For details see:
https://lists.nlnetlabs.nl/pipermail/unbound-users/2022-September/007885.html
"This release fixes CVE-2022-3204 Non-Responsive Delegation
Attack. It was reported by Yehuda Afek from Tel-Aviv
University and Anat Bremler-Barr and Shani Stajnrod from
Reichman University.
This fixes for better performance when under load, by cutting
promiscuous queries for nameserver discovery and limiting the
number of times a delegation point can look in the cache for
missing records.
Bug Fixes
- Patch for CVE-2022-3204 Non-Responsive Delegation Attack."
Signed-off-by: Matthias Fischer <matthias.fischer@ipfire.org>
Reviewed-by: Peter Müller <peter.mueller@ipfire.org>
According to the kernel's documentation,
> debugfs is a virtual file system that kernel developers use to put
> debugging files into. Enable this option to be able to read and
> write to these files.
There is no legitimate reason why one has to do so on an IPFire machine.
Further, the vast debugging options (i.e. related to various drivers)
have never been enabled, limiting the use of this virtual file system
even further.
This patch therefore proposes to disable it entirely, since its
potential security impact outweights its benefits. Due to operational
constraints, changes to ARM kernel configurations will be made if this
patch is approved for x86_64.
Signed-off-by: Peter Müller <peter.mueller@ipfire.org>
Reviewed-by: Michael Tremer <michael.tremer@ipfire.org>
Since we replace Perl, users most likely get to see some nasty "Internal
Server Error" messages during the upgrade. To suppres them, and to limit
the chance of side effects, stop Apache before applying the update, and
start it again afterwards.
Signed-off-by: Peter Müller <peter.mueller@ipfire.org>
Changes in version 0.4.7.10 - 2022-08-12
This version updates the geoip cache that we generate from IPFire location
database to use the August 9th, 2022 one. Everyone MUST update to this
latest release else circuit path selection and relay metrics are badly
affected.
o Major bugfixes (geoip data):
- IPFire informed us on August 12th that databases generated after
(including) August 10th did not have proper ARIN network allocations. We
are updating the database to use the one generated on August 9th, 2022.
Fixes bug 40658; bugfix on 0.4.7.9.
Changes in version 0.4.7.9 - 2022-08-11
This version contains several major fixes aimed at reducing memory pressure on
relays and possible side-channel. It also contains a major bugfix related to
congestion control also aimed at reducing memory pressure on relays.
Finally, there is last one major bugfix related to Vanguard L2 layer node
selection.
We strongly recommend to upgrade to this version especially for Exit relays
in order to help the network defend against this ongoing DDoS.
o Major bugfixes (congestion control):
- Implement RFC3742 Limited Slow Start. Congestion control was
overshooting the congestion window during slow start, particularly
for onion service activity. With this fix, we now update the
congestion window more often during slow start, as well as dampen
the exponential growth when the congestion window grows above a
capping parameter. This should reduce the memory increases guard
relays were seeing, as well as allow us to set lower queue limits
to defend against ongoing DoS attacks. Fixes bug 40642; bugfix
on 0.4.7.5-alpha.
o Major bugfixes (relay):
- Remove OR connections btrack subsystem entries when the connections
close normally. Before this, we would only remove the entry on error and
thus leaking memory for each normal OR connections. Fixes bug 40604;
bugfix on 0.4.0.1-alpha.
- Stop sending TRUNCATED cell and instead close the circuit from which we
received a DESTROY cell. This makes every relay in the circuit path to
stop queuing cells. Fixes bug 40623; bugfix on 0.1.0.2-rc.
o Major bugfixes (vanguards):
- We had omitted some checks for whether our vanguards (second layer
guards from proposal 333) overlapped. Now make sure to pick each
of them to be independent. Also, change the design to allow them
to come from the same family. Fixes bug 40639; bugfix
on 0.4.7.1-alpha.
o Minor features (dirauth):
- Add a torrc option to control the Guard flag bandwidth threshold
percentile. Closes ticket 40652.
