The "openssl speed -testmode -seconds 1 -bytes 1 aes-128-cbc" test
revealed that the assembly code is crashing if length is less than 16.
The code shifts the provided length by 4 and than subtracts one until
the length hits zero. If it was already zero then it underflows the
counter and continues until it segfaults on reading or writing.
Replace the check against 0 with less than 15.
Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25637)
(cherry picked from commit c71c65b922)
When $label == "0", $label is not truthy, so `if ($label)` thinks there isn't
a label. Correct this by looking at the result of the s/// command.
Verified that there are no changes in the .S files created during a normal
build, and that the "0:" labels appear in the translation given in the error
report (and they are the only difference in the before and after output).
Fixes#21647
Change-Id: I5f2440100c62360bf4bdb7c7ece8dddd32553c79
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/21653)
(cherry picked from commit 9607f5ccf2)
an assembler for Windows on Arm builds and also clang-cl as the compiler
as well. Make appropriate changes to armcap source and peralsm scripts.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19523)
(cherry picked from commit b863e1e4c6)
Decoration prefix for some assembler labels in aes-gcm-avx512.pl was
fixed for mingw64 build.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17868)
(cherry picked from commit 224ea84b40)
The VIA Nehemiah CPU is a x86-32 CPU without SSE2 support. It does not
support multi byte nops and considers the endb32 opcode as an invalid
instruction.
Add an ifdef around the endbr32 opcode on x86-32.
Fixes: #18334
Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18353)
(cherry picked from commit 272138795f)
vsr2vr1() fails on OS X because the main loop doesn't strip the
non-numeric register prefixes for OS X.
Strip any non-numeric prefix (likely just "v") from registers before
doing numeric calculation, then put the prefix back on the result.
Fixes: #16995
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17026)
(cherry picked from commit e67edf60f2)
This commit adds an architecture named aix64-gcc-as which can generate
assembler source code compatible with AIX assembler (as) instead of the
GNU Assembler (gas). This architecture name is then used in a callback
for the .p2align directive which is not available in AIX as.
The motivation for this addition came out of an issue we ran into when
working on upgrading OpenSSL in Node.js. We ran into the following
compilation error on one of the CI machines that uses AIX:
05:39:05 Assembler:
05:39:05 crypto/bn/ppc64-mont-fixed.s: line 4: Error In Syntax
This machine is using AIX Version 7.2 and does not have gas installed
and the .p2align directive is causing this error. After asking around if
it would be possible to install GAS on this machine I learned that AIX
GNU utils are not maintained as well as the native AIX ones and we
(Red Hat/IBM) have run into issues with the GNU utils in the past and if
possible it would be preferable to be able to use the AIX native
assembler.
Refs: https://github.com/nodejs/node/pull/38512
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15638)
Power has 2 numbering systems for vector registers:
* VR: Vector Registers are numbered from 0 to 31
* VSR: Vector-Scalar registers are numbers from 32 to 63
These refer to the same registers. Some instructions use VR numbering
for their operands, while others use VSR numbering.
When using Perl to provide a meaningful name for a register it makes
sense to use the same variable for both VR and VSR instructions. This
makes the code more readable.
However, providing a VSR number (i.e. >=32) to an instruction that
expects a VR number will cause an assembler error.
So, for instructions that require VR numbering, map VSR numbers
(i.e. >=32) to VR numbers. This also allows existing code that uses
VR numbering to remain unchanged.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15401)
CLA: trivial
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12161)
A small number of files contain references to the "OpenSSL license"
which has been deprecated and replaced by the "Apache License 2.0".
Amend the occurences.
Fixes#11649
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/11663)
It turns out that GNU as and Solaris as don't have compatible ideas on
the .section syntax, so we need to check if we're using GNU as or
another assembler and adapt this .section syntax accordingly.
Fixes#11132
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/11191)
Since pointer in x32 is 4 bytes, add x86_64-support.pl to define
pointer_size and pointer_register based on flavour to support
stuctures like:
struct { void *ptr; int blocks; }
This fixes 90-test_sslapi.t on x32. Verified with
$ ./Configure shared linux-x86_64
$ make
$ make test
and
$ ./Configure shared linux-x32
$ make
$ make test
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10988)
In https://github.com/openssl/openssl/pull/10883, I'd meant to exclude
the perlasm drivers since they aren't opening pipes and do not
particularly need it, but I only noticed x86_64-xlate.pl, so
arm-xlate.pl and ppc-xlate.pl got the change.
