To indicate that a patch fixes a vulnerability in Buildroot, the convention is:
1. In the patch file, add a tag 'CVE: <cve id>'
2. In <pkg>.mk, and an entry to <PKG>_IGNORE_CVES, and add a comment above
that new entry to reference the patch file(s)
However, as packages get bumped and their patches are added, removed or
rebased; it happens that IGNORE_CVES get outdated. One important issue is
marking a CVE as ignored, while the corresponding patch is not in Buildroot.
To detect such cases, add a new checker to checkpackagelib that finds
occurences of:
# 000x-some-patch.patch
PKG_IGNORE_CVES += CVE-XXXX-YYYY
For each one of them, ensure that the mentioned patch files actually exist
and contain the `CVE: ...` tag.
Assisted-by: Claude:claude-opus-4.8
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit b00ac4e346)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Prior to improving check-package to verify that the comment preceding
a <pkg>_IGNORE_CVES entry mentions an existing patch, and that the
patch itself contains a CVE: tag, we fix all problematic cases that
currently exist in Buildroot:
- In the case of binutils: the CVE was only applicable to binutils
2.43/2.44, and the oldest version now supported is 2.45, so the
patch doesn't exist anymore in Buildroot
- For x11vnc, fix a typo in the patch name
- For gpsd the patches were dropped in [1] as they are included in the
version bump
- Similarly for micropython, the patches were dropped in [2] along with
the version bump
- For util-linux, strip the prefix "package/util-linux/", so that the patch
is relative to the .mk file and can be found by the new check
- Add missing 'CVE:' tag to net-tools patch 0001
[1] 37ef4f862f package/gpsd: bump version to 3.27.2
[2] 28eeca9a98 package/micropython: bump to version 1.28.0
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 636f69ab45)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
https://github.com/PCRE2Project/pcre2/releases/tag/pcre2-10.48
Fixes the following security issues:
(Security fix for specific API usage, GHSA-2p8c-ff85-vh9x)
If pcre2_jit_compile() is called with options for some match modes, and
then pcre2_match() is used to perform a match for a different match
mode, an out-of-bounds read can occur if the match is attempted against
invalid UTF input.
(Security fix for pattern conversion, GHSA-q8g2-wprr-34m9)
If pcre2_convert() is called on untrusted input on platforms with
32-bit size_t, an out-of-bounds heap write can occur.
(Security fix, GHSA-3r4p-g7gg-ppmf) Fixed an out-of-bounds write in DFA
matching when using a heap limit; also fixed possible integer overflows
which could cause under-allocation of the workspace.
(Security fix, GHSA-fmgr-6ggq-9859) Added bounds checks for several
integer overflows while compiling patterns on 32-bit CPUs, which could
cause under-allocation followed by out-of-bounds writes.
(Security fix, GHSA-9qww-pwc4-77qq) Applied lower buffer bound to
prevent two out-of-bounds reads while scanning backwards through
invalid UTF data with PCRE2_MATCH_INVALID_UTF.
(Security fix for specific API usage, #937) Fixed a leak and later
invalid free when calling the fast-path pcre2_jit_match() function with
a match data object previously used with pcre2_match() and
PCRE2_COPY_MATCHED_SUBJECT.
(Low-severity security fix, GHSA-q7rw-r7qq-2hx6) Fixed exposure of two
uninitialised bytes from malloc() via pcre2_serialize_encode().
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 664db5d62c)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
https://github.com/kernelsauce/turbo/releases/tag/v2.1.5https://github.com/kernelsauce/turbo/releases/tag/v2.1.4
Security fixes:
HTTP header injection: header values were only checked for a literal
\r\n, so a lone \r or \n could still split a header. Now rejected on
either character.
Transfer-Encoding requests are now rejected with 501 instead of silently
mishandled, closing a request smuggling avenue.
A real default request body size cap (128 MB) with a 413 response,
previously unbounded.
Secure cookie signature now binds the cookie name, so a value signed for
one cookie can no longer be replayed under a different name.
Verification failures return the default value instead of raising.
Constant-time comparison for the secure cookie HMAC, previously a
timing-leaky ==.
util.secure_random_bytes reads real OS entropy (/dev/urandom,
BCryptGenRandom on Windows) for WebSocket masks and util.rand_str,
previously math.random.
WebSocket: unmasked client frames are rejected per RFC 6455, and
fragmented message reassembly is capped to max_buffer_size to close a
memory exhaustion path.
StaticFileHandler decodes the request path before the traversal check,
closing a bypass.
Fixed a 32-byte-per-malformed-request memory leak in the C header parser
wrapper (found via libFuzzer).
