As described in https://gitlab.com/buildroot.org/buildroot/-/issues/160, the
github mirror is getting shut down - So move to the sourceware.org git repo.
The github mirror was originally used because of performance and reliability
issues with sourceware, but that seems be resolved now after server/RAM
upgrades - E.G. from the sourceware news:
April 22, 2024
server2.sourceware.org now has 512GB RAM, thanks Red Hat.
https://sourceware.org/
So change back to fetch glibc (and localedef) from sourceware.org over git.
Notice: The git archiving leads to slightly different paths and permissions
in the tarball, but the file content is identical:
mkdir a && tar -C a -x --strip-components=1 -f \
path/to/glibc-2.42-51-gcbf39c26b25801e9bc88499b4fd361ac172d4125.tar.gz
mkdir b && tar -C b -x --strip-components=1 -f \
path/to/glibc-2.42-51-gcbf39c26b25801e9bc88499b4fd361ac172d4125-git4.tar.gz
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
[Julien:
- add missing SoB line
- fix command lines in commit log
]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 807b0bab37)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Commit [1] bumped glibc from 2.42-3-gbc13db739 to 2.42-51-gcbf39c26b
to fix some CVEs, but forgot to add those CVEs to GLIBC_IGNORE_CVES.
This was needed because the GLIBC_CPE_ID_VERSION used for CVE checks
remains to the same value "2.42" which is marked as vulnerable to
those CVEs.
This commit adds those _IGNORE_CVES with the corresponding upstream
commit references, to make sure they will not be reported by the
"make pkg-stats" command.
Fixes:
- [1]
[1] 18de297a5a
Cc: Waldemar Brodkorb <wbx@openadk.org>
Cc: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 574aa2cfee)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The IGNORE_CVES entries introduced in [1] no longer match to the glibc
package following the bump to v2.42 in [2]. The version boundaries
specified on the NVD DB are specific to 2.40 & 2.41.
The CVE-2025-8058 though don't have any information available on the NVD
DB and will remain on the IGNORE_CVES then.
[1] feaf53585a package/glibc: security bump to version 2.41-70
[2] fb6256c0ef package/{glibc, localdef}: bump to version 2.42
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 6fc37d7c6e)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Since Buildroot commit [1] the CVEs are no longer matched to CPEs with
versions using '-'.
These IGNORE_CVES entries introduced in [2] are then no longer matched to
the glibc package.
For more information, see the explanation in commit [1].
[1] 35f376d88e support/scripts/cve.py: fix CPE matching
[2] adaae82c58 package/glibc: ignore CVEs not considered as security issues by upstream
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 9383a3a726)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Building a simple configuration such as:
BR2_aarch64=y
BR2_FORCE_HOST_BUILD=y
Under our reference Docker container, as an out-of-tree build:
./utils/docker-run make O=output-test
will fail during the glibc staging installation step with:
/usr/bin/install -c -m 644 /home/thomas/buildroot/br/output-glibc/build/glibc-2.42-3-gbc13db73937730401d592b33092db6df806d193e/build/libc.a /usr/lib64/libc.a
/usr/bin/install: cannot create regular file '/usr/lib64/libc.a': Permission denied
The problem being that the install_root variable is for some reason
lost along the way. We definitely pass it during the staging
installation step, but then glibc calls a sub-make and in that
submake, install_root is empty.
Observations:
- Building the same configuration inside the Docker container, but
in-tree, doesn't exhibit the problem.
- Building outside of the Docker container, but with the same make
version as the one we build due to BR2_FORCE_HOST_BUILD=y doesn't
exhibit the problem.
However, it turns out that glibc sets install_root to $(DESTDIR). And
passing DESTDIR just works. So we drop our custom
GLIBC_INSTALL_STAGING_OPTS to rely on the default one, and this fixes
the problem. It's not extremely satisfying to not fully understand
what's going on, but after spending quite some time on this, and
having a trivial and actually pretty nice workaround, we simply
propose to use this solution.
