add r6,sp,#96
vshr.u64 d10,d13,#16
vzip.16 d12,d13
b LNEON_tail_entry
.align 4
LNEON_8n:
veor q6,q6,q6
sub r7,sp,#128
veor q7,q7,q7
sub r7,r7,r5,lsl#4
veor q8,q8,q8
and r7,r7,#-64
veor q9,q9,q9
mov sp,r7 @ alloca
veor q10,q10,q10
add r7,r7,#256
veor q11,q11,q11
sub r8,r5,#8
veor q12,q12,q12
veor q13,q13,q13
LNEON_8n_init:
vst1.64 {q6,q7},[r7,:256]!
subs r8,r8,#8
vst1.64 {q8,q9},[r7,:256]!
vst1.64 {q10,q11},[r7,:256]!
vst1.64 {q12,q13},[r7,:256]!
bne LNEON_8n_init
add r6,sp,#256
vld1.32 {d0,d1,d2,d3},[r1]!
add r10,sp,#8
vld1.32 {d30[0]},[r4,:32]
mov r9,r5
b LNEON_8n_outer
.align 4
LNEON_8n_outer:
vld1.32 {d28[0]},[r2,:32]! @ *b++
veor d8,d8,d8
vzip.16 d28,d8
add r7,sp,#128
vld1.32 {d4,d5,d6,d7},[r3]!
vmlal.u32 q6,d28,d0[0]
vmlal.u32 q7,d28,d0[1]
veor d8,d8,d8
vmlal.u32 q8,d28,d1[0]
vshl.i64 d29,d13,#16
vmlal.u32 q9,d28,d1[1]
vadd.u64 d29,d29,d12
vmlal.u32 q10,d28,d2[0]
vmul.u32 d29,d29,d30
vmlal.u32 q11,d28,d2[1]
vst1.32 {d28},[sp,:64] @ put aside smashed b[8*i+0]
vmlal.u32 q12,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q13,d28,d3[1]
vld1.32 {d28[0]},[r2,:32]! @ *b++
vmlal.u32 q6,d29,d4[0]
veor d10,d10,d10
vmlal.u32 q7,d29,d4[1]
vzip.16 d28,d10
vmlal.u32 q8,d29,d5[0]
vshr.u64 d12,d12,#16
vmlal.u32 q9,d29,d5[1]
vmlal.u32 q10,d29,d6[0]
vadd.u64 d12,d12,d13
vmlal.u32 q11,d29,d6[1]
vshr.u64 d12,d12,#16
vmlal.u32 q12,d29,d7[0]
vmlal.u32 q13,d29,d7[1]
vadd.u64 d14,d14,d12
vst1.32 {d29},[r10,:64]! @ put aside smashed m[8*i+0]
vmlal.u32 q7,d28,d0[0]
vld1.64 {q6},[r6,:128]!
vmlal.u32 q8,d28,d0[1]
veor d8,d8,d8
vmlal.u32 q9,d28,d1[0]
vshl.i64 d29,d15,#16
vmlal.u32 q10,d28,d1[1]
vadd.u64 d29,d29,d14
vmlal.u32 q11,d28,d2[0]
vmul.u32 d29,d29,d30
vmlal.u32 q12,d28,d2[1]
vst1.32 {d28},[r10,:64]! @ put aside smashed b[8*i+1]
vmlal.u32 q13,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q6,d28,d3[1]
vld1.32 {d28[0]},[r2,:32]! @ *b++
vmlal.u32 q7,d29,d4[0]
veor d10,d10,d10
vmlal.u32 q8,d29,d4[1]
vzip.16 d28,d10
vmlal.u32 q9,d29,d5[0]
vshr.u64 d14,d14,#16
vmlal.u32 q10,d29,d5[1]
vmlal.u32 q11,d29,d6[0]
vadd.u64 d14,d14,d15
vmlal.u32 q12,d29,d6[1]
vshr.u64 d14,d14,#16
vmlal.u32 q13,d29,d7[0]
vmlal.u32 q6,d29,d7[1]
vadd.u64 d16,d16,d14
vst1.32 {d29},[r10,:64]! @ put aside smashed m[8*i+1]
vmlal.u32 q8,d28,d0[0]
vld1.64 {q7},[r6,:128]!
vmlal.u32 q9,d28,d0[1]
veor d8,d8,d8
vmlal.u32 q10,d28,d1[0]
vshl.i64 d29,d17,#16
vmlal.u32 q11,d28,d1[1]
vadd.u64 d29,d29,d16
vmlal.u32 q12,d28,d2[0]
vmul.u32 d29,d29,d30
vmlal.u32 q13,d28,d2[1]
vst1.32 {d28},[r10,:64]! @ put aside smashed b[8*i+2]
vmlal.u32 q6,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q7,d28,d3[1]
vld1.32 {d28[0]},[r2,:32]! @ *b++
vmlal.u32 q8,d29,d4[0]
veor d10,d10,d10
vmlal.u32 q9,d29,d4[1]
vzip.16 d28,d10
vmlal.u32 q10,d29,d5[0]
vshr.u64 d16,d16,#16
vmlal.u32 q11,d29,d5[1]
vmlal.u32 q12,d29,d6[0]
vadd.u64 d16,d16,d17
vmlal.u32 q13,d29,d6[1]
vshr.u64 d16,d16,#16
vmlal.u32 q6,d29,d7[0]
vmlal.u32 q7,d29,d7[1]
vadd.u64 d18,d18,d16
vst1.32 {d29},[r10,:64]! @ put aside smashed m[8*i+2]
vmlal.u32 q9,d28,d0[0]
vld1.64 {q8},[r6,:128]!
vmlal.u32 q10,d28,d0[1]
veor d8,d8,d8
vmlal.u32 q11,d28,d1[0]
vshl.i64 d29,d19,#16
vmlal.u32 q12,d28,d1[1]
vadd.u64 d29,d29,d18
vmlal.u32 q13,d28,d2[0]
vmul.u32 d29,d29,d30
vmlal.u32 q6,d28,d2[1]
vst1.32 {d28},[r10,:64]! @ put aside smashed b[8*i+3]
vmlal.u32 q7,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q8,d28,d3[1]
vld1.32 {d28[0]},[r2,:32]! @ *b++
vmlal.u32 q9,d29,d4[0]
veor d10,d10,d10
vmlal.u32 q10,d29,d4[1]
vzip.16 d28,d10
vmlal.u32 q11,d29,d5[0]
vshr.u64 d18,d18,#16
vmlal.u32 q12,d29,d5[1]
vmlal.u32 q13,d29,d6[0]
vadd.u64 d18,d18,d19
vmlal.u32 q6,d29,d6[1]
vshr.u64 d18,d18,#16
vmlal.u32 q7,d29,d7[0]
vmlal.u32 q8,d29,d7[1]
vadd.u64 d20,d20,d18
vst1.32 {d29},[r10,:64]! @ put aside smashed m[8*i+3]
vmlal.u32 q10,d28,d0[0]
vld1.64 {q9},[r6,:128]!
vmlal.u32 q11,d28,d0[1]
veor d8,d8,d8
vmlal.u32 q12,d28,d1[0]
vshl.i64 d29,d21,#16
vmlal.u32 q13,d28,d1[1]
vadd.u64 d29,d29,d20
vmlal.u32 q6,d28,d2[0]
vmul.u32 d29,d29,d30
vmlal.u32 q7,d28,d2[1]
vst1.32 {d28},[r10,:64]! @ put aside smashed b[8*i+4]
vmlal.u32 q8,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q9,d28,d3[1]
vld1.32 {d28[0]},[r2,:32]! @ *b++
vmlal.u32 q10,d29,d4[0]
veor d10,d10,d10
vmlal.u32 q11,d29,d4[1]
vzip.16 d28,d10
vmlal.u32 q12,d29,d5[0]
vshr.u64 d20,d20,#16
vmlal.u32 q13,d29,d5[1]
vmlal.u32 q6,d29,d6[0]
vadd.u64 d20,d20,d21
vmlal.u32 q7,d29,d6[1]
vshr.u64 d20,d20,#16
vmlal.u32 q8,d29,d7[0]
vmlal.u32 q9,d29,d7[1]
vadd.u64 d22,d22,d20
vst1.32 {d29},[r10,:64]! @ put aside smashed m[8*i+4]
vmlal.u32 q11,d28,d0[0]
vld1.64 {q10},[r6,:128]!
vmlal.u32 q12,d28,d0[1]
veor d8,d8,d8
vmlal.u32 q13,d28,d1[0]
vshl.i64 d29,d23,#16
vmlal.u32 q6,d28,d1[1]
vadd.u64 d29,d29,d22
vmlal.u32 q7,d28,d2[0]
vmul.u32 d29,d29,d30
vmlal.u32 q8,d28,d2[1]
vst1.32 {d28},[r10,:64]! @ put aside smashed b[8*i+5]
vmlal.u32 q9,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q10,d28,d3[1]
vld1.32 {d28[0]},[r2,:32]! @ *b++
vmlal.u32 q11,d29,d4[0]
veor d10,d10,d10
vmlal.u32 q12,d29,d4[1]
vzip.16 d28,d10
vmlal.u32 q13,d29,d5[0]
vshr.u64 d22,d22,#16
vmlal.u32 q6,d29,d5[1]
vmlal.u32 q7,d29,d6[0]
vadd.u64 d22,d22,d23
vmlal.u32 q8,d29,d6[1]
vshr.u64 d22,d22,#16
vmlal.u32 q9,d29,d7[0]
vmlal.u32 q10,d29,d7[1]
vadd.u64 d24,d24,d22
vst1.32 {d29},[r10,:64]! @ put aside smashed m[8*i+5]
vmlal.u32 q12,d28,d0[0]
vld1.64 {q11},[r6,:128]!
vmlal.u32 q13,d28,d0[1]
veor d8,d8,d8
vmlal.u32 q6,d28,d1[0]
vshl.i64 d29,d25,#16
vmlal.u32 q7,d28,d1[1]
vadd.u64 d29,d29,d24
vmlal.u32 q8,d28,d2[0]
vmul.u32 d29,d29,d30
vmlal.u32 q9,d28,d2[1]
vst1.32 {d28},[r10,:64]! @ put aside smashed b[8*i+6]
vmlal.u32 q10,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q11,d28,d3[1]
vld1.32 {d28[0]},[r2,:32]! @ *b++
vmlal.u32 q12,d29,d4[0]
veor d10,d10,d10
vmlal.u32 q13,d29,d4[1]
vzip.16 d28,d10
vmlal.u32 q6,d29,d5[0]
vshr.u64 d24,d24,#16
vmlal.u32 q7,d29,d5[1]
vmlal.u32 q8,d29,d6[0]
vadd.u64 d24,d24,d25
vmlal.u32 q9,d29,d6[1]
vshr.u64 d24,d24,#16
vmlal.u32 q10,d29,d7[0]
vmlal.u32 q11,d29,d7[1]
vadd.u64 d26,d26,d24
vst1.32 {d29},[r10,:64]! @ put aside smashed m[8*i+6]
vmlal.u32 q13,d28,d0[0]
vld1.64 {q12},[r6,:128]!
vmlal.u32 q6,d28,d0[1]
veor d8,d8,d8
vmlal.u32 q7,d28,d1[0]
vshl.i64 d29,d27,#16
vmlal.u32 q8,d28,d1[1]
vadd.u64 d29,d29,d26
vmlal.u32 q9,d28,d2[0]
vmul.u32 d29,d29,d30
vmlal.u32 q10,d28,d2[1]
vst1.32 {d28},[r10,:64]! @ put aside smashed b[8*i+7]
vmlal.u32 q11,d28,d3[0]
vzip.16 d29,d8
vmlal.u32 q12,d28,d3[1]
vld1.32 {d28},[sp,:64] @ pull smashed b[8*i+0]
vmlal.u32 q13,d29,d4[0]
vld1.32 {d0,d1,d2,d3},[r1]!
vmlal.u32 q6,d29,d4[1]
vmlal.u32 q7,d29,d5[0]
vshr.u64 d26,d26,#16
vmlal.u32 q8,d29,d5[1]
vmlal.u32 q9,d29,d6[0]
vadd.u64 d26,d26,d27
vmlal.u32 q10,d29,d6[1]
vshr.u64 d26,d26,#16
vmlal.u32 q11,d29,d7[0]
vmlal.u32 q12,d29,d7[1]
vadd.u64 d12,d12,d26
vst1.32 {d29},[r10,:64] @ put aside smashed m[8*i+7]
add r10,sp,#8 @ rewind
sub r8,r5,#8
b LNEON_8n_inner
.align 4
LNEON_8n_inner:
subs r8,r8,#8
vmlal.u32 q6,d28,d0[0]
vld1.64 {q13},[r6,:128]
vmlal.u32 q7,d28,d0[1]
vld1.32 {d29},[r10,:64]! @ pull smashed m[8*i+0]
vmlal.u32 q8,d28,d1[0]
vld1.32 {d4,d5,d6,d7},[r3]!