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 15a422cee1)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Upstream started using std::string_view
https://github.com/search?q=repo%3APurpleI2P%2Fi2pd+string_view&type=commits&s=committer-date&o=asc
with commit
a3e0b3710c
first released in version 2.54.0 which was added to buildroot with
commit dea4f02bbb.
Building the package with the gcc6-based defconfig
bootlin-aarch64-glibc-old is broken:
/builds/bkuhls/buildroot/br-test-pkg/bootlin-aarch64-glibc-old/build/i2pd-2.61.0/libi2pd/Base.h:14:23:
fatal error: string_view: No such file or directory
BR2_TOOLCHAIN_HAS_GCC_BUG_64735 can be removed as well now as it depends
on gcc < 7.
A backport to LTS branches should be considered.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit a6bb4d52c0)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
https://github.com/the-tcpdump-group/libpcap/blob/libpcap-1.10.7/CHANGES
Fixes the following CVEs:
CVE-2026-0799: Access M[] safely in the BPF interpreter.
CVE-2026-31912: Mind the program bounds in pcap_offline_filter().
CVE-2026-31911: Fail opcodes safely in the BPF interpreter.
CVE-2026-6244: Avoid division by zero via pcap_offline_filter().
CVE-2026-6554: Limit "ja L" looping in pcap_offline_filter().
CVE-2026-18313: Fix a memory leak in rpcapd.
CVE-2026-18238: Fix RPCAP_MSG_PACKET validation.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 446c0f85b8)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
igh-ethercat is failing while attempting to apply patches,
with error:
Applying 0002-Linux-6.19.0-support.patch using patch:
patching file devices/generic.c
Reversed (or previously applied) patch detected! Skipping patch.
1 out of 1 hunk ignored -- saving rejects to file devices/generic.c.rej
The package patch 0002 was added in [1] in branch "master" while it
was in release client cycle. It was cherry-picked in [2] in branch
"next" to apply the bump [3] (which removes the package patches 0001
and 0002). When the branch "next" was merged in "master" in commit [4],
the patch 0002 was kept.
This commit removes this stale patch.
[1] e4cf512c39
[2] 8a5fc970b4
[3] 0a91e760f4
[4] 5f26877955
Fixes:
- https://autobuild.buildroot.org/results/4f509f1f788c1b8dc5a840ffa2e435c5ed7b6eea/
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit d071817969)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
https://github.com/strukturag/libde265/releases/tag/v1.1.2
Security fixes:
(CVE numbers will be added when assigned.)
CVE-2026-XXXXX (GHSA-xp3h-6f5r-8cxp) Heap use-after-free and double free
in multi-threaded (WPP) decoding. A crafted stream whose slice segments
repeat or rewind their slice_segment_address within a picture re-ran
CTB rows that were already marked finished, so the CABAC context handoff
between rows was no longer ordered and the shared context table was
released twice. Slice segments that do not follow the previous one in
tile-scan order are now rejected with the new warning
DE265_WARNING_SLICE_SEGMENT_ADDRESS_NOT_INCREASING, and the WPP row
progress is reset for each slice segment. (medium)
CVE-2026-XXXXX (GHSA-mm7m-v26f-wf8x) Heap use-after-free after
de265_reset(): the pointer to the previous slice header was left
dangling when the DPB was cleared, and a dependent slice pushed after
the reset copied from freed memory. (medium)
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit e55cb31085)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
https://github.com/strukturag/libheif/releases/tag/v1.23.3
Fixes the following CVEs:
(CVE numbers will be added when assigned.)
