Since Buildroot commit 0e3ddc9dc8,
linux hash files are split between pre- and post-6.17.
Since that commit, the script that automatically updates kernel
versions in Buildroot was broken, as it assumed only a single linux
hash file at a static location. Update the script to find all relevant
files, even if a new split occurs in the future.
In addition, this patch carries additional minor changes:
- Fix some minor shellcheck issues found with a newer shellcheck version
than in the container (quoting, test -a -> -e)
- Move from wget to curl, as I had some issue on my home connection
with the former (possibly ipv6 related or something)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Signed-off-by: Arnout Vandecappelle <arnout@rnout.be>
(cherry picked from commit da01b7271a)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
uutils install 0.8.0 as used in Ubuntu 26.04 has a bug in the install
applet, breaking a number of packages:
https://github.com/uutils/coreutils/pull/11505https://bugs.launchpad.net/ubuntu/+source/rust-coreutils/+bug/2151454
The fix has been merged upstream but not yet released or packaged in Ubuntu,
so detect and reject the buggy version and explain how to change to the
coreutils version. Once fixed the version output will hopefully change.
For simplicity, only check for the exact 0.8.0 version string. Hopefully
when it is fixed in Ubuntu, they also update the version string. Note
that earlier versions of uutils have the same issue of course, but those
versions were never the default "install" on Ubuntu (or anywhere else).
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Arnout Vandecappelle <arnout@rnout.be>
(cherry picked from commit 64cb69f155)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
This reverts commit 6b56e0b4f0.
commit 6b56e0b4f0 ("linux: disable -Werror") said we needed to disable
CONFIG_WERROR in the kernel because it failed to build the esp-hosted
kernel module. This was very much the hammer approach because a) it'd be
better to fix the warnings in the kernel module or b) disable the
kernel's CONFIG_WERROR *only* if esp-hosted is to be built, via the
ESP_HOSTED_LINUX_CONFIG_FIXUPS variable in
package/esp-hosted/esp-hosted.mk.
This is a simple revert without making any change to any kernel module
(including esp-hosted) because all upstream Buildroot defconfigs with a
kernel module enabled build fine today (albeit with warnings). The two
defconfigs that were pointed at by the now-reverted commit all have been
(updated for and) run on master without failing the build.
Suggested-by: Arnout Vandecappelle <arnout@rnout.be>
Signed-off-by: Quentin Schulz <quentin.schulz@cherry.de>
Signed-off-by: Arnout Vandecappelle <arnout@rnout.be>
(cherry picked from commit a966f5cde5)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Fixes build errors seen with Gitlab pipeline and this defconfig:
support/config-fragments/autobuild/bootlin-aarch64-glibc-old.config
https://gitlab.com/bkuhls/buildroot/-/jobs/14570080455
aarch64-linux-gcc.br_real: error: unrecognized command line option
‘-Wimplicit-fallthrough=5’; did you mean ‘-Wno-fallthrough’?
aarch64-linux-gcc.br_real: error: unrecognized command line option
‘-Wshadow=local’; did you mean ‘-Wshadow-ivar’?
Removing all the unrecognized command line options from CMakeLists.txt
showed more build errors during compile:
output/build/odhcp6c-24485bb4b35ab84c17c2e87bd561d026d4c15c00/src/odhcp6c.h:47:46:
error: expected ‘)’ before ‘__VA_OPT__’
#define error(fmt, ...) __iflog(LOG_ERR, fmt __VA_OPT__(, ) __VA_ARGS__)
Another pipeline with gcc 9 showed other errors:
https://gitlab.com/bkuhls/buildroot/-/jobs/14570080479
/builds/bkuhls/buildroot/br-test-pkg/br-arm-basic/build/odhcp6c-24485bb4b35ab84c17c2e87bd561d026d4c15c00/src/dhcpv6.c:1126:4:
error: a label can only be part of a statement and a declaration is not a statement
1126 | struct dhcpv6_auth *r = (void*)&odata[-DHCPV6_OPT_HDR_SIZE];
so we raise the minimum gcc version to 10.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Arnout Vandecappelle <arnout@rnout.be>
(cherry picked from commit f8ea3ae2a5)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Buildroot commit ce5d173571 removed
0001-fix-build-with-gcc-14.patch which the version bump to 1.4.3.
