llvm embeds the repository location in a header file. We just changed
the SRC_URI to fetch using https instead of git. This has started giving
errors in reproducibility testing when one of the RPMs to be compared is
fetched from sstate.
Bump HASHEQUIV_HASH_VERSION and PR so its rebuilt.
(From OE-Core rev: 2fd327ae5b3da8841ebed00bfc60b7bb8d6a64c5)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Add branch name explicitly to SRC_URI where it's not defined and switch
to using https protocol for Github projects.
The change was made using convert-srcuri script in scripts/contrib.
(From OE-Core rev: ab781d4e3fa7425d96ea770ddfd0f01f62018c5b)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Add fcntl64 wrapper which hopefully fixes issues seen in findutils and the find
command in the libtool removal code when built with LFS compile flags on Gentoo.
(From OE-Core rev: 64b68a7e2ebc1a7775f5fda64d7024879181aa7f)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit f26867fe4daec7299f59a82ae4a0d70cceb3e082)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Pulls in:
pseudo_db: Flush DB if there is a shutdown request
fcntl: Add support for fcntl F_GETPIPE_SZ and F_SETPIPE_SZ (test fix)
(From OE-Core rev: 7b9123cac297275245c2fef78c286c17ca3690cf)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 0882095d608ce3abbcc9814517434c21ea549063)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
"make install" may attempt to compile the python modules but it uses the host python
and host paths which means the binaries are not reproducbile. Make things consistent.
If anyone needs compiling, it will beed to be fixed to be cross compile compatible.
(From OE-Core rev: b06105ec1a33bb37d42a10222e07b7f8eb69baff)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 1189f95e05c80286e009e1ab46a603ee5b7ca239)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
On an aarch64 build host, vendor is found to be "unknown", on x86 systems
it is "pc". This filters through to the PLATFORM tag in target rpms.
We saw reproducibility test failures where the PLATFORM tags in noarch
rpms were changing depending upon which host built them. Forcing the
vendor value to a consistent one makes things deterministic.
(From OE-Core rev: ef37fca2f434eba22918e5eab4f2d5e2d499fc14)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit f6434075b2bdfc23c683d22281b674b1e6abde77)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Apply some changes on the Ruby makefiles in order to fix the reproducibility:
- use a fixed timestamp,
- sort linked objects,
- doesn't use the current date,
- and use UTC date.
[YOCTO #14268]
(From OE-Core rev: eea2f854fdd0f036dae2671ae52ba4695a83df68)
Signed-off-by: Thomas Perrot <thomas.perrot@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 59b07ab51ff932a4632a31675445ba4192bae36b)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Pull in the following changes:
* ports/linux/guts: Add closefrom support for glibc 2.34
* pseudo_client: Make msg static in pseudo_op_client
* ports/linux/guts: Add close_range wrapper for glibc 2.34
* pseudo_client: Do not pass null argument to pseudo_diag()
* test-openat: Consider device as well as inode number
* test: Add missing test-statx test case
* fcntl: Add support for fcntl F_GETPIPE_SZ and F_SETPIPE_SZ
(From OE-Core rev: 190ed5d766b6a3922c3f841eb5fd04cf603cf76f)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 71b549924a7fa7973a8e03e11f3db45fdc29889d)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The merge of libdl into libc in glibc 2.34 causes problems for pseudo. Add a fix
that works around this issue.
(From OE-Core rev: c9203671d638edb6a063d7522b29b8a8e776c2f3)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit dd3e46a043c81cd4d81731a0f691868d3c059742)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Add a fix for reproducibility issues where pyc files for python-config.py
may not always be generated.
(From OE-Core rev: 94aedf7d2b43ecdf0da1cf6b848f6e95ee253abf)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit d1c3a87c48b598b6e5624d0affe8bd89320631bf)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We don't want the compression thread numbers to be encoded into the rpm
since this results in the rpm not being deterministic. Add a patch
from Alex Kanavin which addresses this issue (was queued for rpm 4.17
but we need to fix this with 4.16 too).
(From OE-Core rev: 16d6f01eced9e6de5068056aea07a08ec9dfb659)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 1ba0bf50c72f2506dfa507559c49a70e16cd5124)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
CVE is effectively disputed - yes there is stack exhaustion but no bug and it
is building the parser, not running it, effectively similar to a compiler ICE.
Upstream no plans to address and there is no security issue.
https://github.com/westes/flex/issues/414
(From OE-Core rev: e2de2e5e977d84dab6cb1461800d4c29436da5c9)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 0cae5d7a24bedf6784781b62cbb3795a44bab4d1)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
CVE-2021-31810:
A malicious FTP server can use the PASV response to trick Net::FTP into
connecting back to a given IP address and port. This potentially makes
Net::FTP extract information about services that are otherwise private
and not disclosed (e.g., the attacker can conduct port scans and service
banner extractions).
CVE-2021-32066:
Net::IMAP does not raise an exception when StartTLS fails with an
unknown response, which might allow man-in-the-middle attackers to
bypass the TLS protections by leveraging a network position between the
client and the registry to block the StartTLS command, aka a “StartTLS
stripping attack.”