CVE-2026-XXXXX (GHSA-x8r2-mggj-j6wr) Heap buffer overflow (write) in the
uncompressed (unci) mixed-interleave decoder when the two chroma
components declare different bit depths. Both the written bytes and the
overflow length are controlled by the file. (critical)
CVE-2026-XXXXX (GHSA-8fmq-r4pf-7m57) Permanent decoder deadlock through
a reference cycle between an image and its alpha auxiliary image. The
alpha edge was not covered by the cycle guard and re-entered a held
mutex. (high)
CVE-2026-XXXXX (GHSA-w7mc-p8jc-p853) Heap out-of-bounds read in the
YCbCr 4:2:0 to 16-bit interleaved RGB conversion when the chroma
planes have a lower bit depth than luma. Heap memory could end up in
the decoded image. YCbCr conversions with mismatched luma and chroma
bit depths are now rejected. (high)
CVE-2026-XXXXX (GHSA-4jqm-2x34-6f6r) Heap buffer overflow in the SVT-AV1
encoder plugin when encoding a high-bit-depth alpha channel, and a
double free on its send-picture error path. (high)
CVE-2026-84451 (GHSA-hh47-fhqr-cj2r) Incomplete fix for
GHSA-73p7-m7gg-w2jv: the tile range check of the unci decoder (without
icef) could still overflow, allowing an out-of-bounds read. (medium)
CVE-2026-XXXXX (GHSA-4h82-g446-83fm) Heap out-of-bounds read when
converting odd-height 4:2:0 frames of an uncompressed (uncv) image
sequence to RGB. (medium)
CVE-2026-XXXXX (GHSA-9rj8-5mp5-26c9) Out-of-bounds read in the RGB to
YCbCr identity-matrix color conversion when the R, G, and B planes
have different bit depths. (medium)
CVE-2026-84450 (GHSA-gh5q-69gg-c964) A clap property combined with an
oversized ispe reached an assert() in the Fraction arithmetic and
aborted the process (incomplete fix for GHSA-jc8f-p23p-5hjg). An error
is returned instead. (medium)
(GHSA-mw6f-29j3-76f4) Several smaller findings:
heif_image_handle_get_depth_image_handle() and
heif_image_handle_get_depth_image_representation_info() dereferenced a
null pointer on files without a depth image; the TIFF input decoder of
the example tools had an unbounded EXIF tag allocation and a division
by zero on zero YCbCr subsampling; assert()s in the PNG input decoder
are now error returns; integer overflow in the Go binding's
ImageAccess.GetPlane(); heif-view now verifies the decoded frame size
before display. (medium)
(GHSA-8857-r8x5-7499) Undefined behavior (negative shift) in the HDR
bit-depth up-conversion for target bit depths above 16. Such
conversions are now rejected. (low)
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit d148168e20)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
Buildroot commit 9a43bf6593 bumped the gcc
dependency from 9 to 10 but forgot to propagate this change to the
libcamera-apps package.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 64ed9e6c43)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
Buildroot 262a7f6d2f added the package but
forgot to provide the hash for LICENSE.GPL3-EXCEPT, instead a hash for
a non-existing file was added to qt5knx.hash.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit d2b7199dea)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
Regenerate the Bootlin toolchain Kconfig and test configurations using
support/scripts/gen-bootlin-toolchains.
This adds BR2_USE_MMU to the affected uClibc entries and to the
architecture support conditions, and updates the generated tests.
The glibc and musl changes only reorder their existing BR2_USE_MMU
dependencies.
Signed-off-by: Fengwei Tan <tfx2001@outlook.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 9c6eed9ec0)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
The Bootlin uClibc toolchains for m68k-68xxx, riscv32-ilp32d, and
xtensa-lx60 require an MMU. However, their generated Kconfig entries
lack a BR2_USE_MMU dependency, allowing them to be selected for noMMU
configurations. External toolchain validation then fails with:
MMU support available in C library, please enable BR2_USE_MMU
Add the missing BR2_USE_MMU dependencies for these architectures to
prevent them from being selected for noMMU targets.
Signed-off-by: Fengwei Tan <tfx2001@outlook.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 3469c6793c)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
Regenerate the Bootlin toolchain Kconfig file with
support/scripts/gen-bootlin-toolchains to remove duplicate
BR2_TOOLCHAIN_HAS_THREADS selections.
Commit 184d47a7ad ("support/scripts/gen-bootlin-toolchains: add new
script to support Bootlin toolchains") initially introduced this issue.
Although commit a33e1af4a0 ("support/scripts/gen-bootlin-toolchains:
avoid selecting _HAS_THREADS multiple times") fixed the generator script,
the Config.in.options file was not regenerated accordingly.
This is a non-functional cleanup, as repeated Kconfig select statements
are harmless.
Signed-off-by: Fengwei Tan <tfx2001@outlook.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 9556895e78)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
dtui depends on tui-textarea which unconditionally imports AtomicU64 in
src/widget.rs to pack a viewport rectangle into a single atomic word:
use std::sync::atomic::{AtomicU64, Ordering};
pub struct Viewport(AtomicU64);
As there is no cfg(target_has_atomic) guard in tui-textarea, its
build fails on any target for which rustc does not provide 64-bit
atomics with:
Compiling tui-textarea v0.7.0
error[E0432]: unresolved import `std::sync::atomic::AtomicU64`
--> .../dtui-3.0.0/VENDOR/tui-textarea/src/widget.rs:10:25
|
10 | use std::sync::atomic::{AtomicU64, Ordering};
| ^^^^^^^^^ no `AtomicU64` in `sync::atomic`
|
help: a similar name exists in the module
|
10 - use std::sync::atomic::{AtomicU64, Ordering};
10 + use std::sync::atomic::{AtomicU32, Ordering};
This has been reported to tui-textarea upstream, but unfortunately the
project seems to be unmaintained (issue linked below). A sane workaround
is to disable the package on targets which lack 64-bit atomic support,
which is exactly what BR2_PACKAGE_HOST_RUSTC_TARGET_HAS_ATOMIC_U64
describes: it is n for armv5te-unknown-linux-{gnu,musl}eabi and
powerpc-unknown-linux-gnu, the only rust targets Buildroot can generate
which lack 64-bit atomics, and y everywhere else.
