redhat-rpm-config/buildflags.md
2018-01-22 14:28:15 +01:00

11 KiB

This document contains documentation of the individual compiler flags and how to use them.

[TOC]

Using RPM build flags

At present, the only supported way is to use the %configure macro to obtain the full complement of flags, like this:

%configure

This will invoke the ./configure with arguments (such as --prefix=/usr) to adjust the paths to the packaging defaults.

As a side effect, this will set the environment variables CFLAGS, CXXFLAGS, FFLAGS, FCFLAGS, and LDFLAGS, so they can be used by makefiles and other build tools.

For some other build tools besides autoconf, separate mechanisms exist:

  • CMake builds use the the %cmake macro from the cmake-rpm-macros package.

Care must be taking not to compile the current selection of compiler flags into any RPM package besides redhat-rpm-config, so that flag changes are picked up automatically once redhat-rpm-config is updated.

Flag selection for the build type

The default flags are suitable for building applications.

For building shared objects, you must compile with -fPIC in (CFLAGS or CXXFLAGS) and link with -shared (in LDFLAGS).

For other considerations involving shared objects, see:

Customizing compiler flags

It is possible to set RPM macros to change some aspects of the compiler flags. Changing these flags should be used as a last recourse if other workarunds are not available.

Disable hardened builds

%undefine _hardened_build

This turns off certain hardening features, as described in detail below. The main difference is that executables will be position-dependent (no full ASLR) and use lazy binding.

Disable annotated builds/watermarking

%undefine _annotated_build

This turns off watermarking, making it impossible to do full hardening coverage analysis for any binaries produced.

Individual compiler flags

Compiler flags end up in the environment variables CFLAGS, CXXFLAGS, FFLAGS, and FCFLAGS.

The general (architecture-independent) build flags are:

  • -O2: Turn on various GCC optimizations. See the GCC manual. Optimization improves performance, the accuracy of warnings, and the reach of toolchain-based hardening, but it makes debugging harder.
  • -g: Generate debugging information (DWARF). In Fedora, this data is separated into -debuginfo RPM packages whose installation is optional, so debuging information does not increase the size of installed binaries by default.
  • -pipe: Run compiler and assembler in parallel and do not use a temporary file for the assembler input. This can improve compilation performance. (This does not affect code generation.)
  • -Wall: Turn on various GCC warnings. See the GCC manual.
  • -Werror=format-security: Turn on format string warnings and treat them as errors. See the GCC manual. This can occasionally result in compilation errors. In this case, the best option is to rewrite the source code so that only constant format strings (string literals) are used.
  • -Wp,-D_FORTIFY_SOURCE=2: Source fortification activates various hardening features in glibc:
    • String functions such as memcpy attempt to detect buffer lengths and terminate the process if a buffer overflow is detected.
    • printf format strings may only contain the %n format specifier if the format string resides in read-only memory.
    • open and openat flags are checked for consistency with the presence of a mode argument.
    • Plus other minor hardening changes. (These changes can occasionally break valid programs.)
  • -fexceptions: Provide exception unwinding support for C programs. See the -fexceptions option in the GCC manual and the cleanup variable attribute. This also hardens cancellation handling in C programs because it is not required to use an on-stack jump buffer to install a cancellation handler with pthread_cleanup_push. It also makes it possible to unwind the stack (using C++ throw or Rust panics) from C callback functions if a C library supports non-local exits from them (e.g., via longjmp).
  • -Wp,-D_GLIBCXX_ASSERTIONS: Enable lightweight assertions in the C++ standard library, such as bounds checking for the subscription operator on vectors. (This flag is added to both CFLAGS and CXXFLAGS; C compilations will simply ignore it.)
  • -fstack-protector-strong: Instrument functions to detect stack-based buffer overflows before jumping to the return address on the stack. The strong variant only performs the instrumentation for functions whose stack frame contains addressable local variables. (If the address of a variable is never taken, it is not possible that a buffer overflow is caused by incorrect pointer arithmetic involving a pointer to that variable.)
  • -grecord-gcc-switches: Include select GCC command line switches in the DWARF debugging information. This is useful for detecting the presence of certain build flags and general hardening coverage.

