Compare commits

..

1 Commits

Author SHA1 Message Date
Rasmus Wriedt Larsen
0019cb10d8 Revert "Python: Use new dataflow API" 2023-09-05 09:47:32 +02:00
1786 changed files with 96593 additions and 156946 deletions

6
.github/labeler.yml vendored
View File

@@ -45,7 +45,11 @@ documentation:
# Since these are all shared files that need to be synced, just pick _one_ copy of each.
"DataFlow Library":
- "shared/dataflow/**/*"
- "java/ql/lib/semmle/code/java/dataflow/internal/DataFlowImpl.qll"
- "java/ql/lib/semmle/code/java/dataflow/internal/DataFlowImplCommon.qll"
- "java/ql/lib/semmle/code/java/dataflow/internal/tainttracking1/TaintTrackingImpl.qll"
- "java/ql/lib/semmle/code/java/dataflow/internal/DataFlowImplConsistency.qll"
- "java/ql/lib/semmle/code/java/dataflow/internal/FlowSummaryImpl.qll"
"ATM":
- javascript/ql/experimental/adaptivethreatmodeling/**/*

View File

@@ -13,7 +13,7 @@ jobs:
check:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Check that implicit this warnings is enabled for all packs
shell: bash
run: |

View File

@@ -15,7 +15,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
with:
fetch-depth: 2

View File

@@ -16,6 +16,6 @@ jobs:
name: Check query IDs
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Check for duplicate query IDs
run: python3 misc/scripts/check-query-ids.py

View File

@@ -33,7 +33,7 @@ jobs:
dotnet-version: 7.0.102
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v3
# Initializes the CodeQL tools for scanning.
- name: Initialize CodeQL

View File

@@ -13,7 +13,7 @@ jobs:
runs-on: ubuntu-latest-xl
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Setup CodeQL
uses: ./.github/actions/fetch-codeql
with:

View File

@@ -29,7 +29,7 @@ jobs:
qlupgrade:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./.github/actions/fetch-codeql
- name: Check DB upgrade scripts
run: |
@@ -52,7 +52,8 @@ jobs:
matrix:
slice: ["1/2", "2/2"]
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./.github/actions/fetch-codeql
- uses: ./csharp/actions/create-extractor-pack
- name: Cache compilation cache
id: query-cache
@@ -61,41 +62,25 @@ jobs:
key: csharp-qltest-${{ matrix.slice }}
- name: Run QL tests
run: |
codeql test run --threads=0 --ram 50000 --slice ${{ matrix.slice }} --search-path extractor-pack --check-databases --check-undefined-labels --check-repeated-labels --check-redefined-labels --consistency-queries ql/consistency-queries ql/test --compilation-cache "${{ steps.query-cache.outputs.cache-dir }}"
CODEQL_PATH=$(gh codeql version --format=json | jq -r .unpackedLocation)
# The legacy ASP extractor is not in this repo, so take the one from the nightly build
mv "$CODEQL_PATH/csharp/tools/extractor-asp.jar" "${{ github.workspace }}/csharp/extractor-pack/tools"
# Safe guard against using the bundled extractor
rm -rf "$CODEQL_PATH/csharp"
codeql test run --threads=0 --ram 50000 --slice ${{ matrix.slice }} --search-path "${{ github.workspace }}/csharp/extractor-pack" --check-databases --check-undefined-labels --check-repeated-labels --check-redefined-labels --consistency-queries ql/consistency-queries ql/test --compilation-cache "${{ steps.query-cache.outputs.cache-dir }}"
env:
GITHUB_TOKEN: ${{ github.token }}
unit-tests:
strategy:
matrix:
os: [ubuntu-latest, windows-2019]
runs-on: ${{ matrix.os }}
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Setup dotnet
uses: actions/setup-dotnet@v3
with:
dotnet-version: 7.0.102
- name: Extractor unit tests
run: |
dotnet test -p:RuntimeFrameworkVersion=7.0.2 extractor/Semmle.Util.Tests
dotnet test -p:RuntimeFrameworkVersion=7.0.2 extractor/Semmle.Extraction.Tests
dotnet test -p:RuntimeFrameworkVersion=7.0.2 autobuilder/Semmle.Autobuild.CSharp.Tests
dotnet test -p:RuntimeFrameworkVersion=7.0.2 "${{ github.workspace }}/csharp/extractor/Semmle.Util.Tests"
dotnet test -p:RuntimeFrameworkVersion=7.0.2 "${{ github.workspace }}/csharp/extractor/Semmle.Extraction.Tests"
dotnet test -p:RuntimeFrameworkVersion=7.0.2 "${{ github.workspace }}/csharp/autobuilder/Semmle.Autobuild.CSharp.Tests"
dotnet test -p:RuntimeFrameworkVersion=7.0.2 "${{ github.workspace }}/cpp/autobuilder/Semmle.Autobuild.Cpp.Tests"
shell: bash
stubgentest:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: ./csharp/actions/create-extractor-pack
- name: Run stub generator tests
run: |
# Generate (Asp)NetCore stubs
STUBS_PATH=stubs_output
python3 ql/src/Stubs/make_stubs_nuget.py webapp Swashbuckle.AspNetCore.Swagger latest "$STUBS_PATH"
rm -rf ql/test/resources/stubs/_frameworks
# Update existing stubs in the repo with the freshly generated ones
mv "$STUBS_PATH/output/stubs/_frameworks" ql/test/resources/stubs/
git status
codeql test run --threads=0 --search-path extractor-pack --check-databases --check-undefined-labels --check-repeated-labels --check-redefined-labels --consistency-queries ql/consistency-queries -- ql/test/library-tests/dataflow/flowsources/aspremote
env:
GITHUB_TOKEN: ${{ github.token }}

View File

@@ -19,7 +19,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v3
- name: Setup CodeQL
uses: ./.github/actions/fetch-codeql
- name: Create empty database
@@ -47,7 +47,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v3
- name: Setup CodeQL
uses: ./.github/actions/fetch-codeql
- name: Create empty database

View File

@@ -31,11 +31,11 @@ jobs:
GITHUB_CONTEXT: ${{ toJSON(github.event) }}
run: echo "$GITHUB_CONTEXT"
- name: Clone self (github/codeql) - MERGE
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: merge
- name: Clone self (github/codeql) - BASE
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
fetch-depth: 2
path: base

View File

@@ -20,7 +20,7 @@ jobs:
GITHUB_CONTEXT: ${{ toJSON(github.event) }}
run: echo "$GITHUB_CONTEXT"
- name: Clone self (github/codeql)
uses: actions/checkout@v4
uses: actions/checkout@v3
- name: Set up Python 3.8
uses: actions/setup-python@v4
with:

View File

@@ -9,11 +9,11 @@ jobs:
steps:
- name: Clone self (github/codeql)
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: script
- name: Clone self (github/codeql) for analysis
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: codeqlModels
fetch-depth: 0

View File

@@ -17,7 +17,7 @@ jobs:
GITHUB_CONTEXT: ${{ toJSON(github.event) }}
run: echo "$GITHUB_CONTEXT"
- name: Clone self (github/codeql)
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: ql
fetch-depth: 0

View File

@@ -13,11 +13,11 @@ jobs:
steps:
- name: Clone self (github/codeql)
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: script
- name: Clone self (github/codeql) for analysis
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: codeqlModels
ref: ${{ github.event.inputs.qlModelShaOverride || github.ref }}

View File

@@ -25,7 +25,7 @@ jobs:
exit 1
- name: Checkout
uses: actions/checkout@v4
uses: actions/checkout@v3
- name: Git config
shell: bash

View File

@@ -21,7 +21,7 @@ jobs:
id: go
- name: Check out code
uses: actions/checkout@v4
uses: actions/checkout@v2
- name: Set up CodeQL CLI
uses: ./.github/actions/fetch-codeql
@@ -56,7 +56,7 @@ jobs:
id: go
- name: Check out code
uses: actions/checkout@v4
uses: actions/checkout@v2
- name: Set up CodeQL CLI
uses: ./.github/actions/fetch-codeql

View File

@@ -29,7 +29,7 @@ jobs:
id: go
- name: Check out code
uses: actions/checkout@v4
uses: actions/checkout@v2
- name: Set up CodeQL CLI
uses: ./.github/actions/fetch-codeql

65
.github/workflows/js-ml-tests.yml vendored Normal file
View File

@@ -0,0 +1,65 @@
name: JS ML-powered queries tests
on:
push:
paths:
- "javascript/ql/experimental/adaptivethreatmodeling/**"
- .github/workflows/js-ml-tests.yml
- .github/actions/fetch-codeql/action.yml
- codeql-workspace.yml
branches:
- main
- "rc/*"
pull_request:
paths:
- "javascript/ql/experimental/adaptivethreatmodeling/**"
- .github/workflows/js-ml-tests.yml
- .github/actions/fetch-codeql/action.yml
- codeql-workspace.yml
workflow_dispatch:
defaults:
run:
working-directory: javascript/ql/experimental/adaptivethreatmodeling
jobs:
qltest:
name: Test QL
runs-on: ubuntu-latest-xl
steps:
- uses: actions/checkout@v3
- uses: ./.github/actions/fetch-codeql
- name: Install pack dependencies
run: |
for pack in modelbuilding src test; do
codeql pack install --mode verify -- "${pack}"
done
- name: Cache compilation cache
id: query-cache
uses: ./.github/actions/cache-query-compilation
with:
key: js-ml-test
- name: Check QL compilation
run: |
codeql query compile \
--check-only \
--ram 50000 \
--additional-packs "${{ github.workspace }}" \
--threads=0 \
--compilation-cache "${{ steps.query-cache.outputs.cache-dir }}" \
-- \
lib modelbuilding src
- name: Run QL tests
run: |
codeql test run \
--threads=0 \
--ram 50000 \
--additional-packs "${{ github.workspace }}" \
--compilation-cache "${{ steps.query-cache.outputs.cache-dir }}" \
-- \
test

View File

@@ -27,12 +27,12 @@ jobs:
slug: ${{fromJson(github.event.inputs.projects || '["apache/commons-codec", "apache/commons-io", "apache/commons-beanutils", "apache/commons-logging", "apache/commons-fileupload", "apache/commons-lang", "apache/commons-validator", "apache/commons-csv", "apache/dubbo"]' )}}
steps:
- name: Clone github/codeql from PR
uses: actions/checkout@v4
uses: actions/checkout@v3
if: github.event.pull_request
with:
path: codeql-pr
- name: Clone github/codeql from main
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: codeql-main
ref: main

View File

@@ -27,11 +27,11 @@ jobs:
ref: "placeholder"
steps:
- name: Clone self (github/codeql)
uses: actions/checkout@v4
uses: actions/checkout@v3
- name: Setup CodeQL binaries
uses: ./.github/actions/fetch-codeql
- name: Clone repositories
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: repos/${{ matrix.ref }}
ref: ${{ matrix.ref }}

View File

@@ -43,7 +43,7 @@ jobs:
if-no-files-found: error
retention-days: 1
- uses: actions/checkout@v4
- uses: actions/checkout@v3
with:
fetch-depth: 2
persist-credentials: false

View File

@@ -14,7 +14,7 @@ jobs:
runs-on: ubuntu-latest-xl
steps:
### Build the queries ###
- uses: actions/checkout@v4
- uses: actions/checkout@v3
with:
fetch-depth: 0
- name: Find codeql

View File

@@ -21,7 +21,7 @@ jobs:
- github/codeql
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Find codeql
id: find-codeql
@@ -42,7 +42,7 @@ jobs:
env:
CODEQL: ${{ steps.find-codeql.outputs.codeql-path }}
- name: Checkout ${{ matrix.repo }}
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
repository: ${{ matrix.repo }}
path: ${{ github.workspace }}/repo
@@ -71,7 +71,7 @@ jobs:
runs-on: ubuntu-latest
needs: measure
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: actions/download-artifact@v3
with:
name: measurements

View File

@@ -21,7 +21,7 @@ jobs:
qltest:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Find codeql
id: find-codeql
uses: github/codeql-action/init@v2
@@ -61,7 +61,7 @@ jobs:
needs: [qltest]
runs-on: ${{ matrix.os }}
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Install GNU tar
if: runner.os == 'macOS'
run: |

View File

@@ -20,7 +20,7 @@ jobs:
steps:
- name: Clone self (github/codeql)
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
path: codeql
- name: Set up Python 3.8

View File

@@ -42,7 +42,7 @@ jobs:
runs-on: ${{ matrix.os }}
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Install GNU tar
if: runner.os == 'macOS'
run: |
@@ -113,7 +113,7 @@ jobs:
compile-queries:
runs-on: ubuntu-latest-xl
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Fetch CodeQL
uses: ./.github/actions/fetch-codeql
- name: Cache compilation cache
@@ -145,7 +145,7 @@ jobs:
runs-on: ubuntu-latest
needs: [build, compile-queries]
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: actions/download-artifact@v3
with:
name: ruby.dbscheme
@@ -206,7 +206,7 @@ jobs:
runs-on: ${{ matrix.os }}
needs: [package]
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Fetch CodeQL
uses: ./.github/actions/fetch-codeql

View File

@@ -27,14 +27,14 @@ jobs:
repo: [rails/rails, discourse/discourse, spree/spree, ruby/ruby]
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./.github/actions/fetch-codeql
- uses: ./ruby/actions/create-extractor-pack
- name: Checkout ${{ matrix.repo }}
uses: actions/checkout@v4
uses: actions/checkout@v3
with:
repository: ${{ matrix.repo }}
path: ${{ github.workspace }}/repo
@@ -59,7 +59,7 @@ jobs:
runs-on: ubuntu-latest
needs: measure
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: actions/download-artifact@v3
with:
name: measurements

View File

@@ -33,7 +33,7 @@ jobs:
qlupgrade:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./.github/actions/fetch-codeql
- name: Check DB upgrade scripts
run: |
@@ -54,7 +54,7 @@ jobs:
strategy:
fail-fast: false
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./.github/actions/fetch-codeql
- uses: ./ruby/actions/create-extractor-pack
- name: Cache compilation cache

View File

@@ -39,31 +39,31 @@ jobs:
build-and-test-macos:
runs-on: macos-12-xl
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./swift/actions/build-and-test
build-and-test-linux:
runs-on: ubuntu-latest-xl
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./swift/actions/build-and-test
qltests-linux:
needs: build-and-test-linux
runs-on: ubuntu-latest-xl
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./swift/actions/run-ql-tests
qltests-macos:
if : ${{ github.event_name == 'pull_request' }}
needs: build-and-test-macos
runs-on: macos-12-xl
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./swift/actions/run-ql-tests
integration-tests-linux:
needs: build-and-test-linux
runs-on: ubuntu-latest-xl
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./swift/actions/run-integration-tests
integration-tests-macos:
if : ${{ github.event_name == 'pull_request' }}
@@ -71,13 +71,13 @@ jobs:
runs-on: macos-12-xl
timeout-minutes: 60
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./swift/actions/run-integration-tests
codegen:
if : ${{ github.event_name == 'pull_request' }}
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: bazelbuild/setup-bazelisk@v2
- uses: actions/setup-python@v4
with:
@@ -102,6 +102,6 @@ jobs:
if : ${{ github.event_name == 'pull_request' }}
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- uses: ./.github/actions/fetch-codeql
- uses: ./swift/actions/database-upgrade-scripts

View File

@@ -14,7 +14,7 @@ jobs:
sync:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Check synchronized files
run: python config/sync-files.py
- name: Check dbscheme fragments

View File

@@ -27,7 +27,7 @@ jobs:
test:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Check formatting
run: cargo fmt --all -- --check
- name: Run tests
@@ -35,12 +35,12 @@ jobs:
fmt:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Check formatting
run: cargo fmt --check
clippy:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Run clippy
run: cargo clippy -- --no-deps -D warnings -A clippy::new_without_default -A clippy::too_many_arguments

View File

@@ -20,7 +20,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v3
- name: Setup CodeQL
uses: ./.github/actions/fetch-codeql

View File

@@ -28,6 +28,8 @@
"python/ql/lib/semmle/python/dataflow/new/internal/DataFlowImpl4.qll",
"ruby/ql/lib/codeql/ruby/dataflow/internal/DataFlowImpl1.qll",
"ruby/ql/lib/codeql/ruby/dataflow/internal/DataFlowImpl2.qll",
"ruby/ql/lib/codeql/ruby/dataflow/internal/DataFlowImplForHttpClientLibraries.qll",
"ruby/ql/lib/codeql/ruby/dataflow/internal/DataFlowImplForPathname.qll",
"swift/ql/lib/codeql/swift/dataflow/internal/DataFlowImpl1.qll"
],
"TaintTracking Legacy Configuration Java/C++/C#/Go/Python/Ruby/Swift": [
@@ -550,4 +552,4 @@
"python/ql/test/experimental/dataflow/model-summaries/InlineTaintTest.ext.yml",
"python/ql/test/experimental/dataflow/model-summaries/NormalDataflowTest.ext.yml"
]
}
}

View File

@@ -1,17 +1,12 @@
load("@rules_pkg//:mappings.bzl", "pkg_filegroup")
package(default_visibility = ["//visibility:public"])
load("@rules_pkg//:mappings.bzl", "pkg_filegroup")
alias(
name = "dbscheme",
actual = "//cpp/ql/lib:dbscheme",
)
alias(
name = "dbscheme-stats",
actual = "//cpp/ql/lib:dbscheme-stats",
)
pkg_filegroup(
name = "db-files",
srcs = [

View File

@@ -11,12 +11,12 @@
<ItemGroup>
<PackageReference Include="System.IO.FileSystem" Version="4.3.0" />
<PackageReference Include="System.IO.FileSystem.Primitives" Version="4.3.0" />
<PackageReference Include="xunit" Version="2.4.2" />
<PackageReference Include="xunit.runner.visualstudio" Version="2.4.5">
<PackageReference Include="xunit" Version="2.5.0" />
<PackageReference Include="xunit.runner.visualstudio" Version="2.5.0">
<PrivateAssets>all</PrivateAssets>
<IncludeAssets>runtime; build; native; contentfiles; analyzers</IncludeAssets>
</PackageReference>
<PackageReference Include="Microsoft.NET.Test.Sdk" Version="17.4.0" />
<PackageReference Include="Microsoft.NET.Test.Sdk" Version="17.7.1" />
</ItemGroup>
<ItemGroup>

View File

@@ -17,7 +17,7 @@
</ItemGroup>
<ItemGroup>
<PackageReference Include="Microsoft.Build" Version="17.3.2" />
<PackageReference Include="Microsoft.Build" Version="17.7.2" />
</ItemGroup>
<ItemGroup>

View File

@@ -1,2 +0,0 @@
description: Make __is_trivial a builtin operation
compatibility: full

View File

@@ -1,7 +1,7 @@
load("@rules_pkg//:mappings.bzl", "pkg_files")
package(default_visibility = ["//cpp:__pkg__"])
load("@rules_pkg//:mappings.bzl", "pkg_files")
pkg_files(
name = "dbscheme",
srcs = ["semmlecode.cpp.dbscheme"],

View File

@@ -1,22 +1,3 @@
## 0.9.3
No user-facing changes.
## 0.9.2
### Deprecated APIs
* `getAllocatorCall` on `DeleteExpr` and `DeleteArrayExpr` has been deprecated. `getDeallocatorCall` should be used instead.
### New Features
* Added `DeleteOrDeleteArrayExpr` as a super type of `DeleteExpr` and `DeleteArrayExpr`
### Minor Analysis Improvements
* `delete` and `delete[]` are now modeled as calls to the relevant `operator delete` in the IR. In the case of a dynamic delete call a new instruction `VirtualDeleteFunctionAddress` is used to represent a function that dispatches to the correct delete implementation.
* Only the 2 level indirection of `argv` (corresponding to `**argv`) is consided for `FlowSource`.
## 0.9.1
No user-facing changes.