The patch originally not only fixed share/utf8.c but also
ogginfo/codec_skeleton.c. Upstream only applied the fix for
ogginfo/codec_skeleton.c:
68c5a33685
so the build still fails:
utf8.c: In function 'convert_buffer':
utf8.c:271:9: error: implicit declaration of function 'charset_convert'
[-Wimplicit-function-declaration]
271 | ret = charset_convert(fromcode, tocode, from, fromlen, to, tolen);
Added upstream commit to fix the problem.
Fixes:
https://autobuild.buildroot.net/results/135/13520e3cbf6b6d169985f4d1378c2a1baab852e0/
The build error can be reproduced by this minimal defconfig:
BR2_TOOLCHAIN_BUILDROOT_UCLIBC=y
BR2_PACKAGE_VORBIS_TOOLS=y
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Arnout Vandecappelle <arnout@rnout.be>
(cherry picked from commit 8b79a9fec5)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
For host packages, this commit adds the same `ignore_cves` list as their
target counterpart and make it available from the `show-info` output.
When generating a CycloneDX SBOM with `make show-info-all |
utils/generate-cyclonedx` and running an analysis over it with
`support/script/cve-check`, multiple vulnerabilities
entries would be created with different analysis for packages that have
both a host and target variant that include IGNORE_CVES entries.
This is the case for the grub2 package that include ignored
vulnerabilities that patch both the target and host package but aren't
declared as ignored for the host package. This resulted in
vulnerabilities marked as 'exploitable' for the host variant while it
is patched.
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 7a5817137b)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Before this commit, only one entry per vulnerability ID was added to the
output. In CycloneDX, if you need to provide different analyses for
different affected components with the same vulnerability ID, you must
create multiple entries with the same ID.
When running `cve-check` with the `--include-resolved` argument, the
analysis of some vulnerabilities would get overwritten, which led to
undefined analysis results.
This is especially true when running the analysis on multiple components
with the same name but different versions. For instance, if the input
SBOM includes both the `gnupg` and `gnupg2` packages, CVE-2025-68973
could be included. This CVE might be exploitable for the `gnupg` package
but resolved for `gnupg2`. Therefore, a single analysis entry cannot
cover both cases.
This commit fixes the logic for adding vulnerabilities to the output
SBOM. A vulnerability is now added as a new entry if:
1. A vulnerability with the same ID doesn't exist yet.
2. The affect of the new vulnerability is not the same as the one
already present.
For the CVE-2025-68973 example this would result in the following
output:
```json
[
{
"id": "CVE-2025-68973",
"analysis": {
"state": "exploitable"
}
"affects": [
{"ref": "gnupg"}
]
},
{
"id": "CVE-2025-68973",
"analysis": {
"state": "resolved"
}
"affects": [
{"ref": "gnupg2"}
]
}
]
```
45 vulnerabilities were concerned by this bug over the Buildroot tree.
Co-Authored-By: Tim Soubry <tim.soubry@mind.be>
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit d4ff747a2b)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The 'bom-ref' are optional and since we don't reference the
vulnerabilities from anywhere else in the SBOM they are not necessary in
this case.
In the following commit, cve-check will potentially emit multiple
vulnerabilities that have the same id. So using the vulnerability id
as 'bom-ref' won't be correct as the 'bom-ref' needs to be unique
unlike the id property.
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit af55c1a39b)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Always run this script from the output of 'generate-cyclonedx'. Do not re-run
this script over an already analysed SBOM.
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 646356162b)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Normalize vulnerability timestamps to RFC 3339 format with explicit UTC
timezone suffix for CycloneDX 1.6 compliance.