That seems to have been fine, so be consistent and also apply the change
to x86_64-xlate.pl. Checking for errors is generally a good idea.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: David Benjamin <davidben@google.com>
(Merged from https://github.com/openssl/openssl/pull/10930)
We should always generate .note.gnu.property section in x86_64 assembly
codes for ELF outputs to mark Intel CET support since all input files
must be marked with Intel CET support in order for linker to mark output
with Intel CET support. Also .note.gnu.property section in x32 should
be aligned to 4 bytes, not 8 bytes and .p2align should be used
consistently.
Verified with
$ CC="gcc -Wl,-z,cet-report=error" ./Configure shared linux-x86_64 -fcf-protection
$ make
$ make test
and
$ CC="gcc -mx32 -Wl,-z,cet-report=error" ./Configure shared linux-x32 -fcf-protection
$ make
$ make test # <<< 90-test_sslapi.t failed because 8-byte pointer size.
Fix#10896
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10985)
We should always generate .note.gnu.property section in x86 assembly
codes for ELF outputs to mark Intel CET support since all input files
must be marked with Intel CET support in order for linker to mark output
with Intel CET support.
Verified with
$ CC="gcc -Wl,-z,cet-report=error" ./Configure shared linux-x86 -fcf-protection
$ make
$ make test
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11044)
To support Intel CET, all indirect branch targets must start with
endbranch. Here is a patch to add endbranch to all function entries
in x86 assembly codes which are indirect branch targets as discovered
by running openssl testsuite on Intel CET machine and visual inspection.
Since x86 cbc.pl uses indirect branch with a jump table, we also need
to add endbranch to all jump targets.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10984)
If one of the perlasm xlate drivers crashes, OpenSSL's build will
currently swallow the error and silently truncate the output to however
far the driver got. This will hopefully fail to build, but better to
check such things.
Handle this by checking for errors when closing STDOUT (which is a pipe
to the xlate driver).
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10883)
This appears to be emitted with gcc and clang with -fcf-protection
selected, so we should do the same.
We're trying to be smart, and only emit this when the 'endbranch'
pseudo-mnemonic has been used at least once.
This is inspired by and owes to work done by @hjl-tools (github)
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/10875)
This is a big endian ELFv2 configuration. ELFv2 was already being
used for little endian, and big endian was traditionally ELFv1
but there are practical configurations that use ELFv2 with big
endian nowadays (Adélie Linux, Void Linux, possibly Gentoo, etc.)
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8883)
- add instructions: clfi, stck, stckf, kdsa
- clfi and clgfi belong to extended-immediate (not long-displacement)
- some cleanup
Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10346)
clang imposes some restrictions on the assembler code that
gcc does not.
Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10330)
Add non-base instructions which are used by the chacha20 and
poly1305 modules.
Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8181)
z14 introduced alignment hints to help vector load/store
performance. For its predecessors, alignment hint defaults
to 0 (no alignment indicated).
Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8181)
The add/double shortcut in ecp_nistz256-x86_64.pl left one instruction
point that did not unwind, and the "slow" path in AES_cbc_encrypt was
not annotated correctly. For the latter, add
.cfi_{remember,restore}_state support to perlasm.
Next, fill in a bunch of functions that are missing no-op .cfi_startproc
and .cfi_endproc blocks. libunwind cannot unwind those stack frames
otherwise.
Finally, work around a bug in libunwind by not encoding rflags. (rflags
isn't a callee-saved register, so there's not much need to annotate it
anyway.)
These were found as part of ABI testing work in BoringSSL.
Reviewed-by: Richard Levitte <levitte@openssl.org>
GH: #8109
"Windows friendliness" means a) unified PIC-ification, unified across
all platforms; b) unified commantary delimiter; c) explicit ldur/stur,
as Visual Studio assembler can't automatically encode ldr/str as
ldur/stur when needed.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8256)
"Windows friendliness" means a) flipping .thumb and .text directives,
b) always generate Thumb-2 code when asked(*); c) Windows-specific
references to external OPENSSL_armcap_P.
(*) so far *some* modules were compiled as .code 32 even if Thumb-2
was targeted. It works at hardware level because processor can alternate
between the modes with no overhead. But clang --target=arm-windows's
builtin assembler just refuses to compile .code 32...
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8252)
Added crypto/perlasm/s390x.pm Perl module. Its primary use is to be
independent of binutils version, that is to write byte codes of
instructions that are not part of the base instruction set.
Currently only gas format is supported.
Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6919)
Sebastian Andrzej Siewior
Tomas Mraz
Richard Levitte
Tom Cosgrove
Everton Constantino
Andrey Matyukov
Deepankar Bhattacharjee
Matt Caswell
Martin Schwenke
Daniel Bevenius
Dr. David von Oheimb
haykam821
Shourya Shukla
Philippe Antoine
H.J. Lu
David Benjamin
Dr. Matthias St. Pierre
Andy Polyakov
Patrick Steuer
Rich Salz