View File

@@ -0,0 +1,4 @@
---
category: minorAnalysis
---
* Only the 2 level indirection of `argv` (corresponding to `**argv`) is consided for `FlowSource`.

View File

@@ -0,0 +1,4 @@
---
category: feature
---
* Added `DeleteOrDeleteArrayExpr` as a super type of `DeleteExpr` and `DeleteArrayExpr`

View File

@@ -0,0 +1,4 @@
---
category: deprecated
---
* `getAllocatorCall` on `DeleteExpr` and `DeleteArrayExpr` has been deprecated. `getDeallocatorCall` should be used instead.

View File

@@ -0,0 +1,4 @@
---
category: minorAnalysis
---
* `delete` and `delete[]` are now modeled as calls to the relevant `operator delete` in the IR. In the case of a dynamic delete call a new instruction `VirtualDeleteFunctionAddress` is used to represent a function that dispatches to the correct delete implementation.

View File

@@ -1,4 +0,0 @@
---
category: minorAnalysis
---
* The `DataFlow::asDefiningArgument` predicate now takes its argument from the range starting at `1` instead of `2`. Queries that depend on the single-parameter version of `DataFlow::asDefiningArgument` should have their arguments updated accordingly.

View File

@@ -1,5 +0,0 @@
---
category: minorAnalysis
---
* Treat functions that reach the end of the function as returning in the IR.
They used to be treated as unreachable but it is allowed in C.

View File

@@ -1,5 +0,0 @@
---
category: minorAnalysis
---
* Functions that do not return due to calling functions that don't return (e.g. `exit`) are now detected as
non-returning in the IR and dataflow.

View File

@@ -1,14 +0,0 @@
## 0.9.2
### Deprecated APIs
* `getAllocatorCall` on `DeleteExpr` and `DeleteArrayExpr` has been deprecated. `getDeallocatorCall` should be used instead.
### New Features
* Added `DeleteOrDeleteArrayExpr` as a super type of `DeleteExpr` and `DeleteArrayExpr`
### Minor Analysis Improvements
* `delete` and `delete[]` are now modeled as calls to the relevant `operator delete` in the IR. In the case of a dynamic delete call a new instruction `VirtualDeleteFunctionAddress` is used to represent a function that dispatches to the correct delete implementation.
* Only the 2 level indirection of `argv` (corresponding to `**argv`) is consided for `FlowSource`.

View File

@@ -1,3 +0,0 @@
## 0.9.3
No user-facing changes.

View File

@@ -1,2 +1,2 @@
---
lastReleaseVersion: 0.9.3
lastReleaseVersion: 0.9.1

View File

@@ -1,3 +0,0 @@
import experimental.cryptography.CryptoArtifact
import experimental.cryptography.CryptoAlgorithmNames
import experimental.cryptography.modules.OpenSSL as OpenSSL

View File

@@ -1,239 +0,0 @@
/**
* Names of known cryptographic algorithms.
* The names are standardized into upper-case, no spaces, dashes or underscores.
*/
/**
* Returns a string to represent generally unknown algorithms.
* Predicate is to be used to get a consistent string representation
* for unknown algorithms.
*/
string unknownAlgorithm() { result = "UNKNOWN" }
string getHashType() { result = "HASH" }
string getSymmetricEncryptionType() { result = "SYMMETRIC_ENCRYPTION" }
string getAsymmetricEncryptionType() { result = "ASYMMETRIC_ENCRYPTION" }
string getKeyDerivationType() { result = "KEY_DERIVATION" }
string getCipherBlockModeType() { result = "BLOCK_MODE" }
string getSymmetricPaddingType() { result = "SYMMETRIC_PADDING" }
string getAsymmetricPaddingType() { result = "ASYMMETRIC_PADDING" }
string getEllipticCurveType() { result = "ELLIPTIC_CURVE" }
string getSignatureType() { result = "SIGNATURE" }
string getKeyExchangeType() { result = "KEY_EXCHANGE" }
string getAsymmetricType() {
result in [
getAsymmetricEncryptionType(), getSignatureType(), getKeyExchangeType(),
getEllipticCurveType()
]
}
predicate isKnownType(string algType) {
algType in [
getHashType(), getSymmetricEncryptionType(), getAsymmetricEncryptionType(),
getKeyDerivationType(), getCipherBlockModeType(), getSymmetricPaddingType(),
getAsymmetricPaddingType(), getEllipticCurveType(), getSignatureType(), getKeyExchangeType()
]
}
predicate isKnownAlgorithm(string name) { isKnownAlgorithm(name, _) }
predicate isKnownAlgorithm(string name, string algType) {
isHashingAlgorithm(name) and algType = "HASH"
or
isEncryptionAlgorithm(name, algType) and
algType in ["SYMMETRIC_ENCRYPTION", "ASYMMETRIC_ENCRYPTION"]
or
isKeyDerivationAlgorithm(name) and algType = "KEY_DERIVATION"
or
isCipherBlockModeAlgorithm(name) and algType = "BLOCK_MODE"
or
isPaddingAlgorithm(name, algType) and algType in ["SYMMETRIC_PADDING", "ASYMMETRIC_PADDING"]
or
isEllipticCurveAlgorithm(name) and algType = "ELLIPTIC_CURVE"
or
isSignatureAlgorithm(name) and algType = "SIGNATURE"
or
isKeyExchangeAlgorithm(name) and algType = "KEY_EXCHANGE"
}
/**
* Holds if `name` is a known hashing algorithm in the model/library.
*/
predicate isHashingAlgorithm(string name) {
name =
[
"BLAKE2", "BLAKE2B", "BLAKE2S", "SHA2", "SHA224", "SHA256", "SHA384", "SHA512", "SHA512224",
"SHA512256", "SHA3", "SHA3224", "SHA3256", "SHA3384", "SHA3512", "SHAKE128", "SHAKE256",
"SM3", "WHIRLPOOL", "POLY1305", "HAVEL128", "MD2", "MD4", "MD5", "PANAMA", "RIPEMD",
"RIPEMD128", "RIPEMD256", "RIPEMD160", "RIPEMD320", "SHA0", "SHA1", "SHA", "MGF1", "MGF1SHA1",
"MDC2", "SIPHASH"
]
}
predicate isEncryptionAlgorithm(string name, string algType) {
isAsymmetricEncryptionAlgorithm(name) and algType = "ASYMMETRIC_ENCRYPTION"
or
isSymmetricEncryptionAlgorithm(name) and algType = "SYMMETRIC_ENCRYPTION"
}
predicate isEncryptionAlgorithm(string name) { isEncryptionAlgorithm(name, _) }
/**
* Holds if `name` corresponds to a known symmetric encryption algorithm.
*/
predicate isSymmetricEncryptionAlgorithm(string name) {
// NOTE: AES is meant to caputure all possible key lengths
name =
[
"AES", "AES128", "AES192", "AES256", "ARIA", "BLOWFISH", "BF", "ECIES", "CAST", "CAST5",
"CAMELLIA", "CAMELLIA128", "CAMELLIA192", "CAMELLIA256", "CHACHA", "CHACHA20",
"CHACHA20POLY1305", "GOST", "GOSTR34102001", "GOSTR341094", "GOSTR341194", "GOST2814789",
"GOSTR341194", "GOST2814789", "GOST28147", "GOSTR341094", "GOST89", "GOST94", "GOST34102012",
"GOST34112012", "IDEA", "RABBIT", "SEED", "SM4", "DES", "DESX", "3DES", "TDES", "2DES",
"DES3", "TRIPLEDES", "TDEA", "TRIPLEDEA", "ARC2", "RC2", "ARC4", "RC4", "ARCFOUR", "ARC5",
"RC5", "MAGMA", "KUZNYECHIK"
]
}
/**
* Holds if `name` corresponds to a known key derivation algorithm.
*/
predicate isKeyDerivationAlgorithm(string name) {
name =
[
"ARGON2", "CONCATKDF", "CONCATKDFHASH", "CONCATKDFHMAC", "KBKDFCMAC", "BCRYPT", "HKDF",
"HKDFEXPAND", "KBKDF", "KBKDFHMAC", "PBKDF1", "PBKDF2", "PBKDF2HMAC", "PKCS5", "SCRYPT",
"X963KDF", "EVPKDF"
]
}
/**
* Holds if `name` corresponds to a known cipher block mode
*/
predicate isCipherBlockModeAlgorithm(string name) {
name = ["CBC", "GCM", "CCM", "CFB", "OFB", "CFB8", "CTR", "OPENPGP", "XTS", "EAX", "SIV", "ECB"]
}
/**
* Holds if `name` corresponds to a known padding algorithm
*/
predicate isPaddingAlgorithm(string name, string algType) {
isSymmetricPaddingAlgorithm(name) and algType = "SYMMETRIC_PADDING"
or
isAsymmetricPaddingAlgorithm(name) and algType = "ASYMMETRIC_PADDING"
}
/**
* holds if `name` corresponds to a known symmetric padding algorithm
*/
predicate isSymmetricPaddingAlgorithm(string name) { name = ["PKCS7", "ANSIX923"] }
/**
* Holds if `name` corresponds to a known asymmetric padding algorithm
*/
predicate isAsymmetricPaddingAlgorithm(string name) { name = ["OAEP", "PKCS1V15", "PSS", "KEM"] }
predicate isBrainpoolCurve(string curveName, int keySize) {
// ALL BRAINPOOL CURVES
keySize in [160, 192, 224, 256, 320, 384, 512] and
(
curveName = "BRAINPOOLP" + keySize.toString() + "R1"
or
curveName = "BRAINPOOLP" + keySize.toString() + "T1"
)
}
predicate isSecCurve(string curveName, int keySize) {
// ALL SEC CURVES
keySize in [112, 113, 128, 131, 160, 163, 192, 193, 224, 233, 239, 256, 283, 384, 409, 521, 571] and
exists(string suff | suff in ["R1", "R2", "K1"] |
curveName = "SECT" + keySize.toString() + suff or
curveName = "SECP" + keySize.toString() + suff
)
}
predicate isC2Curve(string curveName, int keySize) {
// ALL C2 CURVES
keySize in [163, 176, 191, 208, 239, 272, 304, 359, 368, 431] and
exists(string pre, string suff |
pre in ["PNB", "ONB", "TNB"] and suff in ["V1", "V2", "V3", "V4", "V5", "W1", "R1"]
|
curveName = "C2" + pre + keySize.toString() + suff
)
}
predicate isPrimeCurve(string curveName, int keySize) {
// ALL PRIME CURVES
keySize in [192, 239, 256] and
exists(string suff | suff in ["V1", "V2", "V3"] | curveName = "PRIME" + keySize.toString() + suff)
}
predicate isEllipticCurveAlgorithm(string curveName) { isEllipticCurveAlgorithm(curveName, _) }
/**
* Holds if `name` corresponds to a known elliptic curve.
*/
predicate isEllipticCurveAlgorithm(string curveName, int keySize) {
isSecCurve(curveName, keySize)
or
isBrainpoolCurve(curveName, keySize)
or
isC2Curve(curveName, keySize)
or
isPrimeCurve(curveName, keySize)
or
curveName = "ES256" and keySize = 256
or
curveName = "CURVE25519" and keySize = 255
or
curveName = "X25519" and keySize = 255
or
curveName = "ED25519" and keySize = 255
or
curveName = "CURVE448" and keySize = 448 // TODO: need to check the key size
or
curveName = "ED448" and keySize = 448
or
curveName = "X448" and keySize = 448
or
curveName = "NUMSP256T1" and keySize = 256
or
curveName = "NUMSP384T1" and keySize = 384
or
curveName = "NUMSP512T1" and keySize = 512
or
curveName = "SM2" and keySize in [256, 512]
}
/**
* Holds if `name` corresponds to a known signature algorithm.
*/
predicate isSignatureAlgorithm(string name) {
name =
[
"DSA", "ECDSA", "EDDSA", "ES256", "ES256K", "ES384", "ES512", "ED25519", "ED448", "ECDSA256",
"ECDSA384", "ECDSA512"
]
}
/**
* Holds if `name` is a key exchange algorithm.
*/
predicate isKeyExchangeAlgorithm(string name) {
name = ["ECDH", "DH", "DIFFIEHELLMAN", "X25519", "X448"]
}
/**
* Holds if `name` corresponds to a known asymmetric encryption.
*/
predicate isAsymmetricEncryptionAlgorithm(string name) { name = ["RSA"] }

View File

@@ -1,316 +0,0 @@
import cpp
private import experimental.cryptography.CryptoAlgorithmNames
import semmle.code.cpp.ir.dataflow.TaintTracking
/*
* A cryptographic artifact is a DataFlow::Node associated with some
* operation, algorithm, or any other aspect of cryptography.
*/
abstract class CryptographicArtifact extends Expr { }
// /**
// * Associates a symmetric encryption algorithm with a block mode.
// * The DataFlow::Node representing this association should be the
// * point where the algorithm and block mode are combined.
// * This may be at the call to encryption or in the construction
// * of an object prior to encryption.
// */
// abstract class SymmetricCipher extends CryptographicArtifact{
// abstract SymmetricEncryptionAlgorithm getEncryptionAlgorithm();
// abstract BlockMode getBlockMode();
// final predicate hasBlockMode(){
// exists(this.getBlockMode())
// }
// }
// /**
// * A cryptographic operation is a method call that invokes a cryptographic
// * algorithm (encrypt/decrypt) or a function in support of a cryptographic algorithm
// * (key generation).
// *
// * Since operations are related to or in support of algorithms, operations must
// * provide a reference to their associated algorithm. Often operataions themselves
// * encapsulate algorithms, so operations can also extend CryptographicAlgorithm
// * and refer to themselves as the target algorithm.
// */
// abstract class CryptographicOperation extends CryptographicArtifact, Call{
// // bindingset[paramName, ind]
// // final DataFlow::Node getParameterSource(int ind, string paramName){
// // result = Utils::getUltimateSrcFromApiNode(this.(API::CallNode).getParameter(ind, paramName))
// // }
// final string getAlgorithmName(){
// if exists(this.getAlgorithm().getName())
// then result = this.getAlgorithm().getName()
// else result = unknownAlgorithm()
// }
// final predicate hasAlgorithm(){
// exists(this.getAlgorithm())
// }
// final predicate isUnknownAlgorithm(){
// this.getAlgorithmName() = unknownAlgorithm()
// or
// not this.hasAlgorithm()
// }
// // TODO: this might have to be parameterized by a configuration source for
// // situations where an operation is passed an algorithm
// abstract CryptographicAlgorithm getAlgorithm();
// }
// /** A key generation operation for asymmetric keys */
// abstract class KeyGen extends CryptographicOperation{
// int getAKeySizeInBits(){
// result = getKeySizeInBits(_)
// }
// final predicate hasKeySize(Expr configSrc){
// exists(this.getKeySizeInBits(configSrc))
// }
// final predicate hasKeySize(){
// exists(this.getAKeySizeInBits())
// }
// abstract Expr getKeyConfigSrc();
// abstract int getKeySizeInBits(Expr configSrc);
// }
abstract class CryptographicOperation extends CryptographicArtifact, Call { }
abstract class KeyGeneration extends CryptographicOperation {
// TODO: what if the algorithm is UNKNOWN?
abstract Expr getKeyConfigurationSource(CryptographicAlgorithm alg);
abstract CryptographicAlgorithm getAlgorithm();
int getKeySizeInBits(CryptographicAlgorithm alg) {
result = this.getKeyConfigurationSource(alg).(Literal).getValue().toInt()
}
predicate hasConstantKeySize(CryptographicAlgorithm alg) { exists(this.getKeySizeInBits(alg)) }
predicate hasKeyConfigurationSource(CryptographicAlgorithm alg) {
exists(this.getKeyConfigurationSource(alg))
}
Expr getAKeyConfigurationSource() { result = this.getKeyConfigurationSource(_) }
}
abstract class AsymmetricKeyGeneration extends KeyGeneration { }
abstract class SymmetricKeyGeneration extends KeyGeneration { }
/**
* A cryptographic algorithm is a `CryptographicArtifact`
* representing a cryptographic algorithm (see `CryptoAlgorithmNames.qll`).
* Cryptographic algorithms can be functions referencing common crypto algorithms (e.g., hashlib.md5)
* or strings that are used in cryptographic operation configurations (e.g., hashlib.new("md5")).
* Cryptogrpahic algorithms may also be operations that wrap or abstract one or
* more algorithms (e.g., cyrptography.fernet.Fernet and AES, CBC and PKCS7).
*
* In principle, this class should model the location where an algorithm enters the program, not
* necessarily where it is used.
*/
abstract class CryptographicAlgorithm extends CryptographicArtifact {
abstract string getName();
abstract string getAlgType();
// string getAlgType(){
// if this instanceof HashAlgorithm then result = getHashType()
// else if this instanceof KeyDerivationAlgorithm then result = getKeyDerivationType()
// else if this instanceof SymmetricEncryptionAlgorithm then result = getSymmetricEncryptionType()
// else if this instanceof AsymmetricEncryptionAlgorithm then result = getAsymmetricEncryptionType()
// else if this instanceof SymmetricEncryptionAlgorithm then result = getSymmetricPaddingType()
// else if this instanceof AsymmetricEncryptionAlgorithm then result = getAsymmetricPaddingType()
// else if this instanceof EllipticCurveAlgorithm then result = getEllipticCurveType()
// else if this instanceof BlockMode then result = getCipherBlockModeType()
// else if this instanceof KeyExchangeAlgorithm then result = getKeyExchangeType()
// else if this instanceof SigningAlgorithm then result = getSignatureType()
// else result = unknownAlgorithm()
// }
// TODO: handle case where name isn't known, not just unknown?
/**
* Normalizes a raw name into a normalized name as found in `CryptoAlgorithmNames.qll`.
* Subclassess should override for more api-specific normalization.
* By deafult, converts a raw name to upper-case with no hyphen, underscore, hash, or space.
*/
bindingset[s]
string normalizeName(string s) {
exists(string normStr | normStr = s.toUpperCase().regexpReplaceAll("[-_ ]|/", "") |
result = normStr and isKnownAlgorithm(result)
or
result = unknownAlgorithm() and not isKnownAlgorithm(normStr)
)
}
abstract Expr configurationSink();
predicate hasConfigurationSink() { exists(this.configurationSink()) }
}
abstract class HashAlgorithm extends CryptographicAlgorithm {
final string getHashName() {
if exists(string n | n = this.getName() and isHashingAlgorithm(n))
then isHashingAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
override string getAlgType() { result = getHashType() }
}
abstract class KeyDerivationAlgorithm extends CryptographicAlgorithm {
final string getKDFName() {
if exists(string n | n = this.getName() and isKeyDerivationAlgorithm(n))
then isKeyDerivationAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
override string getAlgType() { result = getKeyDerivationType() }
}
// abstract class KeyDerivationOperation extends CryptographicOperation{
// DataFlow::Node getIterationSizeSrc(){
// none()
// }
// DataFlow::Node getSaltConfigSrc(){
// none()
// }
// DataFlow::Node getHashConfigSrc(){
// none()
// }
// // TODO: get encryption algorithm for CBC-based KDF?
// DataFlow::Node getDerivedKeySizeSrc(){
// none()
// }
// DataFlow::Node getModeSrc(){
// none()
// }
// // TODO: add more to cover all the parameters of most KDF operations? Perhaps subclass for each type?
// abstract predicate requiresIteration();
// abstract predicate requiresSalt();
// abstract predicate requiresHash();
// //abstract predicate requiresKeySize(); // Going to assume all requires a size
// abstract predicate requiresMode();
// }
abstract class EncryptionAlgorithm extends CryptographicAlgorithm {
final predicate isAsymmetric() { this instanceof AsymmetricEncryptionAlgorithm }
final predicate isSymmetric() { not this.isAsymmetric() }
// NOTE: DO_NOT add getEncryptionName here, we rely on the fact the parent
// class does not have this common predicate.
}
/**
* A parent class to represent any algorithm for which
* asymmetric cryptography is involved.
* Intended to be distinct from AsymmetricEncryptionAlgorithm
* which is intended only for asymmetric algorithms that specifically encrypt.
*/
abstract class AsymmetricAlgorithm extends CryptographicAlgorithm { }
/**
* Algorithms directly or indirectly related to asymmetric encryption,
* e.g., RSA, DSA, but also RSA padding algorithms
*/
abstract class AsymmetricEncryptionAlgorithm extends AsymmetricAlgorithm, EncryptionAlgorithm {
final string getEncryptionName() {
if exists(string n | n = this.getName() and isAsymmetricEncryptionAlgorithm(n))
then isAsymmetricEncryptionAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
override string getAlgType() { result = getAsymmetricEncryptionType() }
}
/**
* Algorithms directly or indirectly related to symmetric encryption,
* e.g., AES, DES, but also block modes and padding
*/
abstract class SymmetricEncryptionAlgorithm extends EncryptionAlgorithm {
final string getEncryptionName() {
if exists(string n | n = this.getName() and isSymmetricEncryptionAlgorithm(n))
then isSymmetricEncryptionAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
// TODO: add a stream cipher predicate?
override string getAlgType() { result = getSymmetricEncryptionType() }
}
// Used only to categorize all padding into a single object,
// DO_NOT add predicates here. Only for categorization purposes.
abstract class PaddingAlgorithm extends CryptographicAlgorithm { }
abstract class SymmetricPadding extends PaddingAlgorithm {
final string getPaddingName() {
if exists(string n | n = this.getName() and isSymmetricPaddingAlgorithm(n))
then isSymmetricPaddingAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
override string getAlgType() { result = getSymmetricPaddingType() }
}
abstract class AsymmetricPadding extends PaddingAlgorithm {
final string getPaddingName() {
if exists(string n | n = this.getName() and isAsymmetricPaddingAlgorithm(n))
then isAsymmetricPaddingAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
override string getAlgType() { result = getAsymmetricPaddingType() }
}
abstract class EllipticCurveAlgorithm extends AsymmetricAlgorithm {
final string getCurveName() {
if exists(string n | n = this.getName() and isEllipticCurveAlgorithm(n))
then isEllipticCurveAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
final int getCurveBitSize() { isEllipticCurveAlgorithm(this.getCurveName(), result) }
override string getAlgType() { result = getEllipticCurveType() }
}
abstract class BlockModeAlgorithm extends CryptographicAlgorithm {
final string getBlockModeName() {
if exists(string n | n = this.getName() and isCipherBlockModeAlgorithm(n))
then isCipherBlockModeAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
/**
* Gets the source of the IV configuration.
*/
abstract Expr getIVorNonce();
final predicate hasIVorNonce() { exists(this.getIVorNonce()) }
override string getAlgType() { result = getCipherBlockModeType() }
}
// abstract class KeyWrapOperation extends CryptographicOperation{
// }
abstract class AuthenticatedEncryptionAlgorithm extends SymmetricEncryptionAlgorithm {
final string getAuthticatedEncryptionName() {
if exists(string n | n = this.getName() and isSymmetricEncryptionAlgorithm(n))
then isSymmetricEncryptionAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
}
abstract class KeyExchangeAlgorithm extends AsymmetricAlgorithm {
final string getKeyExchangeName() {
if exists(string n | n = this.getName() and isKeyExchangeAlgorithm(n))
then isKeyExchangeAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
override string getAlgType() { result = getKeyExchangeType() }
}
abstract class SigningAlgorithm extends AsymmetricAlgorithm {
final string getSigningName() {
if exists(string n | n = this.getName() and isSignatureAlgorithm(n))
then isSignatureAlgorithm(result) and result = this.getName()
else result = unknownAlgorithm()
}
override string getAlgType() { result = getSignatureType() }
}

