-Wno-error=tautological-constant-compare again (this flag is now out of
-Wextra after upstream https://reviews.llvm.org/rL322901). Otherwise
the MK_SYSTEM_COMPILER logic will not build a cross-tools compiler.
Reported by: jpaetzel, tuexen, Stefan Hagen
6.0.0 (branches/release_60 r324090).
This introduces retpoline support, with the -mretpoline flag. The
upstream initial commit message (r323155 by Chandler Carruth) contains
quite a bit of explanation. Quoting:
Introduce the "retpoline" x86 mitigation technique for variant #2 of
the speculative execution vulnerabilities disclosed today,
specifically identified by CVE-2017-5715, "Branch Target Injection",
and is one of the two halves to Spectre.
Summary:
First, we need to explain the core of the vulnerability. Note that
this is a very incomplete description, please see the Project Zero
blog post for details:
https://googleprojectzero.blogspot.com/2018/01/reading-privileged-memory-with-side.html
The basis for branch target injection is to direct speculative
execution of the processor to some "gadget" of executable code by
poisoning the prediction of indirect branches with the address of
that gadget. The gadget in turn contains an operation that provides a
side channel for reading data. Most commonly, this will look like a
load of secret data followed by a branch on the loaded value and then
a load of some predictable cache line. The attacker then uses timing
of the processors cache to determine which direction the branch took
*in the speculative execution*, and in turn what one bit of the
loaded value was. Due to the nature of these timing side channels and
the branch predictor on Intel processors, this allows an attacker to
leak data only accessible to a privileged domain (like the kernel)
back into an unprivileged domain.
The goal is simple: avoid generating code which contains an indirect
branch that could have its prediction poisoned by an attacker. In
many cases, the compiler can simply use directed conditional branches
and a small search tree. LLVM already has support for lowering
switches in this way and the first step of this patch is to disable
jump-table lowering of switches and introduce a pass to rewrite
explicit indirectbr sequences into a switch over integers.
However, there is no fully general alternative to indirect calls. We
introduce a new construct we call a "retpoline" to implement indirect
calls in a non-speculatable way. It can be thought of loosely as a
trampoline for indirect calls which uses the RET instruction on x86.
Further, we arrange for a specific call->ret sequence which ensures
the processor predicts the return to go to a controlled, known
location. The retpoline then "smashes" the return address pushed onto
the stack by the call with the desired target of the original
indirect call. The result is a predicted return to the next
instruction after a call (which can be used to trap speculative
execution within an infinite loop) and an actual indirect branch to
an arbitrary address.
On 64-bit x86 ABIs, this is especially easily done in the compiler by
using a guaranteed scratch register to pass the target into this
device. For 32-bit ABIs there isn't a guaranteed scratch register
and so several different retpoline variants are introduced to use a
scratch register if one is available in the calling convention and to
otherwise use direct stack push/pop sequences to pass the target
address.
This "retpoline" mitigation is fully described in the following blog
post: https://support.google.com/faqs/answer/7625886
We also support a target feature that disables emission of the
retpoline thunk by the compiler to allow for custom thunks if users
want them. These are particularly useful in environments like
kernels that routinely do hot-patching on boot and want to hot-patch
their thunk to different code sequences. They can write this custom
thunk and use `-mretpoline-external-thunk` *in addition* to
`-mretpoline`. In this case, on x86-64 thu thunk names must be:
```
__llvm_external_retpoline_r11
```
or on 32-bit:
```
__llvm_external_retpoline_eax
__llvm_external_retpoline_ecx
__llvm_external_retpoline_edx
__llvm_external_retpoline_push
```
And the target of the retpoline is passed in the named register, or in
the case of the `push` suffix on the top of the stack via a `pushl`
instruction.
There is one other important source of indirect branches in x86 ELF
binaries: the PLT. These patches also include support for LLD to
generate PLT entries that perform a retpoline-style indirection.
The only other indirect branches remaining that we are aware of are
from precompiled runtimes (such as crt0.o and similar). The ones we
have found are not really attackable, and so we have not focused on
them here, but eventually these runtimes should also be replicated for
retpoline-ed configurations for completeness.