vmlal.u32 q9,d28,d1[1]
it ne
addne r6,r6,#16 @ don't advance in last iteration
vmlal.u32 q10,d28,d2[0]
vmlal.u32 q11,d28,d2[1]
vmlal.u32 q12,d28,d3[0]
vmlal.u32 q13,d28,d3[1]
vld1.32 {d28},[r10,:64]! @ pull smashed b[8*i+1]
vmlal.u32 q6,d29,d4[0]
vmlal.u32 q7,d29,d4[1]
vmlal.u32 q8,d29,d5[0]
vmlal.u32 q9,d29,d5[1]
vmlal.u32 q10,d29,d6[0]
vmlal.u32 q11,d29,d6[1]
vmlal.u32 q12,d29,d7[0]
vmlal.u32 q13,d29,d7[1]
vst1.64 {q6},[r7,:128]!
vmlal.u32 q7,d28,d0[0]
vld1.64 {q6},[r6,:128]
vmlal.u32 q8,d28,d0[1]
vld1.32 {d29},[r10,:64]! @ pull smashed m[8*i+1]
vmlal.u32 q9,d28,d1[0]
it ne
addne r6,r6,#16 @ don't advance in last iteration
vmlal.u32 q10,d28,d1[1]
vmlal.u32 q11,d28,d2[0]
vmlal.u32 q12,d28,d2[1]
vmlal.u32 q13,d28,d3[0]
vmlal.u32 q6,d28,d3[1]
vld1.32 {d28},[r10,:64]! @ pull smashed b[8*i+2]
vmlal.u32 q7,d29,d4[0]
vmlal.u32 q8,d29,d4[1]
vmlal.u32 q9,d29,d5[0]
vmlal.u32 q10,d29,d5[1]
vmlal.u32 q11,d29,d6[0]
vmlal.u32 q12,d29,d6[1]
vmlal.u32 q13,d29,d7[0]
vmlal.u32 q6,d29,d7[1]
vst1.64 {q7},[r7,:128]!
vmlal.u32 q8,d28,d0[0]
vld1.64 {q7},[r6,:128]
vmlal.u32 q9,d28,d0[1]
vld1.32 {d29},[r10,:64]! @ pull smashed m[8*i+2]
vmlal.u32 q10,d28,d1[0]
it ne
addne r6,r6,#16 @ don't advance in last iteration
vmlal.u32 q11,d28,d1[1]
vmlal.u32 q12,d28,d2[0]
vmlal.u32 q13,d28,d2[1]
vmlal.u32 q6,d28,d3[0]
vmlal.u32 q7,d28,d3[1]
vld1.32 {d28},[r10,:64]! @ pull smashed b[8*i+3]
vmlal.u32 q8,d29,d4[0]
vmlal.u32 q9,d29,d4[1]
vmlal.u32 q10,d29,d5[0]
vmlal.u32 q11,d29,d5[1]
vmlal.u32 q12,d29,d6[0]
vmlal.u32 q13,d29,d6[1]
vmlal.u32 q6,d29,d7[0]
vmlal.u32 q7,d29,d7[1]
vst1.64 {q8},[r7,:128]!
vmlal.u32 q9,d28,d0[0]
vld1.64 {q8},[r6,:128]
vmlal.u32 q10,d28,d0[1]
vld1.32 {d29},[r10,:64]! @ pull smashed m[8*i+3]
vmlal.u32 q11,d28,d1[0]
it ne
addne r6,r6,#16 @ don't advance in last iteration
vmlal.u32 q12,d28,d1[1]
vmlal.u32 q13,d28,d2[0]
vmlal.u32 q6,d28,d2[1]
vmlal.u32 q7,d28,d3[0]
vmlal.u32 q8,d28,d3[1]
vld1.32 {d28},[r10,:64]! @ pull smashed b[8*i+4]
vmlal.u32 q9,d29,d4[0]
vmlal.u32 q10,d29,d4[1]
vmlal.u32 q11,d29,d5[0]
vmlal.u32 q12,d29,d5[1]
vmlal.u32 q13,d29,d6[0]
vmlal.u32 q6,d29,d6[1]
vmlal.u32 q7,d29,d7[0]
vmlal.u32 q8,d29,d7[1]
vst1.64 {q9},[r7,:128]!
vmlal.u32 q10,d28,d0[0]
vld1.64 {q9},[r6,:128]
vmlal.u32 q11,d28,d0[1]
vld1.32 {d29},[r10,:64]! @ pull smashed m[8*i+4]
vmlal.u32 q12,d28,d1[0]
it ne
addne r6,r6,#16 @ don't advance in last iteration
vmlal.u32 q13,d28,d1[1]
vmlal.u32 q6,d28,d2[0]
vmlal.u32 q7,d28,d2[1]
vmlal.u32 q8,d28,d3[0]
vmlal.u32 q9,d28,d3[1]
vld1.32 {d28},[r10,:64]! @ pull smashed b[8*i+5]
vmlal.u32 q10,d29,d4[0]
vmlal.u32 q11,d29,d4[1]
vmlal.u32 q12,d29,d5[0]
vmlal.u32 q13,d29,d5[1]
vmlal.u32 q6,d29,d6[0]
vmlal.u32 q7,d29,d6[1]
vmlal.u32 q8,d29,d7[0]
vmlal.u32 q9,d29,d7[1]
vst1.64 {q10},[r7,:128]!
vmlal.u32 q11,d28,d0[0]
vld1.64 {q10},[r6,:128]
vmlal.u32 q12,d28,d0[1]
vld1.32 {d29},[r10,:64]! @ pull smashed m[8*i+5]
vmlal.u32 q13,d28,d1[0]
it ne
addne r6,r6,#16 @ don't advance in last iteration
vmlal.u32 q6,d28,d1[1]
vmlal.u32 q7,d28,d2[0]
vmlal.u32 q8,d28,d2[1]
vmlal.u32 q9,d28,d3[0]
vmlal.u32 q10,d28,d3[1]
vld1.32 {d28},[r10,:64]! @ pull smashed b[8*i+6]
vmlal.u32 q11,d29,d4[0]
vmlal.u32 q12,d29,d4[1]
vmlal.u32 q13,d29,d5[0]
vmlal.u32 q6,d29,d5[1]
vmlal.u32 q7,d29,d6[0]
vmlal.u32 q8,d29,d6[1]
vmlal.u32 q9,d29,d7[0]
vmlal.u32 q10,d29,d7[1]
vst1.64 {q11},[r7,:128]!
vmlal.u32 q12,d28,d0[0]
vld1.64 {q11},[r6,:128]
vmlal.u32 q13,d28,d0[1]
vld1.32 {d29},[r10,:64]! @ pull smashed m[8*i+6]
vmlal.u32 q6,d28,d1[0]
it ne
addne r6,r6,#16 @ don't advance in last iteration
vmlal.u32 q7,d28,d1[1]
vmlal.u32 q8,d28,d2[0]
vmlal.u32 q9,d28,d2[1]
vmlal.u32 q10,d28,d3[0]
vmlal.u32 q11,d28,d3[1]
vld1.32 {d28},[r10,:64]! @ pull smashed b[8*i+7]
vmlal.u32 q12,d29,d4[0]
vmlal.u32 q13,d29,d4[1]
vmlal.u32 q6,d29,d5[0]
vmlal.u32 q7,d29,d5[1]
vmlal.u32 q8,d29,d6[0]
vmlal.u32 q9,d29,d6[1]
vmlal.u32 q10,d29,d7[0]
vmlal.u32 q11,d29,d7[1]
vst1.64 {q12},[r7,:128]!
vmlal.u32 q13,d28,d0[0]
vld1.64 {q12},[r6,:128]
vmlal.u32 q6,d28,d0[1]
vld1.32 {d29},[r10,:64]! @ pull smashed m[8*i+7]
vmlal.u32 q7,d28,d1[0]
it ne
addne r6,r6,#16 @ don't advance in last iteration
vmlal.u32 q8,d28,d1[1]
vmlal.u32 q9,d28,d2[0]
vmlal.u32 q10,d28,d2[1]
vmlal.u32 q11,d28,d3[0]
vmlal.u32 q12,d28,d3[1]
it eq
subeq r1,r1,r5,lsl#2 @ rewind
vmlal.u32 q13,d29,d4[0]
vld1.32 {d28},[sp,:64] @ pull smashed b[8*i+0]
vmlal.u32 q6,d29,d4[1]
vld1.32 {d0,d1,d2,d3},[r1]!
vmlal.u32 q7,d29,d5[0]
add r10,sp,#8 @ rewind
vmlal.u32 q8,d29,d5[1]
vmlal.u32 q9,d29,d6[0]
vmlal.u32 q10,d29,d6[1]
vmlal.u32 q11,d29,d7[0]
vst1.64 {q13},[r7,:128]!
vmlal.u32 q12,d29,d7[1]
bne LNEON_8n_inner
add r6,sp,#128
vst1.64 {q6,q7},[r7,:256]!
veor q2,q2,q2 @ d4-d5
vst1.64 {q8,q9},[r7,:256]!
veor q3,q3,q3 @ d6-d7
vst1.64 {q10,q11},[r7,:256]!
vst1.64 {q12},[r7,:128]
subs r9,r9,#8
vld1.64 {q6,q7},[r6,:256]!
vld1.64 {q8,q9},[r6,:256]!
vld1.64 {q10,q11},[r6,:256]!
vld1.64 {q12,q13},[r6,:256]!
itt ne
subne r3,r3,r5,lsl#2 @ rewind
bne LNEON_8n_outer
add r7,sp,#128
vst1.64 {q2,q3}, [sp,:256]! @ start wiping stack frame
vshr.u64 d10,d12,#16
vst1.64 {q2,q3},[sp,:256]!
vadd.u64 d13,d13,d10
vst1.64 {q2,q3}, [sp,:256]!
vshr.u64 d10,d13,#16
vst1.64 {q2,q3}, [sp,:256]!
vzip.16 d12,d13
mov r8,r5
b LNEON_tail_entry
.align 4
LNEON_tail:
vadd.u64 d12,d12,d10
vshr.u64 d10,d12,#16
vld1.64 {q8,q9}, [r6, :256]!
vadd.u64 d13,d13,d10
vld1.64 {q10,q11}, [r6, :256]!
vshr.u64 d10,d13,#16
vld1.64 {q12,q13}, [r6, :256]!
vzip.16 d12,d13
LNEON_tail_entry:
vadd.u64 d14,d14,d10
vst1.32 {d12[0]}, [r7, :32]!
vshr.u64 d10,d14,#16
vadd.u64 d15,d15,d10
vshr.u64 d10,d15,#16
vzip.16 d14,d15
vadd.u64 d16,d16,d10
vst1.32 {d14[0]}, [r7, :32]!
vshr.u64 d10,d16,#16
vadd.u64 d17,d17,d10
vshr.u64 d10,d17,#16
vzip.16 d16,d17
vadd.u64 d18,d18,d10
vst1.32 {d16[0]}, [r7, :32]!
vshr.u64 d10,d18,#16
vadd.u64 d19,d19,d10
vshr.u64 d10,d19,#16
vzip.16 d18,d19
vadd.u64 d20,d20,d10
vst1.32 {d18[0]}, [r7, :32]!
vshr.u64 d10,d20,#16
vadd.u64 d21,d21,d10
vshr.u64 d10,d21,#16
vzip.16 d20,d21
vadd.u64 d22,d22,d10
vst1.32 {d20[0]}, [r7, :32]!
vshr.u64 d10,d22,#16
vadd.u64 d23,d23,d10
vshr.u64 d10,d23,#16
vzip.16 d22,d23
vadd.u64 d24,d24,d10
vst1.32 {d22[0]}, [r7, :32]!
vshr.u64 d10,d24,#16
vadd.u64 d25,d25,d10
vshr.u64 d10,d25,#16
vzip.16 d24,d25
vadd.u64 d26,d26,d10
vst1.32 {d24[0]}, [r7, :32]!
vshr.u64 d10,d26,#16
vadd.u64 d27,d27,d10
vshr.u64 d10,d27,#16
vzip.16 d26,d27
vld1.64 {q6,q7}, [r6, :256]!
subs r8,r8,#8
vst1.32 {d26[0]}, [r7, :32]!