- Add an AuthDirVoteGuard torrc option that can allow authorities to
assign the Guard flag to the given fingerprints/country code/IPs.
This is a needed feature mostly for defense purposes in case a DoS
hits the network and relay start losing the Guard flags too fast.
- Make UPTIME_TO_GUARANTEE_STABLE, MTBF_TO_GUARANTEE_STABLE,
TIME_KNOWN_TO_GUARANTEE_FAMILIAR WFU_TO_GUARANTEE_GUARD tunable
from torrc.
o Minor features (fallbackdir):
- Regenerate fallback directories generated on August 11, 2022.
o Minor features (geoip data):
- Update the geoip files to match the IPFire Location Database, as
retrieved on 2022/08/11.
o Minor bugfixes (congestion control):
- Add a check for an integer underflow condition that might happen
in cases where the system clock is stopped, the ORconn is blocked,
and the endpoint sends more than a congestion window worth of non-
data control cells at once. This would cause a large congestion
window to be calculated instead of a small one. No security
impact. Fixes bug 40644; bugfix on 0.4.7.5-alpha.
o Minor bugfixes (defense in depth):
- Change a test in the netflow padding code to make it more
_obviously_ safe against remotely triggered crashes. (It was safe
against these before, but not obviously so.) Fixes bug 40645;
bugfix on 0.3.1.1-alpha.
o Minor bugfixes (relay):
- Do not propagate either forward or backward a DESTROY remote reason when
closing a circuit in order to avoid a possible side channel. Fixes bug
40649; bugfix on 0.1.2.4-alpha.
Signed-off-by: Peter Müller <peter.mueller@ipfire.org>
Since we disabled Bluetooth support in the kernel a long time ago due to
security reasons, these do not serve any purpose anymore. Therefore, do
not ship them and delete them on existing installations.
Signed-off-by: Peter Müller <peter.mueller@ipfire.org>
For some reason, stripper crashes processing this directory:
strip: error: the input file '/lib/firmware/qcom/sc8280xp/LENOVO/21BX/qccdsp8280.mbn' has no sections
Signed-off-by: Peter Müller <peter.mueller@ipfire.org>
- The lcd2usb portion of the hd44780 driver in in the latest release version of
lcdproc (0.5.9) are only coded for libusb-0.1, which was removed from IPFire in recent
times.
- Commits have been merged into the lcdproc repository that enable lcd2usb to work with
the libusb-1.0 series but no release has been made since 2017.
- This patch downloaded a zip archive from the status of the lcdproc repository at commit
0e2ce9b. This zip archive was then converted into a tar.gx archive. The lfs and
rootfile have been updated in line with this.
- The lcdproc-0e2ce9b-4.ipfire file created by this build has been tested by the bug
reporter, Rolf Schreiber, and confirmed to fix the issue raised with the bug.
- This patch brings lcdproc upto date with the 149 commits that have been made between
2017 and Dec 2021, the date of the last commit.
- The version number has been defined as the last commit number.
- The -enable-libusb option has to be left in place as it turned out that
-enable-libusb-1-0 only works if -enable-libusb is also set. It looks like this was
identified in the lcdproc issues list but has not yet been fixed.
Fixes: Bug#12920
Signed-off-by: Adolf Belka <adolf.belka@ipfire.org>
Reviewed-by: Peter Müller <peter.mueller@ipfire.org>
Since we have extended services.cgi that it reads the Services field
from the Pakfire metadata, we will need to make sure that that metadata
is going to be on those systems.
Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>
- Update from version 37 to 38
- Update of rootfile
- mandoc is now a build dependency for efivar
- Old compile fixes patches are no longer required with version 38
- Details for lfs build of version 38 obtained from Beyond Linux From Scratch
- Changelog
bug fixes
Rework some makefile bits to make overriding some options simpler. by @vathpela in #140
Handle /sys/devices/virtual/{nvme-fabrics,nvme-subsystem} devices by @vathpela in #139
guids.S: Include <cet.h> when CET is enabled by @hjl-tools in #149
Fix /sys/block sysfs parsing for eMMC-s by @jwrdegoede in #150
Properly check mmap return error by @hannob in #152
Fix s{yt,ty}le typo in efi_get_variable(3) by @nabijaczleweli in #162
Handle NULL set_variable() by @lcp in #159
Fix parsing for nvme-subsystem devices by @dannf in #158
Attempt to fix the identified thread safety bugs by @vathpela in #155
Make thread-test depend on libefivar.so by @hjl-tools in #176
Upstream a local patch from rawhide by @frozencemetery in #177
Fix conversion from UTF8 to UCS2 by @freedge in #171
efivar: make docs match current code for 'efivar -A' by @vathpela in #178
Migrate CI to Github actions by @frozencemetery in #179
Add code of conduct by @frozencemetery in #180
Misc minor fixes by @vathpela in #182
Add efi_time_t declarations and helper functions. by @vathpela in #183
More misc fixes by @vathpela in #185
Run CI on more targets by @vathpela in #187
Coverity fixes 20211208 by @vathpela in #189
CI: run abicheck by @frozencemetery in #190
Fix linux virtual root device parsing by @vathpela in #188
efivar.spec.in: fix license to be valid SPDX by @frozencemetery in #192
Add efisecdb tooling by @vathpela in #184
Fix linker string comparison for dash by @frozencemetery in #194
Full changelog diff between version 37 and 38 is available in github repo
https://github.com/rhboot/efivar/compare/37...38
Signed-off-by: Adolf Belka <adolf.belka@ipfire.org>
Reviewed-by: Peter Müller <peter.mueller@ipfire.org>
- Update from version 3.7.3 to 3.8.1
- Update of rootfile
- Changelog
3.8.1 release
This is a bugfix release, fixing a few portability issues
reported for Nettle-3.8.
Bug fixes:
* Avoid non-posix m4 argument references in the chacha
implementation for arm64, powerpc64 and s390x. Reported by
Christian Weisgerber, fix contributed by Mamone Tarsha.
* Use explicit .machine pseudo-ops where needed in s390x
assembly files. Bug report by Andreas K. Huettel, fix
contributed by Mamone Tarsha.
Optimizations:
* Implemented runtime detection of cpu features for OpenBSD on
arm64. Contributed by Christian Weisgerber.
The new version is intended to be fully source and binary
compatible with Nettle-3.6. The shared library names are
libnettle.so.8.6 and libhogweed.so.6.6, with sonames
libnettle.so.8 and libhogweed.so.6.
3.8 release
This release includes a couple of new features, and many
performance improvements. It adds assembly code for two more
architectures: ARM64 and S390x.
The new version is intended to be fully source and binary
compatible with Nettle-3.6. The shared library names are
libnettle.so.8.5 and libhogweed.so.6.5, with sonames
libnettle.so.8 and libhogweed.so.6.
New features:
* AES keywrap (RFC 3394), contributed by Nicolas Mora.
* SM3 hash function, contributed by Tianjia Zhang.
* New functions cbc_aes128_encrypt, cbc_aes192_encrypt,
cbc_aes256_encrypt.
On processors where AES is fast enough, e.g., x86_64 with
aesni instructions, the overhead of using Nettle's general
cbc_encrypt can be significant. The new functions can be
implemented in assembly, to do multiple blocks with reduced
per-block overhead.
Note that there's no corresponding new decrypt functions,
since the general cbc_decrypt doesn't suffer from the same
performance problem.
Bug fixes:
* Fix fat builds for x86_64 windows, these appear to never
have worked.
Optimizations:
* New ARM64 implementation of AES, GCM, Chacha, SHA1 and
SHA256, for processors supporting crypto extensions. Great
speedups, and fat builds are supported. Contributed by
Mamone Tarsha.