References:
https://www.ruby-lang.org/en/news/2021/07/07/trusting-pasv-responses-in-net-ftp/https://www.ruby-lang.org/en/news/2021/07/07/starttls-stripping-in-net-imap/
Patches from:
bf4d05173ce2ac25d0eb
(From OE-Core rev: e14761916290c01683d72eb8e3de530f944fdfab)
Signed-off-by: Yi Zhao <yi.zhao@windriver.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It is not safe to call the 'file' command from multiple threads. When a
file is checked with multiple threads in parallel, the file descriptors
might get shared which makes the pipe handling lock up, leading to lock
up in rpmbuild. And may lead to rarely deadlock on random recipes's
do_package task.
(From OE-Core rev: 167814b81ddac3934077b0ee91c0c6015fc02bfe)
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
e2fsprogs calls filesystems larger than 3MB but smaller than 512MB
"small", which has some implications:
- blocksize 1024 instead of 4096
- inode_ratio 4096 instead of 16384
- inode_size 128 instead of 256
The outcome of the inode size dropping to 128 bytes is that they cannot
store 64-bit timestamps, so are not Y2038-safe.
A previous attempt to solve this problem[1] changed some of the canned
wic files to pass -T default to mkfs.ext4, but this only covered wic
images and not traditional images. Also, actually small filesystems,
for example a core-image-minimal, will happily be tens of megabytes and
with the "default" options will result in an image which runs out of
blocks before it runs out of space:
mkfs.ext4: Could not allocate block in ext2 filesystem while populating file system
Considering that many OpenEmbedded images are in fact "small", being
2038-safe is worth the marginal increase is disk usage. This patch
alters the small configuration in native builds so that it also has
256-byte inodes. Target is unchanged so that standard behaviour is
maintained outside of the build.
This is actually the same underlying patch that Mathieu Dubois-Briand
sent in April, but the wic change in [1] was accepted instead. I believe
that is the wrong approach and this approach covers more cases.
[ YOCTO #14478 ]
[1] openembedded-core eecbe62
[2] https://lists.openembedded.org/g/openembedded-core/message/150298
(From OE-Core rev: e89bac87c91e943060662be04775a1ff8e4c4f22)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 9ab0ae83a24ee99e69f8ac54256b253a122aef8a)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Fix a slew of CVEs (CVE-2021-3544, CVE-2021-3545, CVE-2021-3546) by
backporting the relevant patches from qemu's git.
(From OE-Core rev: ce850a5ce84f949d3114024c89ae3dd98fcbef41)
(From OE-Core rev: 8eb55f9eba667ab509baeb4328f9a080aa10e3fe)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit ce850a5ce84f949d3114024c89ae3dd98fcbef41)
Signed-off-by: Lee Chee Yang <chee.yang.lee@intel.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In sdk, call createrepo-c failed with:
...
$ createrepo_c --update ./test_repo/rpm
Directory walk started Critical: Failed to detect compression for file
./test_repo/rpm/cortexa72/hello-2.10-r0.cortexa72.rpm: magic_load() failed: could not find any valid magic files!
...
Since commit [ea666fbc74 createrepo-c: set path to magic database for
native and nativesdk] applied, the MAGIC is incorrectly assigned.
The variable datadir will be expanded automatically for nativesdk,
do not need to add prefix ${SDKPATHNATIVE} to MAGIC
(From OE-Core rev: d99b4dac74add826aa63ecb20c427d2884985329)
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 54368f1b02e1ac4aa068515730a8c8bcd3683eb3)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Recently, the none/tests/amd64/fb_test_amd64 test had been flaky and
causing failures on the auto-builder. Until we can get to the root cause
of the issue, we are going to skip the test to reduce the noise from the
ptests.
(From OE-Core rev: 697eb7f2a5d20aa0ad7389efd68420ec59daca32)
Signed-off-by: Tony Tascioglu <tony.tascioglu@windriver.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit a365cd7a358db96791033e6dc6e45d2e816d3e4c)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This updates to the latest pseudo version which includes:
Revert "client: Fix some compiler warnings"
ports/linux: Always build statx support
makewrappers: Handle parameters marked as nonnull
client: Fix some compiler warnings
wrappers: Avoid -Wcast-function-type warning
In particular, this pseudo version always has statx enabled which means
we can then remove the need to make pseudo-native host distro specific
which fixes an eSDK issue.
(From OE-Core rev: bcf74e0d048754ed46bda90cd582320d0df2a4ad)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 58cc70940ff998be49a9b89e1ad0538242cb7998)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When building pseudo-native to work with uninative, we need to ensure the
configuration will work on all supported target systems. This means
"new clone" semantics, xattr and statvfs support in particular. It is
extremely unlikely we'd run on a system without any of these but add
a check just to be sure when uninative is enabled.
(From OE-Core rev: 496f0cd63aaeedbff625eb687f20ace951faf6f1)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit ea5b208ee25752bea6037cd0f3b28da7d2c9905e)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Backport the body of a fix for CVE-2021-3572 since hardknott carries
20.0.2, and the delta between it and the latest 21.1.3 is more than just
bugfixes.