bne LNEON_tail
vst1.32 {d10[0]}, [r7, :32] @ top-most bit
sub r3,r3,r5,lsl#2 @ rewind r3
subs r1,sp,#0 @ clear carry flag
add r2,sp,r5,lsl#2
LNEON_sub:
ldmia r1!, {r4,r5,r6,r7}
ldmia r3!, {r8,r9,r10,r11}
sbcs r8, r4,r8
sbcs r9, r5,r9
sbcs r10,r6,r10
sbcs r11,r7,r11
teq r1,r2 @ preserves carry
stmia r0!, {r8,r9,r10,r11}
bne LNEON_sub
ldr r10, [r1] @ load top-most bit
mov r11,sp
veor q0,q0,q0
sub r11,r2,r11 @ this is num*4
veor q1,q1,q1
mov r1,sp
sub r0,r0,r11 @ rewind r0
mov r3,r2 @ second 3/4th of frame
sbcs r10,r10,#0 @ result is carry flag
LNEON_copy_n_zap:
ldmia r1!, {r4,r5,r6,r7}
ldmia r0, {r8,r9,r10,r11}
it cc
movcc r8, r4
vst1.64 {q0,q1}, [r3,:256]! @ wipe
itt cc
movcc r9, r5
movcc r10,r6
vst1.64 {q0,q1}, [r3,:256]! @ wipe
it cc
movcc r11,r7
ldmia r1, {r4,r5,r6,r7}
stmia r0!, {r8,r9,r10,r11}
sub r1,r1,#16
ldmia r0, {r8,r9,r10,r11}
it cc
movcc r8, r4
vst1.64 {q0,q1}, [r1,:256]! @ wipe
itt cc
movcc r9, r5
movcc r10,r6
vst1.64 {q0,q1}, [r3,:256]! @ wipe
it cc
movcc r11,r7
teq r1,r2 @ preserves carry
stmia r0!, {r8,r9,r10,r11}
bne LNEON_copy_n_zap
mov sp,ip
vldmia sp!,{d8,d9,d10,d11,d12,d13,d14,d15}
ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11}
bx lr @ bx lr
#endif
.byte 77,111,110,116,103,111,109,101,114,121,32,109,117,108,116,105,112,108,105,99,97,116,105,111,110,32,102,111,114,32,65,82,77,118,52,47,78,69,79,78,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0
.align 2
#endif // !OPENSSL_NO_ASM && defined(OPENSSL_ARM) && defined(__APPLE__)
#endif // defined(__arm__) && defined(__APPLE__)
#if defined(__linux__) && defined(__ELF__)
.section .note.GNU-stack,"",%progbits
#endif
@@ -0,0 +1,278 @@
/* Copyright (c) 2017, Google Inc.
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
* SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
#if !defined(_GNU_SOURCE)
#define _GNU_SOURCE // needed for syscall() on Linux.
#endif
#include <CCryptoBoringSSL_crypto.h>
#include <stdlib.h>
#if defined(BORINGSSL_FIPS)
#include <sys/mman.h>
#include <unistd.h>
#endif
#include <CCryptoBoringSSL_digest.h>
#include <CCryptoBoringSSL_hmac.h>
#include <CCryptoBoringSSL_sha.h>
#include "bcm_interface.h"
#include "../bcm_support.h"
#include "../internal.h"
// TODO(crbug.com/362530616): When delocate is removed, build these files as
// separate compilation units again.
#include "aes/aes.c.inc"
#include "aes/aes_nohw.c.inc"
#include "aes/key_wrap.c.inc"
#include "aes/mode_wrappers.c.inc"
#include "bn/add.c.inc"
#include "bn/asm/x86_64-gcc.c.inc"
#include "bn/bn.c.inc"
#include "bn/bytes.c.inc"
#include "bn/cmp.c.inc"
#include "bn/ctx.c.inc"
#include "bn/div.c.inc"
#include "bn/div_extra.c.inc"
#include "bn/exponentiation.c.inc"
#include "bn/gcd.c.inc"
#include "bn/gcd_extra.c.inc"
#include "bn/generic.c.inc"
#include "bn/jacobi.c.inc"
#include "bn/montgomery.c.inc"
#include "bn/montgomery_inv.c.inc"
#include "bn/mul.c.inc"
#include "bn/prime.c.inc"
#include "bn/random.c.inc"
#include "bn/rsaz_exp.c.inc"
#include "bn/shift.c.inc"
#include "bn/sqrt.c.inc"
#include "cipher/aead.c.inc"
#include "cipher/cipher.c.inc"
#include "cipher/e_aes.c.inc"
#include "cipher/e_aesccm.c.inc"
#include "cmac/cmac.c.inc"
#include "dh/check.c.inc"
#include "dh/dh.c.inc"
#include "digest/digest.c.inc"
#include "digest/digests.c.inc"
#include "digestsign/digestsign.c.inc"
#include "ecdh/ecdh.c.inc"
#include "ecdsa/ecdsa.c.inc"
#include "ec/ec.c.inc"
#include "ec/ec_key.c.inc"
#include "ec/ec_montgomery.c.inc"
#include "ec/felem.c.inc"
#include "ec/oct.c.inc"
#include "ec/p224-64.c.inc"
#include "ec/p256.c.inc"
#include "ec/p256-nistz.c.inc"
#include "ec/scalar.c.inc"
#include "ec/simple.c.inc"
#include "ec/simple_mul.c.inc"
#include "ec/util.c.inc"
#include "ec/wnaf.c.inc"
#include "hkdf/hkdf.c.inc"
#include "hmac/hmac.c.inc"
#include "modes/cbc.c.inc"
#include "modes/cfb.c.inc"
#include "modes/ctr.c.inc"
#include "modes/gcm.c.inc"
#include "modes/gcm_nohw.c.inc"
#include "modes/ofb.c.inc"
#include "modes/polyval.c.inc"
#include "rand/ctrdrbg.c.inc"
#include "rand/rand.c.inc"
#include "rsa/blinding.c.inc"
#include "rsa/padding.c.inc"
#include "rsa/rsa.c.inc"
#include "rsa/rsa_impl.c.inc"
#include "self_check/fips.c.inc"
#include "self_check/self_check.c.inc"
#include "service_indicator/service_indicator.c.inc"
#include "sha/sha1.c.inc"
#include "sha/sha256.c.inc"
#include "sha/sha512.c.inc"
#include "tls/kdf.c.inc"
#if defined(BORINGSSL_FIPS)
#if !defined(OPENSSL_ASAN)
// These symbols are filled in by delocate.go (in static builds) or a linker
// script (in shared builds). They point to the start and end of the module, and
// the location of the integrity hash, respectively.
extern const uint8_t BORINGSSL_bcm_text_start[];
extern const uint8_t BORINGSSL_bcm_text_end[];
extern const uint8_t BORINGSSL_bcm_text_hash[];
#if defined(BORINGSSL_SHARED_LIBRARY)
extern const uint8_t BORINGSSL_bcm_rodata_start[];
extern const uint8_t BORINGSSL_bcm_rodata_end[];
#endif
// assert_within is used to sanity check that certain symbols are within the
// bounds of the integrity check. It checks that start <= symbol < end and
// aborts otherwise.
static void assert_within(const void *start, const void *symbol,
const void *end) {
const uintptr_t start_val = (uintptr_t) start;
const uintptr_t symbol_val = (uintptr_t) symbol;
const uintptr_t end_val = (uintptr_t) end;
if (start_val <= symbol_val && symbol_val < end_val) {
return;
}
fprintf(
CRYPTO_get_stderr(),
"FIPS module doesn't span expected symbol. Expected %p <= %p < %p\n",
start, symbol, end);
BORINGSSL_FIPS_abort();
}
#if defined(OPENSSL_ANDROID) && defined(OPENSSL_AARCH64)
static void BORINGSSL_maybe_set_module_text_permissions(int permission) {
// Android may be compiled in execute-only-memory mode, in which case the
// .text segment cannot be read. That conflicts with the need for a FIPS
// module to hash its own contents, therefore |mprotect| is used to make
// the module's .text readable for the duration of the hashing process. In
// other build configurations this is a no-op.
const uintptr_t page_size = getpagesize();
const uintptr_t page_start =
((uintptr_t)BORINGSSL_bcm_text_start) & ~(page_size - 1);
if (mprotect((void *)page_start,
((uintptr_t)BORINGSSL_bcm_text_end) - page_start,
permission) != 0) {
perror("BoringSSL: mprotect");
}
}
#else
static void BORINGSSL_maybe_set_module_text_permissions(int permission) {}
#endif // !ANDROID
#endif // !ASAN
static void __attribute__((constructor))
BORINGSSL_bcm_power_on_self_test(void) {
#if !defined(OPENSSL_ASAN)
// Integrity tests cannot run under ASAN because it involves reading the full
// .text section, which triggers the global-buffer overflow detection.
if (!BORINGSSL_integrity_test()) {
goto err;
}
#endif // OPENSSL_ASAN
if (!boringssl_self_test_startup()) {
goto err;
}
return;
err:
BORINGSSL_FIPS_abort();
}
#if !defined(OPENSSL_ASAN)
int BORINGSSL_integrity_test(void) {
const uint8_t *const start = BORINGSSL_bcm_text_start;
const uint8_t *const end = BORINGSSL_bcm_text_end;
assert_within(start, AES_encrypt, end);
assert_within(start, RSA_sign, end);
assert_within(start, BCM_rand_bytes, end);
assert_within(start, EC_GROUP_cmp, end);
assert_within(start, BCM_sha256_update, end);
assert_within(start, ecdsa_verify_fixed, end);
assert_within(start, EVP_AEAD_CTX_seal, end);
#if defined(BORINGSSL_SHARED_LIBRARY)
const uint8_t *const rodata_start = BORINGSSL_bcm_rodata_start;
const uint8_t *const rodata_end = BORINGSSL_bcm_rodata_end;
#else
// In the static build, read-only data is placed within the .text segment.
const uint8_t *const rodata_start = BORINGSSL_bcm_text_start;
const uint8_t *const rodata_end = BORINGSSL_bcm_text_end;
#endif
assert_within(rodata_start, kPrimes, rodata_end);
assert_within(rodata_start, kP256Field, rodata_end);
assert_within(rodata_start, kPKCS1SigPrefixes, rodata_end);
uint8_t result[SHA256_DIGEST_LENGTH];
const EVP_MD *const kHashFunction = EVP_sha256();
if (!boringssl_self_test_sha256() ||
!boringssl_self_test_hmac_sha256()) {
return 0;
}
static const uint8_t kHMACKey[64] = {0};
unsigned result_len;
HMAC_CTX hmac_ctx;
HMAC_CTX_init(&hmac_ctx);
if (!HMAC_Init_ex(&hmac_ctx, kHMACKey, sizeof(kHMACKey), kHashFunction,
NULL /* no ENGINE */)) {
fprintf(CRYPTO_get_stderr(), "HMAC_Init_ex failed.\n");
return 0;
}
BORINGSSL_maybe_set_module_text_permissions(PROT_READ | PROT_EXEC);
#if defined(BORINGSSL_SHARED_LIBRARY)
uint64_t length = end - start;
HMAC_Update(&hmac_ctx, (const uint8_t *) &length, sizeof(length));
HMAC_Update(&hmac_ctx, start, length);
length = rodata_end - rodata_start;
HMAC_Update(&hmac_ctx, (const uint8_t *) &length, sizeof(length));
HMAC_Update(&hmac_ctx, rodata_start, length);
#else
HMAC_Update(&hmac_ctx, start, end - start);
#endif
BORINGSSL_maybe_set_module_text_permissions(PROT_EXEC);
if (!HMAC_Final(&hmac_ctx, result, &result_len) ||
result_len != sizeof(result)) {
fprintf(CRYPTO_get_stderr(), "HMAC failed.\n");
return 0;
}
HMAC_CTX_cleanse(&hmac_ctx); // FIPS 140-3, AS05.10.
const uint8_t *expected = BORINGSSL_bcm_text_hash;
if (!check_test(expected, result, sizeof(result), "FIPS integrity test")) {
#if !defined(BORINGSSL_FIPS_BREAK_TESTS)
return 0;
#endif
}
OPENSSL_cleanse(result, sizeof(result)); // FIPS 140-3, AS05.10.
return 1;
}
const uint8_t* FIPS_module_hash(void) {
return BORINGSSL_bcm_text_hash;
}
#endif // OPENSSL_ASAN
void BORINGSSL_FIPS_abort(void) {
for (;;) {
abort();
exit(1);
}
}
#endif // BORINGSSL_FIPS
@@ -0,0 +1,244 @@
/* Copyright (c) 2024, Google Inc.
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
* SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
#ifndef OPENSSL_HEADER_CRYPTO_BCM_INTERFACE_H
#define OPENSSL_HEADER_CRYPTO_BCM_INTERFACE_H
#include <CCryptoBoringSSL_bcm_public.h>
// This header will eventually become the interface between BCM and the
// rest of libcrypto. More cleanly separating the two is still a work in
// progress (see https://crbug.com/boringssl/722) so, at the moment, we
// consider this no different from any other header in BCM.
//
// Over time, calls from libcrypto to BCM will all move to this header
// and the separation will become more meaningful.