* New s390x implementation of AES, GCM, Chacha, memxor, SHA1,
SHA256, SHA512 and SHA3. Great speedups, and fat builds are
supported. Contributed by Mamone Tarsha.
* New PPC64 assembly for ecc modulo/redc operations,
contributed by Amitay Isaacs, Martin Schwenke and Alastair
D´Silva.
* The x86_64 AES implementation using aesni instructions has
been reorganized with one separate function per key size,
each interleaving the processing of two blocks at a time
(when the caller processes multiple blocks with each call).
This gives a modest performance improvement on some
processors.
* Rewritten and faster x86_64 poly1305 assembly.
Known issues:
* Nettle's testsuite doesn't work out-of-the-box on recent
MacOS, due to /bin/sh discarding the DYLD_LIBRARY_PATH
environment variable. Nettle's test scripts handle this in
some cases, but currently fails the test cases that are
themselves written as /bin/sh scripts. As a workaround, use
make check EMULATOR='env DYLD_LIBRARY_PATH=$(TEST_SHLIB_DIR)'
Miscellaneous:
* Updated manual to current makeinfo conventions, with no
explicit node pointers. Generate pdf version with texi2pdf,
to get working hyper links.
* Added square root functions for NIST ecc curves, as a
preparation for supporting compact point representation.
* Reworked internal GCM/ghash interfaces, simplifying assembly
implementations. Deleted unused GCM C implementation
variants with less than 8-bit lookup table.
Signed-off-by: Adolf Belka <adolf.belka@ipfire.org>
Reviewed-by: Peter Müller <peter.mueller@ipfire.org>
- Update from version 6.4.19 to 6.4.32
- Update of rootfile not required
- Changelog - range of security and bug fixes
fetchmail-6.4.32 (released 2022-07-30, 31696 LoC):
# FIXES:
* Use configure to find rst2html, some systems install it only with .py suffix,
others only without, and some install both.
* Update README.maintainer
# TRANSLATIONS: language translations were updated by these fine people:
(in alphabetical order of language codes so as not to prefer people):
* cs: Petr Pisar [Czech]
* es: Cristian Othón Martínez Vera [Spanish]
* ja: Takeshi Hamasaki [Japanese]
* pl: Jakub Bogusz [Polish]
* ro: Remus-Gabriel Chelu [Romanian]
* sq: Besnik Bleta [Albanian]
* sv: Göran Uddeborg [Swedish]
fetchmail-6.4.31 (released 2022-07-16, 31694 LoC):
# BUG FIXES:
* Try to fix ./configure --with-ssl=... for systems that have multiple OpenSSL
versions installed. Issues reported by Dennis Putnam.
* The netrc parser now reports its errors to syslog or logfile when appropriate,
previously it would always log to stderr.
* Add error checking to .netrc parser.
# CHANGES:
* manpage: use .UR/.UE macros instead of .URL for URIs.
* manpage: fix contractions. Found with FreeBSD's igor tool.
* manpage: HTML now built with pandoc -> python-docutils
(manServer.pl was dropped)
fetchmail-6.4.30 (released 2022-04-26, 31666 LoC):
# BREAKING CHANGES:
* Bump wolfSSL minimum required version to 5.2.0 to pull in security fix.
# CHANGES:
* Using OpenSSL 1.* before 1.1.1n elicits a compile-time warning.
* Using OpenSSL 3.* before 3.0.2 elicits a compile-time warning.
* configure.ac was tweaked in order to hopefully fix cross-compilation issues
report, and different patch suggested, by Fabrice Fontaine,
https://gitlab.com/fetchmail/fetchmail/-/merge_requests/42
# TRANSLATIONS: language translations were updated by this fine person:
* ro: Remus-Gabriel Chelu [Romanian]
fetchmail-6.4.29 (released 2022-03-20, 31661 LoC):
# TRANSLATIONS: language translations were updated by this fine person:
* vi: Trần Ngọc Quân [Vietnamese]
fetchmail-6.4.28 (released 2022-03-05, 31661 LoC):
# DOCUMENTATION:
* Fix a typo in the manual page, courtesy of Jeremy Petch.