This fixes validation errors in sbom-utility and makes the generated
SBOM with vulnerabilities compatible with DependencyTrack VEX parsers.
The NVD JSON data feeds provide timestamps in ISO 8601 format without timezone
information (e.g., "1999-01-01T05:00:00.000"), but CycloneDX 1.6 requires
RFC 3339 format with explicit timezone designation (e.g.,
"1999-01-01T05:00:00.000Z").
Add nvd_datetime_to_rfc3339() helper function to convert timestamps before
serialization.
Validation results:
Before fix:
$ sbom-utility validate -i cve/cve_report_current.json
[INFO] BOM valid against JSON schema: 'false'
[INFO] (234) schema errors detected.
Error example:
{
"type": "format",
"field": "vulnerabilities.0.updated",
"context": "(root).vulnerabilities.0.updated",
"description": "Does not match format 'date-time'",
"value": "2025-04-03T01:03:51.193"
}
After fix:
$ sbom-utility validate -i cve/cve_report_update.json
[INFO] BOM valid against JSON schema: 'true'
Tested-with: sbom-utility v0.18.1
Co-authored-by: Fabien Lehoussel <fabien.lehoussel@smile.fr>
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit e46783d3a0)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Replace Gentoo patch from 2018 with rebased version of an upstream
commit from 2022 which also fixes build errors introduced by gcc 14.x:
access/rdp.c: In function 'postConnectHandler':
access/rdp.c:238:39: error: assignment to 'pDesktopResize' {aka 'int
(*)(struct rdp_context *)'} from incompatible pointer type 'void
(*)(rdpContext *)' {aka 'void (*)(struct rdp_context *)'}
[-Wincompatible-pointer-types]
238 | p_instance->update->DesktopResize = desktopResizeHandler;
| ^
No autobuilder error was recorded, the build error can be reproduced
with this defconfig:
BR2_x86_64=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_PACKAGE_VLC=y
BR2_PACKAGE_FREERDP=y
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 9097605d29)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Commit 324612d68e fixed several compiler warnings,
but actually introduced a new one by increasing the buffer size in confdata.c that gets passed
along to file_write_dep in util.c, because buf2's size wasn't increased along with it.
./util.c: In function ‘file_write_dep’:
./util.c:86:26: warning: ‘%s’ directive writing 10 or more bytes into a region of size between 1 and 4097 [-Wformat-overflow=]
86 | sprintf(buf2, "%s%s", dir, name);
| ^~
./util.c:86:9: note: ‘sprintf’ output 11 or more bytes (assuming 4107) into a destination of size 4097
86 | sprintf(buf2, "%s%s", dir, name);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Fix this by increasing the size of buf2 to match the passed buffer size.
Signed-off-by: Devreese Jorik <jorik.devreese@barco.com>
Signed-off-by: Thomas Devoogdt <thomas.devoogdt@barco.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 1f5095c1d3)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The commit [1] introduced fixes for the following CVEs:
- CVE-2017-8372: The mad_layer_III function in layer3.c in Underbit MAD
libmad 0.15.1b, if NDEBUG is omitted, allows remote attackers to cause
a denial of service (assertion failure and application exit) via a
crafted audio file.
- CVE-2017-8373: The mad_layer_III function in layer3.c in Underbit MAD
libmad 0.15.1b allows remote attackers to cause a denial of service
(heap-based buffer overflow and application crash) or possibly have
unspecified other impact via a crafted audio file.
- CVE-2017-8374: The mad_bit_skip function in bit.c in Underbit MAD
libmad 0.15.1b allows remote attackers to cause a denial of service
(heap-based buffer over-read and application crash) via a crafted
audio file.
In commit [2], the patches ended up not being applied anymore because
the APPLY_PATCHES step was called before the patch content exists.
This commit import the fixes in Buildroot.
[1] 858df3643f package/libmad: switch to debian to fix CVEs
[2] b21184a877 package/libmad: update the patches to be applied with fuzz 0
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 9c9c531d0b)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
In Buildroot there are multiple way to apply patches on a package [1]
- Adding `.patch` file in the package directory.