#if defined(__cplusplus)
extern "C" {
#endif
// Enumerated types for return values from bcm functions, both infallible
// and fallible functions. Two success values are used to correspond to the
// FIPS service indicator. For the moment, the official service indicator
// remains the counter, not these values. Once we fully transition to
// these return values from bcm we will change that.
enum bcm_infallible_t {
bcm_infallible_approved,
bcm_infallible_not_approved,
};
enum bcm_status_t {
bcm_status_approved,
bcm_status_not_approved,
// Failure codes, which must all be negative.
bcm_status_failure,
};
typedef enum bcm_status_t bcm_status;
typedef enum bcm_infallible_t bcm_infallible;
OPENSSL_INLINE int bcm_success(bcm_status status) {
return status == bcm_status_approved || status == bcm_status_not_approved;
}
// Random number generator.
#if defined(BORINGSSL_FIPS)
// We overread from /dev/urandom or RDRAND by a factor of 10 and XOR to whiten.
// TODO(bbe): disentangle this value which is used to calculate the size of the
// stack buffer in RAND_need entropy based on a calculation.
#define BORINGSSL_FIPS_OVERREAD 10
#endif // BORINGSSL_FIPS
// BCM_rand_load_entropy supplies |entropy_len| bytes of entropy to the BCM
// module. The |want_additional_input| parameter is true iff the entropy was
// obtained from a source other than the system, e.g. directly from the CPU.
bcm_infallible BCM_rand_load_entropy(const uint8_t *entropy, size_t entropy_len,
int want_additional_input);
// BCM_rand_bytes is the same as the public |RAND_bytes| function, other
// than returning a bcm_infallible status indicator.
OPENSSL_EXPORT bcm_infallible BCM_rand_bytes(uint8_t *out, size_t out_len);
// BCM_rand_bytes_hwrng attempts to fill |out| with |len| bytes of entropy from
// the CPU hardware random number generator if one is present.
// bcm_status_approved is returned on success, and a failure status is
// returned otherwise.
bcm_status BCM_rand_bytes_hwrng(uint8_t *out, size_t len);
// BCM_rand_bytes_with_additional_data samples from the RNG after mixing 32
// bytes from |user_additional_data| in.
bcm_infallible BCM_rand_bytes_with_additional_data(
uint8_t *out, size_t out_len, const uint8_t user_additional_data[32]);
// SHA-1
// BCM_SHA_DIGEST_LENGTH is the length of a SHA-1 digest.
#define BCM_SHA_DIGEST_LENGTH 20
// BCM_sha1_init initialises |sha|.
bcm_infallible BCM_sha1_init(SHA_CTX *sha);
// BCM_SHA1_transform is a low-level function that performs a single, SHA-1
// block transformation using the state from |sha| and |SHA_CBLOCK| bytes from
// |block|.
bcm_infallible BCM_sha1_transform(SHA_CTX *c,
const uint8_t data[BCM_SHA_CBLOCK]);
// BCM_sha1_update adds |len| bytes from |data| to |sha|.
bcm_infallible BCM_sha1_update(SHA_CTX *c, const void *data, size_t len);
// BCM_sha1_final adds the final padding to |sha| and writes the resulting
// digest to |out|, which must have at least |SHA_DIGEST_LENGTH| bytes of space.
bcm_infallible BCM_sha1_final(uint8_t out[BCM_SHA_DIGEST_LENGTH], SHA_CTX *c);
// BCM_fips_186_2_prf derives |out_len| bytes from |xkey| using the PRF
// defined in FIPS 186-2, Appendix 3.1, with change notice 1 applied. The b
// parameter is 160 and seed, XKEY, is also 160 bits. The optional XSEED user
// input is all zeros.
//
// The PRF generates a sequence of 320-bit numbers. Each number is encoded as a
// 40-byte string in big-endian and then concatenated to form |out|. If
// |out_len| is not a multiple of 40, the result is truncated. This matches the
// construction used in Section 7 of RFC 4186 and Section 7 of RFC 4187.
//
// This PRF is based on SHA-1, a weak hash function, and should not be used
// in new protocols. It is provided for compatibility with some legacy EAP
// methods.
bcm_infallible BCM_fips_186_2_prf(uint8_t *out, size_t out_len,
const uint8_t xkey[BCM_SHA_DIGEST_LENGTH]);
// SHA-224
// SHA224_DIGEST_LENGTH is the length of a SHA-224 digest.
#define BCM_SHA224_DIGEST_LENGTH 28
// BCM_sha224_unit initialises |sha|.
bcm_infallible BCM_sha224_init(SHA256_CTX *sha);
// BCM_sha224_update adds |len| bytes from |data| to |sha|.
bcm_infallible BCM_sha224_update(SHA256_CTX *sha, const void *data, size_t len);
// BCM_sha224_final adds the final padding to |sha| and writes the resulting
// digest to |out|, which must have at least |SHA224_DIGEST_LENGTH| bytes of
// space. It aborts on programmer error.
bcm_infallible BCM_sha224_final(uint8_t out[BCM_SHA224_DIGEST_LENGTH],
SHA256_CTX *sha);
// SHA-256
// BCM_SHA256_DIGEST_LENGTH is the length of a SHA-256 digest.
#define BCM_SHA256_DIGEST_LENGTH 32
// BCM_sha256_init initialises |sha|.
bcm_infallible BCM_sha256_init(SHA256_CTX *sha);
// BCM_sha256_update adds |len| bytes from |data| to |sha|.
bcm_infallible BCM_sha256_update(SHA256_CTX *sha, const void *data, size_t len);
// BCM_sha256_final adds the final padding to |sha| and writes the resulting
// digest to |out|, which must have at least |BCM_SHA256_DIGEST_LENGTH| bytes of
// space. It aborts on programmer error.
bcm_infallible BCM_sha256_final(uint8_t out[BCM_SHA256_DIGEST_LENGTH],
SHA256_CTX *sha);
// BCM_sha256_transform is a low-level function that performs a single, SHA-256
// block transformation using the state from |sha| and |BCM_SHA256_CBLOCK| bytes
// from |block|.
bcm_infallible BCM_sha256_transform(SHA256_CTX *sha,
const uint8_t block[BCM_SHA256_CBLOCK]);
// BCM_sha256_transform_blocks is a low-level function that takes |num_blocks| *
// |BCM_SHA256_CBLOCK| bytes of data and performs SHA-256 transforms on it to
// update |state|.
bcm_infallible BCM_sha256_transform_blocks(uint32_t state[8],
const uint8_t *data,
size_t num_blocks);
// SHA-384.
// BCM_SHA384_DIGEST_LENGTH is the length of a SHA-384 digest.
#define BCM_SHA384_DIGEST_LENGTH 48
// BCM_sha384_init initialises |sha|.
bcm_infallible BCM_sha384_init(SHA512_CTX *sha);
// BCM_sha384_update adds |len| bytes from |data| to |sha|.
bcm_infallible BCM_sha384_update(SHA512_CTX *sha, const void *data, size_t len);
// BCM_sha384_final adds the final padding to |sha| and writes the resulting
// digest to |out|, which must have at least |BCM_sha384_DIGEST_LENGTH| bytes of
// space. It may abort on programmer error.
bcm_infallible BCM_sha384_final(uint8_t out[BCM_SHA384_DIGEST_LENGTH],
SHA512_CTX *sha);
// SHA-512.
// BCM_SHA512_DIGEST_LENGTH is the length of a SHA-512 digest.
#define BCM_SHA512_DIGEST_LENGTH 64
// BCM_sha512_init initialises |sha|.
bcm_infallible BCM_sha512_init(SHA512_CTX *sha);
// BCM_sha512_update adds |len| bytes from |data| to |sha|.
bcm_infallible BCM_sha512_update(SHA512_CTX *sha, const void *data, size_t len);
// BCM_sha512_final adds the final padding to |sha| and writes the resulting
// digest to |out|, which must have at least |BCM_sha512_DIGEST_LENGTH| bytes of
// space.
bcm_infallible BCM_sha512_final(uint8_t out[BCM_SHA512_DIGEST_LENGTH],
SHA512_CTX *sha);
// BCM_sha512_transform is a low-level function that performs a single, SHA-512
// block transformation using the state from |sha| and |BCM_sha512_CBLOCK| bytes
// from |block|.
bcm_infallible BCM_sha512_transform(SHA512_CTX *sha,
const uint8_t block[BCM_SHA512_CBLOCK]);
// SHA-512-256
//
// See https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.180-4.pdf section 5.3.6
#define BCM_SHA512_256_DIGEST_LENGTH 32
// BCM_sha512_256_init initialises |sha|.
bcm_infallible BCM_sha512_256_init(SHA512_CTX *sha);
// BCM_sha512_256_update adds |len| bytes from |data| to |sha|.
bcm_infallible BCM_sha512_256_update(SHA512_CTX *sha, const void *data,
size_t len);
// BCM_sha512_256_final adds the final padding to |sha| and writes the resulting
// digest to |out|, which must have at least |BCM_sha512_256_DIGEST_LENGTH|
// bytes of space. It may abort on programmer error.
bcm_infallible BCM_sha512_256_final(uint8_t out[BCM_SHA512_256_DIGEST_LENGTH],
SHA512_CTX *sha);
#if defined(__cplusplus)
} // extern C
#endif
#endif // OPENSSL_HEADER_CRYPTO_BCM_INTERFACE_H
@@ -1,989 +0,0 @@
#define BORINGSSL_PREFIX CCryptoBoringSSL
#if defined(__i386__) && defined(_WIN32)
; This file is generated from a similarly-named Perl script in the BoringSSL
; source tree. Do not edit by hand.
