Older ARC700 processors had atomic instructions (AKA llock/scond)
as an option and so quite some "atomic" operations were not possible
w/o OS support, which we implemented - see arc_usr_cmpxchg() in the
Linux kernel.
And in uClibc, which was the only Linux libc back in the day of ARC700
era, it is well supported. Well, uClibc could be configured to support it.
Which is done with CONFIG_ARC_HAS_ATOMICS Kconfig option.
But the problem is there's no check for ARC ISA version in uClibc when
this option gets enabled. That leads to a funny situation when even for
ARCv2 processors (ARC HS3x & HS4x) uClibc tries to utilize
arc_usr_cmpxchg() syscall which is not supported for this newer ISA since
ARCv2 processors have atomic instructions built-in all the time.
So what was happening here we didn't specify additional "-matomic"
CFLAG unless we were targeting exactly those ancient ARC770 processors
(ARC700 + MMUv3 + atomics) and so even for ARCv2 we forced uClibc
to not use built-in atomics.
And even though there're ways to add a smarter solution here to handle
that pretty rare by now case of ARC750 (ARC700 + MMUv2 - atomics),
I suggest we just remove this part completely, leaving a possibility
to add needed option in uClibc-ng's configuration
(I mean "packages/uClibc-ng/config").
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Add sample configuration for building cross toolchain for the TI PRU.
PRU cores are present in many of the BeagleBone single board computers.
More information about the PRU can be found in https://bbb.io/pru
Signed-off-by: Dimitar Dimitrov <dimitar@dinux.eu>
Similar to commit 57679b5e ("Disable context functions for Thumb") when
building for thumb we need to unset UCLIBC_HAS_CONTEXT_FUNCS.
Fixes#1397
Signed-off-by: Chris Packham <judge.packham@gmail.com>
picolibc is another bare-metal C library, and so should be mapped
to CT_TARGET_SYS just like newlib does.
Signed-off-by: Keith Packard <keithp@keithp.com>
-- c6x: Add support for c6x product families to pass on to uClibC-ng
-- c6x: Fix multilib support
-- c6x: Add patch fix internal instruction error (GCC 57295)
Signed-off-by: Dan Tejada <dan.tejada@cantada.com>
GLIBC 2.31 needs --with-cpu=ultrasparc for both 32/64-bits now, and
--with-cpu only sets the CPU model for the "primary" bitness.
Signed-off-by: Alexey Neyman <stilor@att.net>
From GCC's standpoint ARC's multilib items are defined by "mcpu" values
which we have quite a few and for all of them might be built optimized
cross-toolchain.
From Glibc's standpoint multilib is just multi-ABI [1] and so very limited
versions are supposed to co-exist (e.g. arc700 & archs).
Here we force Glibc to install libraries in GCC's multilib folder to create
a universal cross-toolchain that has libs optimized for multiple CPU types.
But note we only need to mess with installation paths in case of real
multilib, otherwise we keep default "lib/" paths so that GCC finds default
(the one and only) libs where it expects them to be.
Also here we add a sample which allows to build universal Glibc Linux
toolchain for ARC.
[1] https://sourceware.org/ml/libc-alpha/2019-06/msg00018.html
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
... parts of the config tuple. While here, remove parts that are
setting portions of the target tuple to a value that's already
the default.
Signed-off-by: Alexey Neyman <stilor@att.net>
This required some rework of the libc selection, as moxiebox is a layer on
top of another libc - newlib.
Also, moxiebox'es host VM (`sandbox`) needs a libcrypto on the host. We will
not have it if we're cross-compiling a canadian cross. Fortunately, all moxiebox
needs from libcrypto is SHA256, and it already includes a standalone implementation
of SHA256 in its runtime. Provide a little wrapper that allows moxiebox use
that implementation for the host binary, too.
Also, automate collecting/printing the list of all packages in a given category
(e.g. LIBC or COMP_TOOLS), generate a list of all Kconfig symbols for a given
category.
Signed-off-by: Alexey Neyman <stilor@att.net>
They're written in ARM dialect, and `ldmia r14, {r14, pc}` is not accepted in T2
encoding. GCC8 changed the list of multilibs for arm-*, which now includes a
variant with CPU that supports T2 but not A1 encoding.
Signed-off-by: Alexey Neyman <stilor@att.net>
In case we build for ARC core which has no support of atomic ops among
other things we need to configure libc to use Linux kernel helper to emulate
HS atomic ops. This is done with disabling of CONFIG_ARC_HAS_ATOMICS in uClibc.
Currently we __remove__ this option from .config but this makes no sense as
its default state is "y" so we need to explicitly disable it instead.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Synopsys' DesignWare ARC Processors are a family of 32-bit CPUs
that SoC designers can optimize for a wide range of uses,
from deeply embedded to high-performance host applications in a variety
of market segments.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
- Glibc configure args and tuple need adjustment on SuperH
- Only allow "both endian" and "with CPU" for unspecified arch
variant. May reconsider endianness (was breaking things before
adjusting glibc tuple)
- Retire non-multilib sample, it should be a subset of the
multilib one now.
Signed-off-by: Alexey Neyman <stilor@att.net>
1. On SuperH, configuring GCC with explicit variant of the CPU
(like "sh4") limits the default set of multilibs to just that CPU
and requires --with-multilib-list to change. Allow for "unspecified"
variant, so that we can defer to GCC to determine the list.