- Define `<pkg>_PATCH` variable with the location of the patch tar.gz.
It used to download Debian patches tarball.
- Implement custom patching logic with `PRE`/`POST` patches hooks.
The libmad package is downloading a diff file from the Debian
mirror with the `<pkg>_PATCH` method [2] and then apply that diff to
create a directory containing patches. The patches are then applied in
the `PRE_PATCH_HOOK`.
The Debian patches were integrated in commit [4], in commit [5] the
application of the patches was moved to the PRE_PATCH_HOOK.
The problem is that in the PRE_PATCH step the `_PATCH` downloaded from
Debian don't exist yet and end up not being applied at all.
Since this is not used, remove the Debian patches.
[1] https://buildroot.org/downloads/manual/manual.html#patch-policy
[2] http://snapshot.debian.org/archive/debian/20190310T213528Z/pool/main/libm/libmad/libmad_0.15.1b-10.diff.gz
[3] 1167d0ff3d docs/manual: mention CVE trailer
[4] 858df3643f package/libmad: switch to debian to fix CVEs
[5] b21184a877 package/libmad: update the patches to be applied with fuzz 0
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 103f84b30f)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
In Buildroot there are multiple way to apply patches on a package [1]
- Adding `.patch` file in the package directory.
- Define `<pkg>_PATCH` variable with the location of the patch tar.gz.
It used to download Debian patches tarball.
- Implement custom patching logic with `PRE`/`POST` patches hooks.
To make the CycloneDX SBOM generation not dependant on downloading the
packages, the two last options have the downside of not appearing on the
generated SBOM.
The heirloom-mailx package is downloading a tarball from the Debian
mirror with the `<pkg>_PATCH` method [2].
To improve the tracking of the patched vulnerabilities for the
heirloom-mailx package this commit import the patches previously
downloaded with the `_PATCH` variable in the Buildroot tree. This allows
to add the `CVE:` trailer [3] on the patches that fix vulnerabilities to
better track which patch is fixing the vulnerability.
[1] https://buildroot.org/downloads/manual/manual.html#patch-policy
[2] http://snapshot.debian.org/archive/debian/20150815T155609Z/pool/main/h/heirloom-mailx/heirloom-mailx_12.5-5.debian.tar.xz
[3] 1167d0ff3d docs/manual: mention CVE trailer
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit c9659fd9e8)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
In Buildroot there are multiple way to apply patches on a package [1]
- Adding `.patch` file in the package directory.
- Define `<pkg>_PATCH` variable with the location of the patch tar.gz.
It used to download Debian patches tarball.
- Implement custom patching logic with `PRE`/`POST` patches hooks.
To make the CycloneDX SBOM generation not dependant on downloading the
packages, the two last options have the downside of not appearing on the
generated SBOM.
The unzip package is downloading a tarball from the Debian mirror with
the `<pkg>_PATCH` method [2].
To improve the tracking of the patched vulnerabilities for the unzip
package this commit import the patches previously downloaded with the
`_PATCH` variable in the Buildroot tree.
This allows to add the `CVE:` trailer [3] on the patches that fix
vulnerabilities to better track which patch is fixing the vulnerability.
[1] https://buildroot.org/downloads/manual/manual.html#patch-policy
[2] https://snapshot.debian.org/archive/debian/20250311T215724Z/pool/main/u/unzip/unzip_6.0-29.debian.tar.xz
[3] 1167d0ff3d docs/manual: mention CVE trailer
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit fb8958e3dc)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
See here for a changelog:
https://downloads.asterisk.org/pub/telephony/asterisk/ChangeLog-22-current.html
Pjsip was updated to 2.16.