%ifdef BORINGSSL_PREFIX
%include "boringssl_prefix_symbols_nasm.inc"
%endif
%ifidn __OUTPUT_FORMAT__, win32
%ifidn __OUTPUT_FORMAT__,obj
section code use32 class=code align=64
%elifidn __OUTPUT_FORMAT__,win32
$@feat.00 equ 1
section .text code align=64
%else
section .text code
%endif
;extern _OPENSSL_ia32cap_P
global _bn_mul_add_words
align 16
_bn_mul_add_words:
L$_bn_mul_add_words_begin:
lea eax,[_OPENSSL_ia32cap_P]
bt DWORD [eax],26
jnc NEAR L$000maw_non_sse2
mov eax,DWORD [4+esp]
mov edx,DWORD [8+esp]
mov ecx,DWORD [12+esp]
movd mm0,DWORD [16+esp]
pxor mm1,mm1
jmp NEAR L$001maw_sse2_entry
align 16
L$002maw_sse2_unrolled:
movd mm3,DWORD [eax]
paddq mm1,mm3
movd mm2,DWORD [edx]
pmuludq mm2,mm0
movd mm4,DWORD [4+edx]
pmuludq mm4,mm0
movd mm6,DWORD [8+edx]
pmuludq mm6,mm0
movd mm7,DWORD [12+edx]
pmuludq mm7,mm0
paddq mm1,mm2
movd mm3,DWORD [4+eax]
paddq mm3,mm4
movd mm5,DWORD [8+eax]
paddq mm5,mm6
movd mm4,DWORD [12+eax]
paddq mm7,mm4
movd DWORD [eax],mm1
movd mm2,DWORD [16+edx]
pmuludq mm2,mm0
psrlq mm1,32
movd mm4,DWORD [20+edx]
pmuludq mm4,mm0
paddq mm1,mm3
movd mm6,DWORD [24+edx]
pmuludq mm6,mm0
movd DWORD [4+eax],mm1
psrlq mm1,32
movd mm3,DWORD [28+edx]
add edx,32
pmuludq mm3,mm0
paddq mm1,mm5
movd mm5,DWORD [16+eax]
paddq mm2,mm5
movd DWORD [8+eax],mm1
psrlq mm1,32
paddq mm1,mm7
movd mm5,DWORD [20+eax]
paddq mm4,mm5
movd DWORD [12+eax],mm1
psrlq mm1,32
paddq mm1,mm2
movd mm5,DWORD [24+eax]
paddq mm6,mm5
movd DWORD [16+eax],mm1
psrlq mm1,32
paddq mm1,mm4
movd mm5,DWORD [28+eax]
paddq mm3,mm5
movd DWORD [20+eax],mm1
psrlq mm1,32
paddq mm1,mm6
movd DWORD [24+eax],mm1
psrlq mm1,32
paddq mm1,mm3
movd DWORD [28+eax],mm1
lea eax,[32+eax]
psrlq mm1,32
sub ecx,8
jz NEAR L$003maw_sse2_exit
L$001maw_sse2_entry:
test ecx,4294967288
jnz NEAR L$002maw_sse2_unrolled
align 4
L$004maw_sse2_loop:
movd mm2,DWORD [edx]
movd mm3,DWORD [eax]
pmuludq mm2,mm0
lea edx,[4+edx]
paddq mm1,mm3
paddq mm1,mm2
movd DWORD [eax],mm1
sub ecx,1
psrlq mm1,32
lea eax,[4+eax]
jnz NEAR L$004maw_sse2_loop
L$003maw_sse2_exit:
movd eax,mm1
emms
ret
align 16
L$000maw_non_sse2:
push ebp
push ebx
push esi
push edi
;
xor esi,esi
mov edi,DWORD [20+esp]
mov ecx,DWORD [28+esp]
mov ebx,DWORD [24+esp]
and ecx,4294967288
mov ebp,DWORD [32+esp]
push ecx
jz NEAR L$005maw_finish
align 16
L$006maw_loop:
; Round 0
mov eax,DWORD [ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [edi]
adc edx,0
mov DWORD [edi],eax
mov esi,edx
; Round 4
mov eax,DWORD [4+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [4+edi]
adc edx,0
mov DWORD [4+edi],eax
mov esi,edx
; Round 8
mov eax,DWORD [8+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [8+edi]
adc edx,0
mov DWORD [8+edi],eax
mov esi,edx
; Round 12
mov eax,DWORD [12+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [12+edi]
adc edx,0
mov DWORD [12+edi],eax
mov esi,edx
; Round 16
mov eax,DWORD [16+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [16+edi]
adc edx,0
mov DWORD [16+edi],eax
mov esi,edx
; Round 20
mov eax,DWORD [20+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [20+edi]
adc edx,0
mov DWORD [20+edi],eax
mov esi,edx
; Round 24
mov eax,DWORD [24+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [24+edi]
adc edx,0
mov DWORD [24+edi],eax
mov esi,edx
; Round 28
mov eax,DWORD [28+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [28+edi]
adc edx,0
mov DWORD [28+edi],eax
mov esi,edx
;
sub ecx,8
lea ebx,[32+ebx]
lea edi,[32+edi]
jnz NEAR L$006maw_loop
L$005maw_finish:
mov ecx,DWORD [32+esp]
and ecx,7
jnz NEAR L$007maw_finish2
jmp NEAR L$008maw_end
L$007maw_finish2:
; Tail Round 0
mov eax,DWORD [ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [edi]
adc edx,0
dec ecx
mov DWORD [edi],eax
mov esi,edx
jz NEAR L$008maw_end
; Tail Round 1
mov eax,DWORD [4+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [4+edi]
adc edx,0
dec ecx
mov DWORD [4+edi],eax
mov esi,edx
jz NEAR L$008maw_end
; Tail Round 2
mov eax,DWORD [8+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [8+edi]
adc edx,0
dec ecx
mov DWORD [8+edi],eax
mov esi,edx
jz NEAR L$008maw_end
; Tail Round 3
mov eax,DWORD [12+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [12+edi]
adc edx,0
dec ecx
mov DWORD [12+edi],eax
mov esi,edx
jz NEAR L$008maw_end
; Tail Round 4
mov eax,DWORD [16+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [16+edi]
adc edx,0
dec ecx
mov DWORD [16+edi],eax
mov esi,edx
jz NEAR L$008maw_end
; Tail Round 5
mov eax,DWORD [20+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [20+edi]
adc edx,0
dec ecx
mov DWORD [20+edi],eax
mov esi,edx
jz NEAR L$008maw_end
; Tail Round 6
mov eax,DWORD [24+ebx]
mul ebp
add eax,esi
adc edx,0
add eax,DWORD [24+edi]
adc edx,0
mov DWORD [24+edi],eax
mov esi,edx
L$008maw_end:
mov eax,esi
pop ecx
pop edi
pop esi
pop ebx
pop ebp
ret
global _bn_mul_words
align 16
_bn_mul_words:
L$_bn_mul_words_begin:
lea eax,[_OPENSSL_ia32cap_P]
bt DWORD [eax],26
jnc NEAR L$009mw_non_sse2
mov eax,DWORD [4+esp]
mov edx,DWORD [8+esp]
mov ecx,DWORD [12+esp]
movd mm0,DWORD [16+esp]
pxor mm1,mm1
align 16
L$010mw_sse2_loop:
movd mm2,DWORD [edx]
pmuludq mm2,mm0
lea edx,[4+edx]
paddq mm1,mm2
movd DWORD [eax],mm1
sub ecx,1
psrlq mm1,32
lea eax,[4+eax]
jnz NEAR L$010mw_sse2_loop
movd eax,mm1
emms
ret
align 16
L$009mw_non_sse2:
push ebp
push ebx
push esi
push edi
;
xor esi,esi
mov edi,DWORD [20+esp]
mov ebx,DWORD [24+esp]
mov ebp,DWORD [28+esp]
mov ecx,DWORD [32+esp]
and ebp,4294967288
jz NEAR L$011mw_finish
L$012mw_loop:
; Round 0
mov eax,DWORD [ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [edi],eax
mov esi,edx
; Round 4
mov eax,DWORD [4+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [4+edi],eax
mov esi,edx
; Round 8
mov eax,DWORD [8+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [8+edi],eax
mov esi,edx
; Round 12
mov eax,DWORD [12+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [12+edi],eax
mov esi,edx
; Round 16
mov eax,DWORD [16+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [16+edi],eax
mov esi,edx
; Round 20
mov eax,DWORD [20+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [20+edi],eax
mov esi,edx
; Round 24
mov eax,DWORD [24+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [24+edi],eax
mov esi,edx
; Round 28
mov eax,DWORD [28+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [28+edi],eax
mov esi,edx
;
add ebx,32
add edi,32
sub ebp,8
jz NEAR L$011mw_finish
jmp NEAR L$012mw_loop
L$011mw_finish:
mov ebp,DWORD [28+esp]
and ebp,7
jnz NEAR L$013mw_finish2
jmp NEAR L$014mw_end
L$013mw_finish2:
; Tail Round 0
mov eax,DWORD [ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [edi],eax
mov esi,edx
dec ebp
jz NEAR L$014mw_end
; Tail Round 1
mov eax,DWORD [4+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [4+edi],eax
mov esi,edx
dec ebp
jz NEAR L$014mw_end
; Tail Round 2
mov eax,DWORD [8+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [8+edi],eax
mov esi,edx
dec ebp
jz NEAR L$014mw_end
; Tail Round 3
mov eax,DWORD [12+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [12+edi],eax
mov esi,edx
dec ebp
jz NEAR L$014mw_end
; Tail Round 4
mov eax,DWORD [16+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [16+edi],eax
mov esi,edx
dec ebp
jz NEAR L$014mw_end
; Tail Round 5
mov eax,DWORD [20+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [20+edi],eax
mov esi,edx
dec ebp
jz NEAR L$014mw_end
; Tail Round 6
mov eax,DWORD [24+ebx]
mul ecx
add eax,esi
adc edx,0
mov DWORD [24+edi],eax
mov esi,edx
L$014mw_end:
mov eax,esi
pop edi
pop esi
pop ebx
pop ebp
ret
global _bn_sqr_words
align 16
_bn_sqr_words:
L$_bn_sqr_words_begin:
lea eax,[_OPENSSL_ia32cap_P]
bt DWORD [eax],26
jnc NEAR L$015sqr_non_sse2
mov eax,DWORD [4+esp]
mov edx,DWORD [8+esp]
mov ecx,DWORD [12+esp]
align 16
L$016sqr_sse2_loop:
movd mm0,DWORD [edx]
pmuludq mm0,mm0
lea edx,[4+edx]
movq [eax],mm0
sub ecx,1
lea eax,[8+eax]
jnz NEAR L$016sqr_sse2_loop
emms
ret
align 16
L$015sqr_non_sse2:
push ebp
push ebx
push esi
push edi
;
mov esi,DWORD [20+esp]
mov edi,DWORD [24+esp]
mov ebx,DWORD [28+esp]
and ebx,4294967288
jz NEAR L$017sw_finish
L$018sw_loop:
; Round 0
mov eax,DWORD [edi]
mul eax
mov DWORD [esi],eax
mov DWORD [4+esi],edx
; Round 4
mov eax,DWORD [4+edi]
mul eax
mov DWORD [8+esi],eax
mov DWORD [12+esi],edx
; Round 8
mov eax,DWORD [8+edi]
mul eax
mov DWORD [16+esi],eax
mov DWORD [20+esi],edx
; Round 12
mov eax,DWORD [12+edi]
mul eax
mov DWORD [24+esi],eax
mov DWORD [28+esi],edx
; Round 16
mov eax,DWORD [16+edi]
mul eax
mov DWORD [32+esi],eax
mov DWORD [36+esi],edx
; Round 20
mov eax,DWORD [20+edi]
mul eax
mov DWORD [40+esi],eax
mov DWORD [44+esi],edx
; Round 24
mov eax,DWORD [24+edi]
mul eax
mov DWORD [48+esi],eax
mov DWORD [52+esi],edx
; Round 28
mov eax,DWORD [28+edi]
mul eax
mov DWORD [56+esi],eax
mov DWORD [60+esi],edx
;
add edi,32
add esi,64
sub ebx,8
jnz NEAR L$018sw_loop
L$017sw_finish:
mov ebx,DWORD [28+esp]
and ebx,7
jz NEAR L$019sw_end
; Tail Round 0
mov eax,DWORD [edi]
mul eax
mov DWORD [esi],eax
dec ebx
mov DWORD [4+esi],edx
jz NEAR L$019sw_end
; Tail Round 1
mov eax,DWORD [4+edi]
mul eax
mov DWORD [8+esi],eax
dec ebx
mov DWORD [12+esi],edx
jz NEAR L$019sw_end
; Tail Round 2
mov eax,DWORD [8+edi]
mul eax
mov DWORD [16+esi],eax
dec ebx
mov DWORD [20+esi],edx
jz NEAR L$019sw_end
; Tail Round 3
mov eax,DWORD [12+edi]
mul eax
mov DWORD [24+esi],eax
dec ebx
mov DWORD [28+esi],edx
jz NEAR L$019sw_end
; Tail Round 4
mov eax,DWORD [16+edi]
mul eax
mov DWORD [32+esi],eax
dec ebx
mov DWORD [36+esi],edx
jz NEAR L$019sw_end
; Tail Round 5
mov eax,DWORD [20+edi]
mul eax
mov DWORD [40+esi],eax
dec ebx
mov DWORD [44+esi],edx
jz NEAR L$019sw_end
; Tail Round 6
mov eax,DWORD [24+edi]
mul eax
mov DWORD [48+esi],eax
mov DWORD [52+esi],edx
L$019sw_end:
pop edi
pop esi
pop ebx
pop ebp
ret
global _bn_div_words
align 16
_bn_div_words:
L$_bn_div_words_begin:
mov edx,DWORD [4+esp]
mov eax,DWORD [8+esp]
mov ecx,DWORD [12+esp]
div ecx
ret
global _bn_add_words
align 16
_bn_add_words:
L$_bn_add_words_begin:
push ebp
push ebx
push esi
push edi
;
mov ebx,DWORD [20+esp]
mov esi,DWORD [24+esp]
mov edi,DWORD [28+esp]
mov ebp,DWORD [32+esp]
xor eax,eax
and ebp,4294967288
jz NEAR L$020aw_finish
L$021aw_loop:
; Round 0
mov ecx,DWORD [esi]
mov edx,DWORD [edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [ebx],ecx
; Round 1
mov ecx,DWORD [4+esi]
mov edx,DWORD [4+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [4+ebx],ecx
; Round 2
mov ecx,DWORD [8+esi]
mov edx,DWORD [8+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [8+ebx],ecx
; Round 3
mov ecx,DWORD [12+esi]
mov edx,DWORD [12+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [12+ebx],ecx
; Round 4
mov ecx,DWORD [16+esi]
mov edx,DWORD [16+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [16+ebx],ecx
; Round 5
mov ecx,DWORD [20+esi]
mov edx,DWORD [20+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [20+ebx],ecx
; Round 6
mov ecx,DWORD [24+esi]
mov edx,DWORD [24+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [24+ebx],ecx
; Round 7
mov ecx,DWORD [28+esi]
mov edx,DWORD [28+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [28+ebx],ecx
;
add esi,32
add edi,32
add ebx,32
sub ebp,8
jnz NEAR L$021aw_loop
L$020aw_finish:
mov ebp,DWORD [32+esp]
and ebp,7
jz NEAR L$022aw_end
; Tail Round 0
mov ecx,DWORD [esi]
mov edx,DWORD [edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
dec ebp
mov DWORD [ebx],ecx
jz NEAR L$022aw_end
; Tail Round 1
mov ecx,DWORD [4+esi]
mov edx,DWORD [4+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
dec ebp
mov DWORD [4+ebx],ecx
jz NEAR L$022aw_end
; Tail Round 2