CVE: CVE-2021-3572
(From OE-Core rev: fb7a2af241795b82f121381cea6f4b56ce948ebf)
Signed-off-by: Trevor Gamblin <trevor.gamblin@windriver.com>
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We use dwarfsrcfiles in package.bbclass to list the source files used by a binary.
This is done before they're stripped and linked to debug symbols in separate files.
It is possible a binary may already have a link to separate debug symbols, e.g.
some of the test binaries in lttng-tools ptest. In those cases, the linked binary
may be changed by package.bbclass code whilst dwarfsrcfiles is reading it. That
would result in a rare SIGBUS race causing the binary to fail.
To avoid this, break the debug file search path so no other binaries are found.
Also fix a segfault if no binary is specified while here.
[YOCTO #14400]
(From OE-Core rev: 52382a03c10a6985ecb6ada24523cb9daf9c6201)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit efef732859e265533acf16f2f4da3b29d50e0df4)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Previouly there is a logic as below used to set libpth in config.sh.
libpth='@LIBDIR@ @BASELIBDIR@'
But after the below commits introduced, the above logic is dropped.
52f2828314 perl: add a version that builds the recipe using perl-cross, and update to 5.28.1
68552c3532 perl: remove the previous version of the recipe
So correct the value of libpth and glibpth to add the dropped logic
back to avoid confusing.
Before the patch(on 64bits system):
# perl -V:libpth
libpth='/usr/lib /lib';
After the patch(on 64bits system):
# perl -V:libpth
libpth='/usr/lib64 /lib64';
(From OE-Core rev: afe58be55b4efc360420a00cbcf60dd5d99ed556)
Signed-off-by: Mingli Yu <mingli.yu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit a494de43c3ccdcf7af988765ae5c3a95bc20c567)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Install list of non-deterministic threaded ptests to be run using taskset
to force them to a single core. This commit works with b318944d7, which
updated the testing script to run the non-deterministic tests separately
but didn't install the list of tests, so these tests were being run
without taskset.
The taskset_nondeterministic_tests file is the list of tests that will
be run separately with taskset, and ignored during the other tests. This
is installed to /usr/lib/valgrind/ptest similar to the 2 existing lists
for tests to skip on ARM and all architectures.
Removed bar_bad and bar_bad_xml to be included separately as they cause
issues on non-kvm QEMU instances.
See:
b318944dd7 valgrind: Improve non-deterministic ptest reliability
for more info.
(From OE-Core rev: f076edb7515ba2ecfc0adbfdf30ae5a9aa96e231)
Signed-off-by: Tony Tascioglu <tony.tascioglu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 3d23985d0d653844863ed513d75d93a36359992f)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This commit removes the stack_changes ptest from aarch64 devices.
This test is buggy and fails almost 100% of the time in qemuarm64.
In general, many of the valgrind tests are more likely to fail on
qemuarm64 vs native x86_64.
This test previously worked on gatesgarth and dunfell, but has
been failing since hardknott. It might be due to a recent change
in the cross-compiler or glibc.
The test runs fine when running natively on arm on a Raspberry Pi.
Until we can find the root cause for the failures, this shorter
term solution should clear up some of the noise from the autobuilder
from a known failure.
(From OE-Core rev: 731d0e9c6921a3ac82e5172e5c7a6088e80243ef)
Signed-off-by: Tony Tascioglu <tony.tascioglu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 82d6411b80a46d8ec0258ca75c3c80dc6128d44e)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Several of the valgrind tests (particulaly helgrind) are unreliable and
can fail with a different output.
Particularly, there is a higher chance of failure on QEMU instances with
SMP enabled and on systems with more interrupts such as laptops on powersave.
The tests have been reported upstream as being unreliable dating back
over 5 years, due in part to the ordering of threads during
an "unwinding" process in helgrind.
https://bugs.kde.org/show_bug.cgi?id=345121https://bugs.kde.org/show_bug.cgi?id=430321
A workaround to improve the reliability of such tests is to force them
to run on a single CPU core using taskset. This greatly reduces the
chance of a failure.
>From my testing, I have found it can help reduce the rate of failures
on both a laptop and QEMU by over 5x. Stress-testing in QEMU for several
hours did not result in a failure while running the test normally did.
The flaky or undeterministic thread-based tests are defined in the
taskset_nondeterministic_tests file. These test cases will be run with
taskset 0x00000001 to run on a single CPU core rather then the regular
test.
The edited run-ptest executes the flaky tests first, then ignores them
to not duplicate the results from the main tests. Everything modified is
restored when testing is complete.
The drawback is that this isn't a foolproof solution. It helps the tests
fail much less frequently, and considering how this issue has been documented
for a long time, a workaround such as this is needed.
(From OE-Core rev: 79ec1d73a107277586d3d8e9c0d46dfc0ac2b0d8)
Signed-off-by: Tony Tascioglu <tony.tascioglu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit b318944dd72ca7b0408e955f3599381ab3ac3ba8)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The git repo for pkg-config was changed, so update the
SRC_URI accordingly with the new link.
(From OE-Core rev: 07f223048a5b8ac3cb828a68b6069825c8d656ae)
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
