--------------------------------------------------------------------
- Update to version v5.7
  + Fix incorrect "FATAL" error message during garbage collection
  + Fix incorrect "FATAL" error message during garbage collection
  + Move fd into frec_src, fixes 15b60ddf935a531269bb8c68198de012a4967156
  + Fix to 75e2f0aec33e58ef5b8d4d107d821c215a52827c
  + Optimise sort_rrset for the case where the RR type no canonicalisation.
  + Fix for 12af2b171de0d678d98583e2190789e544440e02
  + Don't display unrelated CNAME queries when filtering for specific domain
  + dnsmasq-v2.83 forwards multiple queries to the same destination only once and stores the other queries as duplicates. They do receive the answer later on, however, this is usually not logged (when log-queries=extra is enabled, there will be a warning about the duplicate). This commit handles such duplicates and introduces a new reply type 14 = "already forwarded"
  + When seeing duplicated queries, the original query may have been blocked during CNAME inspection. In this case, we need to change the status from "OK (already forwarded)" to the correspondig blocked status. The "already forwarded" information is lost but that seems okay.
  + Check source query for its status when checking if we need to update the duplicated ones
  + Tidy initialisation in hash_questions.c
  + Change HAVE_NETTLEHASH compile-time to HAVE_CRYPTOHASH.
  + Bump copyright notices for 2021. Happy New Year!
  + Fix possible free-memory ref in e75069f79aa6b8a61034a9a4db9b6265b8be8ae4
  + Fixes incorrect "Found unknown status 14 in long term database" warning in the logs. We change the code to use a enum-based struct so we cannot forget to update this in the future when adding further query status types.
  + Add per-client rate-limiting. The default limit is 1000 queries in 60 seconds.
  + Add output of how much memory in /dev/shm is used by FTL itself
  + Try to create shared memory objects before reading the settings
  + Do not try to delete existing shmem objects on start - that may cause running FTL instances to crash when it tries to access them. Instead, new instances should properly fail to start.
  + Tests: Running a second instance is detected and prevented, FTL continues to work as expected afterwards
  + Do not explicitly request a lease time in our DHCPREQUEST as this may lead to incorrect responses. Also, when sending a request to lo, we should send it to the interface address instead of the broadcast (lo doesn't support broadcast destinations).
  + Fix queries sent upstream being counted incorrectly when modified later on (blocked externally, blocked during CNAME inspection). This also applies to queries loaded from the database.
  + Increment forward counters when immporting QUERY_RETRIED or QUERY_RETRIED_DNSSEC fromthe database
  + Retain EDNS0 bits from incoming queries when blocking requests

• Files Changed
• Browse Source