There is no autobuilder issue as we're not using our Docker container
to do autobuild builds.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Release e-mail:
https://inbox.sourceware.org/libc-alpha/5906001.DvuYhMxLoT@pinacolada/
Major new features:
* The following ISO C23 function families (introduced in TS
18661-4:2015) are now supported in <math.h>. Each family includes
functions for float, double, long double, _FloatN and _FloatNx, and a
type-generic macro in <tgmath.h>.
- Power and absolute-value functions: compoundn, pown, powr, rootn,
rsqrt.
* On Linux, the pthread_gettid_np function has been added.
* The ISO C2Y family of unsigned abs functions, i.e. uabs, ulabs,
ullabs, and uimaxabs, is now supported.
* On Linux, the <termios.h> interface now supports arbitrary baud rates;
speed_t is redefined to simply be the baud rate specified as an
unsigned int, which matches the kernel interface.
* The thread-local cache in malloc (tcache) now supports caching of
large blocks. This feature can be enabled by setting the tunable
glibc.malloc.tcache_max to a larger value (max 4194304). Tcache is
also significantly faster for small sizes.
* A new configure option, "--enable-sframe", can be used to enable
SFrame support of the GNU C Libraries. SFrame is a new stack trace
information format which can be used by backtrace. It requires
binutils with a minimum version of 2.45.
* Support for lightweight stack guard pages via madvise and the
MADV_GUARD_INSTALL flag has been added to pthread_create.
* Additional optimized and correctly rounded mathematical functions have
been imported from the CORE-MATH project, in particular acospif,
asinpif, atanpif, atan2pif, cospif, sinpif, tanpif.
* The testsuite has been significantly extended, including coverage of
the functionality of the printf and scanf function families in many
variants.
* The manual has been significantly extended and updated, particularly
the threads, terminal, filesystem, resource, and math chapters.
* Code has been added to detect the x86-64 Intel Arrow Lake, Panther
Lake, Clearwater Forest, and Diamond Rapids microarchitectures.
* Regarding S390, support for the new z17 platform has been added.
Deprecated and removed features, and other changes affecting compatibility:
* The glibc.rtld.execstack tunable now supports a compatibility mode to
allow programs that require an executable stack through dynamically
loaded shared libraries.
* On Linux, the <termio.h> header and the definition of struct termio
in <sys/ioctl.h> have been removed. The termio interface has been
obsolete since the very first version of POSIX.1 in 1988, replaced
with <termios.h>.
* The support for TX lock elision of pthread mutexes has been deprecated
on all architectures and will be removed in the next release.
* On AArch64 Linux targets supporting the Scalable Matrix Extension
(SME), setjmp and sigsetjmp will disable the ZA state of SME.
Changes to build and runtime requirements:
* GCC 12.1 or later is now required to build the GNU C Library.
* GNU Binutils 2.39 or later is now required to build the GNU C Library.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
[Julien: resolve conflict with commit feaf53585a]
Signed-off-by: Julien Olivain <ju.o@free.fr>
Glibc introduced LoongArch64 support in 2.36, it requires kernel
5.19 to build.
Unselect BR2_ARCH_HAS_NO_TOOLCHAIN_BUILDROOT to reflect we can
build the toolchain.
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Julien Olivain <ju.o@free.fr>
glibc fails to build on RISC-V 32-bit with the ilp32f ABI and on
RISC-V 64-bit with the lp64f: both use single-point precision floating
point, which glibc doesn't support, failing during the configure step
with:
configure: error: glibc does not yet support the single floating-point ABI
Fix that by disabling glibc support on those configurations.
Fixes:
https://autobuild.buildroot.org/results/fe8d569cab507992978ef0da649278dd3a9e0b23/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Commit 6ecab31fad bumped glibc to
version 2.41, but forgot to update the associated legal information.