View File

@@ -1,718 +0,0 @@
import cpp
import experimental.cryptography.CryptoAlgorithmNames
import experimental.cryptography.CryptoArtifact
import experimental.cryptography.utils.OpenSSL.CryptoFunction
import experimental.cryptography.utils.OpenSSL.AlgorithmSink
import experimental.cryptography.utils.OpenSSL.PassthroughFunction
import experimental.cryptography.utils.OpenSSL.CryptoAlgorithm
import experimental.cryptography.CryptoArtifact
// import semmle.code.cpp.ir.dataflow.TaintTracking
import semmle.code.cpp.ir.dataflow.DataFlow
/**
* Problematic case in OpenSSL speed.c
* static const char *names[ALGOR_NUM] = {
* "md2", "mdc2", "md4", "md5", "sha1", "rmd160",
* "sha256", "sha512", "whirlpool", "hmac(md5)",
* "des-cbc", "des-ede3", "rc4", "idea-cbc", "seed-cbc",
* "rc2-cbc", "rc5-cbc", "blowfish", "cast-cbc",
* "aes-128-cbc", "aes-192-cbc", "aes-256-cbc",
* "camellia-128-cbc", "camellia-192-cbc", "camellia-256-cbc",
* "evp", "ghash", "rand", "cmac"
* };
*
* Every entry is considered a block mode, hash, and symmetric encryption algorithm
* getEncryptionName for example, will return unknown
*/
predicate nodeToExpr(DataFlow::Node node, Expr e) {
e = node.asExpr() or e = node.asIndirectArgument()
}
Expr getExprFromNode(DataFlow::Node node) { nodeToExpr(node, result) }
DataFlow::Node getNodeFromExpr(Expr e) { nodeToExpr(result, e) }
predicate isEVP_PKEY_CTX(Type t) { t.getUnderlyingType().stripType().getName() = "evp_pkey_ctx_st" }
/**
* An expression representing an EVP_PKEY_CTX* at the location of a
* known AlgorithmSinkArgument.
* The EVP_PKEY_CTX* represents the location where the CTX is tied to the algorithm,
* and can be used as a source for tracing EVP_PKEY_CTX to other operations.
*/
class Known_EVP_PKEY_CTX_Ptr_Source extends Expr {
Known_EVP_PKEY_CTX_Ptr_Source() {
isEVP_PKEY_CTX(this.getUnderlyingType()) and
this.getUnderlyingType() instanceof PointerType and
exists(AlgorithmSinkArgument arg, Call sinkCall |
arg.getSinkCall() = sinkCall and
sinkCall.getAnArgument() = this
or
this = sinkCall
)
}
}
// module CTXFlow implements DataFlow::ConfigSig{
// predicate isSource(DataFlow::Node source) {
// // ASSUMPTION: at a sink, an algorithm is converted into a CTX through a return of the call only
// // and is the primary source of interest for CTX tracing
// source.asExpr() instanceof AlgorithmSinkArgument
// }
// predicate isSink(DataFlow::Node sink){
// sink.asExpr() instanceof CTXSink
// }
// predicate isAdditionalFlowStep(DataFlow::Node node1, DataFlow::Node node2) {
// // cls.getName() = "asn1_object_st" flow out on any EVP_PKEY_CTX which is "evp_pkey_ctx_st"
// exists(Call c |
// isEVP_PKEY_CTX(c.getUnderlyingType()) and
// node1.asExpr() = c.getAnArgument() and c = node2.asExpr())
// }
// }
// module CTXFlowConfig = DataFlow::Global<CTXFlow>;
// TODO: currently only handles tracing from literals to sinks
module LiteralAlgorithmTracerConfig implements DataFlow::ConfigSig {
predicate isSource(DataFlow::Node source) {
source.asExpr() instanceof Literal and
// Optimization to reduce literal tracing on integers to only those that are known/relevant NIDs.
(
exists(source.asExpr().getValue().toInt())
implies
source.asExpr().getValue().toInt() < getNIDMax()
) and
// False positives observed inside OBJ_nid2* and OBJ_sn2* functions where NULL is a possible assignment.
// While this is a concern, it only occurs if the object being referenced is NULL to begin with
// Perhaps a different query should be used to find these caes if they represent a threat.
// Filter out any open ssl function source in a function namae Obj_*
// False positives in OpenSSL also observed for CRYPTO_strndup (filtering any CRYPTO_* function)
// due to setting a null byte in the string
(
isPossibleOpenSSLFunction(source.getEnclosingCallable())
implies
(
not source.getEnclosingCallable().getName().matches("OBJ_%") and
not source.getEnclosingCallable().getName().matches("CRYPTO_%")
)
)
}
predicate isSink(DataFlow::Node sink) {
// A sink is a call to a function that takes an algorithm as an argument
// must include checks for asIndirectArgument since the input may be a pointer to an object
// and the member of the object holds the algorithm on the trace.
getExprFromNode(sink) instanceof AlgorithmSinkArgument
}
predicate isAdditionalFlowStep(DataFlow::Node node1, DataFlow::Node node2) {
knownPassThroughStep(node1, node2)
}
predicate isBarrier(DataFlow::Node node) {
// If the node is the 'next' argument of a isCallPassThrough, it is only allowed if it is an out parameter
// i.e., a defining argument. This barrier says that if the node is an expression not an out parameter, it is filtered.
// Out arguments will not be filtered.
exists(Call c | knownPassthoughCall(c, _, node.asExpr()) and c.getAnArgument() = node.asExpr())
or
// False positive reducer, don't flow out through argv
node.asVariable().hasName("argv")
or
node.asIndirectVariable().hasName("argv")
}
predicate allowImplicitRead(DataFlow::Node node, DataFlow::ContentSet c) {
// Assume a read on crypto identifying field for any object of type asn1_object_st (i.e., ASN1_OBJECT)
exists(Class cls | cls.getName() = "asn1_object_st" |
node.getType().getUnspecifiedType().stripType() = cls and
c.(DataFlow::FieldContent).getField() = cls.getAMember() and
c.(DataFlow::FieldContent).getField().getName() in ["nid", "sn", "ln"]
)
}
}
module LiteralAlgorithmTracer = DataFlow::Global<LiteralAlgorithmTracerConfig>;
/**
* `source` is an expression that is a source of an algorithm of type `algType`.
* `algType` may be `UNKONWN`.
* See CryptoAlgorithmNames for other possible values of `algType`.
*/
bindingset[sinkAlgType]
predicate hasLiteralPathToAlgSink(DataFlow::Node source, DataFlow::Node sink, string sinkAlgType) {
LiteralAlgorithmTracer::flow(source, sink) and
getExprFromNode(sink).(AlgorithmSinkArgument).algType() = sinkAlgType
}
private predicate knownTracedAlgorithm(Literal e, string srcSinkType) {
knownTracedAlgorithm(e, srcSinkType, srcSinkType)
}
private predicate knownTracedAlgorithm(Literal e, string srcType, string sinkType) {
resolveAlgorithmFromLiteral(e, _, srcType) and
hasLiteralPathToAlgSink(DataFlow::exprNode(e), _, sinkType) and
isKnownType(sinkType) and
isKnownType(srcType)
}
private predicate unknownTracedLiteralAlgorithm(Literal e, string srcSinkType) {
// Asymmetric special case:
// Since asymmetric algorithm sinks are used for various categories of asymmetric algorithms
// an asymmetric algorithm is only unknown if there is no trace from any asymmetric type to the given srcSinkType sink
if getAsymmetricType() = srcSinkType
then forall(string t | t = getAsymmetricType() | unknownTracedLiteralAlgorithm(e, t, srcSinkType))
else unknownTracedLiteralAlgorithm(e, srcSinkType, srcSinkType)
}
private predicate unknownTracedLiteralAlgorithm(Literal e, string srcType, string sinkType) {
// the literal resolves to an algorithm, but not to the sinktype
// or generally doesn't resolve to any algorithm type
// this case covers 'nonsense' cases e.g., use RSA for symmetric encryption
not resolveAlgorithmFromLiteral(e, _, srcType) and
isValidAlgorithmLiteral(e) and
hasLiteralPathToAlgSink(DataFlow::exprNode(e), _, sinkType) and
isKnownType(sinkType) and
isKnownType(srcType)
}
private predicate unknownTracedNonLiteralAlgorithm(AlgorithmSinkArgument e, string srcSinkType) {
// Asymmetric special case:
// Since asymmetric algorithm sinks are used for various categories of asymmetric algorithms
// an asymmetric algorithm is only unknown if there is no trace from any asymmetric type to the given srcSinkType sink
if getAsymmetricType() = srcSinkType
then
forall(string t | t = getAsymmetricType() | unknownTracedNonLiteralAlgorithm(e, t, srcSinkType))
else unknownTracedNonLiteralAlgorithm(e, srcSinkType, srcSinkType)
}
private predicate unknownTracedNonLiteralAlgorithm(
AlgorithmSinkArgument e, string srcType, string sinkType
) {
not hasLiteralPathToAlgSink(_, getNodeFromExpr(e), srcType) and
LiteralAlgorithmTracerConfig::isSink(getNodeFromExpr(e)) and
e.algType() = sinkType and
isKnownType(srcType) and
isKnownType(sinkType)
}
private predicate functionAlgorithm(Call c, string algType) {
isOpenSSLCryptoFunctionCall(c, _, algType)
}
abstract class OpenSSLTracedAlgorithm extends CryptographicAlgorithm {
override string getName() { resolveAlgorithmFromLiteral(this, result, this.getAlgType()) }
override Expr configurationSink() {
exists(DataFlow::Node sink |
hasLiteralPathToAlgSink(DataFlow::exprNode(this), sink, this.getAlgType())
|
result = getExprFromNode(sink)
)
}
}
abstract class OpenSSLFunctionAlgorithm extends CryptographicAlgorithm {
override string getName() { isOpenSSLCryptoFunctionCall(this, result, this.getAlgType()) }
override Expr configurationSink() { result = this }
}
abstract class OpenSSLUnknownTracedLiteralAlgorithm extends CryptographicAlgorithm {
override string getName() { result = unknownAlgorithm() }
override Expr configurationSink() {
exists(DataFlow::Node sink |
hasLiteralPathToAlgSink(DataFlow::exprNode(this), sink, this.getAlgType())
|
result = getExprFromNode(sink)
)
}
}
abstract class OpenSSLUnknownTracedNonLiteralAlgorithm extends CryptographicAlgorithm {
override string getName() { result = unknownAlgorithm() }
override Expr configurationSink() { result = this }
}
module SymmetricEncryption {
abstract class OpenSSLSymmetricEncryptionAlgorithm extends SymmetricEncryptionAlgorithm { }
class OpenSSLSymmetricEncryptionTracedAlgorithm extends OpenSSLTracedAlgorithm,
OpenSSLSymmetricEncryptionAlgorithm
{
OpenSSLSymmetricEncryptionTracedAlgorithm() {
knownTracedAlgorithm(this, getSymmetricEncryptionType())
}
}
class OpenSSLSymmetricEncryptionFunctionAlgorithm extends OpenSSLFunctionAlgorithm,
OpenSSLSymmetricEncryptionAlgorithm
{
OpenSSLSymmetricEncryptionFunctionAlgorithm() {
functionAlgorithm(this, getSymmetricEncryptionType())
}
}
class OpenSSLSymmetricEncryptionTracedUnknownLiteralAlgorithm extends OpenSSLUnknownTracedLiteralAlgorithm,
OpenSSLSymmetricEncryptionAlgorithm
{
OpenSSLSymmetricEncryptionTracedUnknownLiteralAlgorithm() {
unknownTracedLiteralAlgorithm(this, getSymmetricEncryptionType())
}
}
class OpenSSLSymmetricEncryptionUnknownNonLiteralTracedAlgorithm extends OpenSSLUnknownTracedNonLiteralAlgorithm,
OpenSSLSymmetricEncryptionAlgorithm
{
OpenSSLSymmetricEncryptionUnknownNonLiteralTracedAlgorithm() {
unknownTracedNonLiteralAlgorithm(this, getSymmetricEncryptionType())
}
}
}
module BlockModes {
/**
* In OpenSSL, block modes are associated directly with symmetric encryption algorithms.
* As such, OpenSSLBLockModes are modeled as extensions of any openssl symmetric encryption algorithm
*/
class OpenSSLBlockModeAlgorithm extends BlockModeAlgorithm, Expr instanceof SymmetricEncryption::OpenSSLSymmetricEncryptionAlgorithm
{
OpenSSLBlockModeAlgorithm() {
//two cases, either the block mode is a literal or it is a function call
resolveAlgorithmFromLiteral(this, _, "BLOCK_MODE")
or
isOpenSSLCryptoFunctionCall(this, _, "BLOCK_MODE")
}
override string getName() {
resolveAlgorithmFromLiteral(this, result, "BLOCK_MODE")
or
isOpenSSLCryptoFunctionCall(this, result, "BLOCK_MODE")
}
override Expr configurationSink() {
result = this.(SymmetricEncryption::OpenSSLSymmetricEncryptionAlgorithm).configurationSink()
}
override Expr getIVorNonce() {
// TODO
none()
}
}
class UnknownOpenSSLBlockModeAlgorithm extends BlockModeAlgorithm, Expr instanceof SymmetricEncryption::OpenSSLSymmetricEncryptionAlgorithm
{
UnknownOpenSSLBlockModeAlgorithm() {
//two cases, either the block mode is a literal or it is a function call
not resolveAlgorithmFromLiteral(this, _, "BLOCK_MODE") and
not isOpenSSLCryptoFunctionCall(this, _, "BLOCK_MODE")
}
override string getName() { result = unknownAlgorithm() }
override Expr configurationSink() {
result = this.(SymmetricEncryption::OpenSSLSymmetricEncryptionAlgorithm).configurationSink()
}
override Expr getIVorNonce() { none() }
}
}
module Hashes {
abstract class OpenSSLHashAlgorithm extends HashAlgorithm { }
class OpenSSLHashTracedAlgorithm extends OpenSSLTracedAlgorithm, OpenSSLHashAlgorithm {
OpenSSLHashTracedAlgorithm() { knownTracedAlgorithm(this, getHashType()) }
}
class OpenSSLHashFunctionAlgorithm extends OpenSSLFunctionAlgorithm, OpenSSLHashAlgorithm {
OpenSSLHashFunctionAlgorithm() { functionAlgorithm(this, getHashType()) }
}
class OpenSSLHashTracedUnknownLiteralAlgorithm extends OpenSSLUnknownTracedLiteralAlgorithm,
OpenSSLHashAlgorithm
{
OpenSSLHashTracedUnknownLiteralAlgorithm() {
unknownTracedLiteralAlgorithm(this, getHashType())
}
}
class OpenSSLHashUnknownNonLiteralTracedAlgorithm extends OpenSSLUnknownTracedNonLiteralAlgorithm,
OpenSSLHashAlgorithm
{
OpenSSLHashUnknownNonLiteralTracedAlgorithm() {