2. Support toolchains with both endiannesses at the same time.
3. Add a SuperH/newlib sample
4. Add more flags processing for uClibc
Signed-off-by: Alexey Neyman <stilor@att.net>
Must have eabi suffix for GCC to accept it. Also:
- We only have one glibc now, no need to account for eglibc.
- Rename aarch64 samples, eabi suffix does not apply to them
(and ct-ng saveconfig was saving them into a different directory).
Fixes#772.
Signed-off-by: Alexey Neyman <stilor@att.net>
The generated sysnum.h is different for o32/n32/64 ABIs.
This needs to be revisited; either do this for all architecutres or
perhaps, compare the headers for various multilibs and combine them
if the are identical.
Signed-off-by: Alexey Neyman <stilor@att.net>
(see the comments in the code for details on the issue)
Old workaround in 100-gcc.sh stopped working (probably, due to one
of GCC version upgrades), so switch to the other approach originally
described there: adjust the list of multilibs to not include the
default target explicitly.
Signed-off-by: Alexey Neyman <stilor@att.net>
To build uClibc correctly we need correct endianness selected in the
crosstool-NG. Xtensa cores may be little- or big-endian, but this
property is static. The toolchain knows the core endianness and doesn't
need options to select it.
Enable ARCH_SUPPORTS_BOTH_ENDIAN and select LE by default. Specify empty
CT_ARCH_ENDIAN_CFLAG so that -m{big,little}-endian don't get added to
the TARGET_CFLAGS, as it's not supported by gcc. Specify empty
CT_ARCH_ENDIAN_LDFLAG so that -EB/-EL don't get added to the
TARGET_LDFLAGS as they are ignored. Select big-endian in the example
xtensa-unknown-linux-uclibc configuration.
This fixes uClibc toolchain build for little-endian cores.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Create a separate 'libc_backend_once', install headers into a
subdirectory (different sets of headers are installed for 32- and 64-bit
architectures), and create a symlink for the dynamic linker location
expected by GCC.
Signed-off-by: Alexey Neyman <stilor@att.net>
Rather than echo-ing the new value, set the value into the variable with
the name passed as an argument (similar to CT_SanitizeVarDir). This
allows to use CT_DoLog in these functions.
Signed-off-by: Alexey Neyman <stilor@att.net>
This step was only used in uClibc. However, with upcoming multilib, the
config management will have to be done for each variant differently,
anyway.
uClibc was the only user of libc_check_config step, as well as
CT_CONFIG_DIR directory. Retire these.
Two other clean-ups in uClibc.sh:
- KERNEL_HEADERS check seems to be bogus, this config option is not
present even in 0.9.30 - which is not supported already.
- SHARED_LIB_LOADER_PREFIX was renamed to MULTILIB_DIR in 0.9.31,
according to ChangeLog - and MULTILIB_DIR is passed from command line
instead.
Signed-off-by: Alexey Neyman <stilor@att.net>
i[34567]86-*-gnux32 is not a valid tuple.
Signed-off-by: Bryan Hundven <bryanhundven@gmail.com>
Signed-off-by: Ray Donnelly <ray.donnelly@gmail.com>
Signed-off-by: Alexey Neyman <stilor@att.net>
This code was abstracted out of Cody P Schafer's multilib patch.
It doesn't seem right having architecture dependent code in a
specific libc implementation script. So this patch breaks it out into
scripts/build/arch/<arch>.sh in a function:
multilib_target_to_build="$(CT_DoArchMultilibTarget 'multi_flags'
'target-in')"
Note that this function gets called on each multilib variant with
different sets of compiler flags supplied in 'multi_flags'. The caller
will first filter the flags so that there is no conflicting flags (e.g.,
no '-m32 -m64') supplied.
Changed by Alexey Neyman:
- make option analysis check specific option rather than match global
options string as a whole. Moreover, old code did not handle multiple
options in the same multilib, e.g. '-m64 -mlittle'.
- fixed substitutions in powerpc.sh (*le variants did not match the
pattern in the shell parameter expansion)
- make s390.sh actually apply the flags it gathered from the options.
- straighten the spaghetti in x86.sh by setting two flags, arch & abi.
Also, do not depend on "gnu" being the last part - we can have
'*-uclibcx32', for example.
Signed-off-by: Bryan Hundven <bryanhundven@gmail.com>
Signed-off-by: Ray Donnelly <ray.donnelly@gmail.com>
Signed-off-by: Alexey Neyman <stilor@att.net>
By default, sparc64-*-linux is configured with -mcpu=v9. However,
according to https://sourceware.org/ml/libc-alpha/2005-12/msg00027.html:
"There is no Linux sparc64 port that runs on non-UltraSPARC-I+ ISA
CPUs."
v9 is such a "non-UltraSPARC-I+ ISA CPU", so it makes no sense to
default to v9 when targetting Linux.
Change the default to ultrasparc, even though it can suboptimally
schedule instructions for newer SPARC CPUs. See the pending patch:
https://patchwork.ozlabs.org/patch/409424/
Signed-off-by: Alexey Neyman <stilor@att.net>
