This change takes the idea of handling strings as pairs in the form
(start, len) by Robert Elz from
https://mail-index.netbsd.org/source-changes-d/2021/08/20/msg013427.html
and expands it by avoiding one more memory allocation, for iterating the
PATH environment variable.
No functional change.
Previously, findcc modified its argument string, even though it had been
declared as 'const char *'. This triggered a lint warning that "strchr
effectively discards 'const char *' from argument", in fact, this code
caused the lint check to be implemented in the first place.
The first attempt at fixing it by removing the 'const' from the
parameter type was a bad idea since it made the API of that function
more complicated.
Revert back to making the parameter a 'const char *' and duplicate that
string internally as necessary. Add a few more tests for absolute
pathnames since these had been missing before. There are no tests yet
for snprintf with too long strings, but the current change does not
modify that part of the code.
A function that modifies a string argument must not declare that
argument as 'const char *', even if all callers (mkdep and lint) always
pass it a modifiable string.
No functional change.
Since November 2001, there is a comment above the function 'fold' that
suggests there are a few bugs concerning overflow detection. Add some
first 'proper regression tests' to prove these bugs.
When lexing an integer constant, it does not make sense to check for
'struct', 'void' or other types. Furthermore, the constants were listed
in some random-looking order, which made the code hard to read.
Only support the 4-wire mode for now. Support for 3-wire mode is
theoretically possible but most (all?) modules in circulation have the
BS0 / IM0 3-wire config pin internally grounded within the flat-flex
assembly so it is not easy to enable & test.
--- 9.16.20 released ---
5689. [security] An assertion failure occurred when named attempted to
send a UDP packet that exceeded the MTU size, if
Response Rate Limiting (RRL) was enabled.
(CVE-2021-25218) [GL #2856]
5688. [bug] Zones using KASP and inline-signed zones failed to apply
changes from the unsigned zone to the signed zone under
certain circumstances. This has been fixed. [GL #2735]
5687. [bug] "rndc reload <zonename>" could trigger a redundant
reload for an inline-signed zone whose zone file was not
modified since the last "rndc reload". This has been
fixed. [GL #2855]
5686. [func] The number of internal data structures allocated for
each zone was reduced. [GL #2829]
5685. [bug] named failed to check the opcode of responses when
performing zone refreshes, stub zone updates, and UPDATE
forwarding. This has been fixed. [GL #2762]
5682. [bug] Some changes to "zone-statistics" settings were not
properly processed by "rndc reconfig". This has been
fixed. [GL #2820]
5681. [func] Relax the checks in the dns_zone_cdscheck() function to
allow CDS and CDNSKEY records in the zone that do not
match an existing DNSKEY record, as long as the
algorithm matches. This allows a clean rollover from one
provider to another in a multi-signer DNSSEC
configuration. [GL #2710]
5679. [func] Thread affinity is no longer set. [GL #2822]
5678. [bug] The "check DS" code failed to release all resources upon
named shutdown when a refresh was in progress. This has
been fixed. [GL #2811]
5672. [bug] Authentication of rndc messages could fail if a
"controls" statement was configured with multiple key
algorithms for the same listener. This has been fixed.
[GL #2756]
--- 9.16.19 released ---
5671. [bug] A race condition could occur where two threads were
competing for the same set of key file locks, leading to
a deadlock. This has been fixed. [GL #2786]
5670. [bug] create_keydata() created an invalid placeholder keydata
record upon a refresh failure, which prevented the
database of managed keys from subsequently being read
back. This has been fixed. [GL #2686]
5669. [func] KASP support was extended with the "check DS" feature.
Zones with "dnssec-policy" and "parental-agents"
configured now check for DS presence and can perform
automatic KSK rollovers. [GL #1126]
5668. [bug] Rescheduling a setnsec3param() task when a zone failed
to load on startup caused a hang on shutdown. This has
been fixed. [GL #2791]
5667. [bug] The configuration-checking code failed to account for
the inheritance rules of the "dnssec-policy" option.
