Another synthesized function which may be referenced directly.
Change-Id: Ic75fe66ce7244246a2d4a707b6a5fee24cac6941
Reviewed-on: https://boringssl-review.googlesource.com/15831
Commit-Queue: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
When code wants to push a pointer from the GOT onto the stack, we don't
have any registers to play with. We do, however, know that the stack is
viable and thankfully Intel has an “xchg” instruction that avoids the
need for an intermediate register.
Change-Id: Iba7e4f0f4c9b43b3d994cf6cfc92837b312c7728
Reviewed-on: https://boringssl-review.googlesource.com/15625
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
OPENSSL_ia32cap_addr avoids any relocations within the module, at the
cost of a runtime TEXTREL, which causes problems in some cases.
(Notably, if someone links us into a binary which uses the GCC "ifunc"
attribute, the loader crashes.)
We add a OPENSSL_ia32cap_addr_delta symbol (which is reachable
relocation-free from the module) stores the difference between
OPENSSL_ia32cap_P and its own address. Next, reference
OPENSSL_ia32cap_P in code as usual, but always doing LEAQ (or the
equivalent GOTPCREL MOVQ) into a register first. This pattern we can
then transform into a LEAQ and ADDQ on OPENSSL_ia32cap_addr_delta.
ADDQ modifies the FLAGS register, so this is only a safe transformation
if we safe and restore flags first. That, in turn, is only a safe
transformation if code always uses %rsp as a stack pointer (specifically
everything below the stack must be fair game for scribbling over). Linux
delivers signals on %rsp, so this should already be an ABI requirement.
Further, we must clear the red zone (using LEAQ to avoid touching FLAGS)
which signal handlers may not scribble over.
This also fixes the GOTTPOFF logic to clear the red zone.
Change-Id: I4ca6133ab936d5a13d5c8ef265a12ab6bd0073c9
Reviewed-on: https://boringssl-review.googlesource.com/15545
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Past the first word, the remaining arguments are usually separated by
commas. This avoids some of the awkward fixing up needed to extract
target registers, etc.
Change-Id: Id99b99e5160abf80e60afea96f2b46b53b55c9c5
Reviewed-on: https://boringssl-review.googlesource.com/15544
Reviewed-by: Adam Langley <agl@google.com>
OPENSSL_ia32cap_addr avoids any relocations within the module, at the
cost of a runtime TEXTREL, which causes problems in some cases.
(Notably, if someone links us into a binary which uses the GCC "ifunc"
attribute, the loader crashes.)
Fix C references of OPENSSL_ia32cap_addr with a function. This is
analogous to the BSS getters. A follow-up commit will fix perlasm with a
different scheme which avoids calling into a function (clobbering
registers and complicating unwind directives.)
Change-Id: I09d6cda4cec35b693e16b5387611167da8c7a6de
Reviewed-on: https://boringssl-review.googlesource.com/15525
Reviewed-by: Adam Langley <agl@google.com>
Not requiring the list of assembly sources to be comma-separated is
helpful to environments where the list would more naturally be
treated as a list.
Change-Id: I43b18cdbeed1dc7ad217ff61557ac55860f40733
Reviewed-on: https://boringssl-review.googlesource.com/15585
Reviewed-by: Matt Braithwaite <mab@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Change-Id: I4e34dabe302f7dacdf04a89052ad9fe9254a1b81
Reviewed-on: https://boringssl-review.googlesource.com/15404
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
In some modes the compiler will emit a section for BSS symbols and
construct the values with labels, alignment and data instructions. This
change parses these sections and emits the local versions of each symbol
needed to make this work.
Change-Id: I8d43ffe4b5b734950aa4287a3dd7c0d2f191f2e4
Reviewed-on: https://boringssl-review.googlesource.com/15206
Reviewed-by: Adam Langley <agl@google.com>
We might want to back off on this in the future so that we don't upset
future compiler work but, for now, it's useful to know when we hit
something that we don't understand.
Change-Id: I763830b0ddcf5da20061fad673265d4a5855479c
Reviewed-on: https://boringssl-review.googlesource.com/15205
Reviewed-by: Adam Langley <agl@google.com>
In order to better handle BSS sections, rather than having a single loop
over the lines and state flags, pull lines as needed. This means that
subfunctions can process sections of the input.
Also, stop bothering to move the init_array to the end, it's already put
into its own section.
Change-Id: I0e62930c65d29baecb39ba0d8bbc21f2da3bde56
Reviewed-on: https://boringssl-review.googlesource.com/15204
Reviewed-by: Adam Langley <agl@google.com>
With some optimisation settings, Clang was loading
BORINGSSL_bcm_text_hash with AVX2 instructions, which weren't getting
translated correctly. This seems to work and is less fragile.
The compiler just emits an leaq here. This is because it knows the
symbol is hidden (in the shared library sense), so it needn't go through
GOTPCREL. The assembler would have added a relocation, were the symbol
left undefined, but since we define the symbol later on, it all works
out without a relocation.
Were the symbol not hidden, the compiler would have emitted a movq by
way of GOTPCREL, but we can now translate those away anyway.