The same problem was hit by package/dust and worked around in commit
3abc3b97ba ("package/dust: bump to version 1.1.2") by bumping to a
version in which upstream had added the missing guard. That is not an
option here as tui-textarea 0.7.0 is the latest release.
Note that a runtime test for dtui cannot use the default
infra.basetest.BASIC_TOOLCHAIN_CONFIG, since that builds with
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_ARMV5_EABI_GLIBC_STABLE, where dtui is now
disabled; such a test would need an armv7 or aarch64 toolchain instead.
Build tested with utils/test-pkg against:
- BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_ARMV5_EABI_GLIBC_STABLE
- BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_ARMV5_EABI_MUSL_STABLE
- BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_POWERPC_E500MC_GLIBC_STABLE
all three fail with the above error before this change and are skipped
after it, while armv7 (glibc and musl), aarch64, powerpc64le and x86-64
still select and build the package.
Link: https://github.com/rhysd/tui-textarea/issues/66
Fixes: https://autobuild.buildroot.org/results/188f6442371500731453f75983590c922eab6d57
Fixes: https://autobuild.buildroot.org/results/e254db2654f18f1d2110eb8b1a32b43ad0f2a3d6
Signed-off-by: Christopher Obbard <chris.obbard@oss.qualcomm.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 99529edaef)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
Rust does not provide 64-bit atomics on every target Buildroot can
generate. rustc sets max_atomic_width = 32 for three of the 25 targets
listed in RUST_TARGETS in utils/update-rust, so
core::sync::atomic::AtomicU64 and AtomicI64 simply do not exist there:
$ rustc --print cfg --target <target> | grep target_has_atomic
armv5te-unknown-linux-gnueabi "16" "32" "8" "ptr"
armv5te-unknown-linux-musleabi "16" "32" "8" "ptr"
powerpc-unknown-linux-gnu "16" "32" "8" "ptr"
Every other supported target, including armv6, armv7, aarch64, all the
x86 variants, riscv64, s390x, sparc64 and both 64-bit powerpcs, has
them, e.g.:
arm-unknown-linux-gnueabi "16" "32" "64" "8" "ptr"
armv7-unknown-linux-gnueabihf "16" "32" "64" "8" "ptr"
A crate that uses 64-bit atomics without a cfg(target_has_atomic = "64")
guard therefore fails to build on those three targets with:
error[E0432]: unresolved import `std::sync::atomic::AtomicU64`
|
| atomic::{AtomicU64, AtomicU8, AtomicUsize, Ordering},
| ^^^^^^^^^ no `AtomicU64` in `sync::atomic`
This has been hit at least twice already: by package/dust, worked around
in commit 3abc3b97ba ("package/dust: bump to version 1.1.2") by moving
to a release in which upstream had added the guard and by package/dtui,
which has no such release available and had to open-code the affected
architectures instead.
It is likely to keep recurring: infra.basetest.BASIC_TOOLCHAIN_CONFIG
builds with BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_ARMV5_EABI_GLIBC_STABLE, so
every runtime test that does not override the toolchain compiles for
armv5te, one of the three affected targets. That is exactly how the two
failures above were found.
Add a hidden symbol so packages can express this constraint once, rather
than each open-coding BR2_ARM_CPU_ARMV5 and BR2_powerpc and needing to
update whenever rust gains or changes a target.
Note that armv5te and 32-bit powerpc are only supported by rust for
glibc and musl, so the uclibc variants of those architectures are
already excluded by BR2_PACKAGE_HOST_RUSTC_TARGET_ARCH_SUPPORTS.
Signed-off-by: Christopher Obbard <chris.obbard@oss.qualcomm.com>
Reviewed-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 698535473f)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
DEJAVU_LICENSE is primarily BitstreamVera. The license file also
specifies that DejaVu-specific changes and certain math extensions are
in the Public Domain. This matches the licensing logic used by
OpenEmbedded/Yocto.
Signed-off-by: Martin Bachmann <martin.bachmann@designwerk.com>
[Fiona: wrap lines in commit message]
Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
(cherry picked from commit df61b7e9bb)
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
Was missed when the download page was updated for 2026.08-rc3 in commit
e6b06b8d9c ("Update for 2026.08-rc3").