# TRANSLATIONS: language translations were updated by this fine person:
* es: Cristian Othón Martínez Vera [Spanish]
fetchmail-6.4.27 (released 2022-01-26, 31661 LoC):
# BREAKING CHANGES:
* Bump wolfSSL minimum required version to 5.1.1 to pull in security fix.
# TRANSLATIONS: language translations were updated by this fine person:
* ro: Remus-Gabriel Chelu [Romanian]
fetchmail-6.4.26 (released 2021-12-26, 31661 LoC):
# FIXES:
* When using wolfSSL 5.0.0, work around a bug that appears to hit wolfSSL when
receiving handshake records while still in SSL_peek(). Workaround is to read
1 byte and cache it, then call SSL_peek() again.
This affects only some servers. https://github.com/wolfSSL/wolfssl/issues/4593
# TRANSLATIONS: language translations were updated by this fine person:
* sr: Мирослав Николић (Miroslav Nikolić) [Serbian]
fetchmail-6.4.25 (released 2021-12-10, 31653 LoC):
# BREAKING CHANGES:
* Since distributions continue patching for LibreSSL use, which cannot be
linked legally, block out LibreSSL in configure.ac and socket.c, and
refer to COPYING, unless on OpenBSD (which ships it in the base system).
OpenSSL and wolfSSL 5 can be used. SSL-related documentation was updated, do
re-read COPYING, INSTALL, README, README.packaging, README.SSL.
* Bump OpenSSL version requirement to 1.0.2f in order to safely remove
the obsolete OpenSSL flag SSL_OP_SINGLE_DH_USE. This blocks out 1.0.2e and
older 1.0.2 versions. 1.0.2f was a security fix release, and 1.0.2u is
publicly available from https://www.openssl.org/source/old/1.0.2/
* Some of the configure.ac fiddling MIGHT have broken cross-compilation
again. The maintainer does not test cross-compiling fetchmail; if you
have difficulties, try setting PKG_CONFIG_LIBDIR to the pkg-config path
containing your target/host libraries, or see if --with-ssl-prefix or
--with-wolfssl-prefix, or overriding LDFLAGS/LIBS/CPPFLAGS, can help.
Feedback solicited on compliant systems that are before end-of-life.
# BUG FIXES:
* 6.4.24's workaround for OpenSSL 1.0.2's X509_V_FLAG_TRUSTED_FIRST flag
contained a typo and would not kick in properly.
* Library and/or rpath setting from configure.ac was fixed.
# ADDITIONS:
* Added an example systemd unit file and instructions to contrib/systemd/
which runs fetchmail as a daemon with 5-minute poll intervals.
Courteously contributed by Barak A. Pearlmutter, Debian Bug#981464.
* fetchmail can now be used with wolfSSL 5's OpenSSL compatibility layer,
see INSTALL and README.SSL. This is considered experimental.
Feedback solicited.
# CHANGES:
* The getstats.py dist-tool now counts lines of .ac and .am files.
* ./configure --with-ssl now supports pkg-config module names, too. See INSTALL.
# TRANSLATIONS: language translations were updated by these fine people:
(in reverse alphabetical order of language codes so as not to prefer people):
* sv: Göran Uddeborg [Swedish]
* sq: Besnik Bleta [Albanian]
* pl: Jakub Bogusz [Polish]
* ja: Takeshi Hamasaki [Japanese]
* fr: Frédéric Marchal [French]
* eo: Keith Bowes [Esperanto]
* cs: Petr Pisar [Czech]
fetchmail-6.4.24 (released 2021-11-20, 30218 LoC):
# OPENSSL AND LICENSING NOTE:
> see fetchmail-6.4.22 below, and the file COPYING.