For hardened builds (which are enabled by default, see above for how to disable them), the flag -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 is added to the command line. It adds the following flag to the command line:

  • -fPIE: Compile for a position-independent executable (PIE), enabling full address space layout randomization (ASLR). This is similar to -fPIC, but avoids run-time indirections on certain architectures, resulting in improved performance and slightly smaller executables. However, compared to position-dependent code (the default generated by GCC), there is still a measurable performance impact.

    If the command line also contains -r (producing a relocatable object file), -fpic or -fPIC, this flag is automatically dropped. (-fPIE can only be used for code which is linked into the main program.) Code which goes into static libraries should be compiled with -fPIE, except when this code is expected to be linked into DSOs, when -fPIC must be used.

    To be effective, -fPIE must be used with the -pie linker flag when producing an executable, see below.

To support binary watermarks for ELF objects using annobin, the -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 flag is added by default. This can be switched off by undefining the %_annotated_build RPM macro (see above).

Architecture-specific compiler flags

These compiler flags are enabled for all builds (hardened/annotated or not), but their selection depends on the architecture:

  • -fstack-clash-protection: Turn on instrumentation to avoid skipping the guard page in large stack frames. (Without this flag, vulnerabilities can result where the stack overlaps with the heap, or thread stacks spill into other regions of memory.) This flag is fully ABI-compatible and has adds very little run-time overhead, but is only on certain architectures (currently aarch64, i386, ppc64, ppc64le, s390x, x86-64).
  • -m64 and -m32: Some GCC builds support both 32-bit and 64-bit in the same compilation. For such architectures, the RPM build process explicitly selects the architecture variant by passing this compiler flag.
  • -fasynchronous-unwind-tables: Generate full unwind information covering all program points. This is required for support of asynchronous cancellation and proper unwinding from signal handlers. It also makes performance and debugging tools more useful because unwind information is available without having to install (and load) debugging information. This flag is enabled explictly for i686 (and derived architectures), and implicity for x86-64.

In addition, redhat-rpm-config re-selects the built-in default tuning in the gcc package. These settings are:

  • armhfp: -march=armv7-a -mfpu=vfpv3-d16 -mfloat-abi=hard selects an Arm subarchitecture based on the ARMv7-A architecture with 16 64-bit floating point registers.
  • i686: -march=i686 is used to select a minmum support CPU level of i686 (corresponding to the Pentium Pro).
  • ppc64le: -mcpu=power8 -mtune=power8 selects a minimum supported CPU level of POWER8 (the first CPU with ppc64le support) and tunes for POWER8.
  • s390x: -march=zEC12 -mtune=z13 specifies a minimum supported CPU level of zEC12, while optimizing for a subsequent CPU generation (z13).
  • x86-64: -mtune=generic selects tuning which is expected to beneficial for a broad range of current CPUs.
  • ppc64 and aarch64 do not have any architecture-specific tuning.

Individual linker flags

Linker flags end up in the environment variable LDFLAGS.

The linker flags listed below are injected. Note that they are prefixed with -Wl because it is expected that these flags are passed to the compiler driver gcc, and not directly to the link editor ld.

  • -z relro: Activate the read-only after relocation feature. Constant data and relocations are placed on separate pages, and the dynamic linker is instructed to revoke write permissions after dynamic linking. Full protection of relocation data requires the -z now flag (see below).

For hardened builds, the -specs=/usr/lib/rpm/redhat/redhat-hardened-ld flag is added to the compiler driver command line. (This can be disabled by undefining the %_hardened_build macro; see above) This activates the following linker flags:

  • -pie: Produce a PIE binary. This is only activated for the main executable, and only if it is dynamically linked. This requires that all objects which are linked in the main executable have been compiled with -fPIE or -fPIC (or -fpie or -fpic; see above). By itself, -pie has only a slight performance impact because it disables some link editor optimization, however the -fPIE compiler flag has some overhead.
  • -z now: Disable lazy binding and turn on the BIND_NOW dynamic linker feature. Lazy binding involves an array of function pointers which is writable at run time (which could be overwritten as part of security exploits, redirecting execution). Therefore, it is preferable to turn of lazy binding, although it increases startup time.