For kernels or other freestanding or fully static executables, the
compiler switch `-mretpoline` is sufficient to fully mitigate this
particular attack. For dynamic executables, you must compile *all*
libraries with `-mretpoline` and additionally link the dynamic
executable and all shared libraries with LLD and pass `-z
retpolineplt` (or use similar functionality from some other linker).
We strongly recommend also using `-z now` as non-lazy binding allows
the retpoline-mitigated PLT to be substantially smaller.
When manually apply similar transformations to `-mretpoline` to the
Linux kernel we observed very small performance hits to applications
running typic al workloads, and relatively minor hits (approximately
2%) even for extremely syscall-heavy applications. This is largely
due to the small number of indirect branches that occur in
performance sensitive paths of the kernel.
When using these patches on statically linked applications,
especially C++ applications, you should expect to see a much more
dramatic performance hit. For microbenchmarks that are switch,
indirect-, or virtual-call heavy we have seen overheads ranging from
10% to 50%.
However, real-world workloads exhibit substantially lower performance
impact. Notably, techniques such as PGO and ThinLTO dramatically
reduce the impact of hot indirect calls (by speculatively promoting
them to direct calls) and allow optimized search trees to be used to
lower switches. If you need to deploy these techniques in C++
applications, we *strongly* recommend that you ensure all hot call
targets are statically linked (avoiding PLT indirection) and use both
PGO and ThinLTO. Well tuned servers using all of these techniques saw
5% - 10% overhead from the use of retpoline.
We will add detailed documentation covering these components in
subsequent patches, but wanted to make the core functionality
available as soon as possible. Happy for more code review, but we'd
really like to get these patches landed and backported ASAP for
obvious reasons. We're planning to backport this to both 6.0 and 5.0
release streams and get a 5.0 release with just this cherry picked
ASAP for distros and vendors.
This patch is the work of a number of people over the past month:
Eric, Reid, Rui, and myself. I'm mailing it out as a single commit
due to the time sensitive nature of landing this and the need to
backport it. Huge thanks to everyone who helped out here, and
everyone at Intel who helped out in discussions about how to craft
this. Also, credit goes to Paul Turner (at Google, but not an LLVM
contributor) for much of the underlying retpoline design.
Reviewers: echristo, rnk, ruiu, craig.topper, DavidKreitzer
Subscribers: sanjoy, emaste, mcrosier, mgorny, mehdi_amini, hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D41723
MFC after: 3 months
X-MFC-With: r327952
PR: 224669
Allow usage of X86-prefixes as separate instrs.
Differential Revision: https://reviews.llvm.org/D42102
This should fix parse errors when x86 prefixes (such as 'lock' and
'rep') are followed by various non-mnemonic tokens, e.g. comments, .byte
directives and labels.
PR: 224669,225054
This allows the _SKIP_DEPEND optimization to work, avoiding reading
the files when not needed. It also fixes META_MODE incorrectly
reading these files when not needed.
Sponsored by: Dell EMC Isilon
The Makefile.inc1 TARGET_TRIPLE is for specifying which -target is used
during the build of world.
MFC after: 2 weeks
Reviewed by: dim, imp
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D12792
Make armv7 as a new MACHINE_ARCH.
Copy all the places we do armv6 and add armv7 as basically an
alias. clang appears to generate code for armv7 by default. armv7 hard
float isn't supported by the the in-tree gcc, so it hasn't been
updated to have a new default.
Support armv7 as a new valid MACHINE_ARCH (and by extension
TARGET_ARCH).
Add armv7 to the universe build.
Differential Revision: https://reviews.freebsd.org/D12010
the upstream release_50 branch. This corresponds to 5.0.0 rc4.
As of this version, the cad/stepcode port should now compile in a more
reasonable time on i386 (see bug 221836 for more information).
PR: 221836
MFC after: 2 months
X-MFC-with: r321369
the upstream release_50 branch.
As of this version, lib/msun's trig test should also work correctly
again (see bug 220989 for more information).