unknownTracedNonLiteralAlgorithm(this, getHashType())
}
}
class OpenSSLNullHash extends HashAlgorithm {
OpenSSLNullHash() {
exists(Call c |
this = c and
isPossibleOpenSSLFunction(c.getTarget()) and
c.getTarget().getName() in ["EVP_md_null"]
)
}
override string getName() { result = unknownAlgorithm() }
override Expr configurationSink() { result = this }
}
}
module EllipticCurves {
// TODO: need to address EVP_PKEY_Q_keygen where the type is "EC" but the curve is UNKNOWN?
class OpenSSLEllipticCurveTracedAlgorithm extends OpenSSLTracedAlgorithm, EllipticCurveAlgorithm {
OpenSSLEllipticCurveTracedAlgorithm() { knownTracedAlgorithm(this, getEllipticCurveType()) }
}
class OpenSSLEllipticCurveFunctionAlgorithm extends OpenSSLFunctionAlgorithm,
EllipticCurveAlgorithm
{
OpenSSLEllipticCurveFunctionAlgorithm() { functionAlgorithm(this, getEllipticCurveType()) }
}
class OpenSSLEllipticCurveTracedUnknownLiteralAlgorithm extends OpenSSLUnknownTracedLiteralAlgorithm,
EllipticCurveAlgorithm
{
OpenSSLEllipticCurveTracedUnknownLiteralAlgorithm() {
unknownTracedLiteralAlgorithm(this, getEllipticCurveType())
}
}
class OpenSSLEllipticCurvehUnknownNonLiteralTracedAlgorithm extends OpenSSLUnknownTracedNonLiteralAlgorithm,
EllipticCurveAlgorithm
{
OpenSSLEllipticCurvehUnknownNonLiteralTracedAlgorithm() {
unknownTracedNonLiteralAlgorithm(this, getEllipticCurveType())
}
}
// https://www.openssl.org/docs/manmaster/man3/EC_KEY_new_ex.html
class OpenSSLNullEllipticCurve extends EllipticCurveAlgorithm {
OpenSSLNullEllipticCurve() {
exists(Call c |
this = c and
isPossibleOpenSSLFunction(c.getTarget()) and
c.getTarget().getName() in ["EC_KEY_new", "EC_KEY_new_ex"]
)
}
override string getName() { result = unknownAlgorithm() }
override Expr configurationSink() { result = this }
}
}
module AsymmetricEncryption {
class OpenSSLAsymmetricEncryptionTracedAlgorithm extends OpenSSLTracedAlgorithm,
AsymmetricEncryptionAlgorithm
{
OpenSSLAsymmetricEncryptionTracedAlgorithm() {
knownTracedAlgorithm(this, getAsymmetricEncryptionType())
}
}
class OpenSSLAsymmetricEncryptionFunctionAlgorithm extends OpenSSLFunctionAlgorithm,
AsymmetricEncryptionAlgorithm
{
OpenSSLAsymmetricEncryptionFunctionAlgorithm() {
functionAlgorithm(this, getAsymmetricEncryptionType())
}
}
class OpenSSLAsymmetricEncryptionTracedUnknownLiteralAlgorithm extends OpenSSLUnknownTracedLiteralAlgorithm,
AsymmetricEncryptionAlgorithm
{
OpenSSLAsymmetricEncryptionTracedUnknownLiteralAlgorithm() {
unknownTracedLiteralAlgorithm(this, getAsymmetricEncryptionType())
}
}
class OpenSSLAsymmetricEncryptionUnknownNonLiteralTracedAlgorithm extends OpenSSLUnknownTracedNonLiteralAlgorithm,
AsymmetricEncryptionAlgorithm
{
OpenSSLAsymmetricEncryptionUnknownNonLiteralTracedAlgorithm() {
unknownTracedNonLiteralAlgorithm(this, getAsymmetricEncryptionType())
}
}
}
module SigningAlgorithms {
class OpenSSLSignatureTracedAlgorithm extends OpenSSLTracedAlgorithm, SigningAlgorithm {
OpenSSLSignatureTracedAlgorithm() { knownTracedAlgorithm(this, getSignatureType()) }
}
class OpenSSLSignatureFunctionAlgorithm extends OpenSSLFunctionAlgorithm, SigningAlgorithm {
OpenSSLSignatureFunctionAlgorithm() { functionAlgorithm(this, getSignatureType()) }
}
class OpenSSLSignatureTracedUnknownLiteralAlgorithm extends OpenSSLUnknownTracedLiteralAlgorithm,
SigningAlgorithm
{
OpenSSLSignatureTracedUnknownLiteralAlgorithm() {
unknownTracedLiteralAlgorithm(this, getSignatureType())
}
}
class OpenSSLSignatureUnknownNonLiteralTracedAlgorithm extends OpenSSLUnknownTracedNonLiteralAlgorithm,
SigningAlgorithm
{
OpenSSLSignatureUnknownNonLiteralTracedAlgorithm() {
unknownTracedNonLiteralAlgorithm(this, getSignatureType())
}
}
}
module KeyExchange {
class OpenSSLKeyExchangeTracedAlgorithm extends OpenSSLTracedAlgorithm, KeyExchangeAlgorithm {
OpenSSLKeyExchangeTracedAlgorithm() { knownTracedAlgorithm(this, getKeyExchangeType()) }
}
class OpenSSLKeyExchangeFunctionAlgorithm extends OpenSSLFunctionAlgorithm, KeyExchangeAlgorithm {
OpenSSLKeyExchangeFunctionAlgorithm() { functionAlgorithm(this, getKeyExchangeType()) }
}
class OpenSSLKeyExchangeTracedUnknownLiteralAlgorithm extends OpenSSLUnknownTracedLiteralAlgorithm,
KeyExchangeAlgorithm
{
OpenSSLKeyExchangeTracedUnknownLiteralAlgorithm() {
unknownTracedLiteralAlgorithm(this, getKeyExchangeType())
}
}
class OpenSSLKeyExchangeUnknownNonLiteralTracedAlgorithm extends OpenSSLUnknownTracedNonLiteralAlgorithm,
KeyExchangeAlgorithm
{
OpenSSLKeyExchangeUnknownNonLiteralTracedAlgorithm() {
unknownTracedNonLiteralAlgorithm(this, getKeyExchangeType())
}
}
}
module KeyGeneration {
/**
* Functions that explicitly set key generation parameters.
* `sizeInd` is the parameter specifying the size of the key.
* `outInd` is the parameter or return value that the key is written to.
* `outInd` is -1 if the key is written to the return value.
*/
predicate isAsymmetricKeyGenExplicitAlgorithm(Function func, int sizeInd, int outInd) {
isPossibleOpenSSLFunction(func) and
exists(string name | func.hasGlobalName(name) |
name in [
"EVP_PKEY_CTX_set_dsa_paramgen_bits", "DSA_generate_parameters_ex",
"EVP_PKEY_CTX_set_rsa_keygen_bits", "RSA_generate_key_ex", "RSA_generate_key_fips",
"EVP_PKEY_CTX_set_dh_paramgen_prime_len", "DH_generate_parameters_ex"
] and
sizeInd = 1 and
outInd = 0
or
name in ["DSA_generate_parameters", "RSA_generate_key", "DH_generate_parameters"] and
sizeInd = 0 and
outInd = -1
) and
exists(Type t |
(
if sizeInd = -1
then t = func.getType().getUnderlyingType()
else t = func.getParameter(sizeInd).getUnderlyingType()
) and
t instanceof IntegralType and
not t instanceof CharType
)
}
module AsymExplicitAlgKeyLengthFlowConfig implements DataFlow::ConfigSig {
predicate isSource(DataFlow::Node node) {
// Optimizations to avoid tracing all integers
node.asExpr().(Literal).getValue().toInt() > 0 and // exclude sentinel values
node.asExpr().(Literal).getValue().toInt() < 8500
}
predicate isSink(DataFlow::Node node) {
exists(FunctionCall c, int sizeInd |
isAsymmetricKeyGenExplicitAlgorithm(c.getTarget(), sizeInd, _) and
c.getArgument(sizeInd) = node.asExpr()
)
}
}
module AsymExplicitAlgKeyLengthFlow = DataFlow::Global<AsymExplicitAlgKeyLengthFlowConfig>;
class OpenSSLAsymmetricKeyGenTiedToAlgorithm extends AsymmetricKeyGeneration {
OpenSSLAsymmetricKeyGenTiedToAlgorithm() {
exists(Call c |
this = c and
isPossibleOpenSSLFunction(c.getTarget()) and
isAsymmetricKeyGenExplicitAlgorithm(c.getTarget(), _, _)
)
}
override CryptographicAlgorithm getAlgorithm() { result = this }
override Expr getKeyConfigurationSource(CryptographicAlgorithm alg) {
alg = this and
exists(int sizeInd |
isAsymmetricKeyGenExplicitAlgorithm(this.getTarget(), sizeInd, _) and
AsymExplicitAlgKeyLengthFlow::flow(DataFlow::exprNode(result),
DataFlow::exprNode(this.getArgument(sizeInd)))
)
}
}
module Length_to_RSA_EVP_PKEY_Q_keygen_Config implements DataFlow::ConfigSig {
predicate isSource(DataFlow::Node node) {
// Optimizations to avoid tracing all integers
node.asExpr().(Literal).getValue().toInt() > 0 and // exclude sentinel values
node.asExpr().(Literal).getValue().toInt() < 5000
}
predicate isSink(DataFlow::Node node) {
exists(FunctionCall c |
c.getTarget().getName() = "EVP_PKEY_Q_keygen" and
isPossibleOpenSSLFunction(c.getTarget()) and
c.getArgument(3) = node.asExpr()
)
}
}
module Length_to_RSA_EVP_PKEY_Q_keygen_Flow =
DataFlow::Global<Length_to_RSA_EVP_PKEY_Q_keygen_Config>;
class OpenSSL_RSA_EVP_PKEY_Q_keygen extends AsymmetricKeyGeneration {
OpenSSL_RSA_EVP_PKEY_Q_keygen() {
exists(Call c |
this = c and
isPossibleOpenSSLFunction(c.getTarget()) and
this.getTarget().getName() = "EVP_PKEY_Q_keygen" and
this.getArgument(3).getUnderlyingType() instanceof IntegralType
)
}
override CryptographicAlgorithm getAlgorithm() {
result.configurationSink().(AlgorithmSinkArgument).getSinkCall() = this
}
override Expr getKeyConfigurationSource(CryptographicAlgorithm alg) {
alg = this.getAlgorithm() and
Length_to_RSA_EVP_PKEY_Q_keygen_Flow::flow(DataFlow::exprNode(result),
DataFlow::exprNode(this.getArgument(3)))
}
}
predicate isKeyGenOperationWithNoSize(Function func) {
isPossibleOpenSSLFunction(func) and
exists(string name | func.hasGlobalName(name) |
name in ["EVP_PKEY_keygen", "DSA_generate_key", "DH_generate_key", "EVP_PKEY_generate"]
)
}
module KeyGenKeySizeInitToKeyGenConfig implements DataFlow::ConfigSig {
predicate isSource(DataFlow::Node node) {
exists(Call c, Function func, int outInd |
isAsymmetricKeyGenExplicitAlgorithm(func, _, outInd) and
c.getTarget() = func
|
if outInd = -1 then node.asExpr() = c else node.asExpr() = c.getArgument(outInd)
)
}
predicate isSink(DataFlow::Node node) {
exists(Call c |
isKeyGenOperationWithNoSize(c.getTarget()) and c.getAnArgument() = node.asExpr()
)
}
}
module KeyGenKeySizeInitToKeyGenFlow = DataFlow::Global<KeyGenKeySizeInitToKeyGenConfig>;
predicate isEVP_PKEY_CTX_Source(DataFlow::Node node, CryptographicAlgorithm alg) {
exists(Call c |
alg.configurationSink().(AlgorithmSinkArgument).getSinkCall() = c and
(
node.asExpr() = c
or
node.asExpr() = c.getAnArgument()
or
node.asDefiningArgument() = c.getAnArgument()
)
) and
(
node.asExpr() instanceof Known_EVP_PKEY_CTX_Ptr_Source
or
node.asDefiningArgument() instanceof Known_EVP_PKEY_CTX_Ptr_Source
)
}
predicate isKeyGen_EVP_PKEY_CTX_Sink(DataFlow::Node node, Call c) {
isKeyGenOperationWithNoSize(c.getTarget()) and nodeToExpr(node, c.getAnArgument())
}
/**
* Trace from EVP_PKEY_CTX* at algorithm sink to keygen,
* users can then extrapolatae the matching algorithm from the alg sink to the keygen
*/
module EVP_PKEY_CTX_Ptr_Source_to_KeyGenOperationWithNoSize implements DataFlow::ConfigSig {
predicate isSource(DataFlow::Node source) { isEVP_PKEY_CTX_Source(source, _) }
predicate isSink(DataFlow::Node sink) { isKeyGen_EVP_PKEY_CTX_Sink(sink, _) }
}
module EVP_PKEY_CTX_Ptr_Source_to_KeyGenOperationWithNoSize_Flow =
DataFlow::Global<EVP_PKEY_CTX_Ptr_Source_to_KeyGenOperationWithNoSize>;
/**
* UNKNOWN key sizes to general purpose key generation functions (i.e., that take in no key size and assume
* is it set on context prior to the call). No path from a key configuration to these operations
* means the key size is UNKNOWN, or more precisely the key size is DEFAULT but
* the defaults can change with each version of OpenSSL, we simply assume the size is generally UNKNOWN.
* ASSUMPTION/TODO: we currently model all known locations where a key size is set explicitly.
* When a key is set implicitly, this usually means a key generation operation
* is called where the operation takes in no key size, and no flow to this operation
* initializes the context with a key size.
* Currently, without a definitive source (set of sources) to start tracing from, we cannot determine
* determine if a single path exists that initializes the context with a key size and another that doesn't.
* Rather than attempt to model all possible sources, we assume that if no path
* from a key config location reaches a generic key generation operation, then the key size is not set.
* NOTE: while this is true, it is possible a key size is set in one path, but not in another
* meaning this approach (and other similar approaches used in this model for UNKNOWN)
* can produce false negatives.
*/
class OpenSSLDefaultKeyGeneration extends AsymmetricKeyGeneration {
OpenSSLDefaultKeyGeneration() {
// this is a call to a function matching isKeyGenOperationWithNoSize
// and there is no flow from a key configuration source to this call
exists(Call c |
this = c and
isKeyGenOperationWithNoSize(this.getTarget()) and
not exists(DataFlow::Node src, DataFlow::Node sink |
KeyGenKeySizeInitToKeyGenFlow::flow(src, sink) and
nodeToExpr(sink, this.getAnArgument())
)
)
}
override CryptographicAlgorithm getAlgorithm() {
if this.getTarget().getName() in ["DSA_generate_key", "DH_generate_key"]
then result = this
else
// NOTE/ASSUMPTION: EVP_PKEY_keygen, EVP_PKEY_generate assume only other possibilities,
// each take in a CTX as the first arg, need to trace from an alg sink from this CTX param
// get every alg sink, get the corresponding call, trace out on any CTX type variable
// to the key gen
// NOTE: looking for any cryptographic algorithm tracing to the keygen to handle
// any odd cases we aren't awaare of where keygen can be used for other algorithm types
exists(DataFlow::Node src, DataFlow::Node sink |
EVP_PKEY_CTX_Ptr_Source_to_KeyGenOperationWithNoSize_Flow::flow(src, sink) and
isEVP_PKEY_CTX_Source(src, result) and
isKeyGen_EVP_PKEY_CTX_Sink(sink, this)
// TODO: what if there is no CTX source? then the keygen becomes an UNKNOWN sink
)
}
/**
* For this class, there is no known configuration source for any algorithm
*/
override Expr getKeyConfigurationSource(CryptographicAlgorithm alg) { none() }
}
}