Indeed, the LICENSES file has changed upstream, with a few commits
between 2.40 and 2.41:
- 2843e78b30da0aa743fdfb0ac61435c925182c04
Adds another MIT license. We already list MIT in the licenses
- d421d36582281a62e05f96a1bfb085db9d85f40b
Adds URL of a project included in glibc. No license change.
- 4871ab841b3bfbabdba62cc86dd6b8db84f5c134
Drops Intel License Agreement due to IA64 removal, which was in fact
BSD-3-Clause, but there's still BSD-3-Clause code in glibc, and this
license is therefore still listed in GLIBC_LICENSE as it should be
- 1ed3e4468879b79492dfd453255b9318b3265843
Updates some file locations. No license change.
- aba14403c606fdba6726f76442272d5011c52900
Updates CORE-MATH copyright. No license change.
Fixes:
http://autobuild.buildroot.net/results/74fb3ade56140cf8c0496f1d087f0b33ee6b6ff3/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
When glibc was bumped from 2.39 to 2.40 in commit
74cb378aff ("package/{glibc, localdef}:
bump to version 2.40"), the IGNORE_CVES entries should have been
dropped, as they were applicable to the 2.39 branch. Drop them now.
We did verify the pkg-stats output, and those CVEs don't show up as
affecting glibc even after this change, as expected.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Doing so, we remove nios2 support from the internal toolchain backend.
Signed-off-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Fixes the following security issues:
GLIBC-SA-2024-0004:
ISO-2022-CN-EXT: fix out-of-bound writes when writing escape
sequence (CVE-2024-2961)
GLIBC-SA-2024-0005:
nscd: Stack-based buffer overflow in netgroup cache (CVE-2024-33599)
GLIBC-SA-2024-0006:
nscd: Null pointer crash after notfound response (CVE-2024-33600)
GLIBC-SA-2024-0007:
nscd: netgroup cache may terminate daemon on memory allocation
failure (CVE-2024-33601)
GLIBC-SA-2024-0008:
nscd: netgroup cache assumes NSS callback uses in-buffer strings
(CVE-2024-33602)
In addition, the following bugs are fixed:
[19622] network: Support aliasing with struct sockaddr
[30701] time: getutxent misbehaves on 32-bit x86 when _TIME_BITS=64
[30994] REP MOVSB performance suffers from page aliasing on Zen 4
[31339] libc: arm32 loader crash after cleanup in 2.36
[31325] mips: clone3 is wrong for o32
[31335] math: Compile glibc with -march=x86-64-v3 should disable FMA4
multi-arch version
[31402] libc: clone (NULL, NULL, ...) clobbers %r7 register on
s390{,x}
[31479] libc: Missing #include <sys/rseq.h> in sched_getcpu.c may
result in a loss of rseq acceleration
[31316] build: Fails test misc/tst-dirname "Didn't expect signal from
child: got `Illegal instruction'" on non SSE CPUs
[31371] x86-64: APX and Tile registers aren't preserved in ld.so
trampoline
[31372] dynamic-link: _dl_tlsdesc_dynamic doesn't preserve all caller-
saved registers
[31429] build: Glibc failed to build with -march=x86-64-v3
[31501] dynamic-link: _dl_tlsdesc_dynamic_xsavec may clobber %rbx
[31640] dynamic-link: POWER10 ld.so crashes in
elf_machine_load_address with GCC 14
[31676] Configuring with CC="gcc -march=x86-64-v3"
--with-rtld-early-cflags=-march=x86-64 results in linker failure
[31677] nscd: nscd: netgroup cache: invalid memcpy under low
memory/storage conditions
[31678] nscd: nscd: Null pointer dereferences after failed netgroup
cache insertion
[31679] nscd: nscd: netgroup cache may terminate daemon on memory
allocation failure
[31680] nscd: nscd: netgroup cache assumes NSS callback uses in-buffer
strings
[31686] dynamic-link: Stack-based buffer overflow in
parse_tunables_string
[31719] dynamic-link: --enable-hardcoded-path-in-tests doesn't work
with -Wl,--enable-new-dtags
[31782] Test build failure with recent GCC trunk
(x86/tst-cpu-features-supports.c:69:3: error: parameter to builtin
not valid: avx5124fmaps)
[31798] pidfd_getpid.c is miscompiled by GCC 6.4
[31867] build: "CPU ISA level is lower than required" on SSE2-free
CPUs
[31883] build: ISA level support configure check relies on bashism /
is otherwise broken for arithmetic
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Commit b5680f53d6 (package/glibc: bump to 2.39) forgot to drop the ignores
for the 2.38 specific CVEs. Do that now.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
The 3 patches we had have all been merged upstream as of 2.39, so they
can be dropped.