PR: 220989
MFC after: 2 months
X-MFC-with: r321369
- Include debug symbols in static libraries. This permits binaries
to include debug symbols for functions obtained from static libraries.
- Permit the C/C++ compiler flags added for MK_DEBUG_FILES to be
overridden by setting DEBUG_FILES_CFLAGS. Use this to limit the debug
information for llvm libraries and binaries.
Reviewed by: emaste
Sponsored by: DARPA / AFRL
Differential Revision: https://reviews.freebsd.org/D12025
The files are only ever generated to .OBJDIR, not to WORLDTMP (as a
sysroot) and are only ever included from a compilation. So using
a beforebuild target here removes the file before the compilation
tries to include it.
MFC after: 2 months
X-MFC-With: r321369
linking the lldb executable in some cases. In particular, when the
-ffunction-sections -fdata-sections options are turned off, or
ineffective.
Reported by: Shawn Webb, Mark Millard
MFC after: 2 months
X-MFC-With: r308421
For some reason, we have been inserting the ABI specification into the
middle of the target triple, when building LLVM, like so:
armv6-gnueabi-freebsd12.0
This is the wrong way around. LLVM even auto-canonicalizes it to:
armv6--freebsd12.0-gnueabi
Let's do this the right way in llvm.build.mk instead. While here,
define a proper VENDOR macro which can be overridden easily.
Reviewed by: emaste
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D10846
properly anyway. (Upstream has reorganized this somewhat in the mean
time, but for proper backtraces we would need llvm-symbolizer in base.)
MFC after: 3 days
-mlong-calls was set only in STATIC_CXXFLAGS, but there are some .c
source files in LLVM which also need -mlong-calls.
Unfortunately this is not sufficient to fix linking lldb on ARM,
because LLVM-generated calls to __aeabi_read_tp do not honour the
-mlong-calls flag. See LLVM PR31769 for details.
Reviewed by: dim
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D9348
Previously WITH_LLD_AS_LD installed LLD as /usr/bin/ld in the target
system, but still used the GNU BFD ld to link the binaries in that
target. LLD 4.0.0 can link the FreeBSD/amd64 world and kernel so use
LLD as the build-time linker as well when the knob is set.
Reviewed by: dim
Relnotes: Yes
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D9226
Move llvm-objdump from CLANG_EXTRAS to installed by default
We currently install three tools from binutils 2.17.50: as, ld, and
objdump. Work is underway to migrate to a permissively-licensed
tool-chain, with one goal being the retirement of binutils 2.17.50.
LLVM's llvm-objdump is intended to be compatible with GNU objdump
although it is currently missing some options and may have formatting
differences. Enable it by default for testing and further investigation.
It may later be changed to install as /usr/bin/objdump, it becomes a
fully viable replacement.
Reviewed by: emaste
Differential Revision: https://reviews.freebsd.org/D8879
We currently install three tools from binutils 2.17.50: as, ld, and
objdump. Work is underway to migrate to a permissively-licensed
tool-chain, with one goal being the retirement of binutils 2.17.50.
LLVM's llvm-objdump is intended to be compatible with GNU objdump
although it is currently missing some options and may have formatting
differences. Enable it by default for testing and further investigation.
It may later be changed to install as /usr/bin/objdump, it becomes a
fully viable replacement.
Reviewed by: emaste
Differential Revision: https://reviews.freebsd.org/D8879
PowerPC, add lib/Support/Atomic.cpp. This is needed because upstream
llvm revision r271821 disabled the use of std::call_once, which causes
some fallback functions from Atomic.cpp to be used instead.
Reported by: Mark Millard
PR: 214902
X-MFC-With: 309124
Use this to control inclusion of the libllvm functionality required
by lld. Enable by default on arm64 and amd64, the two platforms where
lld is most usable for testing.
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D7713
We might want to generate this file at build time, but for now just
provide a static copy as done with other version files.
Sponsored by: The FreeBSD Foundation
* Bootstrap llvm-tblgen and clang-tblgen with a minimal llvm static
library, that has no other dependencies.