View File

@@ -1,296 +0,0 @@
/**
* Predicates/classes for identifying algorithm sinks.
* An Algorithm Sink is a function that takes an algorithm as an argument.
* In particular, any function that takes in an algorithm that until the call
* the algorithm is not definitely known to be an algorithm (e.g., an integer used as an identifier to fetch an algorithm)
*/
//TODO: enforce a hierarchy of AlgorithmSinkArgument, e.g., so I can get all Asymmetric SinkArguments that includes all the strictly RSA etc.
import cpp
import experimental.cryptography.utils.OpenSSL.LibraryFunction
import experimental.cryptography.CryptoAlgorithmNames
predicate isAlgorithmSink(AlgorithmSinkArgument arg, string algType) { arg.algType() = algType }
abstract class AlgorithmSinkArgument extends Expr {
AlgorithmSinkArgument() {
exists(Call c | c.getAnArgument() = this and openSSLLibraryFunc(c.getTarget()))
}
/**
* Gets the function call in which the argument exists
*/
Call getSinkCall() { result.getAnArgument() = this }
abstract string algType();
}
// https://www.openssl.org/docs/manmaster/man3/EVP_CIPHER_fetch.html
predicate cipherAlgorithmSink(string funcName, int argInd) {
funcName in ["EVP_get_cipherbyname", "EVP_get_cipherbynid", "EVP_get_cipherbyobj"] and argInd = 0
or
funcName = "EVP_CIPHER_fetch" and argInd = 1
}
class CipherAlgorithmSink extends AlgorithmSinkArgument {
CipherAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
cipherAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = getSymmetricEncryptionType() }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_MAC_fetch
predicate macAlgorithmSink(string funcName, int argInd) {
(funcName = "EVP_MAC_fetch" and argInd = 1)
}
class MACAlgorithmSink extends AlgorithmSinkArgument {
MACAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
macAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = "TBD" }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_MD_fetch
predicate messageDigestAlgorithmSink(string funcName, int argInd) {
funcName in ["EVP_get_digestbyname", "EVP_get_digestbynid", "EVP_get_digestbyobj"] and argInd = 0
or
funcName = "EVP_MD_fetch" and argInd = 1
}
class MessageDigestAlgorithmSink extends AlgorithmSinkArgument {
MessageDigestAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
messageDigestAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = getHashType() }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_KEYEXCH_fetch
// https://www.openssl.org/docs/manmaster/man3/EVP_KEM_fetch
predicate keyExchangeAlgorithmSink(string funcName, int argInd) {
funcName = "EVP_KEYEXCH_fetch" and argInd = 1
or
funcName = "EVP_KEM_fetch" and argInd = 1
}
class KeyExchangeAlgorithmSink extends AlgorithmSinkArgument {
KeyExchangeAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
keyExchangeAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = getKeyExchangeType() }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_KEYMGMT_fetch
predicate keyManagementAlgorithmSink(string funcName, int argInd) {
funcName = "EVP_KEYMGMT_fetch" and argInd = 1
}
class KeyManagementAlgorithmSink extends AlgorithmSinkArgument {
KeyManagementAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
keyManagementAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = "TBD" }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_KDF
predicate keyDerivationAlgorithmSink(string funcName, int argInd) {
funcName = "EVP_KDF_fetch" and argInd = 1
}
class KeyDerivationAlgorithmSink extends AlgorithmSinkArgument {
KeyDerivationAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
keyDerivationAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = getKeyDerivationType() }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_ASYM_CIPHER_fetch
// https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_CTX_new_id
// https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_new_CMAC_key.html
predicate asymmetricCipherAlgorithmSink(string funcName, int argInd) {
funcName = "EVP_ASYM_CIPHER_fetch" and argInd = 1
or
funcName = "EVP_PKEY_new_CMAC_key" and argInd = 3
// NOTE: other cases are handled by AsymmetricAlgorithmSink
}
class AsymmetricCipherAlgorithmSink extends AlgorithmSinkArgument {
AsymmetricCipherAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
asymmetricCipherAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = "ASYMMETRIC_ENCRYPTION" }
}
class AsymmetricCipherAlgorithmSink_EVP_PKEY_Q_keygen extends AlgorithmSinkArgument {
AsymmetricCipherAlgorithmSink_EVP_PKEY_Q_keygen() {
exists(Call c, string funcName |
funcName = c.getTarget().getName() and
this = c.getArgument(3)
|
funcName = "EVP_PKEY_Q_keygen" and
c.getArgument(3).getType().getUnderlyingType() instanceof IntegralType
)
}
override string algType() { result = "ASYMMETRIC_ENCRYPTION" }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_RAND_fetch
predicate randomAlgorithmSink(string funcName, int argInd) {
funcName = "EVP_RAND_fetch" and argInd = 1
}
class RandomAlgorithmSink extends AlgorithmSinkArgument {
RandomAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
randomAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = "TBD" }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_SIGNATURE_fetch
predicate signatureAlgorithmSink(string funcName, int argInd) {
funcName = "EVP_SIGNATURE_fetch" and argInd = 1
}
class SignatureAlgorithmSink extends AlgorithmSinkArgument {
SignatureAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
signatureAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = getSignatureType() }
}
// https://www.openssl.org/docs/manmaster/man3/EC_KEY_new_by_curve_name.html
// https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_CTX_set_ec_paramgen_curve_nid.html
predicate ellipticCurveAlgorithmSink(string funcName, int argInd) {
funcName in ["EC_KEY_new_by_curve_name", "EVP_EC_gen"] and argInd = 0
or
funcName = "EC_KEY_new_by_curve_name_ex" and argInd = 2
or
funcName in ["EVP_PKEY_CTX_set_ec_paramgen_curve_nid"] and argInd = 1
}
class EllipticCurveAlgorithmSink extends AlgorithmSinkArgument {
EllipticCurveAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
ellipticCurveAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = getEllipticCurveType() }
}
/**
* Special cased to address the fact that arg index 3 (zero offset based) is the curve name.
* ASSUMPTION: if the arg ind 3 is a char* assume it is an elliptic curve
*/
class EllipticCurveAlgorithmSink_EVP_PKEY_Q_keygen extends AlgorithmSinkArgument {
EllipticCurveAlgorithmSink_EVP_PKEY_Q_keygen() {
exists(Call c, string funcName |
funcName = c.getTarget().getName() and
this = c.getArgument(3)
|
funcName = "EVP_PKEY_Q_keygen" and
c.getArgument(3).getType().getUnderlyingType() instanceof PointerType and
c.getArgument(3).getType().getUnderlyingType().stripType() instanceof CharType
)
}
override string algType() { result = getEllipticCurveType() }
}
// https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_CTX_new_id.html
// https://www.openssl.org/docs/man1.1.1/man3/EVP_PKEY_new_raw_private_key.html
// https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_new.html
// https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_CTX_ctrl.html
// https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_Q_keygen.html
// https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_CTX_ctrl.html
predicate asymmetricAlgorithmSink(string funcName, int argInd) {
funcName = "EVP_PKEY_CTX_new_id" and argInd = 0
or
funcName = "EVP_PKEY_CTX_new_from_name" and argInd = 1
or
funcName in [
"EVP_PKEY_new_raw_private_key", "EVP_PKEY_new_raw_public_key", "EVP_PKEY_new_mac_key"
] and
argInd = 0
or
funcName in ["EVP_PKEY_new_raw_private_key_ex", "EVP_PKEY_new_raw_public_key_ex"] and argInd = 1
or
// special casing this as arg index 3 must be specified depending on if RSA or ECC, and otherwise not specified for other algs
// funcName = "EVP_PKEY_Q_keygen" and argInd = 2
funcName in ["EVP_PKEY_CTX_ctrl", "EVP_PKEY_CTX_set_group_name"] and argInd = 1
// TODO consider void cases EVP_PKEY_new
}
class AsymmetricAlgorithmSink extends AlgorithmSinkArgument {
AsymmetricAlgorithmSink() {
exists(Call c, string funcName, int argInd |
funcName = c.getTarget().getName() and this = c.getArgument(argInd)
|
asymmetricAlgorithmSink(funcName, argInd)
)
}
override string algType() { result = getAsymmetricType() }
}
class AsymmetricAlgorithmSink_EVP_PKEY_Q_keygen extends AlgorithmSinkArgument {
AsymmetricAlgorithmSink_EVP_PKEY_Q_keygen() {
exists(Call c, string funcName |
funcName = c.getTarget().getName() and
this = c.getArgument(2)
|
funcName = "EVP_PKEY_Q_keygen" and
not exists(c.getArgument(3))
)
}
override string algType() { result = getAsymmetricType() }
}

View File

@@ -1,121 +0,0 @@
import cpp
import experimental.cryptography.utils.OpenSSL.LibraryFunction
import experimental.cryptography.CryptoAlgorithmNames
predicate inferredOpenSSLCryptoFunctionCall(Call c, string normalized, string algType) {
inferredOpenSSLCryptoFunction(c.getTarget(), normalized, algType)
}
predicate inferredOpenSSLCryptoFunction(Function f, string normalized, string algType) {
isPossibleOpenSSLFunction(f) and
normalizeFunctionName(f, algType) = normalized
}
predicate isOpenSSLCryptoFunction(Function f, string normalized, string algType) {
// NOTE: relying on inference as there are thousands of functions for crypto
// enumerating them all and maintaining the list seems problematic.
// For now, we will rely on dynamically inferring algorithms for function names.
// This has been seen to be reasonably efficient and accurate.
inferredOpenSSLCryptoFunction(f, normalized, algType)
}
predicate isOpenSSLCryptoFunctionCall(Call c, string normalized, string algType) {
isOpenSSLCryptoFunction(c.getTarget(), normalized, algType)
}
private string basicNormalizeFunctionName(Function f, string algType) {
isPossibleOpenSSLFunction(f) and
isKnownAlgorithm(result, algType) and
exists(string normStr | normStr = f.getName().toUpperCase().regexpReplaceAll("[-_ ]|/", "") |
normStr.matches("%" + result + "%")
)
}
/**
* Converts a raw OpenSSL algorithm to a normalized algorithm name.
*
* If more than one match occurs for a given algorithm type, normalize attempts to find the "max"
* string (max in terms of string length) e.g., matching AES128 to AES128 and not simply AES.
*
* An unknown algorithm is only identified if there exists no known algorithm found for any algorithm type.
*
* `f` is the function name to normalize.
* `algType` is a string representing the classification of the algorithm (see `CryptoAlgorithmNames`)
*/
private string privateNormalizeFunctionName(Function f, string algType) {
isPossibleOpenSSLFunction(f) and
result = basicNormalizeFunctionName(f, algType) and
not exists(string res2 |
result != res2 and
res2 = basicNormalizeFunctionName(f, algType) and
res2.length() > result.length()
)
}
/**
* Normalizes a function name to a known algorithm name, similar to `normalizeName`.
* A function is not, however, allowed to be UNKNOWN. The function either
* normalizes to a known algorithm name, or the predicate does not hold (no result).
*
* The predicate attempts to restrict normalization to what looks like an openssl
* library by looking for functions only in an openssl path (see `isPossibleOpenSSLFunction`).
* This may give false postive functions if a directory erronously appears to be openssl;
* however, we take the stance that if a function
* exists strongly mapping to a known function name in a directory such as these,
* regardless of whether its actually a part of openSSL or not, we will analyze it as though it were.
*/
private string normalizeFunctionName(Function f, string algType) {
algType != "UNKNOWN" and
isPossibleOpenSSLFunction(f) and
result = privateNormalizeFunctionName(f, algType) and
// Addressing false positives
// For algorithm names less than or equal to 4, we must see the algorithm name
// in the original function as upper case (it can't be split between tokens)
// One exception found is DES_xcbc_encrypt, this is DESX
(
(result.length() <= 4 and result != "DESX")
implies
f.getName().toUpperCase().matches("%" + result + "%")
) and
(
(result.length() <= 4 and result = "DESX")
implies
(f.getName().toUpperCase().matches("%DESX%") or f.getName().toUpperCase().matches("%DES_X%"))
) and
// (result.length() <= 3 implies (not f.getName().toUpperCase().regexpMatch(".*" + result + "[a-zA-Z0-9].*|.*[a-zA-Z0-9]" + result + ".*")))
// and
// DES specific false positives
(
result.matches("DES")
implies
not f.getName().toUpperCase().regexpMatch(".*DES[a-zA-Z0-9].*|.*[a-zA-Z0-9]DES.*")
) and
// ((result.matches("%DES%")) implies not exists(string s | s in ["DESCRIBE", "DESTROY", "DESCRIPTION", "DESCRIPTOR", "NODES"] |
// f.getName().toUpperCase().matches("%" + s + "%"))) and
// SEED specific false positives
(
result.matches("SEED")
implies
not exists(string s |
s in [
"SEED_SRC_GENERATE", "RAND", "NEW_SEED", "GEN_SEED", "SEED_GEN", "SET_SEED", "GET_SEED",
"GET0_SEED", "RESEED", "SEEDING"
]
|
f.getName().toUpperCase().matches("%" + s + "%")
)
) and
// ARIA specific false positives
(result.matches("ARIA") implies not f.getName().toUpperCase().matches("%VARIANT%")) and
// CTR false positives
(result.matches("CTR") implies not f.getName().toUpperCase().matches("%CTRL%")) and
// ES false positives (e.g., ES256 from AES256)
(result.matches("ES%") implies not f.getName().toUpperCase().matches("%AES%")) and
// RSA false positives
(result.matches("RSA") implies not f.getName().toUpperCase().matches("%UNIVERSAL%")) and
//rsaz functions deemed to be too low level, and can be ignored
not f.getLocation().getFile().getBaseName().matches("rsaz_exp.c") and
// General False positives
// Functions that 'get' do not set an algorithm, and therefore are considered ignorable
not f.getName().toLowerCase().matches("%get%")
}

View File

@@ -1,153 +0,0 @@
/**
* This file contains predicates create to build up initial data sets for OpenSSL
* predicates. E.g., These predicates were used to assist in associating all
* openSSL functions with their known crypto algorithms.
*/
import cpp
import experimental.cryptography.CryptoAlgorithmNames
import experimental.cryptography.utils.OpenSSL.CryptoFunction
private string basicNormalizeFunctionName(Function f, string algType) {
isKnownAlgorithm(result, algType) and
exists(string normStr | normStr = f.getName().toUpperCase().regexpReplaceAll("[-_ ]|/", "") |
normStr.matches("%" + result + "%")
)
}
/**
* Converts a raw OpenSSL algorithm to a normalized algorithm name.
*
* If more than one match occurs for a given algorithm type, normalize attempts to find the "max"
* string (max in terms of string length) e.g., matching AES128 to AES128 and not simply AES.
*
* An unknown algorithm is only identified if there exists no known algorithm found for any algorithm type.
*
* `f` is the function name to normalize.
* `algType` is a string representing the classification of the algorithm (see `CryptoAlgorithmNames`)
*/
private string privateNormalizeFunctionName(Function f, string algType) {
result = basicNormalizeFunctionName(f, algType) and
not exists(string res2 |
result != res2 and
res2 = basicNormalizeFunctionName(f, algType) and
res2.length() > result.length()
) and
// Addressing bad normalization case-by-case
// CASE: ES256 being identified when the algorithm is AES256
(
result.matches("ES256")
implies
not exists(string res2 | res2 = basicNormalizeFunctionName(f, _) and res2.matches("AES%"))
)
}
/**
* Normalizes a function name to a known algorithm name, similar to `normalizeName`.
* A function is not, however, allowed to be UNKNOWN. The function either
* normalizes to a known algorithm name, or the predicate does not hold (no result).
*
* The predicate attempts to restrict normalization to what looks like an openssl
* library by looking for functions only in an openssl path (see `isPossibleOpenSSLFunction`).
* This may give false postive functions if a directory erronously appears to be openssl;
* however, we take the stance that if a function
* exists strongly mapping to a known function name in a directory such as these,
* regardless of whether its actually a part of openSSL or not, we will analyze it as though it were.
*/
string normalizeFunctionName(Function f, string algType) {
algType != "UNKNOWN" and
result = privateNormalizeFunctionName(f, algType) and
openSSLLibraryFunc(f) and
// Addressing false positives
// For algorithm names less than or equal to 4, we must see the algorithm name
// in the original function as upper case (it can't be split between tokens)
// One exception found is DES_xcbc_encrypt, this is DESX
(
(result.length() <= 4 and result != "DESX")
implies
f.getName().toUpperCase().matches("%" + result + "%")
) and
(
(result.length() <= 4 and result = "DESX")
implies
(f.getName().toUpperCase().matches("%DESX%") or f.getName().toUpperCase().matches("%DES_X%"))
) and
// (result.length() <= 3 implies (not f.getName().toUpperCase().regexpMatch(".*" + result + "[a-zA-Z0-9].*|.*[a-zA-Z0-9]" + result + ".*")))
// and
// DES specific false positives
(
result.matches("DES")
implies
not f.getName().toUpperCase().regexpMatch(".*DES[a-zA-Z0-9].*|.*[a-zA-Z0-9]DES.*")
) and
// ((result.matches("%DES%")) implies not exists(string s | s in ["DESCRIBE", "DESTROY", "DESCRIPTION", "DESCRIPTOR", "NODES"] |
// f.getName().toUpperCase().matches("%" + s + "%"))) and
// SEED specific false positives
(
result.matches("%SEED%")
implies
not not exists(string s |
s in ["NEW_SEED", "GEN_SEED", "SET_SEED", "GET_SEED", "GET0_SEED", "RESEED", "SEEDING"]
|
f.getName().toUpperCase().matches("%" + s + "%")
)
) and
// ARIA specific false positives
(result.matches("%ARIA%") implies not f.getName().toUpperCase().matches("%VARIANT%"))
}
/**
* Predicate to support name normalization.
* Converts the raw name upper-case with no hyphen, slash, underscore, hash, or space.
* Looks for substrings that are known algorithms, and normalizes the name.
* If the algorithm cannot be determined or is in the ignorable list (`isIgnorableOpenSSLAlgorithm`)
* this predicate will not resolve a name.
*
* Rationale for private: For normalization, we want to get the longest string for a normalized name match
* for a given algorithm type. I found this easier to express if the public normalizeName
* checks that the name is the longest, and that UNKNOWN is reserved if there exists no
* result from this predicate that is known.
*/
bindingset[name]
string privateNormalizeName(string name, string algType) {
//not isIgnorableOpenSSLAlgorithm(name, _, _) and
// targetOpenSSLAlgorithm(name, _) and
isKnownAlgorithm(result, algType) and
exists(string normStr | normStr = name.toUpperCase().regexpReplaceAll("[-_ ]|/", "") |
normStr.matches("%" + result + "%")
)
}
/**
* Converts a raw OpenSSL algorithm to a normalized algorithm name.
*
* If more than one match occurs for a given algorithm type, normalize attempts to find the "max"
* string (max in terms of string length) e.g., matching AES128 to AES128 and not simply AES.
*
* An unknown algorithm is only identified if there exists no known algorithm found for any algorithm type.
*
* `name` is the name to normalize.
* `algType` is a string representing the classification of the algorithm (see `CryptoAlgorithmNames`)
*/
bindingset[name]
string normalizeName(string name, string algType) {
(
if exists(privateNormalizeName(name, _))
then result = privateNormalizeName(name, algType)
else (
result = unknownAlgorithm() and algType = "UNKNOWN"
)
) and
not exists(string res2 |
result != res2 and
res2 = privateNormalizeName(name, algType) and
res2.length() > result.length()
) and
// Addressing bad normalization case-by-case
// CASE: ES256 being identified when the algorithm is AES256
(
result.matches("ES256")
implies
not exists(string res2 | res2 = privateNormalizeName(name, _) and res2.matches("AES%"))
)
}