This has been fixed. [GL #2780]
5666. [doc] The safe "edns-udp-size" value was tweaked to match the
probing value from BIND 9.16 for better compatibility.
[GL #2183]
5665. [bug] If nsupdate sends an SOA request and receives a REFUSED
response, it now fails over to the next available
server. [GL #2758]
5664. [func] For UDP messages larger than the path MTU, named now
sends an empty response with the TC (TrunCated) bit set.
In addition, setting the DF (Don't Fragment) flag on
outgoing UDP sockets was re-enabled. [GL #2790]
5662. [bug] Views with recursion disabled are now configured with a
default cache size of 2 MB unless "max-cache-size" is
explicitly set. This prevents cache RBT hash tables from
being needlessly preallocated for such views. [GL #2777]
5661. [bug] Change 5644 inadvertently introduced a deadlock: when
locking the key file mutex for each zone structure in a
different view, the "in-view" logic was not considered.
This has been fixed. [GL #2783]
5658. [bug] Increasing "max-cache-size" for a running named instance
(using "rndc reconfig") did not cause the hash tables
used by cache databases to be grown accordingly. This
has been fixed. [GL #2770]
5655. [bug] Signed, insecure delegation responses prepared by named
either lacked the necessary NSEC records or contained
duplicate NSEC records when both wildcard expansion and
CNAME chaining were required to prepare the response.
This has been fixed. [GL #2759]
5653. [bug] A bug that caused the NSEC3 salt to be changed on every
restart for zones using KASP has been fixed. [GL #2725]
--- 9.16.18 released ---
5660. [bug] The configuration-checking code failed to account for
the inheritance rules of the "key-directory" option.
[GL #2778]
5659. [bug] When preparing DNS responses, named could replace the
letters 'W' (uppercase) and 'w' (lowercase) with '\000'.
This has been fixed. [GL #2779]
--- 9.16.17 released ---
5652. [bug] A copy-and-paste error in change 5584 caused the
IP_DONTFRAG socket option to be enabled instead of
disabled. This has been fixed. [GL #2746]
5651. [func] Refactor zone dumping to be processed asynchronously via
the uv_work_t thread pool API. [GL #2732]
5650. [bug] Prevent a crash that could occur if serve-stale was
enabled and a prefetch was triggered during a query
restart. [GL #2733]
5649. [bug] If a query was answered with stale data on a server with
DNS64 enabled, an assertion could occur if a non-stale
answer arrived afterward. [GL #2731]
5648. [bug] The calculation of the estimated IXFR transaction size
in dns_journal_iter_init() was invalid. [GL #2685]
5644. [bug] Fix a race condition in reading and writing key files
for zones using KASP and configured in multiple views.
[GL #1875]
5643. [cleanup] "make install" no longer creates an empty
${localstatedir}/run directory. [GL #2709]
5642. [bug] Zones which are configured in multiple views with
different values set for "dnssec-policy" and with
identical values set for "key-directory" are now
detected and treated as a configuration error.
[GL #2463]
5641. [bug] Address a potential memory leak in
dst_key_fromnamedfile(). [GL #2689]
5639. [bug] Check that the first and last SOA record of an AXFR are
consistent. [GL #2528]
5638. [bug] Improvements related to network manager/task manager
integration:
- isc_managers_create() and isc_managers_destroy()
functions were added to handle setup and teardown of
netmgr, taskmgr, timermgr, and socketmgr, since these
require a precise order of operations now.
- Event queue processing is now quantized to prevent
infinite looping.
- The netmgr can now be paused from within a netmgr
thread.
- Deadlocks due to a conflict between netmgr's
pause/resume and listen/stoplistening operations were
fixed.
[GL #2654]
5633. [doc] The "inline-signing" option was incorrectly described as
being inherited from the "options"/"view" levels and was
incorrectly accepted at those levels without effect.