The libcrypt library is gone from glibc, and therefore the
--enable-crypt option is also gone.
The LICENSES file has changed with the addition of a copyright notice
from IBM (some files contributed by IBM have been relicensed), and the
license is MIT, which is already listed in GLIBC_LICENSES, so no
changes are needed on our side.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
[yann.morin.1998@free.fr:
- also update localedef
- rebase localedef patches
]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Commit 8519de517e (package/{glibc, localedef}: security bump to version
glibc-2.38-27-g750a45a783906a19591fb8ff6b7841470f1f5701) correctly mentioned
CVE-2023-4806 in the commit message, but forgot to add an ignore for it.
Fix that.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
5 CVEs affecting glibc according to the NVD database are considered as
not being security issues by upstream glibc developers:
* CVE-2010-4756: The glob implementation in the GNU C Library (aka
glibc or libc6) allows remote authenticated users to cause a denial
of service (CPU and memory consumption) via crafted glob expressions
that do not match any pathnames. glibc maintainers position: "That's
standard POSIX behaviour implemented by (e)glibc. Applications using
glob need to impose limits for themselves"
* CVE-2019-1010022: GNU Libc current is affected by: Mitigation
bypass. The impact is: Attacker may bypass stack guard
protection. The component is: nptl. The attack vector is: Exploit
stack buffer overflow vulnerability and use this bypass
vulnerability to bypass stack guard. NOTE: Upstream comments
indicate "this is being treated as a non-security bug and no real
threat. glibc maintainers position: "Not treated as a security issue
by upstream https://sourceware.org/bugzilla/show_bug.cgi?id=22850"
* CVE-2019-1010023: GNU Libc current is affected by: Re-mapping
current loaded library with malicious ELF file. The impact is: In
worst case attacker may evaluate privileges. The component is:
libld. The attack vector is: Attacker sends 2 ELF files to victim
and asks to run ldd on it. ldd execute code. NOTE: Upstream comments
indicate "this is being treated as a non-security bug and no real
threat. glibc maintainers position: "Not treated as a security issue
by upstream https://sourceware.org/bugzilla/show_bug.cgi?id=22851"
* CVE-2019-1010024: GNU Libc current is affected by: Mitigation
bypass. The impact is: Attacker may bypass ASLR using cache of
thread stack and heap. The component is: glibc. NOTE: Upstream
comments indicate "this is being treated as a non-security bug and
no real threat. glibc maintainers position: "Not treated as a
security issue by upstream
https://sourceware.org/bugzilla/show_bug.cgi?id=22852"
* CVE-2019-1010025: GNU Libc current is affected by: Mitigation
bypass. The impact is: Attacker may guess the heap addresses of
pthread_created thread. The component is: glibc. NOTE: the vendor's
position is "ASLR bypass itself is not a vulnerability. Glibc
maintainers position: "Not treated as a security issue by upstream
https://sourceware.org/bugzilla/show_bug.cgi?id=22853"
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
As reported in bug 15895, the GLIBC_VERSION field having a value
looking like 2.38-27-g750a45a783906a19591fb8ff6b7841470f1f5701, it
prevents the CPE/CVE matching with the NVD database to work correctly.