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
See the changelogs:
- https://www.erlang.org/patches/OTP-26.2.5.16
- https://www.erlang.org/patches/OTP-26.2.5.17
- https://www.erlang.org/patches/OTP-26.2.5.18
- https://www.erlang.org/patches/OTP-26.2.5.19
- https://www.erlang.org/patches/OTP-26.2.5.20
- https://www.erlang.org/patches/OTP-26.2.5.21
This fixes the following vulnerabilities:
- CVE-2026-21620:
Relative Path Traversal, Improper Isolation or Compartmentalization
vulnerability in erlang otp erlang/otp (tftp_file modules), erlang otp
inets (tftp_file modules), erlang otp tftp (tftp_file modules) allows
Relative Path Traversal. This vulnerability is associated with program
files lib/tftp/src/tftp_file.erl, src/tftp_file.erl.
For more information, see:
- https://www.cve.org/CVERecord?id=CVE-2026-21620
- CVE-2026-23941:
Inconsistent Interpretation of HTTP Requests ('HTTP Request
Smuggling') vulnerability in Erlang OTP (inets httpd module) allows
HTTP Request Smuggling. This vulnerability is associated with program
files lib/inets/src/http_server/httpd_request.erl and program routines
httpd_request:parse_headers/7. The server does not reject or
normalize duplicate Content-Length headers. The earliest Content-
Length in the request is used for body parsing while common reverse
proxies (nginx, Apache httpd, Envoy) honor the last Content-Length
value. This violates RFC 9112 Section 6.3 and allows front-end/back-
end desynchronization, leaving attacker-controlled bytes queued as the
start of the next request.
For more information, see:
- https://www.cve.org/CVERecord?id=CVE-2026-23941
- CVE-2026-23942:
Improper Limitation of a Pathname to a Restricted Directory ('Path
Traversal') vulnerability in Erlang OTP (ssh_sftpd module) allows Path
Traversal. This vulnerability is associated with program files
lib/ssh/src/ssh_sftpd.erl and program routines
ssh_sftpd:is_within_root/2. The SFTP server uses string prefix
matching via lists:prefix/2 rather than proper path component
validation when checking if a path is within the configured root
directory. This allows authenticated users to access sibling
directories that share a common name prefix with the configured root
directory. For example, if root is set to /home/user1, paths like
/home/user10 or /home/user1_backup would incorrectly be considered
within the root.
For more information, see:
- https://www.cve.org/CVERecord?id=CVE-2026-23942
- CVE-2026-23943:
Improper Handling of Highly Compressed Data (Compression Bomb)
vulnerability in Erlang OTP ssh (ssh_transport modules) allows Denial
of Service via Resource Depletion. The SSH transport layer advertises
legacy zlib compression by default and inflates attacker-controlled
payloads pre-authentication without any size limit, enabling reliable
memory exhaustion DoS. Two compression algorithms are affected: *
zlib: Activates immediately after key exchange, enabling
unauthenticated attacks * zlib@openssh.com: Activates post-
authentication, enabling authenticated attacks Each SSH packet can
decompress ~255 MB from 256 KB of wire data (1029:1 amplification
ratio). Multiple packets can rapidly exhaust available memory, causing
OOM kills in memory-constrained environments. This vulnerability is
associated with program files lib/ssh/src/ssh_transport.erl and
program routines ssh_transport:decompress/2,
ssh_transport:handle_packet_part/4.
For more information, see:
- https://www.cve.org/CVERecord?id=CVE-2026-23943
- CVE-2026-28810:
Generation of Predictable Numbers or Identifiers vulnerability in
Erlang/OTP kernel (inet_res, inet_db modules) allows DNS Cache
Poisoning. The built-in DNS resolver (inet_res) uses a sequential,
process-global 16-bit transaction ID for UDP queries and does not
implement source port randomization. Response validation relies almost
entirely on this ID, making DNS cache poisoning practical for an
attacker who can observe one query or predict the next ID. This
conflicts with RFC 5452 recommendations for mitigating forged DNS
answers. inet_res is intended for use in trusted network environments
and with trusted recursive resolvers. Earlier documentation did not
clearly state this deployment assumption, which could lead users to
deploy the resolver in environments where spoofed DNS responses are
possible. This vulnerability is associated with program files
lib/kernel/src/inet_db.erl and lib/kernel/src/inet_res.erl.