mov ecx,DWORD [8+esi]
mov edx,DWORD [8+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
dec ebp
mov DWORD [8+ebx],ecx
jz NEAR L$022aw_end
; Tail Round 3
mov ecx,DWORD [12+esi]
mov edx,DWORD [12+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
dec ebp
mov DWORD [12+ebx],ecx
jz NEAR L$022aw_end
; Tail Round 4
mov ecx,DWORD [16+esi]
mov edx,DWORD [16+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
dec ebp
mov DWORD [16+ebx],ecx
jz NEAR L$022aw_end
; Tail Round 5
mov ecx,DWORD [20+esi]
mov edx,DWORD [20+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
dec ebp
mov DWORD [20+ebx],ecx
jz NEAR L$022aw_end
; Tail Round 6
mov ecx,DWORD [24+esi]
mov edx,DWORD [24+edi]
add ecx,eax
mov eax,0
adc eax,eax
add ecx,edx
adc eax,0
mov DWORD [24+ebx],ecx
L$022aw_end:
pop edi
pop esi
pop ebx
pop ebp
ret
global _bn_sub_words
align 16
_bn_sub_words:
L$_bn_sub_words_begin:
push ebp
push ebx
push esi
push edi
;
mov ebx,DWORD [20+esp]
mov esi,DWORD [24+esp]
mov edi,DWORD [28+esp]
mov ebp,DWORD [32+esp]
xor eax,eax
and ebp,4294967288
jz NEAR L$023aw_finish
L$024aw_loop:
; Round 0
mov ecx,DWORD [esi]
mov edx,DWORD [edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [ebx],ecx
; Round 1
mov ecx,DWORD [4+esi]
mov edx,DWORD [4+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [4+ebx],ecx
; Round 2
mov ecx,DWORD [8+esi]
mov edx,DWORD [8+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [8+ebx],ecx
; Round 3
mov ecx,DWORD [12+esi]
mov edx,DWORD [12+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [12+ebx],ecx
; Round 4
mov ecx,DWORD [16+esi]
mov edx,DWORD [16+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [16+ebx],ecx
; Round 5
mov ecx,DWORD [20+esi]
mov edx,DWORD [20+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [20+ebx],ecx
; Round 6
mov ecx,DWORD [24+esi]
mov edx,DWORD [24+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [24+ebx],ecx
; Round 7
mov ecx,DWORD [28+esi]
mov edx,DWORD [28+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [28+ebx],ecx
;
add esi,32
add edi,32
add ebx,32
sub ebp,8
jnz NEAR L$024aw_loop
L$023aw_finish:
mov ebp,DWORD [32+esp]
and ebp,7
jz NEAR L$025aw_end
; Tail Round 0
mov ecx,DWORD [esi]
mov edx,DWORD [edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
dec ebp
mov DWORD [ebx],ecx
jz NEAR L$025aw_end
; Tail Round 1
mov ecx,DWORD [4+esi]
mov edx,DWORD [4+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
dec ebp
mov DWORD [4+ebx],ecx
jz NEAR L$025aw_end
; Tail Round 2
mov ecx,DWORD [8+esi]
mov edx,DWORD [8+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
dec ebp
mov DWORD [8+ebx],ecx
jz NEAR L$025aw_end
; Tail Round 3
mov ecx,DWORD [12+esi]
mov edx,DWORD [12+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
dec ebp
mov DWORD [12+ebx],ecx
jz NEAR L$025aw_end
; Tail Round 4
mov ecx,DWORD [16+esi]
mov edx,DWORD [16+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
dec ebp
mov DWORD [16+ebx],ecx
jz NEAR L$025aw_end
; Tail Round 5
mov ecx,DWORD [20+esi]
mov edx,DWORD [20+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
dec ebp
mov DWORD [20+ebx],ecx
jz NEAR L$025aw_end
; Tail Round 6
mov ecx,DWORD [24+esi]
mov edx,DWORD [24+edi]
sub ecx,eax
mov eax,0
adc eax,eax
sub ecx,edx
adc eax,0
mov DWORD [24+ebx],ecx
L$025aw_end:
pop edi
pop esi
pop ebx
pop ebp
ret
segment .bss
common _OPENSSL_ia32cap_P 16
%else
; Work around https://bugzilla.nasm.us/show_bug.cgi?id=3392738
ret
%endif
#endif // defined(__i386__) && defined(_WIN32)
#if defined(__linux__) && defined(__ELF__)
.section .note.GNU-stack,"",%progbits
#endif
@@ -186,7 +186,7 @@ void bn_assert_fits_in_bytes(const BIGNUM *bn, size_t num) {
void bn_words_to_big_endian(uint8_t *out, size_t out_len, const BN_ULONG *in,
size_t in_len) {
// The caller should have selected an output length without truncation.
assert(fits_in_bytes(in, in_len, out_len));
declassify_assert(fits_in_bytes(in, in_len, out_len));
// We only support little-endian platforms, so the internal representation is
// also little-endian as bytes. We can simply copy it in reverse.
@@ -149,11 +149,11 @@ static inline void bn_div_rem_words(BN_ULONG *quotient_out, BN_ULONG *rem_out,
// * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65668
//
// Clang bugs:
// * https://llvm.org/bugs/show_bug.cgi?id=6397
// * https://llvm.org/bugs/show_bug.cgi?id=12418
// * https://github.com/llvm/llvm-project/issues/6769
// * https://github.com/llvm/llvm-project/issues/12790
//
// These issues aren't specific to x86 and x86_64, so it might be worthwhile
// to add more assembly language implementations.
// These is specific to x86 and x86_64; Arm and RISC-V do not have double-wide
// division instructions.
#if defined(BN_CAN_USE_INLINE_ASM) && defined(OPENSSL_X86)
__asm__ volatile("divl %4"
: "=a"(*quotient_out), "=d"(*rem_out)
@@ -175,44 +175,16 @@ static inline void bn_div_rem_words(BN_ULONG *quotient_out, BN_ULONG *rem_out,
#endif
}
// BN_div computes "quotient := numerator / divisor", rounding towards zero,
// and sets up |rem| such that "quotient * divisor + rem = numerator" holds.
//
// Thus:
//
// quotient->neg == numerator->neg ^ divisor->neg
// (unless the result is zero)
// rem->neg == numerator->neg
// (unless the remainder is zero)
//
// If |quotient| or |rem| is NULL, the respective value is not returned.
//
// This was specifically designed to contain fewer branches that may leak
// sensitive information; see "New Branch Prediction Vulnerabilities in OpenSSL
// and Necessary Software Countermeasures" by Onur Acıçmez, Shay Gueron, and
// Jean-Pierre Seifert.
int BN_div(BIGNUM *quotient, BIGNUM *rem, const BIGNUM *numerator,
const BIGNUM *divisor, BN_CTX *ctx) {
int norm_shift, loop;
BIGNUM wnum;
BN_ULONG *resp, *wnump;
BN_ULONG d0, d1;
int num_n, div_n;
// This function relies on the historical minimal-width |BIGNUM| invariant.
// It is already not constant-time (constant-time reductions should use
// Montgomery logic), so we shrink all inputs and intermediate values to
// retain the previous behavior.
// Invalid zero-padding would have particularly bad consequences.
int numerator_width = bn_minimal_width(numerator);
int divisor_width = bn_minimal_width(divisor);
if ((numerator_width > 0 && numerator->d[numerator_width - 1] == 0) ||
(divisor_width > 0 && divisor->d[divisor_width - 1] == 0)) {
OPENSSL_PUT_ERROR(BN, BN_R_NOT_INITIALIZED);
return 0;
}
// This function implements long division, per Knuth, The Art of Computer
// Programming, Volume 2, Chapter 4.3.1, Algorithm D. This algorithm only
// divides non-negative integers, but we round towards zero, so we divide
// absolute values and adjust the signs separately.
//
// Inputs to this function are assumed public and may be leaked by timing and
// cache side channels. Division with secret inputs should use other
// implementation strategies such as Montgomery reduction.
if (BN_is_zero(divisor)) {
OPENSSL_PUT_ERROR(BN, BN_R_DIV_BY_ZERO);
return 0;
@@ -222,174 +194,168 @@ int BN_div(BIGNUM *quotient, BIGNUM *rem, const BIGNUM *numerator,
BIGNUM *tmp = BN_CTX_get(ctx);
BIGNUM *snum = BN_CTX_get(ctx);
BIGNUM *sdiv = BN_CTX_get(ctx);
BIGNUM *res = NULL;
if (quotient == NULL) {
res = BN_CTX_get(ctx);
} else {
res = quotient;
}
if (sdiv == NULL || res == NULL) {
BIGNUM *res = quotient == NULL ? BN_CTX_get(ctx) : quotient;
if (tmp == NULL || snum == NULL || sdiv == NULL || res == NULL) {
goto err;
}
// First we normalise the numbers
norm_shift = BN_BITS2 - (BN_num_bits(divisor) % BN_BITS2);
if (!BN_lshift(sdiv, divisor, norm_shift)) {
// Knuth step D1: Normalise the numbers such that the divisor's MSB is set.
// This ensures, in Knuth's terminology, that v1 >= b/2, needed for the
// quotient estimation step.
int norm_shift = BN_BITS2 - (BN_num_bits(divisor) % BN_BITS2);
if (!BN_lshift(sdiv, divisor, norm_shift) ||
!BN_lshift(snum, numerator, norm_shift)) {
goto err;
}
// This algorithm relies on |sdiv| being minimal width. We do not use this
// function on secret inputs, so leaking this is fine. Also minimize |snum| to
// avoid looping on leading zeros, as we're not trying to be leak-free.
bn_set_minimal_width(sdiv);
sdiv->neg = 0;
norm_shift += BN_BITS2;
if (!BN_lshift(snum, numerator, norm_shift)) {
goto err;
}
bn_set_minimal_width(snum);
snum->neg = 0;
int div_n = sdiv->width;
const BN_ULONG d0 = sdiv->d[div_n - 1];
const BN_ULONG d1 = (div_n == 1) ? 0 : sdiv->d[div_n - 2];
assert(d0 & (((BN_ULONG)1) << (BN_BITS2 - 1)));
// Since we don't want to have special-case logic for the case where snum is
// larger than sdiv, we pad snum with enough zeroes without changing its
// value.
if (snum->width <= sdiv->width + 1) {
if (!bn_wexpand(snum, sdiv->width + 2)) {
goto err;
}
for (int i = snum->width; i < sdiv->width + 2; i++) {
snum->d[i] = 0;
}
snum->width = sdiv->width + 2;
} else {
if (!bn_wexpand(snum, snum->width + 1)) {
goto err;
}
snum->d[snum->width] = 0;
snum->width++;
}
div_n = sdiv->width;
num_n = snum->width;
loop = num_n - div_n;
// Lets setup a 'window' into snum
// This is the part that corresponds to the current
// 'area' being divided
wnum.neg = 0;
wnum.d = &(snum->d[loop]);
wnum.width = div_n;
// only needed when BN_ucmp messes up the values between width and max
wnum.dmax = snum->dmax - loop; // so we don't step out of bounds
// Get the top 2 words of sdiv
// div_n=sdiv->width;
d0 = sdiv->d[div_n - 1];
d1 = (div_n == 1) ? 0 : sdiv->d[div_n - 2];
// pointer to the 'top' of snum
wnump = &(snum->d[num_n - 1]);
// Setup |res|. |numerator| and |res| may alias, so we save |numerator->neg|
// for later.
const int numerator_neg = numerator->neg;
res->neg = (numerator_neg ^ divisor->neg);
if (!bn_wexpand(res, loop + 1)) {
goto err;
}
res->width = loop - 1;
resp = &(res->d[loop - 1]);
// space for temp
if (!bn_wexpand(tmp, div_n + 1)) {
// Extend |snum| with zeros to satisfy the long division invariants:
// - |snum| must have at least |div_n| + 1 words.
// - |snum|'s most significant word must be zero to guarantee the first loop
// iteration works with a prefix greater than |sdiv|. (This is the extra u0
// digit in Knuth step D1.)
int num_n = snum->width <= div_n ? div_n + 1 : snum->width + 1;
if (!bn_resize_words(snum, num_n)) {
goto err;
}
// if res->width == 0 then clear the neg value otherwise decrease
// the resp pointer
if (res->width == 0) {
res->neg = 0;
} else {
resp--;
// Knuth step D2: The quotient's width is the difference between numerator and
// denominator. Also set up its sign and size a temporary for the loop.
int loop = num_n - div_n;
res->neg = snum->neg ^ sdiv->neg;
if (!bn_wexpand(res, loop) || //
!bn_wexpand(tmp, div_n + 1)) {
goto err;
}
res->width = loop;
for (int i = 0; i < loop - 1; i++, wnump--, resp--) {
BN_ULONG q, l0;
// the first part of the loop uses the top two words of snum and sdiv to
// calculate a BN_ULONG q such that | wnum - sdiv * q | < sdiv
BN_ULONG n0, n1, rm = 0;
n0 = wnump[0];
n1 = wnump[-1];
// Knuth steps D2 through D7: Compute the quotient with a word-by-word long
// division. Note that Knuth indexes words from most to least significant, so
// our index is reversed. Each loop iteration computes res->d[i] of the
// quotient and updates snum with the running remainder. Before each loop
// iteration, the div_n words beginning at snum->d[i+1] must be less than
// snum.
for (int i = loop - 1; i >= 0; i--) {
// The next word of the quotient, q, is floor(wnum / sdiv), where wnum is
// the div_n + 1 words beginning at snum->d[i]. i starts at
// num_n - div_n - 1, so there are at least div_n + 1 words available.