This has been fixed. [GL #2536]
5624. [func] Task manager events are now processed inside network
manager loops. The task manager no longer needs its own
set of worker threads, which improves resolver
performance. [GL #2638]
--- 9.16.16 released ---
5637. [func] Change the default value of the "max-ixfr-ratio" option
to "unlimited". [GL #2671]
5636. [bug] named and named-checkconf did not report an error when
multiple zones with the "dnssec-policy" option set were
using the same zone file. This has been fixed.
[GL #2603]
5635. [bug] Journal compaction could fail when a journal with
invalid transaction headers was not detected at startup.
This has been fixed. [GL #2670]
5634. [bug] If "dnssec-policy" was active and a private key file was
temporarily offline during a rekey event, named could
incorrectly introduce replacement keys and break a
signed zone. This has been fixed. [GL #2596]
5633. [doc] The "inline-signing" option was incorrectly described as
being inherited from the "options"/"view" levels and was
incorrectly accepted at those levels without effect.
This has been fixed. [GL #2536]
5632. [func] Add a new built-in KASP, "insecure", which is used to
transition a zone from a signed to an unsigned state.
The existing built-in KASP "none" should no longer be
used to unsign a zone. [GL #2645]
5631. [protocol] Update the implementation of the ZONEMD RR type to match
RFC 8976. [GL #2658]
5630. [func] Treat DNSSEC responses containing NSEC3 records with
iteration counts greater than 150 as insecure.
[GL #2445]
5629. [func] Reduce the maximum supported number of NSEC3 iterations
that can be configured for a zone to 150. [GL #2642]
5627. [bug] RRSIG(SOA) RRsets placed anywhere other than at the zone
apex were triggering infinite resigning loops. This has
been fixed. [GL #2650]
5626. [bug] When generating zone signing keys, KASP now also checks
for key ID conflicts among newly created keys, rather
than just between new and existing ones. [GL #2628]
5625. [bug] A deadlock could occur when multiple "rndc addzone",
"rndc delzone", and/or "rndc modzone" commands were
invoked simultaneously for different zones. This has
been fixed. [GL #2626]
5622. [cleanup] The lib/samples/ directory has been removed, as export
versions of libraries are no longer maintained.
[GL !4835]
5619. [protocol] Implement draft-vandijk-dnsop-nsec-ttl, updating the
protocol such that NSEC(3) TTL values are set to the
minimum of the SOA MINIMUM value or the SOA TTL.
[GL #2347]
5618. [bug] Change 5149 introduced some inconsistencies in the way
record TTLs were presented in cache dumps. These
inconsistencies have been eliminated. [GL #389]
[GL #2289]
These controllers require D/C# to be deasserted for the entire duration
of the command, contrary to docs and how newer controllers like SSD1353
behave. Probably this is needed for all controllers that do not have the
WRITE_RAM (0x5c) command. Also support using alternate padding commands
for 3-wire mode when WRITE_RAM is not available.
no idea why it was enabled as we can't easily target 32 bit.
if we figure this out for aarch32, the subdir names etc will
probably be different, or we can just revive the deleted
files no longer referenced with this commit.
If an mbuf is ETHER_ALIGNed, the packet size may be shorter than MCLBYTES.
For example, if the max frame size is 2048 by changing the interface's MTU,
an mbuf's max length is 2046. So, don't use rx_mbuf_sz for bus_dmamap_sync.
t_vlan has rarely failed by checking IFF_PROMISC before the
test program do BIOCPROMISC. To solve this, BIOCPROMISC is
done in the foreground.
fixes PR/56357
- Update next_to_refresh at least before ixgbe_rx_unrefresed() to detect
the unrefreshed status correctly in ixgbe_rxeof().
- next_to_refresh points to the previous entry of the first unrefreshed
descriptor, so fix a loop variable to point to the correct one in
ixgbe_refresh_mbufs().
- Without the above two fixes, RX ring may have some unrefreshed entries
which have inconsistent state. On such state, "ifconfig down up" causes
panic in bus_dmamap_sync() on aarch64.
- Tested on amd64 and aarch64. OK'd by knakahara.
rillig