* Roll up all separate llvm libraries into one big static libllvm.
* Similar for all separate clang and lldb static libraries.
* For all these libraries, generate their .inc files only once.
* Link all llvm tools (including extra) against the big libllvm.
* Link clang and clang-format against the big libllvm and libclang.
* Link lldb against the big libllvm, libclang and liblldb.
N.B.: This is work in progress, some details may still be missing.
It also heavily depends on bsd.*.mk's support for SRCS and DPSRCS with
relative pathnames, which apparently does not always work as expected.
For building llvm, clang and lldb though, it seems to work just fine.
The main idea behind this restructuring is maintainability and build
peformance. The previous large number of very small libraries, each
with their own generated files and dependencies was slow to traverse
and hard to understand.
Possible future improvements:
* Only build certain targets, e.g. for most regular users having just
one target will be fine. This will shave off some build time.
* Building the big llvm, clang and lldb libraries as shared (private)
libraries.
* Adding other components from the LLVM project, such as lld.
Fix for pr24346: arm asm label calculation error in sub
Some ARM instructions encode 32-bit immediates as a 8-bit integer
(0-255) and a 4-bit rotation (0-30, even) in its least significant 12
bits. The original fixup, FK_Data_4, patches the instruction by the
value bit-to-bit, regardless of the encoding. For example, assuming
the label L1 and L2 are 0x0 and 0x104 respectively, the following
instruction:
add r0, r0, #(L2 - L1) ; expects 0x104, i.e., 260
would be assembled to the following, which adds 1 to r0, instead of
260:
e2800104 add r0, r0, #4, 2 ; equivalently 1
The new fixup kind fixup_arm_mod_imm takes care of the encoding:
e2800f41 add r0, r0, #260
Patch by Ting-Yuan Huang!
This fixes label calculation for ARM assembly, and is needed to enable
ARM assembly sources for OpenSSL.
Requested by: jkim
MFC after: 3 days
[X86] AMD Bobcat CPU (btver1) doesn't support XSAVE
btver1 is a SSSE3/SSE4a only CPU - it doesn't have AVX and doesn't
support XSAVE.
Differential Revision: http://reviews.llvm.org/D17682
Pull in r262782 from upstream llvm trunk (by Simon Pilgrim):
[X86] AMD Bobcat CPU (btver1) doesn't support XSAVE
btver1 is a SSSE3/SSE4a only CPU - it doesn't have AVX and doesn't
support XSAVE.
Differential Revision: http://reviews.llvm.org/D17683
This ensures clang does not emit AVX instructions for CPUTYPE=btver1.
Reported by: Michel Depeige <demik+freebsd@lostwave.net>
PR: 211864
MFC after: 3 days
file, lib/clang/freebsd_cc_version.h, instead of reusing Version.inc.
The header is only included from one .cpp file in the clang tree.
This minimizes the number of .cpp files that need to be rebuilt if the
version is bumped.
Discussed with: bdrewery
Only attempt to detect AVG if SSE2 is available
Summary:
In PR29973 Sanjay Patel reported an assertion failure when a certain
loop was optimized, for a target without SSE2 support. It turned out
this was because of the AVG pattern detection introduced in rL253952.
Prevent the assertion failure by bailing out early in
`detectAVGPattern()`, if the target does not support SSE2.
Also add a minimized test case.
Reviewers: congh, eli.friedman, spatel
Subscribers: emaste, llvm-commits
Differential Revision: http://reviews.llvm.org/D20905
This should fix assertion failures ("Requires at least SSE2!") when
building the games/0ad port with CPUTYPE=pentium3.
Reported by: madpilot
The WITH_SYSTEM_COMPILER build option will rely on this value to determine what
__FreeBSD_cc_version the source tree will produce. This value will be compared
against the /usr/bin/cc value to determine if a new compiler is needed.
Start with 1100002 which is 1 more than than the value we've had since
3.8.0 to ensure that all changes since then are present.
Reviewed by: dim
Sponsored by: EMC / Isilon Storage Division
As mentioned in the Makefile there's an "atrocious" hack to generate a
different version of Options.inc.h, depending on the library being
built.