For more information, see:
- https://www.cve.org/CVERecord?id=CVE-2026-28810
- CVE-2026-32147:
Improper Limitation of a Pathname to a Restricted Directory ('Path
Traversal') vulnerability in Erlang OTP ssh (ssh_sftpd module) allows
an authenticated SFTP user to modify file attributes outside the
configured chroot directory. The SFTP daemon (ssh_sftpd) stores the
raw, user-supplied path in file handles instead of the chroot-resolved
path. When SSH_FXP_FSETSTAT is issued on such a handle, file
attributes (permissions, ownership, timestamps) are modified on the
real filesystem path, bypassing the root directory boundary entirely.
Any authenticated SFTP user on a server configured with the root
option can modify file attributes of files outside the intended chroot
boundary. The prerequisite is that a target file must exist on the
real filesystem at the same relative path. Note that this
vulnerability only allows modification of file attributes; file
contents cannot be read or altered through this attack vector. If the
SSH daemon runs as root, this enables direct privilege escalation: an
attacker can set the setuid bit on any binary, change ownership of
sensitive files, or make system configuration world-writable. This
vulnerability is associated with program files
lib/ssh/src/ssh_sftpd.erl and program routines ssh_sftpd:do_open/4 and
ssh_sftpd:handle_op/4.
For more information, see:
- https://www.cve.org/CVERecord?id=CVE-2026-32147
- CVE-2026-42789:
Improper Following of a Certificate's Chain of Trust vulnerability in
Erlang OTP public_key (pubkey_cert module) allows a non-CA certificate
to be accepted as an intermediate issuer, enabling certificate chain
forgery. In lib/public_key/src/pubkey_cert.erl,
pubkey_cert:validate_extensions/7 contains two flaws that together
allow a certificate with basicConstraints cA:false and no keyUsage
extension to be used as an intermediate issuer in a chain passed to
public_key:pkix_path_validation/3: the cA:false clause recurses into
the remaining extensions without rejecting the certificate when it is
in issuer position, and the keyUsage check only fires when the
extension is present, so a certificate lacking keyUsage entirely
bypasses the keyCertSign enforcement. Any party holding an end-entity
certificate with basicConstraints cA:false and no keyUsage extension,
issued by any CA in the victim's trust store, can use that
certificate's private key to sign forged leaf certificates for
arbitrary identities. public_key:pkix_path_validation/3 accepts the
resulting chain, and by extension every TLS or mTLS endpoint built on
the OTP ssl application that relies on the default verifier is
affected, including server identity verification on the client side
and client certificate verification on mTLS servers.
For more information, see:
- https://www.cve.org/CVERecord?id=CVE-2026-42789
- CVE-2026-42790:
Improper Certificate Validation vulnerability in Erlang OTP public_key
(pubkey_cert and public_key modules) allows a DNS nameConstraints
bypass via subject CommonName fallback in TLS hostname verification.
Two flaws combine to allow a subordinate CA whose DNS nameConstraints
are restricted (e.g. permitted;DNS:allowed.example.com) to issue a
leaf certificate that an OTP TLS client accepts as a valid identity
for an out-of-scope hostname (e.g. victim.example.com): First,
pubkey_cert:validate_names/6 in lib/public_key/src/pubkey_cert.erl
only checks SAN DNS entries against nameConstraints. Per RFC 5280, a
permitted DNS subtree only restricts certificates that contain a DNS-
typed name. A leaf with no subjectAltName therefore trivially
satisfies any permitted;DNS:... constraint regardless of its subject
commonName. Second, public_key:pkix_verify_hostname/3 in
lib/public_key/src/public_key.erl falls back to the subject commonName
when no subjectAltName is present, extracting id-at-commonName
attributes as presented IDs and matching them against the reference
hostname. The strict pkix_verify_hostname_match_fun(https) matcher
does not suppress this fallback. The result is that path validation
accepts a CN-only leaf under a DNS-constrained intermediate (no SAN
means the nameConstraints are not triggered), and hostname
verification then accepts it via the CN fallback. The bypass is
reachable from stock ssl:connect with verify_peer, a trusted CA, SNI,
and the canonical strict https hostname matcher.