Note that distribution of packages linked with LibreSSL is not feasible
due to a missing GPLv2 clause 2(b) exception.
# COMPATIBILITY:
* Bison 3.8 dropped yytoknum altogether, breaking compilation due to a
warning workaround. Remove the cast of yytoknum to void. This may cause
a compiler warning to reappear with older Bison versions.
* OpenSSL 1.0.2: Workaround for systems that keep the expired DST Root CA X3
certificate in its trust store because OpenSSL by default prefers the
untrusted certificate and fails. Fetchmail now sets the
X509_V_FLAG_TRUSTED_FIRST flag (on OpenSSL 1.0.2 only).
This is workaround #2 from the OpenSSL Blog. For details, see both:
https://www.openssl.org/blog/blog/2021/09/13/LetsEncryptRootCertExpire/https://letsencrypt.org/docs/dst-root-ca-x3-expiration-september-2021/
NOTE: OpenSSL 1.0.2 is end of life, it is assumed that the OpenSSL library
is kept up to date by a distributor or via OpenSSL support contract.
Where this is not the case, please upgrade to a supported OpenSSL version.
# DOCUMENTATION:
* The manual page was revised after re-checking with mandoc -Tlint, aspell,
igor. Some more revisions were made for clarity.
# TRANSLATIONS: language translations were updated by these fine people:
* sv: Göran Uddeborg [Swedish]
* pl: Jakub Bogusz [Polish]
* fr: Frédéric Marchal [French]
* cs: Petr Pisar [Czech]
* eo: Keith Bowes [Esperanto]
* ja: Takeshi Hamasaki [Japanese]
fetchmail-6.4.23 (released 2021-10-31, 30206 LoC):
# USABILITY:
* For common ssh-based IMAP PREAUTH setups (i. e. those that use a plugin
- no matter its contents - and that set auth ssh), change the STARTTLS
error message to suggest sslproto '' instead.
This is a commonly reported issue after the CVE-2021-39272 fix in 6.4.22.
Fixes Redhat Bugzilla 2008160. Fixes GitLab #39.
# TRANSLATIONS: language translations were updated by these fine people:
* ja: Takeshi Hamasaki [Japanese]
* sr: Мирослав Николић (Miroslav Nikolić) [Serbian]
fetchmail-6.4.22 (released 2021-09-13, 30201 LoC):
# OPENSSL AND LICENSING NOTE:
* fetchmail 6.4.22 is compatible with OpenSSL 1.1.1 and 3.0.0.
OpenSSL's licensing changed between these releases from dual OpenSSL/SSLeay
license to Apache License v2.0, which is considered incompatible with GPL v2
by the FSF. For implications and details, see the file COPYING.
# SECURITY FIXES:
* CVE-2021-39272: fetchmail-SA-2021-02: On IMAP connections, without --ssl and
with nonempty --sslproto, meaning that fetchmail is to enforce TLS, and when
the server or an attacker sends a PREAUTH greeting, fetchmail used to continue
an unencrypted connection. Now, log the error and abort the connection.
--Recommendation for servers that support SSL/TLS-wrapped or "implicit" mode on
a dedicated port (default 993): use --ssl, or the ssl user option in an rcfile.
--Reported by: Andrew C. Aitchison, based on the USENIX Security 21 paper "Why
TLS is better without STARTTLS - A Security Analysis of STARTTLS in the Email
Context" by Damian Poddebniak, Fabian Ising, Hanno Böck, and Sebastian
Schinzel. The paper did not mention fetchmail.
* On IMAP and POP3 connections, --auth ssh no longer prevents STARTTLS
negotiation.
* On IMAP connections, fetchmail does not permit overriding a server-side
LOGINDISABLED with --auth password any more.
* On POP3 connections, the possibility for RPA authentication (by probing with
an AUTH command without arguments) no longer prevents STARTTLS negotiation.