This commit fixes that by defining GLIBC_CPE_ID_VERSION, derived from
GLIBC_VERSION, by extracting the base version.
Also, we update GLIBC_IGNORE_CVES to account for the CVEs that have
clearly been fixed between 2.38 and
2.38-27-g750a45a783906a19591fb8ff6b7841470f1f5701. There are a number
of other CVEs still affecting the glibc package, but they are not
related to this
2.38...2.38-27-g750a45a783906a19591fb8ff6b7841470f1f5701 range.
Fixes: #15895
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Fixes the following security issues:
CVE-2023-4527: If the system is configured in no-aaaa mode via
/etc/resolv.conf, getaddrinfo is called for the AF_UNSPEC address
family, and a DNS response is received over TCP that is larger than
2048 bytes, getaddrinfo may potentially disclose stack contents via
the returned address data, or crash.
CVE-2023-4806: When an NSS plugin only implements the
_gethostbyname2_r and _getcanonname_r callbacks, getaddrinfo could use
memory that was freed during buffer resizing, potentially causing a
crash or read or write to arbitrary memory.
CVE-2023-5156: The fix for CVE-2023-4806 introduced a memory leak when
an application calls getaddrinfo for AF_INET6 with AI_CANONNAME,
AI_ALL and AI_V4MAPPED flags set.
CVE-2023-4911: If a tunable of the form NAME=NAME=VAL is passed in the
environment of a setuid program and NAME is valid, it may result in a
buffer overflow, which could be exploited to achieve escalated
privileges. This flaw was introduced in glibc 2.34.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Enable mathvec explicitly on aarch64(be) since it's now enabled by
default [1]. aarch64 mathvec requires at gcc-10 but Buildroot already
provide gcc-11 as minimum version.
Don't use --enable-fortify-source for now in order to keep original
behavior while doing the glibc version bump (and because some
architecture doesn't support well fortify-source, i.e Microblaze).
Postpone this change to a follow up commit.
Keep the "deprecated" libcrypt enabled just in case if some
application are not yet ready to use an alternative such as libxcrypt.
Security related changes:
CVE-2023-25139: When the printf family of functions is called with a
format specifier that uses an <apostrophe> (enable grouping) and a
minimum width specifier, the resulting output could be larger than
reasonably expected by a caller that computed a tight bound on the
buffer size. The resulting larger than expected output could result
in a buffer overflow in the printf family of functions.
See:
https://lists.gnu.org/archive/html/info-gnu/2023-07/msg00010.html
Runtime tested with Qemu on Gitlab-ci:
https://gitlab.com/kubu93/buildroot/-/pipelines/998435203https://gitlab.com/buildroot.org/toolchains-builder/-/pipelines/998926028
[1] https://sourceware.org/git/?p=glibc.git;a=commit;h=cd94326a1326c4e3f1ee7a8d0a161cc0bdcaf07e
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Since it was introduced in 5a6087d62e (toolchain: add powerpc SPE ABI
support), the CLASSIC vs. SPE choice for the ABI was never really a
choice: CPU without SPE could only use the CLASSIC ABI, while CPUs with
SPE could only use the SPE ABI.
Commit b4c824562b (powerpc: add BR2_POWERPC_CPU_HAS_SPE to replace
adhoc deps/checks) added a blind option that CPUs with SPE would select
rather than duplicate the ad-hoc dependencies in both CLASSIC and SPE
ABI options. Since then, it was even more obvious that the ABI choice
was really not a choice, as the two options have mutually exclusive
conditions.
Drop the useless choice, and directly use the blind option as selected
by the specific CPUs.
We don't need legacy handling, because the situation fixes itself.
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Joel Stanley <joel@jms.id.au>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Since glibc 2.33 (upstream commit
7a55dd3fb6d2c307a002a16776be84310b9c8989), headers >= 5.4.0 are needed
to build glibc for RISC-V 32-bit. Indeed
sysdeps/unix/sysv/linux/riscv/configure.ac contains:
if test $libc_cv_riscv_int_abi = ilp32; then
arch_minimum_kernel=5.4.0
fi
In order to take into account this dependency, we add the appropriate
logic in package/glibc/Config.in and
toolchain/toolchain-buildroot/Config.in.