For more information, see:
- https://www.cve.org/CVERecord?id=CVE-2026-42790
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This patch fixes a build error with OpenSSL-enabled libcurl which was
detected by the Gitlab pipelines:
checking for openssl options with pkg-config... found
configure: pkg-config: SSL_LIBS: "-lssl -lcrypto -pthread"
configure: pkg-config: SSL_LDFLAGS: "-L/builds/bkuhls/buildroot/br-test-pkg/bootlin-m68k-5208-uclibc/host/bin/../m68k-buildroot-uclinux-uclibc/sysroot/usr/lib"
configure: pkg-config: SSL_CPPFLAGS: ""
checking for HMAC_Update in -lcrypto... no
checking for HMAC_Init_ex in -lcrypto... no
checking OpenSSL linking with -ldl... no
checking OpenSSL linking with -ldl and -lpthread... no
checking for SSL_set_quic_use_legacy_codepoint... no
checking for SSL_set_quic_tls_cbs... no
configure: OpenSSL version does not speak any known QUIC API
configure: OPT_OPENSSL: /builds/bkuhls/buildroot/br-test-pkg/bootlin-m68k-5208-uclibc/host/m68k-buildroot-uclinux-uclibc/sysroot/usr
configure: OPENSSL_ENABLED:
configure: error: --with-openssl was given but OpenSSL could not be detected
make[1]: *** [package/pkg-generic.mk:263: /builds/bkuhls/buildroot/br-test-pkg/bootlin-m68k-5208-uclibc/build/libcurl-8.21.0/.stamp_configured] Error 1
Although OpenSSL was found using pkg-config the build tests fail.
A local build shows the concrete error in config.log, for example:
configure:27577: checking for HMAC_Update in -lcrypto
configure:27599: /home/bernd/buildroot/output/host/bin/m68k-linux-gcc
-o conftest -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE
-D_FILE_OFFSET_BITS=64 -O2 -g0 -fno-dwarf2-cfi-asm -Wl,-elf2flt=-r
-static -Werror-implicit-function-declaration -Wno-system-headers
-D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64
-D_GNU_SOURCE -Wl,-elf2flt=-r -static
-L/home/bernd/buildroot/output/host/bin/../m68k-buildroot-uclinux-uclibc/sysroot/usr/lib
-L/home/bernd/buildroot/output/host/bin/../m68k-buildroot-uclinux-uclibc/sysroot/usr/lib
conftest.c -lcrypto -lssl -lcrypto -lz -pthread -lz >&5
/home/bernd/buildroot/output/host/opt/ext-toolchain/m68k-buildroot-uclinux-uclibc/bin/ld.real:
/home/bernd/buildroot/output/host/bin/../m68k-buildroot-uclinux-uclibc/sysroot/usr/lib/libcrypto.a(libcrypto-lib-threads_pthread.o):
in function `ossl_rcu_read_lock':
threads_pthread.c:(.text+0xa4): undefined reference to `__atomic_fetch_add_8'
This error occurs many times for various atomic operations:
$ grep "undefined reference to \`__atomic" output/build/libcurl-8.20.0/config.log | sort -u | grep -v real
threads_pthread.c:(.text+0x28a): undefined reference to `__atomic_fetch_sub_8'
threads_pthread.c:(.text+0x3b4): undefined reference to `__atomic_fetch_add_8'
threads_pthread.c:(.text+0x9c8): undefined reference to `__atomic_is_lock_free'
threads_pthread.c:(.text+0xa4): undefined reference to `__atomic_fetch_add_8'
threads_pthread.c:(.text+0xa9c): undefined reference to `__atomic_is_lock_free'
threads_pthread.c:(.text+0xb66): undefined reference to `__atomic_is_lock_free'
threads_pthread.c:(.text+0xc30): undefined reference to `__atomic_is_lock_free'
threads_pthread.c:(.text+0xcdc): undefined reference to `__atomic_is_lock_free'
The build error can be reproduced with the current buildroot tree using
this defconfig:
BR2_m68k=y
BR2_m68k_cf5208=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_M68K_COLDFIRE_UCLIBC_STABLE=y
BR2_PACKAGE_OPENSSL=y
BR2_PACKAGE_LIBCURL=y
Although the toolchain lacks atomics support
$ grep ATOMIC .config
$
it emits atomic-related defines, for example:
$ echo | output/host/bin/m68k-linux-gcc -dM -E - | grep __ATOMIC_ACQ_REL
#define __ATOMIC_ACQ_REL 4
$
This specific define __ATOMIC_ACQ_REL is used in OpenSSL to enable
atomic support at various places:
https://github.com/openssl/openssl/blob/openssl-3.6.3/crypto/threads_pthread.c
causing the build errors we see with the mentioned defconfig.
To fix the problem we use an OpenSSL-provided define to forcefully
disable the usage of atomic intrinsics.
The misdetection of atomic intrinsics for m68k coldfire is not a new
problem:
https://lists.buildroot.org/pipermail/buildroot/2017-May/180841.htmlhttps://lists.buildroot.org/pipermail/buildroot/2026-May/803110.html
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Version 0.24.15 change log:
* protocol
fix crash on "sticker delete" and "sticker find"
* database
upnp: fix crash bug
* input
alsa, curl, nfs: fix stalled transfers
* playlist
asx, pls, rss, xspf: limit to 16 MB
cue: fix problem playing CUE tracks in music directory root
* player
fix noise with replay gain and cross-fade
Signed-off-by: Andreas Ziegler <br025@umbiko.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Floating-point type _Float16 added in gcc-16 on s390 now requires
decimal floating-point support enabled in the toolchain [1][2].