View File

@@ -1,59 +0,0 @@
import cpp
import experimental.cryptography.utils.OpenSSL.LibraryFunction
import semmle.code.cpp.ir.dataflow.DataFlow
// TODO: possible use of extensible predicates here
// NOTE: -1 for outInd represents the return value
predicate knownPassthroughFunction(Function f, int inInd, int outInd) {
// Trace through functions
// See https://www.openssl.org/docs/man1.1.1/man3/OBJ_obj2txt
// https://www.openssl.org/docs/man3.0/man3/EVP_CIPHER_get0_name
openSSLLibraryFunc(f) and
(
f.getName() in [
"OBJ_nid2obj", "OBJ_nid2ln", "OBJ_nid2sn", "OBJ_obj2nid", "OBJ_ln2nid", "OBJ_sn2nid",
"OBJ_txt2nid", "OBJ_txt2obj", "OBJ_dup", "EVP_CIPHER_get0_name"
] and
inInd = 0 and
outInd = -1
or
f.getName() in ["OBJ_obj2txt", "i2t_ASN1_OBJECT"] and
inInd = 2 and
outInd = 0
or
// Dup/copy pattern occurs in more places,
//see: https://www.openssl.org/docs/manmaster/man3/EC_KEY_copy.html and https://www.openssl.org/docs/manmaster/man3/EVP_PKEY_CTX_dup.html
f.getName().matches("%_dup") and inInd = 0 and outInd = -1
or
f.getName().matches("%_copy") and inInd = 0 and outInd = -1
)
}
/**
* `c` is a call to a function that preserves the algorithm but changes its form.
* `onExpr` is the input argument passing through to, `outExpr` is the next expression in a dataflow step associated with `c`
*/
predicate knownPassthoughCall(Call c, Expr inExpr, Expr outExpr) {
exists(int inInd, int outInd |
knownPassthroughFunction(c.getTarget(), inInd, outInd) and
inExpr = c.getArgument(inInd) and
if outInd = -1 then outExpr = c else outExpr = c.getArgument(outInd)
)
}
/*
* Explicitly add flow through openssl functions that preserve the algorithm but alter the form (e.g., from NID to string)
*/
predicate knownPassThroughStep(DataFlow::Node node1, DataFlow::Node node2) {
exists(Expr cur, Expr next |
(cur = node1.asExpr() or cur = node1.asIndirectArgument()) and
(
next = node2.asExpr() or
next = node2.asIndirectArgument() or
next = node2.asDefiningArgument()
)
|
exists(Call c | knownPassthoughCall(c, cur, next))
)
}

View File

@@ -1,5 +1,5 @@
name: codeql/cpp-all
version: 0.10.0-dev
version: 0.9.2-dev
groups: cpp
dbscheme: semmlecode.cpp.dbscheme
extractor: cpp

View File

@@ -5,35 +5,155 @@
import semmle.code.cpp.Element
import semmle.code.cpp.Declaration
import semmle.code.cpp.metrics.MetricFile
private import codeql.util.FileSystem
private module Input implements InputSig {
abstract class ContainerBase extends @container {
abstract string getAbsolutePath();
ContainerBase getParentContainer() {
containerparent(unresolveElement(result), underlyingElement(this))
}
string toString() { result = this.getAbsolutePath() }
}
class FolderBase extends ContainerBase, @folder {
override string getAbsolutePath() { folders(underlyingElement(this), result) }
}
class FileBase extends ContainerBase, @file {
override string getAbsolutePath() { files(underlyingElement(this), result) }
}
predicate hasSourceLocationPrefix = sourceLocationPrefix/1;
}
private module Impl = Make<Input>;
/** A file or folder. */
class Container extends Locatable, Impl::Container {
override string toString() { result = Impl::Container.super.toString() }
class Container extends Locatable, @container {
/**
* Gets the absolute, canonical path of this container, using forward slashes
* as path separator.
*
* The path starts with a _root prefix_ followed by zero or more _path
* segments_ separated by forward slashes.
*
* The root prefix is of one of the following forms:
*
* 1. A single forward slash `/` (Unix-style)
* 2. An upper-case drive letter followed by a colon and a forward slash,
* such as `C:/` (Windows-style)
* 3. Two forward slashes, a computer name, and then another forward slash,
* such as `//FileServer/` (UNC-style)
*
* Path segments are never empty (that is, absolute paths never contain two
* contiguous slashes, except as part of a UNC-style root prefix). Also, path
* segments never contain forward slashes, and no path segment is of the
* form `.` (one dot) or `..` (two dots).
*
* Note that an absolute path never ends with a forward slash, except if it is
* a bare root prefix, that is, the path has no path segments. A container
* whose absolute path has no segments is always a `Folder`, not a `File`.
*/
string getAbsolutePath() { none() } // overridden by subclasses
/**
* Gets the relative path of this file or folder from the root folder of the
* analyzed source location. The relative path of the root folder itself is
* the empty string.
*
* This has no result if the container is outside the source root, that is,
* if the root folder is not a reflexive, transitive parent of this container.
*/
string getRelativePath() {
exists(string absPath, string pref |
absPath = this.getAbsolutePath() and sourceLocationPrefix(pref)
|
absPath = pref and result = ""
or
absPath = pref.regexpReplaceAll("/$", "") + "/" + result and
not result.matches("/%")
)
}
/**
* Gets the base name of this container including extension, that is, the last
* segment of its absolute path, or the empty string if it has no segments.
*
* Here are some examples of absolute paths and the corresponding base names
* (surrounded with quotes to avoid ambiguity):
*
* <table border="1">
* <tr><th>Absolute path</th><th>Base name</th></tr>
* <tr><td>"/tmp/tst.js"</td><td>"tst.js"</td></tr>
* <tr><td>"C:/Program Files (x86)"</td><td>"Program Files (x86)"</td></tr>
* <tr><td>"/"</td><td>""</td></tr>
* <tr><td>"C:/"</td><td>""</td></tr>
* <tr><td>"D:/"</td><td>""</td></tr>
* <tr><td>"//FileServer/"</td><td>""</td></tr>
* </table>
*/
string getBaseName() {
result = this.getAbsolutePath().regexpCapture(".*/(([^/]*?)(?:\\.([^.]*))?)", 1)
}
/**
* Gets the extension of this container, that is, the suffix of its base name
* after the last dot character, if any.
*
* In particular,
*
* - if the name does not include a dot, there is no extension, so this
* predicate has no result;
* - if the name ends in a dot, the extension is the empty string;
* - if the name contains multiple dots, the extension follows the last dot.
*
* Here are some examples of absolute paths and the corresponding extensions
* (surrounded with quotes to avoid ambiguity):
*
* <table border="1">
* <tr><th>Absolute path</th><th>Extension</th></tr>
* <tr><td>"/tmp/tst.js"</td><td>"js"</td></tr>
* <tr><td>"/tmp/.classpath"</td><td>"classpath"</td></tr>
* <tr><td>"/bin/bash"</td><td>not defined</td></tr>
* <tr><td>"/tmp/tst2."</td><td>""</td></tr>
* <tr><td>"/tmp/x.tar.gz"</td><td>"gz"</td></tr>
* </table>
*/
string getExtension() {
result = this.getAbsolutePath().regexpCapture(".*/([^/]*?)(\\.([^.]*))?", 3)
}
/**
* Gets the stem of this container, that is, the prefix of its base name up to
* (but not including) the last dot character if there is one, or the entire
* base name if there is not.
*
* Here are some examples of absolute paths and the corresponding stems
* (surrounded with quotes to avoid ambiguity):
*
* <table border="1">
* <tr><th>Absolute path</th><th>Stem</th></tr>
* <tr><td>"/tmp/tst.js"</td><td>"tst"</td></tr>
* <tr><td>"/tmp/.classpath"</td><td>""</td></tr>
* <tr><td>"/bin/bash"</td><td>"bash"</td></tr>
* <tr><td>"/tmp/tst2."</td><td>"tst2"</td></tr>
* <tr><td>"/tmp/x.tar.gz"</td><td>"x.tar"</td></tr>
* </table>
*/
string getStem() {
result = this.getAbsolutePath().regexpCapture(".*/([^/]*?)(?:\\.([^.]*))?", 1)
}
/** Gets the parent container of this file or folder, if any. */
Container getParentContainer() {
containerparent(unresolveElement(result), underlyingElement(this))
}
/** Gets a file or sub-folder in this container. */
Container getAChildContainer() { this = result.getParentContainer() }
/** Gets a file in this container. */
File getAFile() { result = this.getAChildContainer() }
/** Gets the file in this container that has the given `baseName`, if any. */
File getFile(string baseName) {
result = this.getAFile() and
result.getBaseName() = baseName
}
/** Gets a sub-folder in this container. */
Folder getAFolder() { result = this.getAChildContainer() }
/** Gets the sub-folder in this container that has the given `baseName`, if any. */
Folder getFolder(string baseName) {
result = this.getAFolder() and
result.getBaseName() = baseName
}
/**
* Gets a textual representation of the path of this container.
*
* This is the absolute path of the container.
*/
override string toString() { result = this.getAbsolutePath() }
}
/**
@@ -46,7 +166,9 @@ class Container extends Locatable, Impl::Container {
*
* To get the full path, use `getAbsolutePath`.
*/
class Folder extends Container, Impl::Folder {
class Folder extends Container, @folder {
override string getAbsolutePath() { folders(underlyingElement(this), result) }
override Location getLocation() {
result.getContainer() = this and
result.hasLocationInfo(_, 0, 0, 0, 0)
@@ -67,7 +189,9 @@ class Folder extends Container, Impl::Folder {
* The base name further decomposes into the _stem_ and _extension_ -- see
* `getStem` and `getExtension`. To get the full path, use `getAbsolutePath`.
*/
class File extends Container, Impl::File {
class File extends Container, @file {
override string getAbsolutePath() { files(underlyingElement(this), result) }
override string getAPrimaryQlClass() { result = "File" }
override Location getLocation() {

View File

@@ -26,18 +26,17 @@ predicate callDereferences(FunctionCall fc, int i) {
}
/**
* Holds if evaluation of `op` dereferences `e` directly.
*
* This predicate does not recurse through function calls or arithmetic operations. To find
* such cases, use `dereferencedByOperation`.
* Holds if evaluation of `op` dereferences `e`.
*/
predicate directDereferencedByOperation(Expr op, Expr e) {
predicate dereferencedByOperation(Expr op, Expr e) {
exists(PointerDereferenceExpr deref |
deref.getAChild() = e and
deref = op and
not deref.getParent*() instanceof SizeofOperator
)
or
exists(CrementOperation crement | dereferencedByOperation(e, op) and crement.getOperand() = e)
or
exists(ArrayExpr ae |
(
not ae.getParent() instanceof AddressOfExpr and
@@ -51,24 +50,6 @@ predicate directDereferencedByOperation(Expr op, Expr e) {
)
)
or
// ptr->Field
e = op.(FieldAccess).getQualifier() and isClassPointerType(e.getType())
or
// ptr->method()
e = op.(Call).getQualifier() and isClassPointerType(e.getType())
}
/**
* Holds if evaluation of `op` dereferences `e`.
*
* This includes the set of operations identified via `directDereferencedByOperation`, as well
* as calls to function that are known to dereference an argument.
*/
predicate dereferencedByOperation(Expr op, Expr e) {
directDereferencedByOperation(op, e)
or
exists(CrementOperation crement | dereferencedByOperation(e, op) and crement.getOperand() = e)
or
exists(AddressOfExpr addof, ArrayExpr ae |
dereferencedByOperation(addof, op) and
addof.getOperand() = ae and
@@ -93,6 +74,12 @@ predicate dereferencedByOperation(Expr op, Expr e) {
e = fc.getArgument(i) and
op = fc
)
or
// ptr->Field
e = op.(FieldAccess).getQualifier() and isClassPointerType(e.getType())
or
// ptr->method()
e = op.(Call).getQualifier() and isClassPointerType(e.getType())
}
private predicate isClassPointerType(Type t) {

View File

@@ -240,7 +240,7 @@ private class GuardConditionFromIR extends GuardCondition {
*/
private predicate controlsBlock(BasicBlock controlled, boolean testIsTrue) {
exists(IRBlock irb |
ir.controls(irb, testIsTrue) and
forex(IRGuardCondition inst | inst = ir | inst.controls(irb, testIsTrue)) and
irb.getAnInstruction().getAst().(ControlFlowNode).getBasicBlock() = controlled and
not isUnreachedBlock(irb)
)

View File

@@ -79,3 +79,13 @@ class ArgumentPosition extends int {
/** Holds if arguments at position `apos` match parameters at position `ppos`. */
pragma[inline]
predicate parameterMatch(ParameterPosition ppos, ArgumentPosition apos) { ppos = apos }
/**
* Holds if flow from `call`'s argument `arg` to parameter `p` is permissible.
*
* This is a temporary hook to support technical debt in the Go language; do not use.
*/
pragma[inline]
predicate golangSpecificParamArgFilter(DataFlowCall call, ParameterNode p, ArgumentNode arg) {
any()
}

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -208,8 +208,6 @@ predicate expectsContent(Node n, ContentSet c) { none() }
predicate typeStrongerThan(DataFlowType t1, DataFlowType t2) { none() }
predicate localMustFlowStep(Node node1, Node node2) { none() }
/** Gets the type of `n` used for type pruning. */
Type getNodeType(Node n) {
suppressUnusedNode(n) and
@@ -297,3 +295,12 @@ class ContentApprox = Unit;
/** Gets an approximated value for content `c`. */
pragma[inline]
ContentApprox getContentApprox(Content c) { any() }
/**
* Gets an additional term that is added to the `join` and `branch` computations to reflect
* an additional forward or backwards branching factor that is not taken into account
* when calculating the (virtual) dispatch cost.
*
* Argument `arg` is part of a path from a source to a sink, and `p` is the target parameter.
*/
int getAdditionalFlowIntoCallNodeTerm(ArgumentNode arg, ParameterNode p) { none() }

View File

@@ -1547,21 +1547,3 @@ class BuiltInBitCast extends BuiltInOperation, @builtinbitcast {
override string getAPrimaryQlClass() { result = "BuiltInBitCast" }
}
/**
* A C++ `__is_trivial` built-in operation (used by some implementations of the
* `<type_traits>` header).
*
* Returns `true` if a type is a trivial type.
* ```
* template<typename _Tp>
* struct is_trivial
* : public integral_constant<bool, __is_trivial(_Tp)>
* {};
* ```
*/
class BuiltInIsTrivial extends BuiltInOperation, @istrivialexpr {
override string toString() { result = "__is_trivial" }
override string getAPrimaryQlClass() { result = "BuiltInIsTrivial" }
}

View File

@@ -271,3 +271,13 @@ DataFlowCallable viableImplInCallContext(DataFlowCall call, DataFlowCall ctx) {
/** Holds if arguments at position `apos` match parameters at position `ppos`. */
pragma[inline]
predicate parameterMatch(ParameterPosition ppos, ArgumentPosition apos) { ppos = apos }
/**
* Holds if flow from `call`'s argument `arg` to parameter `p` is permissible.
*
* This is a temporary hook to support technical debt in the Go language; do not use.
*/
pragma[inline]
predicate golangSpecificParamArgFilter(DataFlowCall call, ParameterNode p, ArgumentNode arg) {
any()
}

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -297,10 +297,6 @@ private module Config implements FullStateConfigSig {
predicate isBarrierOut(Node node) { any(Configuration config).isBarrierOut(node) }
predicate isBarrierIn(Node node, FlowState state) { none() }
predicate isBarrierOut(Node node, FlowState state) { none() }
predicate isAdditionalFlowStep(Node node1, Node node2) {
singleConfiguration() and
any(Configuration config).isAdditionalFlowStep(node1, node2)

View File

@@ -18,6 +18,4 @@ module CppDataFlow implements InputSig {
import Public
Node exprNode(DataFlowExpr e) { result = Public::exprNode(e) }
predicate getAdditionalFlowIntoCallNodeTerm = Private::getAdditionalFlowIntoCallNodeTerm/2;
}

View File

@@ -804,8 +804,6 @@ predicate expectsContent(Node n, ContentSet c) { none() }
predicate typeStrongerThan(DataFlowType t1, DataFlowType t2) { none() }
predicate localMustFlowStep(Node node1, Node node2) { none() }
/** Gets the type of `n` used for type pruning. */
DataFlowType getNodeType(Node n) {
suppressUnusedNode(n) and