Remove the catch-all else case and limit it to specific libraries, so
that we don't accidentally use the Options.inc.h from clangdriver if a
future libary also uses Options.inc.h.
Reviewed by: dim
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D6209
At least FAST_DEPEND won't even run 'make depend', so the code was
potentially broken with FAST_DEPEND anyhow. The .dinclude directive
will ignore missing files rather than make them be fatal.
Sponsored by: EMC / Isilon Storage Division
These are no longer needed after the recent 'beforebuild: depend' changes
and hooking DIRDEPS_BUILD into a subset of FAST_DEPEND which supports
skipping 'make depend'.
Sponsored by: EMC / Isilon Storage Division
This will generate dependencies rather than depending on the previous behavior
of depending on the guessed OBJS: *.h dependecies or a user running
'make depend'.
Experimentation showed that depending only on headers was not enough and
prone to .ORDER errors. Downstream users may also have added
dependencies into beforedepend or afterdepend targets. The safest way to
ensure dependencies are generated before build is to run 'make depend'
beforehand rather than just depending on DPSRCS+SRCS.
Note that the OBJS_DEPEND_GUESS mechanism (a.k.a .if !exists(.depend) then
foo.o: *.h) is still useful as it improves incremental builds with missing
.depend.* files and allows 'make foo.o' to usually work, while this
'beforebuild: depend' ensures that the build will always find all dependencies.
The 'make foo.o' case has no means of a 'beforebuild' hook.
This also removes several hacks in the DIRDEPS_BUILD:
- NO_INSTALL_INCLUDES is no longer needed as it mostly was to work around
.ORDER problems with building the needed headers early.
- DIRDEPS_BUILD: It is no longer necesarry to track "local dependencies" in
Makefile.depend.
These were only in Makefile.depend for 'clean builds' since nothing would
generate the files due to skipping 'make depend' and early dependency
bugs that have been fixed, such as adding headers into SRCS for the
OBJS_DEPEND_GUESS mechanism. Normally if a .depend file does not exist then
a dependency is added by bsd.lib.mk/bsd.prog.mk from OBJS: *.h. However,
meta.autodep.mk creates a .depend file from created meta files and inserts
that into Makefile.depend. It also only tracks *.[ch] files though which can
miss some dependencies that are hooked into 'make depend'. This .depend
that is created then breaks incremental builds due to the !exists(.depend)
checks for OBJS_DEPEND_GUESS. The goal was to skip 'make depend' yet it only
really works the first time. After that files are not generated as expected,
which r288966 tried to address but was using buildfiles: rather than
beforebuild: and was reverted in r291725. As noted previously,
depending only on headers in beforebuild: would create .ORDER errors
in some cases.
meta.autodep.mk is still used to generate Makefile.depend though via:
gendirdeps: Makefile.depend
.END: gendirdeps
This commit allows removing all of the "local dependencies" in
Makefile.depend which cuts down on churn and removes some of the
arch-dependent Makefile.depend files.
The "local dependencies" were also problematic for bootstrapping.
Sponsored by: EMC / Isilon Storage Division
linking. These are too large for a branch instruction to branch from an
earlier point in the code to somewhere later.
This will also allow these to be build with Thumb-2 when we get this
infrastructure.
Reviewed by: dim
Differential Revision: https://reviews.freebsd.org/D4855
llvm's LinkAllPasses.h. This caused some of the calls not to be
emitted, if the optimization level was -O2 or higher.
Conversely, if you used -O1 or lower, calls to e.g. RunningOnValgrind()
would be emitted, leading to link failures, because we did not include
Valgrind.cpp into libllvmsupport. Therefore, add it unconditionally.
Noticed by: ian
bugfix-only release, with no new features.
Please note that from 3.5.0 onwards, clang and llvm require C++11
support to build; see UPDATING for more information.
This is not properly respecting WITHOUT or ARCH dependencies in target/.