Without decimal floating-point, gcc 16.2.0 fails to build with:
../../../libgcc/config/s390/_dpd_sd_to_hf.c:27:25: error: decimal floating-point not supported for this target
27 | HFtype __dpd_truncsdhf (_Decimal32);
| ^~~~~~~~~~
../../../libgcc/config/s390/_dpd_sd_to_hf.c:31:16: error: decimal floating-point not supported for this target
31 | force_convert (_Decimal32 x)
| ^~~~~~~~~~
../../../libgcc/config/s390/_dpd_hf_to_td.c:34:1: error: decimal floating-point not supported for this target
34 | _Decimal128
| ^~~~~~~~~~~
../../../libgcc/config/s390/_dpd_sd_to_hf.c:35:18: error: decimal floating-point not supported for this target
35 | __dpd_truncsdhf (_Decimal32 x)
Enable decimal floating-point support as suggested by Alexander
Egorenkov.
Fixes:
https://lore.kernel.org/buildroot/ansJ5dXXblvYeMLB@windsurf/
[1] https://gcc.gnu.org/gcc-16/changes.html#s390
[2] https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=5d6d56d837c3dbeabd382c1fb4f7d21d9891f4b9
Cc: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Julien Olivain <ju.o@free.fr>
The mdnsd runtime test can randomly fail on slow runners.
It's hard to reproduce locally (only one failure after a few attempts)
but we can reproduce it easily by removing the while loop entirely.
It turns out that mdnsd is started by S50mdnsd before the
emulator.login() change the system date:
[BRTEST# date -s @1788032864
Sat Aug 29 19:47:44 UTC 2026
Since the minimal rootfs.cpio generated for TestMdnsd doesn't have any
ntp client installed, it start with "January 1, 1970".
The date change may cause some issue to the mdnsd daemon which blocks
any response from mquery command.
When the problem occurs, "mquery -T _http._tcp" reply is empty:
# mquery -T _http._tcp
Querying _http._tcp.local. for PTR (12) ... press Ctrl-C to stop
To workaround the issue, restart mdnsd manually.
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/16185948555
Signed-off-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Dracut-internal executables were linked against system libraries,
instead of Buildroot host packages. For example:
$ ldd host/lib/dracut/dracut-install
linux-vdso.so.1 (0x00007f578cf06000)
libc.so.6 => /usr/lib/x86_64-linux-gnu/libc.so.6 (0x00007f578cccd000)
libkmod.so.2 => /usr/lib/x86_64-linux-gnu/libkmod.so.2 (0x00007f578ccb1000)
/lib64/ld-linux-x86-64.so.2 (0x00007f578cf08000)
libcrypto.so.3 => /usr/lib/x86_64-linux-gnu/libcrypto.so.3 (0x00007f578c600000)
libz.so.1 => /usr/lib/x86_64-linux-gnu/libz.so.1 (0x00007f578cc92000)
libzstd.so.1 => /usr/lib/x86_64-linux-gnu/libzstd.so.1 (0x00007f578c536000)
The reason is that Dracut is not a "real" autoconf package, and the
hand-written ./configure script does not preserve LDFLAGS for
make. Pass the environment variables directly to fix this.
Signed-off-by: Fiona Klute (othermo GmbH) <fiona.klute@gmx.de>
[Julien: add comment in dracut.mk]
Signed-off-by: Julien Olivain <ju.o@free.fr>
examples/vm_power_manager/meson.build in DPDK detects the presence of
libvirt:
opt_dep = cc.find_library('virt', required : false)
and then builds some examples or not depending on the availability of
libvirt. Let's make this optional dependency explicit in dpdk.mk, even
if there's no explicit enable/disable option for it.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
When libvirt is present before DPDK is built, some additional examples
are compiled. One of them fails to build due to a missing <stdlib.h>
include. Let's import a patch from OpenSuse, that we have submitted
upstream, to fix this issue.
We couldn't find any autobuilder failure for this issue, but the
following defconfig allows to reproduce the failure:
BR2_aarch64=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_AARCH64_GLIBC_STABLE=y
BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_EUDEV=y
BR2_PACKAGE_DPDK=y
BR2_PACKAGE_DPDK_EXAMPLES=y
BR2_PACKAGE_LIBVIRT=y
The problem exists since DPDK v19.11, so it has been in Buildroot
since DPDK was introduced in commit
d17d1b6bde.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