View File

@@ -193,23 +193,13 @@ class Node extends TIRDataFlowNode {
* a `Conversion`, then the result is the underlying non-`Conversion` base
* expression.
*/
Expr asExpr() { result = this.asExpr(_) }
/**
* INTERNAL: Do not use.
*/
Expr asExpr(int n) { result = this.(ExprNode).getExpr(n) }
/**
* INTERNAL: Do not use.
*/
Expr asIndirectExpr(int n, int index) { result = this.(IndirectExprNode).getExpr(n, index) }
Expr asExpr() { result = this.(ExprNode).getExpr() }
/**
* Gets the non-conversion expression that's indirectly tracked by this node
* under `index` number of indirections.
*/
Expr asIndirectExpr(int index) { result = this.asIndirectExpr(_, index) }
Expr asIndirectExpr(int index) { result = this.(IndirectExprNode).getExpr(index) }
/**
* Gets the non-conversion expression that's indirectly tracked by this node
@@ -221,26 +211,15 @@ class Node extends TIRDataFlowNode {
* Gets the expression corresponding to this node, if any. The returned
* expression may be a `Conversion`.
*/
Expr asConvertedExpr() { result = this.asConvertedExpr(_) }
/**
* Gets the expression corresponding to this node, if any. The returned
* expression may be a `Conversion`.
*/
Expr asConvertedExpr(int n) { result = this.(ExprNode).getConvertedExpr(n) }
/**
* INTERNAL: Do not use.
*/
Expr asIndirectConvertedExpr(int n, int index) {
result = this.(IndirectExprNode).getConvertedExpr(n, index)
}
Expr asConvertedExpr() { result = this.(ExprNode).getConvertedExpr() }
/**
* Gets the expression that's indirectly tracked by this node
* behind `index` number of indirections.
*/
Expr asIndirectConvertedExpr(int index) { result = this.asIndirectConvertedExpr(_, index) }
Expr asIndirectConvertedExpr(int index) {
result = this.(IndirectExprNode).getConvertedExpr(index)
}
/**
* Gets the expression that's indirectly tracked by this node behind a
@@ -275,7 +254,9 @@ class Node extends TIRDataFlowNode {
* after the `f` has returned.
*/
Expr asDefiningArgument(int index) {
this.(DefinitionByReferenceNode).getIndirectionIndex() = index and
// Subtract one because `DefinitionByReferenceNode` is defined to be in
// the range `[0 ... n - 1]` for some `n` instead of `[1 ... n]`.
this.(DefinitionByReferenceNode).getIndirectionIndex() = index - 1 and
result = this.(DefinitionByReferenceNode).getArgument()
}
@@ -412,10 +393,9 @@ class Node extends TIRDataFlowNode {
}
private string toExprString(Node n) {
result = n.asExpr(0).toString()
result = n.asExpr().toString()
or
not exists(n.asExpr()) and
result = n.asIndirectExpr(0, 1).toString() + " indirection"
result = n.asIndirectExpr().toString() + " indirection"
}
/**
@@ -955,7 +935,7 @@ class RawIndirectOperand extends Node, TRawIndirectOperand {
}
override string toStringImpl() {
result = operandNode(this.getOperand()).toStringImpl() + " indirection"
result = instructionNode(this.getOperand().getDef()).toStringImpl() + " indirection"
}
}
@@ -1062,130 +1042,77 @@ class RawIndirectInstruction extends Node, TRawIndirectInstruction {
}
}
private module GetConvertedResultExpression {
private import semmle.code.cpp.ir.implementation.raw.internal.TranslatedExpr
private import semmle.code.cpp.ir.implementation.raw.internal.InstructionTag
private Operand getAnInitializeDynamicAllocationInstructionAddress() {
result = any(InitializeDynamicAllocationInstruction init).getAllocationAddressOperand()
}
/**
* Gets the expression that should be returned as the result expression from `instr`.
*
* Note that this predicate may return multiple results in cases where a conversion belongs to a
* different AST element than its operand.
*/
Expr getConvertedResultExpression(Instruction instr, int n) {
// Only fully converted instructions have a result for `asConvertedExpr`
not conversionFlow(unique(Operand op |
// The address operand of a `InitializeDynamicAllocationInstruction` is
// special: we need to handle it during dataflow (since it's
// effectively a store to an indirection), but it doesn't appear in
// source syntax, so dataflow node <-> expression conversion shouldn't
// care about it.
op = getAUse(instr) and not op = getAnInitializeDynamicAllocationInstructionAddress()
|
op
), _, false, false) and
result = getConvertedResultExpressionImpl(instr) and
n = 0
or
// If the conversion also has a result then we return multiple results
exists(Operand operand | conversionFlow(operand, instr, false, false) |
n = 1 and
result = getConvertedResultExpressionImpl(operand.getDef())
or
result = getConvertedResultExpression(operand.getDef(), n - 1)
)
}
private Expr getConvertedResultExpressionImpl0(Instruction instr) {
// For an expression such as `i += 2` we pretend that the generated
// `StoreInstruction` contains the result of the expression even though
// this isn't totally aligned with the C/C++ standard.
exists(TranslatedAssignOperation tao |
result = tao.getExpr() and
instr = tao.getInstruction(any(AssignmentStoreTag tag))
)
or
// Similarly for `i++` and `++i` we pretend that the generated
// `StoreInstruction` is contains the result of the expression even though
// this isn't totally aligned with the C/C++ standard.
exists(TranslatedCrementOperation tco |
result = tco.getExpr() and
instr = tco.getInstruction(any(CrementStoreTag tag))
)
or
// IR construction inserts an additional cast to a `size_t` on the extent
// of a `new[]` expression. The resulting `ConvertInstruction` doesn't have
// a result for `getConvertedResultExpression`. We remap this here so that
// this `ConvertInstruction` maps to the result of the expression that
// represents the extent.
exists(TranslatedNonConstantAllocationSize tas |
result = tas.getExtent().getExpr() and
instr = tas.getInstruction(any(AllocationExtentConvertTag tag))
)
or
// There's no instruction that returns `ParenthesisExpr`, but some queries
// expect this
exists(TranslatedTransparentConversion ttc |
result = ttc.getExpr().(ParenthesisExpr) and
instr = ttc.getResult()
)
}
private Expr getConvertedResultExpressionImpl(Instruction instr) {
result = getConvertedResultExpressionImpl0(instr)
or
not exists(getConvertedResultExpressionImpl0(instr)) and
result = instr.getConvertedResultExpression()
}
}
private import GetConvertedResultExpression
/** Holds if `node` is an `OperandNode` that should map `node.asExpr()` to `e`. */
predicate exprNodeShouldBeOperand(OperandNode node, Expr e, int n) {
predicate exprNodeShouldBeOperand(OperandNode node, Expr e) {
exists(Instruction def |
unique( | | getAUse(def)) = node.getOperand() and
e = getConvertedResultExpression(def, n)
e = def.getConvertedResultExpression()
)
}
private predicate indirectExprNodeShouldBeIndirectOperand0(
VariableAddressInstruction instr, RawIndirectOperand node, Expr e
) {
instr = node.getOperand().getDef() and
e = instr.getAst().(Expr).getUnconverted()
}
/** Holds if `node` should be an `IndirectOperand` that maps `node.asIndirectExpr()` to `e`. */
private predicate indirectExprNodeShouldBeIndirectOperand(
IndirectOperand node, Expr e, int n, int indirectionIndex
) {
exists(Instruction def |
node.hasOperandAndIndirectionIndex(unique( | | getAUse(def)), indirectionIndex) and
e = getConvertedResultExpression(def, n)
private predicate indirectExprNodeShouldBeIndirectOperand(RawIndirectOperand node, Expr e) {
exists(Instruction instr | instr = node.getOperand().getDef() |
exists(Expr e0 |
indirectExprNodeShouldBeIndirectOperand0(instr, node, e0) and
e = e0.getFullyConverted()
)
or
not indirectExprNodeShouldBeIndirectOperand0(_, node, _) and
e = instr.getConvertedResultExpression()
)
}
private predicate exprNodeShouldBeIndirectOutNode(IndirectArgumentOutNode node, Expr e, int n) {
private predicate exprNodeShouldBeIndirectOutNode(IndirectArgumentOutNode node, Expr e) {
exists(CallInstruction call |
call.getStaticCallTarget() instanceof Constructor and
e = getConvertedResultExpression(call, n) and
e = call.getConvertedResultExpression() and
call.getThisArgumentOperand() = node.getAddressOperand()
)
}
/** Holds if `node` should be an instruction node that maps `node.asExpr()` to `e`. */
predicate exprNodeShouldBeInstruction(Node node, Expr e, int n) {
not exprNodeShouldBeOperand(_, e, n) and
not exprNodeShouldBeIndirectOutNode(_, e, n) and
e = getConvertedResultExpression(node.asInstruction(), n)
predicate exprNodeShouldBeInstruction(Node node, Expr e) {
not exprNodeShouldBeOperand(_, e) and
not exprNodeShouldBeIndirectOutNode(_, e) and
(
e = node.asInstruction().getConvertedResultExpression()
or
// The instruction that contains the result of an `AssignOperation` is
// the unloaded left operand (see the comments in `TranslatedAssignOperation`).
// That means that for cases like
// ```cpp
// int x = ...;
// x += 1;
// ```
// the result of `x += 1` is the `VariableAddressInstruction` that represents `x`. But
// that instruction doesn't receive the flow from this `AssignOperation`. So instead we
// map the operation to the `AddInstruction`.
node.asInstruction().getAst() = e.(AssignOperation)
or
// Same story for `CrementOperation`s (cf. the comments in the subclasses
// of `TranslatedCrementOperation`).
node.asInstruction().getAst() = e.(CrementOperation)
)
}
/** Holds if `node` should be an `IndirectInstruction` that maps `node.asIndirectExpr()` to `e`. */
predicate indirectExprNodeShouldBeIndirectInstruction(
IndirectInstruction node, Expr e, int n, int indirectionIndex
) {
not indirectExprNodeShouldBeIndirectOperand(_, e, n, indirectionIndex) and
predicate indirectExprNodeShouldBeIndirectInstruction(IndirectInstruction node, Expr e) {
exists(Instruction instr |
node.hasInstructionAndIndirectionIndex(instr, indirectionIndex) and
e = getConvertedResultExpression(instr, n)
node.hasInstructionAndIndirectionIndex(instr, _) and
not indirectExprNodeShouldBeIndirectOperand(_, e)
|
e = instr.(VariableAddressInstruction).getAst().(Expr).getFullyConverted()
or
not instr instanceof VariableAddressInstruction and
e = instr.getConvertedResultExpression()
)
}
@@ -1194,32 +1121,30 @@ abstract private class ExprNodeBase extends Node {
* Gets the expression corresponding to this node, if any. The returned
* expression may be a `Conversion`.
*/
abstract Expr getConvertedExpr(int n);
abstract Expr getConvertedExpr();
/** Gets the non-conversion expression corresponding to this node, if any. */
final Expr getExpr(int n) { result = this.getConvertedExpr(n).getUnconverted() }
abstract Expr getExpr();
}
private class InstructionExprNode extends ExprNodeBase, InstructionNode {
InstructionExprNode() {
exists(Expr e, int n |
exprNodeShouldBeInstruction(this, e, n) and
not exprNodeShouldBeInstruction(_, e, n + 1)
)
}
InstructionExprNode() { exprNodeShouldBeInstruction(this, _) }
final override Expr getConvertedExpr(int n) { exprNodeShouldBeInstruction(this, result, n) }
final override Expr getConvertedExpr() { exprNodeShouldBeInstruction(this, result) }
final override Expr getExpr() { result = this.getConvertedExpr().getUnconverted() }
final override string toStringImpl() { result = this.getConvertedExpr().toString() }
}
private class OperandExprNode extends ExprNodeBase, OperandNode {
OperandExprNode() {
exists(Expr e, int n |
exprNodeShouldBeOperand(this, e, n) and
not exprNodeShouldBeOperand(_, e, n + 1)
)
}
OperandExprNode() { exprNodeShouldBeOperand(this, _) }
final override Expr getConvertedExpr(int n) { exprNodeShouldBeOperand(this, result, n) }
final override Expr getConvertedExpr() { exprNodeShouldBeOperand(this, result) }
final override Expr getExpr() { result = this.getConvertedExpr().getUnconverted() }
final override string toStringImpl() { result = this.getConvertedExpr().toString() }
}
abstract private class IndirectExprNodeBase extends Node {
@@ -1227,75 +1152,67 @@ abstract private class IndirectExprNodeBase extends Node {
* Gets the expression corresponding to this node, if any. The returned
* expression may be a `Conversion`.
*/
abstract Expr getConvertedExpr(int n, int indirectionIndex);
abstract Expr getConvertedExpr(int indirectionIndex);
/** Gets the non-conversion expression corresponding to this node, if any. */
final Expr getExpr(int n, int indirectionIndex) {
result = this.getConvertedExpr(n, indirectionIndex).getUnconverted()
abstract Expr getExpr(int indirectionIndex);
}
private class IndirectOperandIndirectExprNode extends IndirectExprNodeBase, RawIndirectOperand {
IndirectOperandIndirectExprNode() { indirectExprNodeShouldBeIndirectOperand(this, _) }
final override Expr getConvertedExpr(int index) {
this.getIndirectionIndex() = index and
indirectExprNodeShouldBeIndirectOperand(this, result)
}
final override Expr getExpr(int index) {
this.getIndirectionIndex() = index and
result = this.getConvertedExpr(index).getUnconverted()
}
}
private class IndirectOperandIndirectExprNode extends IndirectExprNodeBase instanceof IndirectOperand
private class IndirectInstructionIndirectExprNode extends IndirectExprNodeBase,
RawIndirectInstruction
{
IndirectOperandIndirectExprNode() {
exists(Expr e, int n, int indirectionIndex |
indirectExprNodeShouldBeIndirectOperand(this, e, n, indirectionIndex) and
not indirectExprNodeShouldBeIndirectOperand(_, e, n + 1, indirectionIndex)
)
IndirectInstructionIndirectExprNode() { indirectExprNodeShouldBeIndirectInstruction(this, _) }
final override Expr getConvertedExpr(int index) {
this.getIndirectionIndex() = index and
indirectExprNodeShouldBeIndirectInstruction(this, result)
}
final override Expr getConvertedExpr(int n, int index) {
indirectExprNodeShouldBeIndirectOperand(this, result, n, index)
}
}
private class IndirectInstructionIndirectExprNode extends IndirectExprNodeBase instanceof IndirectInstruction
{
IndirectInstructionIndirectExprNode() {
exists(Expr e, int n, int indirectionIndex |
indirectExprNodeShouldBeIndirectInstruction(this, e, n, indirectionIndex) and
not indirectExprNodeShouldBeIndirectInstruction(_, e, n + 1, indirectionIndex)
)
}
final override Expr getConvertedExpr(int n, int index) {
indirectExprNodeShouldBeIndirectInstruction(this, result, n, index)
final override Expr getExpr(int index) {
this.getIndirectionIndex() = index and
result = this.getConvertedExpr(index).getUnconverted()
}
}
private class IndirectArgumentOutExprNode extends ExprNodeBase, IndirectArgumentOutNode {
IndirectArgumentOutExprNode() { exprNodeShouldBeIndirectOutNode(this, _, _) }
IndirectArgumentOutExprNode() { exprNodeShouldBeIndirectOutNode(this, _) }
final override Expr getConvertedExpr(int n) { exprNodeShouldBeIndirectOutNode(this, result, n) }
final override Expr getConvertedExpr() { exprNodeShouldBeIndirectOutNode(this, result) }
final override Expr getExpr() { result = this.getConvertedExpr() }
}
/**
* An expression, viewed as a node in a data flow graph.
*/
class ExprNode extends Node instanceof ExprNodeBase {
/**
* INTERNAL: Do not use.
*/
Expr getExpr(int n) { result = super.getExpr(n) }
/**
* Gets the non-conversion expression corresponding to this node, if any. If
* this node strictly (in the sense of `getConvertedExpr`) corresponds to a
* `Conversion`, then the result is that `Conversion`'s non-`Conversion` base
* expression.
*/
final Expr getExpr() { result = this.getExpr(_) }
/**
* INTERNAL: Do not use.
*/
Expr getConvertedExpr(int n) { result = super.getConvertedExpr(n) }
Expr getExpr() { result = super.getExpr() }
/**
* Gets the expression corresponding to this node, if any. The returned
* expression may be a `Conversion`.
*/
final Expr getConvertedExpr() { result = this.getConvertedExpr(_) }
Expr getConvertedExpr() { result = super.getConvertedExpr() }
}
/**
@@ -1308,27 +1225,13 @@ class IndirectExprNode extends Node instanceof IndirectExprNodeBase {
* `Conversion`, then the result is that `Conversion`'s non-`Conversion` base
* expression.
*/
final Expr getExpr(int indirectionIndex) { result = this.getExpr(_, indirectionIndex) }
/**
* INTERNAL: Do not use.
*/
Expr getExpr(int n, int indirectionIndex) { result = super.getExpr(n, indirectionIndex) }
/**
* INTERNAL: Do not use.
*/
Expr getConvertedExpr(int n, int indirectionIndex) {
result = super.getConvertedExpr(n, indirectionIndex)
}
Expr getExpr(int indirectionIndex) { result = super.getExpr(indirectionIndex) }
/**
* Gets the expression corresponding to this node, if any. The returned
* expression may be a `Conversion`.
*/
Expr getConvertedExpr(int indirectionIndex) {
result = this.getConvertedExpr(_, indirectionIndex)
}
Expr getConvertedExpr(int indirectionIndex) { result = super.getConvertedExpr(indirectionIndex) }
}
/**
@@ -1354,9 +1257,6 @@ class ParameterNode extends Node {
* pointer-indirection parameters are at further negative positions.
*/
predicate isParameterOf(Function f, ParameterPosition pos) { none() } // overridden by subclasses
/** Gets the `Parameter` associated with this node, if it exists. */
Parameter getParameter() { none() } // overridden by subclasses
}
/** An explicit positional parameter, including `this`, but not `...`. */
@@ -1379,9 +1279,10 @@ private class ExplicitParameterNode extends ParameterNode, DirectParameterNode {
f.getParameter(pos.(DirectPosition).getIndex()) = instr.getParameter()
}
override string toStringImpl() { result = instr.getParameter().toString() }
/** Gets the `Parameter` associated with this node. */
Parameter getParameter() { result = instr.getParameter() }
override Parameter getParameter() { result = instr.getParameter() }
override string toStringImpl() { result = instr.getParameter().toString() }
}
/** An implicit `this` parameter. */
@@ -1543,7 +1444,7 @@ OperandNode operandNode(Operand operand) { result.getOperand() = operand }
* _out of_ an expression, like when an argument is passed by reference, use
* `definitionByReferenceNodeFromArgument` instead.
*/
ExprNode exprNode(Expr e) { result.getExpr(_) = e }
ExprNode exprNode(Expr e) { result.getExpr() = e }
/**
* Gets the `Node` corresponding to the value of evaluating `e`. Here, `e` may
@@ -1551,7 +1452,7 @@ ExprNode exprNode(Expr e) { result.getExpr(_) = e }
* argument is passed by reference, use
* `definitionByReferenceNodeFromArgument` instead.
*/
ExprNode convertedExprNode(Expr e) { result.getConvertedExpr(_) = e }
ExprNode convertedExprNode(Expr e) { result.getConvertedExpr() = e }
/**
* Gets the `Node` corresponding to the value of `p` at function entry.

View File

@@ -447,16 +447,9 @@ class GlobalUse extends UseImpl, TGlobalUse {
IRFunction getIRFunction() { result = f }
final override predicate hasIndexInBlock(IRBlock block, int index) {
// Similar to the `FinalParameterUse` case, we want to generate flow out of
// globals at any exit so that we can flow out of non-returning functions.
// Obviously this isn't correct as we can't actually flow but the global flow
// requires this if we want to flow into children.
exists(Instruction return |
return instanceof ReturnInstruction or
return instanceof UnreachedInstruction
|
block.getInstruction(index) = return and
return.getEnclosingIRFunction() = f
exists(ExitFunctionInstruction exit |
exit = f.getExitFunctionInstruction() and
block.getInstruction(index) = exit
)
}
@@ -645,24 +638,12 @@ private predicate adjustForPointerArith(PostUpdateNode pun, UseOrPhi use) {
)
}
/**
* Holds if `nodeFrom` flows to `nodeTo` because there is `def-use` or
* `use-use` flow from `defOrUse` to `use`.
*
* `uncertain` is `true` if the `defOrUse` is an uncertain definition.
*/
private predicate localSsaFlow(
SsaDefOrUse defOrUse, Node nodeFrom, UseOrPhi use, Node nodeTo, boolean uncertain
) {
nodeToDefOrUse(nodeFrom, defOrUse, uncertain) and
adjacentDefRead(defOrUse, use) and
useToNode(use, nodeTo) and
nodeFrom != nodeTo
}
private predicate ssaFlowImpl(SsaDefOrUse defOrUse, Node nodeFrom, Node nodeTo, boolean uncertain) {
exists(UseOrPhi use |
localSsaFlow(defOrUse, nodeFrom, use, nodeTo, uncertain)
nodeToDefOrUse(nodeFrom, defOrUse, uncertain) and
adjacentDefRead(defOrUse, use) and
useToNode(use, nodeTo) and
nodeFrom != nodeTo
or
// Initial global variable value to a first use
nodeFrom.(InitialGlobalValue).getGlobalDef() = defOrUse and
@@ -740,62 +721,15 @@ private predicate isArgumentOfCallable(DataFlowCall call, Node n) {
)
}
/**
* Holds if there is use-use flow from `pun`'s pre-update node to `n`.
*/
private predicate postUpdateNodeToFirstUse(PostUpdateNode pun, Node n) {
exists(UseOrPhi use |
adjustForPointerArith(pun, use) and
useToNode(use, n)
)
}
private predicate stepUntilNotInCall(DataFlowCall call, Node n1, Node n2) {
isArgumentOfCallable(call, n1) and
exists(Node mid | localSsaFlow(_, n1, _, mid, _) |
isArgumentOfCallable(call, mid) and
stepUntilNotInCall(call, mid, n2)
or
not isArgumentOfCallable(call, mid) and
mid = n2
)
}
bindingset[n1, n2]
pragma[inline_late]
private predicate isArgumentOfSameCall(DataFlowCall call, Node n1, Node n2) {
isArgumentOfCallable(call, n1) and isArgumentOfCallable(call, n2)
}
/**
* Holds if there is def-use or use-use flow from `pun` to `nodeTo`.
*
* Note: This is more complex than it sounds. Consider a call such as:
* ```cpp
* write_first_argument(x, x);
* sink(x);
* ```
* Assume flow comes out of the first argument to `write_first_argument`. We
* don't want flow to go to the `x` that's also an argument to
* `write_first_argument` (because we just flowed out of that function, and we
* don't want to flow back into it again).
*
* We do, however, want flow from the output argument to `x` on the next line, and
* similarly we want flow from the second argument of `write_first_argument` to `x`
* on the next line.
*/
/** Holds if there is def-use or use-use flow from `pun` to `nodeTo`. */
predicate postUpdateFlow(PostUpdateNode pun, Node nodeTo) {
exists(Node preUpdate, Node mid |
exists(UseOrPhi use, Node preUpdate |
adjustForPointerArith(pun, use) and
useToNode(use, nodeTo) and
preUpdate = pun.getPreUpdateNode() and
postUpdateNodeToFirstUse(pun, mid)
|
exists(DataFlowCall call |
isArgumentOfSameCall(call, preUpdate, mid) and
stepUntilNotInCall(call, mid, nodeTo)
not exists(DataFlowCall call |
isArgumentOfCallable(call, preUpdate) and isArgumentOfCallable(call, nodeTo)
)
or
not isArgumentOfSameCall(_, preUpdate, mid) and
nodeTo = mid
)
}
@@ -825,7 +759,7 @@ predicate fromPhiNode(SsaPhiNode nodeFrom, Node nodeTo) {
or
exists(PhiNode phiTo |
phi != phiTo and
lastRefRedefExt(phi, bb1, i1, phiTo) and
lastRefRedefExt(phi, _, _, phiTo) and
nodeTo.(SsaPhiNode).getPhiNode() = phiTo
)
)