This change means that if headers < 5.4.0 are selected, then no C
library at all will be available for RISC-V 32-bit, as glibc is the
only C library supporting RISC-V 32-bit currently. However, thanks to
the recent addition of BR2_TOOLCHAIN_BUILDROOT_NONE, the
choice...endchoice for the C library selection will not be empty,
allowing the user to see the Config.in comment explaining why glibc
can't be selected.
Therefore, technically this commit does prevent from creating a
configuration with RISC-V 32-bit and headers < 5.4.0, but it will have
BR2_TOOLCHAIN_BUILDROOT_NONE=y, which is catched by
package/Makefile.in, which aborts the build early on pointing out that
the configuration is invalid.
Fixes:
http://autobuild.buildroot.net/results/5ca49b2732f68eccb5276e7112f7f496dcc514ee/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
With "arch/powerpc: Enable powerpc64le only on CPUs that support it"
we can use this symbol to indicate the CPU is supported by glibc.
This also aligns the conditions with the kernel headers dependencies.
Signed-off-by: Joel Stanley <joel@jms.id.au>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Acked-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
Commit [1] enabled glibc on or1k since it's now supported but it
requires a toolchain with linux-headers >= 5.4.
From [2]:
"Here we define the minumum linux kernel version at 5.4.0, as that is the
long term support version where 32-bit architectures start to support
64-bit time API's. The OpenRISC kernel had some bugs up until version 5.8
which caused issues with glibc fork/clone, they have been backported to
5.4 but not previous versions."
Fixes:
checking installed Linux kernel header files... 3.2.0 or later
checking for kernel header at least 5.4.0... too old!
configure: error: *** The available kernel headers are older than the requested
https://gitlab.com/buildroot.org/toolchains-builder/-/jobs/2875256686
[1] 68d0aede59
[2] https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=0c3c62ca7d9ff3bdacdd13e636bc858101e3e288
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Arnout Vandecappelle <arnout@mind.be>
Currently, glibc depends on !BR2_STATIC_LIBS in all the toolchain
variants.
However, for some architectures, glibc is the only supported libc. In
commit 3b3105328e ("Config.in: only
allow BR2_STATIC_LIBS on supported libc/arch"), we implemented a fix
to avoid configurations were BR2_STATIC_LIBS=y with an architecture
already supported by glibc, because these configurations are
impossible. This commit 3b3105328e
prevents from selecting BR2_STATIC_LIBS=y when the C library used for
the internal toolchain backend is glibc.
However, it introduces a discrepency between how this topic is handled
for internal and external toolchains:
- For internal toolchains, we prevent BR2_STATIC_LIBS=y if glibc is
chosen.
- For external toolchains, we allow BR2_STATIC_LIBS=y in all cases,
and it's each glibc toolchain that has !BR2_STATIC_LIBS
This commit addresses this discrepency by preventing BR2_STATIC_LIBS=y
if glibc is chosen in all cases.
Thanks to this, we can remove the !BR2_STATIC_LIBS dependency on both
the glibc package, and all glibc external toolchains.
Fixes: https://bugs.busybox.net/show_bug.cgi?id=14256
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[Thomas: update to master, fix the gen-bootlin-toolchains script, add
a comment in the static/shared choice to indicate that static is
supported only with uclibc or musl]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
In commit fd839aeb7f ("package/glibc:
introduce and use BR2_PACKAGE_GLIBC_ARCH_SUPPORTS and
BR2_PACKAGE_GLIBC_SUPPORTS") we moved the Config.in logic about glibc
dependencies from toolchain/toolchain-buildroot/Config.in into
package/glibc/Config.in.