* For POP3 connections, only attempt RPA if the authentication type is "any".
# BUG FIXES:
* On IMAP connections, when AUTHENTICATE EXTERNAL fails and we have received the
tagged (= final) response, do not send "*".
* On IMAP connections, AUTHENTICATE EXTERNAL without username will properly send
a "=" for protocol compliance.
* On IMAP connections, AUTHENTICATE EXTERNAL will now check if the server
advertised SASL-IR (RFC-4959) support and otherwise refuse (fetchmail <= 6.4
has not supported and does not support the separate challenge/response with
command continuation)
* On IMAP connections, when --auth external is requested but not advertised by
the server, log a proper error message.
* Fetchmail no longer crashes when attempting a connection with --plugin "" or
--plugout "".
* Fetchmail no longer leaks memory when processing the arguments of --plugin or
--plugout on connections.
* On POP3 connections, the CAPAbilities parser is now caseblind.
* Fix segfault on configurations with "defaults ... no envelope". Reported by
Bjørn Mork. Fixes Debian Bug#992400. This is a regression in fetchmail 6.4.3
and happened when plugging memory leaks, which did not account for that the
envelope parameter is special when set as "no envelope". The segfault happens
in a constant strlen(-1), triggered by trusted local input => no vulnerability.
* Fix program abort (SIGABRT) with "internal error" when invalid sslproto is
given with OpenSSL 1.1.0 API compatible SSL implementations.
# CHANGES:
* IMAP: When fetchmail is in not-authenticated state and the server volunteers
CAPABILITY information, use it and do not re-probe. (After STARTTLS, fetchmail
must and will re-probe explicitly.)
* For typical POP3/IMAP ports 110, 143, 993, 995, if port and --ssl option
do not match, emit a warning and continue. Closes Gitlab #31.
(cherry-picked from 6.5 beta branch "legacy_6x")
* fetchmail.man and README.SSL were updated in line with RFC-8314/8996/8997
recommendations to prefer Implicit TLS (--ssl/ssl) and TLS v1.2 or newer,
placing --sslproto tls1.2+ more prominently.
The defaults shall not change between 6.4.X releases for compatibility.
# TRANSLATIONS: language translations were updated by these fine people:
* sq: Besnik Bleta [Albanian]
* cs: Petr Pisar [Czech]
* eo: Keith Bowes [Esperanto]
* fr: Frédéric Marchal [French]
* pl: Jakub Bogusz [Polish]
* sv: Göran Uddeborg [Swedish]
fetchmail-6.4.21 (released 2021-08-09, 30042 LoC):
# REGRESSION FIX:
* The new security fix in 6.4.20 for CVE-2021-36386 caused truncation of
messages logged to buffered outputs, from --logfile and --syslog.
This also caused lines in the logfile to run into one another because
the fragment containing the '\n' line-end character was usually lost.
Reason is that on all modern systems (with <stdarg.h> header and vsnprintf()
interface), the length of log message fragments was added up twice, so
that these ended too deep into a freshly allocated buffer, after the '\0'
byte. Unbuffered outputs flushed the fragments right away, which masked the
bug.
fetchmail-6.4.20 (released 2021-07-28, 30042 LoC):
# SECURITY FIX:
* When a log message exceeds c. 2 kByte in size, for instance, with very long
header contents, and depending on verbosity option, fetchmail can crash or
misreport each first log message that requires a buffer reallocation.
fetchmail then reallocates memory and re-runs vsnprintf() without another
call to va_start(), so it reads garbage. The exact impact depends on
many factors around the compiler and operating system configurations used and
the implementation details of the stdarg.h interfaces of the two functions
mentioned before. To fix CVE-2021-36386.
Signed-off-by: Adolf Belka <adolf.belka@ipfire.org>
Reviewed-by: Peter Müller <peter.mueller@ipfire.org>