//
// Knuth step D3: Compute q', an estimate of q by looking at the top words
// of wnum and sdiv. We must estimate such that q' = q or q' = q + 1.
BN_ULONG q, rm = 0;
BN_ULONG *wnum = snum->d + i;
BN_ULONG n0 = wnum[div_n];
BN_ULONG n1 = wnum[div_n - 1];
if (n0 == d0) {
// Estimate q' = b - 1, where b is the base.
q = BN_MASK2;
// Knuth also runs the fixup routine in this case, but this would require
// computing rm and is unnecessary. q' is already close enough. That is,
// the true quotient, q is either b - 1 or b - 2.
//
// By the loop invariant, q <= b - 1, so we must show that q >= b - 2. We
// do this by showing wnum / sdiv >= b - 2. Suppose wnum / sdiv < b - 2.
// wnum and sdiv have the same most significant word, so:
//
// wnum >= n0 * b^div_n
// sdiv < (n0 + 1) * b^(d_div - 1)
//
// Thus:
//
// b - 2 > wnum / sdiv
// > (n0 * b^div_n) / (n0 + 1) * b^(div_n - 1)
// = (n0 * b) / (n0 + 1)
//
// (n0 + 1) * (b - 2) > n0 * b
// n0 * b + b - 2 * n0 - 2 > n0 * b
// b - 2 > 2 * n0
// b/2 - 1 > n0
//
// This contradicts the normalization condition, so q >= b - 2 and our
// estimate is close enough.
} else {
// n0 < d0
// Estimate q' = floor(n0n1 / d0). Per Theorem B, q' - 2 <= q <= q', which
// is slightly outside of our bounds.
assert(n0 < d0);
bn_div_rem_words(&q, &rm, n0, n1, d0);
// Fix the estimate by examining one more word and adjusting q' as needed.
// This is the second half of step D3 and is sufficient per exercises 19,
// 20, and 21. Although only one iteration is needed to correct q + 2 to
// q + 1, Knuth uses a loop. A loop will often also correct q + 1 to q,
// saving the slightly more expensive underflow handling below.
if (div_n > 1) {
BN_ULONG n2 = wnum[div_n - 2];
#ifdef BN_ULLONG
BN_ULLONG t2 = (BN_ULLONG)d1 * q;
for (;;) {
if (t2 <= ((((BN_ULLONG)rm) << BN_BITS2) | wnump[-2])) {
break;
BN_ULLONG t2 = (BN_ULLONG)d1 * q;
for (;;) {
if (t2 <= ((((BN_ULLONG)rm) << BN_BITS2) | n2)) {
break;
}
q--;
rm += d0;
if (rm < d0) {
// If rm overflows, the true value exceeds BN_ULONG and the next
// t2 comparison should exit the loop.
break;
}
t2 -= d1;
}
q--;
rm += d0;
if (rm < d0) {
break; // don't let rm overflow
#else // !BN_ULLONG
BN_ULONG t2l, t2h;
BN_UMULT_LOHI(t2l, t2h, d1, q);
for (;;) {
if (t2h < rm || (t2h == rm && t2l <= n2)) {
break;
}
q--;
rm += d0;
if (rm < d0) {
// If rm overflows, the true value exceeds BN_ULONG and the next
// t2 comparison should exit the loop.
break;
}
if (t2l < d1) {
t2h--;
}
t2l -= d1;
}
t2 -= d1;
}
#else // !BN_ULLONG
BN_ULONG t2l, t2h;
BN_UMULT_LOHI(t2l, t2h, d1, q);
for (;;) {
if (t2h < rm ||
(t2h == rm && t2l <= wnump[-2])) {
break;
}
q--;
rm += d0;
if (rm < d0) {
break; // don't let rm overflow
}
if (t2l < d1) {
t2h--;
}
t2l -= d1;
}
#endif // !BN_ULLONG
}
l0 = bn_mul_words(tmp->d, sdiv->d, div_n, q);
tmp->d[div_n] = l0;
wnum.d--;
// ingore top values of the bignums just sub the two
// BN_ULONG arrays with bn_sub_words
if (bn_sub_words(wnum.d, wnum.d, tmp->d, div_n + 1)) {
// Note: As we have considered only the leading
// two BN_ULONGs in the calculation of q, sdiv * q
// might be greater than wnum (but then (q-1) * sdiv
// is less or equal than wnum)
q--;
if (bn_add_words(wnum.d, wnum.d, sdiv->d, div_n)) {
// we can't have an overflow here (assuming
// that q != 0, but if q == 0 then tmp is
// zero anyway)
(*wnump)++;
}
}
// store part of the result
*resp = q;
// Knuth step D4 through D6: Now q' = q or q' = q + 1, and
// -sdiv < wnum - sdiv * q < sdiv. If q' = q + 1, the subtraction will
// underflow, and we fix it up below.
tmp->d[div_n] = bn_mul_words(tmp->d, sdiv->d, div_n, q);
if (bn_sub_words(wnum, wnum, tmp->d, div_n + 1)) {
q--;
// The final addition is expected to overflow, canceling the underflow.
wnum[div_n] += bn_add_words(wnum, wnum, sdiv->d, div_n);
}
// q is now correct, and wnum has been updated to the running remainder.
res->d[i] = q;
}
// Trim leading zeros and correct any negative zeros.
bn_set_minimal_width(snum);
bn_set_minimal_width(res);
if (rem != NULL) {
if (!BN_rshift(rem, snum, norm_shift)) {
goto err;
}
if (!BN_is_zero(rem)) {
rem->neg = numerator_neg;
}
// Knuth step D8: Unnormalize. snum now contains the remainder.
if (rem != NULL && !BN_rshift(rem, snum, norm_shift)) {
goto err;
}
bn_set_minimal_width(res);
BN_CTX_end(ctx);
return 1;
@@ -406,8 +372,9 @@ int BN_nnmod(BIGNUM *r, const BIGNUM *m, const BIGNUM *d, BN_CTX *ctx) {
return 1;
}
// now -|d| < r < 0, so we have to set r := r + |d|.
return (d->neg ? BN_sub : BN_add)(r, r, d);
// now -d < r < 0, so we have to set r := r + d. Ignoring the sign bits, this
// is r = d - r.
return BN_usub(r, d, r);
}
BN_ULONG bn_reduce_once(BN_ULONG *r, const BN_ULONG *a, BN_ULONG carry,
@@ -425,7 +392,7 @@ BN_ULONG bn_reduce_once(BN_ULONG *r, const BN_ULONG *a, BN_ULONG carry,
//
// Although |carry| may be one if it was one on input and |bn_sub_words|
// returns zero, this would give |r| > |m|, violating our input assumptions.
assert(carry == 0 || carry == (BN_ULONG)-1);
declassify_assert(carry + 1 <= 1);
bn_select_words(r, carry, a /* r < 0 */, r /* r >= 0 */, num);
return carry;
}
@@ -434,7 +401,7 @@ BN_ULONG bn_reduce_once_in_place(BN_ULONG *r, BN_ULONG carry, const BN_ULONG *m,
BN_ULONG *tmp, size_t num) {
// See |bn_reduce_once| for why this logic works.
carry -= bn_sub_words(tmp, r, m, num);
assert(carry == 0 || carry == (BN_ULONG)-1);
declassify_assert(carry + 1 <= 1);
bn_select_words(r, carry, r /* tmp < 0 */, tmp /* tmp >= 0 */, num);
return carry;
}
@@ -504,7 +471,7 @@ int bn_div_consttime(BIGNUM *quotient, BIGNUM *remainder,
// |divisor_min_bits| bits, the top |divisor_min_bits - 1| can be incorporated
// without reductions. This significantly speeds up |RSA_check_key|. For
// simplicity, we round down to a whole number of words.
assert(divisor_min_bits <= BN_num_bits(divisor));
declassify_assert(divisor_min_bits <= BN_num_bits(divisor));
int initial_words = 0;
if (divisor_min_bits > 0) {
initial_words = (divisor_min_bits - 1) / BN_BITS2;
@@ -39,7 +39,7 @@ static uint16_t mod_u16(uint32_t n, uint16_t d, uint32_t p, uint32_t m) {
// Multiply and subtract to get the remainder.
n -= d * t;
assert(n < d);
declassify_assert(n < d);
return n;
}
@@ -119,6 +119,50 @@
#include "internal.h"
#include "rsaz_exp.h"
#if defined(OPENSSL_BN_ASM_MONT5)
// bn_mul_mont_gather5 multiples loads index |power| of |table|, multiplies it
// by |ap| modulo |np|, and stores the result in |rp|. The values are |num|
// words long and represented in Montgomery form. |n0| is a pointer to the
// corresponding field in |BN_MONT_CTX|. |table| must be aligned to at least
// 16 bytes. |power| must be less than 32 and is treated as secret.
//
// WARNING: This function implements Almost Montgomery Multiplication from
// https://eprint.iacr.org/2011/239. The inputs do not need to be fully reduced.
// However, even if they are fully reduced, the output may not be.
static void bn_mul_mont_gather5(
BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *table, const BN_ULONG *np,
const BN_ULONG *n0, int num, int power) {
if (bn_mulx4x_mont_gather5_capable(num)) {
bn_mulx4x_mont_gather5(rp, ap, table, np, n0, num, power);
} else if (bn_mul4x_mont_gather5_capable(num)) {
bn_mul4x_mont_gather5(rp, ap, table, np, n0, num, power);
} else {
bn_mul_mont_gather5_nohw(rp, ap, table, np, n0, num, power);
}
}
// bn_power5 squares |ap| five times and multiplies it by the value stored at
// index |power| of |table|, modulo |np|. It stores the result in |rp|. The
// values are |num| words long and represented in Montgomery form. |n0| is a
// pointer to the corresponding field in |BN_MONT_CTX|. |num| must be divisible
// by 8. |power| must be less than 32 and is treated as secret.
//
// WARNING: This function implements Almost Montgomery Multiplication from
// https://eprint.iacr.org/2011/239. The inputs do not need to be fully reduced.
// However, even if they are fully reduced, the output may not be.
static void bn_power5(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *table,
const BN_ULONG *np, const BN_ULONG *n0, int num,
int power) {
assert(bn_power5_capable(num));
if (bn_powerx5_capable(num)) {
bn_powerx5(rp, ap, table, np, n0, num, power);
} else {
bn_power5_nohw(rp, ap, table, np, n0, num, power);
}
}
#endif // defined(OPENSSL_BN_ASM_MONT5)
int BN_exp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx) {
int i, bits, ret = 0;
@@ -1013,7 +1057,7 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
// Prepare a^1 in the Montgomery domain.
assert(!a->neg);
assert(BN_ucmp(a, m) < 0);
declassify_assert(BN_ucmp(a, m) < 0);
if (!BN_to_montgomery(&am, a, mont, ctx) ||
!bn_resize_words(&am, top)) {
goto err;
@@ -1079,7 +1123,7 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
// Scan the exponent one window at a time starting from the most
// significant bits.
if (top & 7) {
if (!bn_power5_capable(top)) {
while (bits >= 0) {
for (wvalue = 0, i = 0; i < 5; i++, bits--) {
wvalue = (wvalue << 1) + BN_is_bit_set(p, bits);
@@ -93,7 +93,7 @@ static int bn_gcd_consttime(BIGNUM *r, unsigned *out_shift, const BIGNUM *x,
// At least one of |u| and |v| is now even.
BN_ULONG u_is_odd = word_is_odd_mask(u->d[0]);
BN_ULONG v_is_odd = word_is_odd_mask(v->d[0]);
assert(!(u_is_odd & v_is_odd));
declassify_assert(!(u_is_odd & v_is_odd));
// If both are even, the final GCD gains a factor of two.
shift += 1 & (~u_is_odd & ~v_is_odd);
@@ -106,7 +106,7 @@ static int bn_gcd_consttime(BIGNUM *r, unsigned *out_shift, const BIGNUM *x,
// One of |u| or |v| is zero at this point. The algorithm usually makes |u|
// zero, unless |y| was already zero on input. Fix this by combining the
// values.
assert(BN_is_zero(u) || BN_is_zero(v));
declassify_assert(BN_is_zero(u) | BN_is_zero(v));
for (size_t i = 0; i < width; i++) {
v->d[i] |= u->d[i];
}
@@ -289,7 +289,7 @@ int bn_mod_inverse_consttime(BIGNUM *r, int *out_no_inverse, const BIGNUM *a,
// and |v| is now even.