Unfortunately, it is not possible to move the Config.in comments that
tell the user, within the choice..endchoice for the C library why
glibc is not currently selectable, so we had to keep them in
toolchain/toolchain-buildroot/Config.in.
Turns out that the comments were out of sync with the dependencies,
and two comments were missing. This commit adds the missing ones, and
adds a comment in package/glibc/Config.in explaining that we need to
be careful about updating toolchain/toolchain-buildroot/Config.in as
well.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Contains various fixes some of which may effect pthread fork handling.
Add patch reverting a commit breaking builds.
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
As part of this, the dependency of the comment "glibc needs a
toolchain w/ dynamic library, kernel headers >= 3.2" is changed to use
BR2_PACKAGE_GLIBC_ARCH_SUPPORTS instead of just BR2_USE_MMU, so that
the comment only appears on architectures for which glibc is supported
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Fixes:
Generating locale C.UTF-8
[error] LC_MONETARY: value for field `mon_decimal_point' must not be an empty string
[error] no output file produced because errors were issued
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
https://sourceware.org/pipermail/libc-alpha/2022-February/136040.html
- Remove upstreamed patch 0001. Patches to localedef are not upstream.
- allow to use optimization CFLAGS (not CPPFLAGS) which are nowadays
supported by upstream (except nios2)
- enable support for or1k, which is now included upstream
- runtime tested with qemu-system for aarch64/arm/microblaze/mips/mips64/nios2/
or1k/powerpc/powerpc64/powerpc64le/riscv32/riscv64/s390x/sh4/sparc64/x86/x86_64
Since only a single version is supported (no csky fork any more), move
the hash file out of the version directory. Also, make a symlink from
the localedef to the glibc hash file rather than copying it.
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Reviewed-by: Petr Vorel <petr.vorel@gmail.com>
Tested-by: Petr Vorel <petr.vorel@gmail.com>
[Arnout: make localedef.hash a symlink]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Currently, we configure glibc to not add compatibility support for
kernels older than the one used for the headers. This is on the
expectation that the system will never run on a kernel that is older
than the one used for the headers or, when Buildroot builds the kernel,
on another, older kernel.
However, in some situations, it is possible to build for a generic
system, where the kernel may be a different version. This can be the
case, for example, when Building an image that is to be used in a
container that can run on a range of machines each with different kernel
versions. In such a case, it is interesting to build glibc in a way as
to take better advantage of the newer kernels, and thus using newer
kernel headers, while still allowing running on older kernels, and thus
carrying more compatibility code.
We add an option to glibc to allow the user to enable compatibility
shims. To simplify the case, when that option is enabled, we just let
glibc enable as old compatibility shims as supported by the current
architecture.
The code size increase is very small. For an ARM Cortex-A7, with
gcc-10.3.0, the delta is as follows (other files installed by glibc had
no size delta; sizes in bytes):
file | no compat | compat | delta
----------------------+-----------+-----------+-------
ld-linux-armhf.so.3 | 200216 | 200284 | + 68
libc.so.6 | 1814496 | 1823120 | +8624
------+-------
Total | +8692
No runtime overhead has been measured; the overhead is most probably
in the measurement noise. Indeed, the compatibility shims are very
lightweight. For example, there are 9 arch-generic shims:
renameat2(), execveat(), mlock2(), statx(), faccessat2(),
close_range(), time64-related syscall shenanigans, a waitid()
feature, and a futex operation (LOCK_PI2)
and then each arch may define a few others. i386 has less than 20
(mostly related to socket options, and one for the ordering of the
clone() arguments), while ARM seems to have only two (mlock2() and a
configurable futex feature).
Note: however, as Arnout pointed out, some programs may still actually
fail to run even with such compatibility shim, if they really expect the
shimed syscalls to really exist and have no fallback (and/or no proper
error-handling). Still, in the vast majority of cases, those
compatibility shims are enough to have a system running.
Signed-off-by: Yann E. MORIN <yann.morin@orange.com>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Tested-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