Doing so requires a massive effort to rework targets/ to do so. A
better approach will be to either include the SUBDIR Makefiles directly
and map to DIRDEPS or just dynamically lookup the SUBDIR. These lose
the benefit of having a userland/lib, userland/libexec, etc, though and
results in a massive package. The current implementation of targets/ is
very unmaintainable.
Currently rescue/rescue and sys/modules are still not connected.
Sponsored by: EMC / Isilon Storage Division
- Support more of the toolchain from TOOLSDIR.
- This also improves 'make bootstrap-tools' to pass, for example,
AS=/usr/bin/as to Makefile.inc1, which will tell cross-tools to use
external toolchain support and avoid building things we won't be using
in the build.
- Always set the PATH to contain the staged TOOLSDIR directories when
not building the bootstrap targets.
The previous version was only setting this at MAKE.LEVEL==0 and if the
TOOLSDIR existed. Both of these prevented using staged tools that were
built during the build though as DIRDEPS with .host dependencies, such
as the fix for needing usr.bin/localedef.host in r291311.
This is not a common tool so we must build and use it during the build,
and need to be prepared to change PATH as soon as it appears.
This should also fix the issue of host dependencies disappearing from
Makefile.depend and then reappearing due to the start of the fresh build not
having the directory yet, resulting in the tools that were built not actually
being used.
- Only use LEGACY_TOOLS while building in Makefile.inc1. After r291317
and r291546 there is no need to add LEGACY_TOOLS into the PATH for
the pseudo/targets/toolchain build.
- Because the pseudo/targets/toolchain will now build its own
[clang-]tblgen, the special logic in clang.build.mk is no longer needed.
- LEGACY_TOOLS is no longer used outside of targets/pseudo/bootstrap-tools
so is no longer passed into the environment in its build.
Sponsored by: EMC / Isilon Storage Division
staged.
None of usr.bin/clang/clang-tblgen or its dependencies need
lib/clang/include, so there is no cyclic dependency here to worry about.
The issue came about because of workarounds to dependencies on clang
being optional.
Without this, the clang-tblgen called during the build would change
after it was staged for the host. This would cause lib/clang/include to
rebuild due to changed build commands.
Sponsored by: EMC / Isilon Storage Division
It was not being used outside of META_MODE but this should make it more clear
that it is only for META_MODE.
Sponsored by: EMC / Isilon Storage Division
the target is "make depend". This works around errors during
incremental make depend of some clang libraries, for example "don't know
how to make contrib/llvm/include/llvm/IR/IntrinsicsR600.td".
Reported by: emaste
Notable upstream commits (upstream revision in parens):
- Add a JSON producer to LLDB (228636)
- Don't crash on bad DWARF expression (228729)
- Add support of DWARFv3 DW_OP_form_tls_address (231342)
- Assembly profiler for MIPS64 (232619)
- Handle FreeBSD/arm64 core files (233273)
- Read/Write register for MIPS64 (233685)
- Rework LLDB system initialization (233758)
- SysV ABI for aarch64 (236098)
- MIPS software single stepping (236696)
- FreeBSD/arm live debugging support (237303)
- Assembly profiler for mips32 (237420)
- Parse function name from DWARF DW_AT_abstract_origin (238307)
- Improve LLDB prompt handling (238313)
- Add real time signals support to FreeBSDSignals (238316)
- Fix race in IOHandlerProcessSTDIO (238423)
- MIPS64 Branch instruction emulation for SW single stepping (238820)
- Improve OSType initialization in elf object file's arch_spec (239148)
- Emulation of MIPS64 floating-point branch instructions (239996)
- ABI Plugin for MIPS32 (239997)
- ABI Plugin for MIPS64 (240123)
- MIPS32 branch emulation and single stepping (240373)
- Improve instruction emulation based stack unwinding on ARM (240533)
- Add branch emulation to aarch64 instruction emulator (240769)
Off by default, build behaves normally.
WITH_META_MODE we get auto objdir creation, the ability to
start build from anywhere in the tree.
Still need to add real targets under targets/ to build packages.
Differential Revision: D2796
Reviewed by: brooks imp