BN_ULONG u_is_even = ~word_is_odd_mask(u->d[0]);
BN_ULONG v_is_even = ~word_is_odd_mask(v->d[0]);
assert(u_is_even != v_is_even);
declassify_assert(u_is_even != v_is_even);
// Halve the even one and adjust the corresponding coefficient.
maybe_rshift1_words(u->d, u_is_even, tmp->d, n_width);
@@ -313,8 +313,11 @@ int bn_mod_inverse_consttime(BIGNUM *r, int *out_no_inverse, const BIGNUM *a,
maybe_rshift1_words_carry(D->d, D_carry, v_is_even, tmp->d, a_width);
}
assert(BN_is_zero(v));
if (!BN_is_one(u)) {
declassify_assert(BN_is_zero(v));
// While the inputs and output are secret, this function considers whether the
// input was invertible to be public. It is used as part of RSA key
// generation, where inputs are chosen to already be invertible.
if (constant_time_declassify_int(!BN_is_one(u))) {
*out_no_inverse = 1;
OPENSSL_PUT_ERROR(BN, BN_R_NO_INVERSE);
goto err;
@@ -438,18 +438,26 @@ int bn_mul_mont_nohw(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86_64)
#define OPENSSL_BN_ASM_MONT5
// bn_mul_mont_gather5 multiples loads index |power| of |table|, multiplies it
// by |ap| modulo |np|, and stores the result in |rp|. The values are |num|
// words long and represented in Montgomery form. |n0| is a pointer to the
// corresponding field in |BN_MONT_CTX|. |table| must be aligned to at least
// 16 bytes. |power| must be less than 32 and is treated as secret.
//
// WARNING: This function implements Almost Montgomery Multiplication from
// https://eprint.iacr.org/2011/239. The inputs do not need to be fully reduced.
// However, even if they are fully reduced, the output may not be.
void bn_mul_mont_gather5(BN_ULONG *rp, const BN_ULONG *ap,
const BN_ULONG *table, const BN_ULONG *np,
const BN_ULONG *n0, int num, int power);
// The following functions implement |bn_mul_mont_gather5|. See
// |bn_mul_mont_gather5| for details.
OPENSSL_INLINE int bn_mul4x_mont_gather5_capable(int num) {
return (num & 7) == 0;
}
void bn_mul4x_mont_gather5(BN_ULONG *rp, const BN_ULONG *ap,
const BN_ULONG *table, const BN_ULONG *np,
const BN_ULONG *n0, int num, int power);
OPENSSL_INLINE int bn_mulx4x_mont_gather5_capable(int num) {
return bn_mul4x_mont_gather5_capable(num) && CRYPTO_is_ADX_capable() &&
CRYPTO_is_BMI1_capable() && CRYPTO_is_BMI2_capable();
}
void bn_mulx4x_mont_gather5(BN_ULONG *rp, const BN_ULONG *ap,
const BN_ULONG *table, const BN_ULONG *np,
const BN_ULONG *n0, int num, int power);
void bn_mul_mont_gather5_nohw(BN_ULONG *rp, const BN_ULONG *ap,
const BN_ULONG *table, const BN_ULONG *np,
const BN_ULONG *n0, int num, int power);
// bn_scatter5 stores |inp| to index |power| of |table|. |inp| and each entry of
// |table| are |num| words long. |power| must be less than 32 and is treated as
@@ -463,17 +471,19 @@ void bn_scatter5(const BN_ULONG *inp, size_t num, BN_ULONG *table,
// is treated as secret. |table| must be aligned to at least 16 bytes.
void bn_gather5(BN_ULONG *out, size_t num, const BN_ULONG *table, size_t power);
// bn_power5 squares |ap| five times and multiplies it by the value stored at
// index |power| of |table|, modulo |np|. It stores the result in |rp|. The
// values are |num| words long and represented in Montgomery form. |n0| is a
// pointer to the corresponding field in |BN_MONT_CTX|. |num| must be divisible
// by 8. |power| must be less than 32 and is treated as secret.
//
// WARNING: This function implements Almost Montgomery Multiplication from
// https://eprint.iacr.org/2011/239. The inputs do not need to be fully reduced.
// However, even if they are fully reduced, the output may not be.
void bn_power5(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *table,
const BN_ULONG *np, const BN_ULONG *n0, int num, int power);
// The following functions implement |bn_power5|. See |bn_power5| for details.
void bn_power5_nohw(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *table,
const BN_ULONG *np, const BN_ULONG *n0, int num, int power);
OPENSSL_INLINE int bn_power5_capable(int num) { return (num & 7) == 0; }
OPENSSL_INLINE int bn_powerx5_capable(int num) {
return bn_power5_capable(num) && CRYPTO_is_ADX_capable() &&
CRYPTO_is_BMI1_capable() && CRYPTO_is_BMI2_capable();
}
void bn_powerx5(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *table,
const BN_ULONG *np, const BN_ULONG *n0, int num, int power);
#endif // !OPENSSL_NO_ASM && OPENSSL_X86_64
uint64_t bn_mont_n0(const BIGNUM *n);
@@ -153,7 +153,7 @@ static uint64_t bn_neg_inv_mod_r_u64(uint64_t n) {
// The invariant now shows that u*r - v*n == 1 since r == 2 * alpha.
#if BN_BITS2 == 64 && defined(BN_ULLONG)
assert(1 == ((BN_ULLONG)u * 2 * alpha) - ((BN_ULLONG)v * beta));
declassify_assert(1 == ((BN_ULLONG)u * 2 * alpha) - ((BN_ULLONG)v * beta));
#endif
return v;
@@ -292,7 +292,7 @@ static void bn_mul_recursive(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
}
// The product should fit without carries.
assert(c == 0);
declassify_assert(c == 0);
}
// bn_mul_part_recursive sets |r| to |a| * |b|, using |t| as scratch space. |r|
@@ -406,7 +406,7 @@ static void bn_mul_part_recursive(BN_ULONG *r, const BN_ULONG *a,
}
// The product should fit without carries.
assert(c == 0);
declassify_assert(c == 0);
}
// bn_mul_impl implements |BN_mul| and |bn_mul_consttime|. Note this function
@@ -487,7 +487,10 @@ err:
static int bn_trial_division(uint16_t *out, const BIGNUM *bn) {
const size_t num_primes = num_trial_division_primes(bn);
for (size_t i = 1; i < num_primes; i++) {
if (bn_mod_u16_consttime(bn, kPrimes[i]) == 0) {
// During RSA key generation, |bn| may be secret, but only if |bn| was
// prime, so it is safe to leak failed trial divisions.
if (constant_time_declassify_int(bn_mod_u16_consttime(bn, kPrimes[i]) ==
0)) {
*out = kPrimes[i];
return 1;
}
@@ -573,7 +576,8 @@ int bn_miller_rabin_iteration(const BN_MILLER_RABIN *miller_rabin,
// To avoid leaking |a|, we run the loop to |w_bits| and mask off all
// iterations once |j| = |a|.
for (int j = 1; j < miller_rabin->w_bits; j++) {
if (constant_time_eq_int(j, miller_rabin->a) & ~is_possibly_prime) {
if (constant_time_declassify_w(constant_time_eq_int(j, miller_rabin->a) &
~is_possibly_prime)) {
// If the loop is done and we haven't seen z = 1 or z = w-1 yet, the
// value is composite and we can break in variable time.
break;
@@ -593,12 +597,14 @@ int bn_miller_rabin_iteration(const BN_MILLER_RABIN *miller_rabin,
// Step 4.5.3. If z = 1 and the loop is not done, the previous value of z
// was not -1. There are no non-trivial square roots of 1 modulo a prime, so
// w is composite and we may exit in variable time.
if (BN_equal_consttime(z, miller_rabin->one_mont) & ~is_possibly_prime) {
if (constant_time_declassify_w(
BN_equal_consttime(z, miller_rabin->one_mont) &
~is_possibly_prime)) {
break;
}
}
*out_is_possibly_prime = is_possibly_prime & 1;
*out_is_possibly_prime = constant_time_declassify_w(is_possibly_prime) & 1;
ret = 1;
err:
@@ -736,8 +742,9 @@ int BN_primality_test(int *out_is_probably_prime, const BIGNUM *w, int checks,
crypto_word_t uniform_iterations = 0;
// Using |constant_time_lt_w| seems to prevent the compiler from optimizing
// this into two jumps.
for (int i = 1; (i <= BN_PRIME_CHECKS_BLINDED) |
constant_time_lt_w(uniform_iterations, checks);
for (int i = 1; constant_time_declassify_w(
(i <= BN_PRIME_CHECKS_BLINDED) |
constant_time_lt_w(uniform_iterations, checks));
i++) {
// Step 4.1-4.2
int is_uniform;
@@ -766,7 +773,7 @@ int BN_primality_test(int *out_is_probably_prime, const BIGNUM *w, int checks,
}
}
assert(uniform_iterations >= (crypto_word_t)checks);
declassify_assert(uniform_iterations >= (crypto_word_t)checks);
*out_is_probably_prime = 1;
ret = 1;
@@ -113,10 +113,9 @@
#include <string.h>
#include <CCryptoBoringSSL_err.h>
#include <CCryptoBoringSSL_rand.h>
#include "../../internal.h"
#include "../rand/internal.h"
#include "../bcm_interface.h"
#include "../service_indicator/internal.h"
#include "internal.h"
@@ -157,7 +156,7 @@ int BN_rand(BIGNUM *rnd, int bits, int top, int bottom) {
}
FIPS_service_indicator_lock_state();
RAND_bytes((uint8_t *)rnd->d, words * sizeof(BN_ULONG));
BCM_rand_bytes((uint8_t *)rnd->d, words * sizeof(BN_ULONG));
FIPS_service_indicator_unlock_state();
rnd->d[words - 1] &= mask;
@@ -225,8 +224,7 @@ static int bn_range_to_mask(size_t *out_words, BN_ULONG *out_mask,
while (words > 0 && max_exclusive[words - 1] == 0) {
words--;
}
if (words == 0 ||
(words == 1 && max_exclusive[0] <= min_inclusive)) {
if (words == 0 || (words == 1 && max_exclusive[0] <= min_inclusive)) {
OPENSSL_PUT_ERROR(BN, BN_R_INVALID_RANGE);
return 0;
}
@@ -275,8 +273,8 @@ int bn_rand_range_words(BN_ULONG *out, BN_ULONG min_inclusive,
// Steps 4 and 5. Use |words| and |mask| together to obtain a string of N
// bits, where N is the bit length of |max_exclusive|.
FIPS_service_indicator_lock_state();
RAND_bytes_with_additional_data((uint8_t *)out, words * sizeof(BN_ULONG),
additional_data);
BCM_rand_bytes_with_additional_data(
(uint8_t *)out, words * sizeof(BN_ULONG), additional_data);
FIPS_service_indicator_unlock_state();
out[words - 1] &= mask;
@@ -326,7 +324,7 @@ int bn_rand_secret_range(BIGNUM *r, int *out_is_uniform, BN_ULONG min_inclusive,
// Select a uniform random number with num_bits(max_exclusive) bits.
FIPS_service_indicator_lock_state();
RAND_bytes((uint8_t *)r->d, words * sizeof(BN_ULONG));
BCM_rand_bytes((uint8_t *)r->d, words * sizeof(BN_ULONG));
FIPS_service_indicator_unlock_state();
r->d[words - 1] &= mask;
@@ -339,7 +337,8 @@ int bn_rand_secret_range(BIGNUM *r, int *out_is_uniform, BN_ULONG min_inclusive,
// If the value is not in range, force it to be in range.
r->d[0] |= constant_time_select_w(in_range, 0, min_inclusive);
r->d[words - 1] &= constant_time_select_w(in_range, BN_MASK2, mask >> 1);
assert(bn_in_range_words(r->d, min_inclusive, max_exclusive->d, words));
declassify_assert(
bn_in_range_words(r->d, min_inclusive, max_exclusive->d, words));
r->neg = 0;
r->width = (int)words;
@@ -236,7 +236,7 @@ BIGNUM *BN_mod_sqrt(BIGNUM *in, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx) {
goto end;
}
if (BN_ucmp(y, p) >= 0) {
if (!(p->neg ? BN_add : BN_sub)(y, y, p)) {
if (BN_usub(y, y, p)) {
goto end;
}
}
Loaded 100 of 618 files, more files were not shown because too many files have changed in this diff. Show more