View File

@@ -815,7 +815,7 @@ private module Cached {
) {
indirectionIndex = [1 .. countIndirectionsForCppType(getResultLanguageType(instr))] and
exists(Instruction load, Operand address |
address = unique( | | getAUse(instr)) and
address.getDef() = instr and
isDereference(load, address, false) and
instrRepr = load and
indirectionIndexRepr = indirectionIndex - 1

View File

@@ -405,6 +405,9 @@ predicate hasUnreachedInstruction(IRFunction func) {
exists(Call c |
c.getEnclosingFunction() = func.getFunction() and
any(Options opt).exits(c.getTarget())
) and
not exists(TranslatedUnreachableReturnStmt return |
return.getEnclosingFunction().getFunction() = func.getFunction()
)
}

View File

@@ -824,9 +824,6 @@ abstract class TranslatedElement extends TTranslatedElement {
/** DEPRECATED: Alias for getAst */
deprecated Locatable getAST() { result = this.getAst() }
/** Gets the location of this element. */
Location getLocation() { result = this.getAst().getLocation() }
/**
* Get the first instruction to be executed in the evaluation of this element.
*/

View File

@@ -1906,10 +1906,8 @@ class TranslatedNonConstantAllocationSize extends TranslatedAllocationSize {
final override predicate hasInstruction(Opcode opcode, InstructionTag tag, CppType resultType) {
resultType = getTypeForPRValue(expr.getAllocator().getParameter(0).getType()) and
(
this.extentNeedsConversion() and
// Convert the extent to `size_t`, because the AST doesn't do this already.
tag = AllocationExtentConvertTag() and
opcode instanceof Opcode::Convert
tag = AllocationExtentConvertTag() and opcode instanceof Opcode::Convert
or
tag = AllocationElementSizeTag() and opcode instanceof Opcode::Constant
or
@@ -1920,7 +1918,6 @@ class TranslatedNonConstantAllocationSize extends TranslatedAllocationSize {
final override Instruction getInstructionSuccessor(InstructionTag tag, EdgeKind kind) {
kind instanceof GotoEdge and
(
this.extentNeedsConversion() and
tag = AllocationExtentConvertTag() and
result = this.getInstruction(AllocationElementSizeTag())
or
@@ -1936,9 +1933,7 @@ class TranslatedNonConstantAllocationSize extends TranslatedAllocationSize {
final override Instruction getChildSuccessor(TranslatedElement child) {
child = this.getExtent() and
if this.extentNeedsConversion()
then result = this.getInstruction(AllocationExtentConvertTag())
else result = this.getInstruction(AllocationElementSizeTag())
result = this.getInstruction(AllocationExtentConvertTag())
}
final override string getInstructionConstantValue(InstructionTag tag) {
@@ -1950,31 +1945,19 @@ class TranslatedNonConstantAllocationSize extends TranslatedAllocationSize {
tag = AllocationSizeTag() and
(
operandTag instanceof LeftOperandTag and
(
if this.extentNeedsConversion()
then result = this.getInstruction(AllocationExtentConvertTag())
else result = this.getExtent().getResult()
)
result = this.getInstruction(AllocationExtentConvertTag())
or
operandTag instanceof RightOperandTag and
result = this.getInstruction(AllocationElementSizeTag())
)
or
this.extentNeedsConversion() and
tag = AllocationExtentConvertTag() and
operandTag instanceof UnaryOperandTag and
result = this.getExtent().getResult()
}
TranslatedExpr getExtent() { result = getTranslatedExpr(expr.getExtent().getFullyConverted()) }
/**
* Holds if the result of `expr.getExtent()` does not have the same type as
* the allocator's size parameter.
*/
private predicate extentNeedsConversion() {
expr.getExtent().getFullyConverted().getUnspecifiedType() !=
expr.getAllocator().getParameter(0).getUnspecifiedType()
private TranslatedExpr getExtent() {
result = getTranslatedExpr(expr.getExtent().getFullyConverted())
}
}

View File

@@ -22,6 +22,8 @@ class TranslatedStaticStorageDurationVarInit extends TranslatedRootElement,
final override Declaration getFunction() { result = var }
final Location getLocation() { result = var.getLocation() }
override Instruction getFirstInstruction() { result = this.getInstruction(EnterFunctionTag()) }
override TranslatedElement getChild(int n) {

View File

@@ -442,26 +442,29 @@ class TranslatedReturnVoidStmt extends TranslatedReturnStmt {
/**
* The IR translation of an implicit `return` statement generated by the extractor to handle control
* flow that reaches the end of a non-`void`-returning function body. Such control flow
* produces undefined behavior in C++ but not in C. However even in C using the return value is
* undefined behaviour. We make it return uninitialized memory to get as much flow as possible.
* flow that reaches the end of a non-`void`-returning function body. Since such control flow
* produces undefined behavior, we simply generate an `Unreached` instruction to prevent that flow
* from continuing on to pollute other analysis. The assumption is that the developer is certain
* that the implicit `return` is unreachable, even if the compiler cannot prove it.
*/
class TranslatedNoValueReturnStmt extends TranslatedReturnStmt, TranslatedVariableInitialization {
TranslatedNoValueReturnStmt() {
class TranslatedUnreachableReturnStmt extends TranslatedReturnStmt {
TranslatedUnreachableReturnStmt() {
not stmt.hasExpr() and hasReturnValue(stmt.getEnclosingFunction())
}
final override Instruction getInitializationSuccessor() {
result = this.getEnclosingFunction().getReturnSuccessorInstruction()
override TranslatedElement getChild(int id) { none() }
override Instruction getFirstInstruction() { result = this.getInstruction(OnlyInstructionTag()) }
override predicate hasInstruction(Opcode opcode, InstructionTag tag, CppType resultType) {
tag = OnlyInstructionTag() and
opcode instanceof Opcode::Unreached and
resultType = getVoidType()
}
final override Type getTargetType() { result = this.getEnclosingFunction().getReturnType() }
override Instruction getInstructionSuccessor(InstructionTag tag, EdgeKind kind) { none() }
final override TranslatedInitialization getInitialization() { none() }
final override IRVariable getIRVariable() {
result = this.getEnclosingFunction().getReturnVariable()
}
override Instruction getChildSuccessor(TranslatedElement child) { none() }
}
/**

View File

@@ -10,65 +10,6 @@ predicate isInfeasibleInstructionSuccessor(Instruction instr, EdgeKind kind) {
or
instr.getSuccessor(kind) instanceof UnreachedInstruction and
kind instanceof GotoEdge
or
isCallToNonReturningFunction(instr) and exists(instr.getSuccessor(kind))
}
/**
* Holds if all calls to `f` never return (e.g. they call `exit` or loop forever)
*/
private predicate isNonReturningFunction(IRFunction f) {
// If the function has an instruction with a missing successor then
// the analysis is probably going to be incorrect, so assume they exit.
not hasInstructionWithMissingSuccessor(f) and
(
// If all flows to the exit block are pass through an unreachable then f never returns.
any(UnreachedInstruction instr).getBlock().postDominates(f.getEntryBlock())
or
// If there is no flow to the exit block then f never returns.
not exists(IRBlock entry, IRBlock exit |
exit = f.getExitFunctionInstruction().getBlock() and
entry = f.getEntryBlock() and
exit = entry.getASuccessor*()
)
or
// If all flows to the exit block are pass through a call that never returns then f never returns.
exists(CallInstruction ci |
ci.getBlock().postDominates(f.getEntryBlock()) and
isCallToNonReturningFunction(ci)
)
)
}
/**
* Holds if `f` has an instruction with a missing successor.
* This matches `instructionWithoutSuccessor` from `IRConsistency`, but
* avoids generating the error strings.
*/
predicate hasInstructionWithMissingSuccessor(IRFunction f) {
exists(Instruction missingSucc |
missingSucc.getEnclosingIRFunction() = f and
not exists(missingSucc.getASuccessor()) and
not missingSucc instanceof ExitFunctionInstruction and
// Phi instructions aren't linked into the instruction-level flow graph.
not missingSucc instanceof PhiInstruction and
not missingSucc instanceof UnreachedInstruction
)
}
/**
* Holds if the call `ci` never returns.
*/
private predicate isCallToNonReturningFunction(CallInstruction ci) {
exists(IRFunction callee, Language::Function staticTarget |
staticTarget = ci.getStaticCallTarget() and
staticTarget = callee.getFunction() and
// We can't easily tell if the call is virtual or not
// if the callee is virtual. So assume that the call is virtual
// if the target is.
not staticTarget.isVirtual() and
isNonReturningFunction(callee)
)
}
pragma[noinline]

View File

@@ -1,3 +1,2 @@
import semmle.code.cpp.ir.implementation.raw.IR as IR
import semmle.code.cpp.ir.implementation.raw.constant.ConstantAnalysis as ConstantAnalysis
import semmle.code.cpp.ir.internal.IRCppLanguage as Language

View File

@@ -10,65 +10,6 @@ predicate isInfeasibleInstructionSuccessor(Instruction instr, EdgeKind kind) {
or
instr.getSuccessor(kind) instanceof UnreachedInstruction and
kind instanceof GotoEdge
or
isCallToNonReturningFunction(instr) and exists(instr.getSuccessor(kind))
}
/**
* Holds if all calls to `f` never return (e.g. they call `exit` or loop forever)
*/
private predicate isNonReturningFunction(IRFunction f) {
// If the function has an instruction with a missing successor then
// the analysis is probably going to be incorrect, so assume they exit.
not hasInstructionWithMissingSuccessor(f) and
(
// If all flows to the exit block are pass through an unreachable then f never returns.
any(UnreachedInstruction instr).getBlock().postDominates(f.getEntryBlock())
or
// If there is no flow to the exit block then f never returns.
not exists(IRBlock entry, IRBlock exit |
exit = f.getExitFunctionInstruction().getBlock() and
entry = f.getEntryBlock() and
exit = entry.getASuccessor*()
)
or
// If all flows to the exit block are pass through a call that never returns then f never returns.
exists(CallInstruction ci |
ci.getBlock().postDominates(f.getEntryBlock()) and
isCallToNonReturningFunction(ci)
)
)
}
/**
* Holds if `f` has an instruction with a missing successor.
* This matches `instructionWithoutSuccessor` from `IRConsistency`, but
* avoids generating the error strings.
*/
predicate hasInstructionWithMissingSuccessor(IRFunction f) {
exists(Instruction missingSucc |
missingSucc.getEnclosingIRFunction() = f and
not exists(missingSucc.getASuccessor()) and
not missingSucc instanceof ExitFunctionInstruction and
// Phi instructions aren't linked into the instruction-level flow graph.
not missingSucc instanceof PhiInstruction and
not missingSucc instanceof UnreachedInstruction
)
}
/**
* Holds if the call `ci` never returns.
*/
private predicate isCallToNonReturningFunction(CallInstruction ci) {
exists(IRFunction callee, Language::Function staticTarget |
staticTarget = ci.getStaticCallTarget() and
staticTarget = callee.getFunction() and
// We can't easily tell if the call is virtual or not
// if the callee is virtual. So assume that the call is virtual
// if the target is.
not staticTarget.isVirtual() and
isNonReturningFunction(callee)
)
}
pragma[noinline]

View File

@@ -1,3 +1,2 @@
import semmle.code.cpp.ir.implementation.unaliased_ssa.IR as IR
import semmle.code.cpp.ir.implementation.unaliased_ssa.constant.ConstantAnalysis as ConstantAnalysis
import semmle.code.cpp.ir.internal.IRCppLanguage as Language

View File

@@ -917,46 +917,25 @@ module RangeStage<
bounded(cast.getOperand(), b, delta, upper, fromBackEdge, origdelta, reason)
}
pragma[nomagic]
private predicate initialBoundedUpper(SemExpr e) {
exists(D::Delta d |
initialBounded(e, _, d, false, _, _, _) and
D::toFloat(d) >= 0
)
}
private predicate noOverflow0(SemExpr e, boolean upper) {
exists(boolean lower | lower = upper.booleanNot() |
semExprDoesNotOverflow(lower, e)
or
upper = [true, false] and
not potentiallyOverflowingExpr(lower, e)
)
}
pragma[nomagic]
private predicate initialBoundedLower(SemExpr e) {
exists(D::Delta d |
initialBounded(e, _, d, true, _, _, _) and
D::toFloat(d) <= 0
)
}
pragma[nomagic]
private predicate noOverflow(SemExpr e, boolean upper) {
noOverflow0(e, upper)
or
upper = true and initialBoundedUpper(e)
or
upper = false and initialBoundedLower(e)
}
predicate bounded(
SemExpr e, SemBound b, D::Delta delta, boolean upper, boolean fromBackEdge, D::Delta origdelta,
SemReason reason
) {
initialBounded(e, b, delta, upper, fromBackEdge, origdelta, reason) and
noOverflow(e, upper)
(
semExprDoesNotOverflow(upper.booleanNot(), e)
or
not potentiallyOverflowingExpr(upper.booleanNot(), e)
or
exists(D::Delta otherDelta |
initialBounded(e, _, otherDelta, upper.booleanNot(), _, _, _) and
(
upper = true and D::toFloat(otherDelta) >= 0
or
upper = false and D::toFloat(otherDelta) <= 0
)
)
)
}
predicate potentiallyOverflowingExpr(boolean positively, SemExpr expr) {

View File

@@ -72,7 +72,7 @@ predicate hasSize(HeuristicAllocationExpr alloc, DataFlow::Node n, int state) {
// Compute `delta` as the constant difference between `x` and `x + 1`.
bounded1(any(Instruction instr | instr.getUnconvertedResultExpression() = size),
any(LoadInstruction load | load.getUnconvertedResultExpression() = va), delta) and
n.asExpr() = va and
n.asConvertedExpr() = va.getFullyConverted() and
state = delta
)
}
@@ -210,7 +210,7 @@ private module InterestingPointerAddInstruction {
predicate isSource(DataFlow::Node source) {
// The sources is the same as in the sources for the second
// projection in the `AllocToInvalidPointerConfig` module.
hasSize(source.asExpr(), _, _)
hasSize(source.asConvertedExpr(), _, _)
}
int fieldFlowBranchLimit() { result = allocationToInvalidPointerFieldFlowBranchLimit() }
@@ -243,7 +243,7 @@ private module InterestingPointerAddInstruction {
*/
predicate isInterestingSize(DataFlow::Node n) {
exists(DataFlow::Node alloc |
hasSize(alloc.asExpr(), n, _) and
hasSize(alloc.asConvertedExpr(), n, _) and
flow(alloc, _)
)
}
@@ -268,7 +268,7 @@ private module Config implements ProductFlow::StateConfigSig {
// we use `state2` to remember that there was an offset (in this case an offset of `1`) added
// to the size of the allocation. This state is then checked in `isSinkPair`.
exists(unit) and
hasSize(allocSource.asExpr(), sizeSource, sizeAddend)
hasSize(allocSource.asConvertedExpr(), sizeSource, sizeAddend)
}
int fieldFlowBranchLimit1() { result = allocationToInvalidPointerFieldFlowBranchLimit() }

View File

@@ -1755,7 +1755,6 @@ case @expr.kind of
| @istriviallydestructibleexpr
| @istriviallyassignableexpr
| @isnothrowassignableexpr
| @istrivialexpr
| @isstandardlayoutexpr
| @istriviallycopyableexpr
| @isliteraltypeexpr

View File

@@ -1,2 +0,0 @@
description: Make __is_trivial a builtin operation
compatibility: full

View File

@@ -1,19 +1,3 @@
## 0.7.5
No user-facing changes.
## 0.7.4
### New Queries
* Added a new query, `cpp/invalid-pointer-deref`, to detect out-of-bounds pointer reads and writes.
### Minor Analysis Improvements
* The "Comparison where assignment was intended" query (`cpp/compare-where-assign-meant`) no longer reports comparisons that appear in macro expansions.
* Some queries that had repeated results corresponding to different levels of indirection for `argv` now only have a single result.
* The `cpp/non-constant-format` query no longer considers an assignment on the right-hand side of another assignment to be a source of non-constant format strings. As a result, the query may now produce fewer results.
## 0.7.3
No user-facing changes.

View File

@@ -2,7 +2,7 @@
* @name Potential double free
* @description Freeing a resource more than once can lead to undefined behavior and cause memory corruption.
* @kind path-problem
* @precision high
* @precision medium
* @id cpp/double-free
* @problem.severity warning
* @security-severity 9.3

View File

@@ -98,11 +98,8 @@ module FlowFromFree<isSinkSig/2 isASink, isExcludedSig/2 isExcluded> {
* is being freed by a deallocation expression `dealloc`.
*/
predicate isFree(DataFlow::Node n, Expr e, DeallocationExpr dealloc) {
exists(Expr conv |
e = conv.getUnconverted() and
conv = dealloc.getFreedExpr().getFullyConverted() and
conv = n.asConvertedExpr()
) and
e = dealloc.getFreedExpr() and
e = n.asExpr() and
// Ignore realloc functions
not exists(dealloc.(FunctionCall).getTarget().(AllocationFunction).getReallocPtrArg())
}

Some files were not shown because too many files have changed in this diff Show More