Change-Id: I442a22f4f8afaadaacbab7044f946a963ebfc46c
Reviewed-on: https://boringssl-review.googlesource.com/15384
Reviewed-by: Adam Langley <agl@google.com>
The changes to delocate.go are needed because modes/ does things like
return the address of a module function. Both of these need to be
changed from referencing the GOT to using local symbols.
Rather than testing whether |ghash| is |gcm_ghash_avx|, we can just keep
that information in a flag.
The test for |aesni_ctr32_encrypt_blocks| is more problematic, but I
believe that it's superfluous and can be dropped: if you passed in a
stream function that was semantically different from
|aesni_ctr32_encrypt_blocks| you would already have a bug because
|CRYPTO_gcm128_[en|de]crypt_ctr32| will handle a block at the end
themselves, and assume a big-endian, 32-bit counter anyway.
Change-Id: I68a84ebdab6c6006e11e9467e3362d7585461385
Reviewed-on: https://boringssl-review.googlesource.com/15064
Reviewed-by: Adam Langley <agl@google.com>
If all the inputs are given as assembly files then we can skip rewriting
symbols for the first file. If this file is bcm.s (i.e. the large
compiler output), this can save a few seconds of build time.
Change-Id: I4e4ea114acb86cd93e831b23b58f8c3401bc711c
Reviewed-on: https://boringssl-review.googlesource.com/15149
Reviewed-by: Adam Langley <agl@google.com>
delocate.go was adding redirector functions for the “_bss_get”
functions. (And they were going via the PLT too.)
Change-Id: I86bc9f0516a128a769068182cc280499f89b6c29
Reviewed-on: https://boringssl-review.googlesource.com/15148
Reviewed-by: Adam Langley <agl@google.com>
These relocations can be emitted for thread-local data. BoringSSL itself
doesn't include any thread-local variables that need linker support, but
ASAN and MSAN may inject these references in order to handle their own
bookkeeping.
Change-Id: I0c6e61d244be84d6bee5ccbf7c4ff4ea0f0b90fd
Reviewed-on: https://boringssl-review.googlesource.com/15147
Reviewed-by: Adam Langley <agl@google.com>
I had a brain-fart and had in mind that strings.Index(x[i:], _) would
return a value relative to the beginning of |x|, which is impossible.
Change-Id: I905ea1fa3469ea13f2e3b782c4baf2431b615a2f
Reviewed-on: https://boringssl-review.googlesource.com/15146
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Otherwise the order changes each time, which will make the build
egregiously non-deterministic.
Change-Id: Idd501ecd118c61a27566eafc61157715e48758bc
Reviewed-on: https://boringssl-review.googlesource.com/15026
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
References to global symbols generate relocations, which breaks the
integrity check.
Change-Id: If6fa06d5d924294ab496c32e7f082a1ae60fdb24
Reviewed-on: https://boringssl-review.googlesource.com/15025
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Some assembly code references “OPENSSL_ia32cap_P+4(%rip)” etc, which
slipped by the previous check.
Change-Id: I22c3fbf9883aea695e8584857bf9c0e3113f9a77
Reviewed-on: https://boringssl-review.googlesource.com/15024
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
It's not obvious how to make ASAN happy with the integrity test but this
will let us test FIPS-only code with ASAN at least.
Change-Id: Iac983787e04cb86a158e4416c410d9b2d1e5e03f
Reviewed-on: https://boringssl-review.googlesource.com/14965
Reviewed-by: Adam Langley <agl@google.com>
This restores the original version of delocate.go, with the subsequent
bugfixes patched in. With this, the FIPS module builds with GCC and
Clang, with and without optimizations. I did patch over a variant of the
macro though, since it was otherwise really wordy.
Playing games with sections was a little overly clever and relied on the
compiler not performing a number of optimizations. Clang blew threw all
of those assumptions.
Change-Id: Ib4da468a5925998457994f9e392cf0c04573fe91
Reviewed-on: https://boringssl-review.googlesource.com/14805
Reviewed-by: Adam Langley <agl@google.com>
This fixes two issues in clang.
- clang emits callq instead of call.
- clang emits .cfi_endproc after .size for the dummy functions. This
causes it to get confused as there is no matching .cfi_startproc.
Don't bother trying to omit the dummy functions.
Alas, clang seems to compile the DEFINE_METHOD_FUNCTION hooks in a way
that brings the .rel.ro back AND isn't honoring the noinline. We'll
probably need to go back to the original CL's setup there.
Change-Id: Ic21ea99e54a93cdc739e4f67dc308d83083607d6
Reviewed-on: https://boringssl-review.googlesource.com/14804
Commit-Queue: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
In typical style I forgot to push a new revision before
landing fd49993c3b. That change accidently
dropped patchset eight when I squashed David's changes in, so this
restores that and fixes a couple of 80-char issues in a Python script.
Change-Id: I7e9338a715c68ae5c89d9d1f7d03782b99af2aa8
Reviewed-on: https://boringssl-review.googlesource.com/14784
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