Fixes some security issues in pjsip:
Author: Mike Bradeen Date: 2026-03-23
Address the following pjproject security vulnerabilities
GHSA-j29p-pvh2-pvqp - Buffer overflow in ICE with long username
GHSA-8fj4-fv9f-hjpc - Heap use-after-free in PJSIP presense subscription termination header
GHSA-g88q-c2hm-q7p7 - ICE session use-after-free race conditions
GHSA-x5pq-qrp4-fmrj - Out-of-bounds read in SIP multipart parsing
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 6add6e3cd8)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
In commit 770f915497, libjwt was bumped
from 1.17.2 to 3.2.0, which is a big bump. And the libjwt website
states "Version 3 of LibJWT is a complete overhaul of the code. Please
see documentation for usage."
And indeed, this version bump breaks the build of Asterisk:
res_stir_shaken/attestation.c: In function 'pack_payload':
res_stir_shaken/attestation.c:357:9: error: implicit declaration of function 'jwt_add_grants_json' [-Wimplicit-function-declaration]
357 | jwt_add_grants_json(jwt, payload_str);
| ^~~~~~~~~~~~~~~~~~~
In file included from /home/autobuild/autobuild/instance-11/output-1/build/asterisk-22.3.0/include/asterisk/module.h:39,
from res_stir_shaken/attestation.c:24:
res_stir_shaken/attestation.c: In function '_dtor_jwt':
res_stir_shaken/attestation.c:367:38: error: implicit declaration of function 'jwt_free'; did you mean 'jwks_free'? [-Wimplicit-function-declaration]
367 | RAII_VAR(jwt_t *, jwt, NULL, jwt_free);
| ^~~~~~~~
To fix this, let's use the bundled version of libjwt in Asterisk,
since even Asterisk master hasn't been fixed to be compatible with
libjwt 3.x.
Of course, this raises the question of whether the separate libjwt
package should be removed, as it is anyway only used by Asterisk.
Fixes:
http://autobuild.buildroot.net/results/2d39791c65c5b71b0533d4b2bce5f073f5919552/
Cc: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
[Bernd: Added libopenssl dependency needed by bundled libjwt]
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 61271a6c55)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Buildroot commit d4b8a42fb2 bumped the
package from 1.0.16 to 1.0.18.
Upstream added optional neon support in version 1.0.17:
f22cf00a86
The configure check detects neon support:
-- Performing Test HAVE_NEON
-- Performing Test HAVE_NEON - Success
but causes a build error at linking stage:
[100%] Linking CXX executable dec265
/builds/bkuhls/buildroot/br-test-pkg/bootlin-armv7m-uclibc/host/opt/ext-toolchain/arm-buildroot-uclinux-uclibcgnueabi/bin/ld.real:
error: ../libde265/libde265.a(hevcdsp_qpel_neon.S.o):
conflicting architecture profiles A/M
so we add a configure option to disable neon support.
The build error was not yet detected by the autobuilders but can be
reproduced using this defconfig:
BR2_arm=y
BR2_cortex_m4=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_ARMV7M_UCLIBC_STABLE=y
BR2_PACKAGE_LIBDE265=y
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit f51769b7e9)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The current version was more than 3 years old and had known CVEs (see
CVE-2023-36328). Since this is a host only package, we don't consider
this as a security bump.
This upgrade allows the removal of the two patches currently applied:
- The commit 1b57b62, which fixes the build with autoconf 2.72, is
already included upstream.
- We no longer need to use the JSON-PP module, as it is now
optional (commit 13d3bcf).
Also, since the latest version of Heimdal no longer depends on
e2fsprogs, the host-e2fsprogs dependency has been removed.
Signed-off-by: Guillaume Chaye <guillaume.chaye@zeetim.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 573ecbd44c)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
HEIMDAL_INSTALL_STAGING = YES makes no sense since the package was
introduced in 56258f491b ("heimdal: new
package") since it's a host only package.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit d2d5014a8e)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
wpewebkit depends directly on OpenGL ES and EGL support, but those
dependencies are not imposed by libepoxy itself. Update the inline
comments so they describe the actual dependency source before relaxing
libepoxy's package-level EGL/GL dependency.
Signed-off-by: Joseph Kogut <joseph.kogut@gmail.com>
Signed-off-by: Arnout Vandecappelle <arnout@rnout.be>
(cherry picked from commit a4b761f395)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
For more information about the release, see:
- https://www.postgresql.org/docs/17/release-17-9.html
- https://www.postgresql.org/docs/17/release-17-10.html
Fixes the following vulnerabilities:
- CVE-2026-6479:
Prevent unbounded recursion while processing startup packets
A malicious client could crash the connected backend by alternating
rejected SSL and GSS encryption requests indefinitely.
- CVE-2026-6473
Fix assorted integer overflows in memory-allocation calculations
Various places were incautious about the possibility of integer overflow
in calculations of how much memory to allocate. Overflow would lead to
allocating a too-small buffer which the caller would then write past the
end of. This would at least trigger server crashes, and probably could
be exploited for arbitrary code execution. In many but by no means all
cases, the hazard exists only in 32-bit builds.
- CVE-2026-6476
Properly quote subscription names in pg_createsubscriber
The given subscription name was inserted into SQL commands without
quoting, so that SQL injection could be achieved in the (perhaps
unlikely) case that the subscription name comes from an untrusted
source.
- CVE-2026-6638
Properly quote object names in logical replication origin checks
ALTER SUBSCRIPTION ... REFRESH PUBLICATION interpolated schema and
relation names into SQL commands without quoting them, allowing
execution of arbitrary SQL on the publisher.
- CVE-2026-6473
Reject over-length options in ts_headline()
The StartSel, StopSel and FragmentDelimiter strings must not exceed 32Kb
in length, but this was not checked for. An over-length value would
typically crash the server.
- CVE-2026-6474
Guard against malicious time zone names in timeofday() and pg_strftime()
A crafted time zone setting could pass % sequences to snprintf(),
potentially causing crashes or disclosure of server memory. Another path
to similar results was to overflow the limited-size output buffer used
by pg_strftime().
- CVE-2026-6472
When creating a multirange type, ensure the user has CREATE privilege on
the schema specified for the multirange type.
The multirange type can be put into a different schema than its parent
range type, but we neglected to apply the required privilege check when
doing so.
- CVE-2026-6478
Use timing-safe string comparisons in authentication code.
Use timingsafe_bcmp() instead of memcpy() or strcmp() when checking
passwords, hashes, etc. It is not known whether the data dependency of
those functions is usefully exploitable in any of these places, but in
the interests of safety, replace them.
- CVE-2026-6477
Mark PQfn() as unsafe, and avoid using it within libpq
For a non-integral result type, PQfn() is not passed the size of the
output buffer, so it cannot check that the data returned by the server
will fit. A malicious server could therefore overwrite client memory.
This is unfixable without an API change, so mark the function as
deprecated. Internally to libpq, use a variant version that can apply
the missing check.
- CVE-2026-6475
Prevent path traversal in pg_basebackup and pg_rewind
These applications failed to validate output file paths read from their
input, so that a malicious source could overwrite any file writable by
these applications. Constrain where data can be written by rejecting
paths that are absolute or contain parent-directory references.
- CVE-2026-6473
Guard against field overflow within contrib/intarray's query_int type
and contrib/ltree's ltxtquery type.
Parsing of these query structures did not check for overflow of 16-bit
fields, so that construction of an invalid query tree was possible.
This can crash the server when executing the query.
- CVE-2026-6473
Guard against overly long values of contrib/ltree's lquery type.
Values with more than 64K items caused internal overflows, potentially
resulting in stack smashes or wrong answers.
- CVE-2026-6637
Prevent SQL injection and buffer overruns in contrib/spi.
check_foreign_key() was insufficiently careful about quoting key
values, and also used fixed-length buffers for constructing queries.
While this module is only meant as example code, it still shouldn't
contain such dangerous errors.
(cherry picked from commit 9fb64dfc24)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
