aboutsummaryrefslogtreecommitdiff
path: root/res_cache.cpp
Commit message (Collapse)AuthorAgeFilesLines
* Take metered information from setResolverConfigurationKen Chen2023-11-011-1/+12
| | | | | | | | | | | 1. Uses dnsresolver_aidl_interface_lateststable_version 12. 2. Get the 'metered' parameter from setResolverConfiguration and keep it in NetConfig of each network. 3. Add resolv_is_metered_network() for DnsProxyListener. Bug: 288340533 Test: atest resolv_integration_test resolv_unit_test Change-Id: I390199b93a9f5b3c0abc8f072d91153ef9fac32e
* Merge "Accommodate a change in the type of std::span's size" into mainMaciej Żenczykowski2023-09-261-3/+3
|\
| * Accommodate a change in the type of std::span's sizeRyan Prichard2023-09-211-3/+3
| | | | | | | | | | | | | | | | | | | | | | The WIP version of std::span in external/libcxx uses a ptrdiff_t size, but the final standardized version of std::span uses size_t instead. Use std::span() constructor calls rather than {}-syntax, which will convert the signed length to unsigned and works with either the old or the new libc++. Test: treehugger Change-Id: I5b5a16d0949e77a74269b9f6cf24382dd69a5973
* | Add a function to query enforceDnsUid settingKen Chen2023-09-211-1/+9
|/ | | | | | | | | | The "DNS query fail-fast when network access is restricted" feature needs to know whether the enforceDnsUid is set or not in DnsProxyListener. Bug: 288340533 Test: atest resolv_unit_test:ResolvCacheTest#IsEnforceDnsUidEnabled Change-Id: I8e7a5d5d030602eced05c6f7f3809a57bfabebc3
* Change MAX_ENTRIES_LOWER_BOUND from 0 to 1Ken Chen2023-07-101-1/+1
| | | | | | | | 0 is a common default value of experiment configuration. Uses 1 to prevent misconfiguration from the server. Test: presubmit Change-Id: I198bbea3543b0318dd25aaca0843100ad7c5b1b8
* Allow setting values lower than MAX_ENTRIES_DEFAULT in max_cache_entriesKen Chen2023-06-131-1/+2
| | | | | | | | | | Adjust the lower bound of max_cache_entries from MAX_ENTRIES_DEFAULT(640) to 0. So we can do experiments with values like 320. Bug: 241953569 Test: atest resolv_unit_test Change-Id: I4da710dccf4efe0edc625221b2519050069e0ba6
* [Code Health] Fix misc-const-correctness build warningKen Chen2023-03-271-1/+1
| | | | | | | Fix lint on Ica3f7d3c5a262a797a6bdc99147c63d5a9108dff Change-Id: I90dbc91d5c0437ac8556277ff79e39f8f936a10f Test: TH
* [Code Health] Reduce nestingKen Chen2023-03-221-24/+23
| | | | | Change-Id: Ica3f7d3c5a262a797a6bdc99147c63d5a9108dff Test: TH
* Revert "Add debug trace for MdnsIPv6LinkLocalWithDefaultRoute"Ken Chen2022-10-191-6/+1
| | | | | | | | This reverts commit 46079f7260a0bd7ad59ab8074e5e4ef4195e0d33. Reason for revert: b/247693272 is clarified Change-Id: Ib3f2d7b756e795dd44c2924a8e4b45c0e2c7e44b
* Add debug trace for MdnsIPv6LinkLocalWithDefaultRouteKen Chen2022-09-271-1/+6
| | | | | | | | | | | The test is flaky and not local reproducible. Add logs with DEBUG level and set logging level to DEBUG in the test. This commit is for debugging purpose, which should be reverted when root cause is found. Bug: 247693272 Bug: 246688231 Test: atest Change-Id: I7bee788381ff245ae321f4816b81f71087bbeec2
* Add experiment flag 'max_cache_entries'Ken Chen2022-08-111-7/+35
| | | | | | | | Make max cache entries of DNS resolver cache adjustable by server push. Test: atest Bug: 241953569 Change-Id: Iaa733782f8407b9ba1bb3725395745921b526ba0
* Add more error log in resolv_stats_set_addrsMike Yu2022-07-251-23/+7
| | | | | | Bug: 239659682 Test: atest Change-Id: I9896ceb902883ac45d3627dbc358cc30e2c18ed3
* Address comment from aosp/2138392.Wayne Ma2022-07-051-3/+3
| | | | | Test: atest ResolverTest Change-Id: Id2328ae1ebb90fe93f8852df61e9dcf09cd5b646
* Add retry_count and retransmission_time_interval to Experiment class.Wayne Ma2022-07-011-12/+14
| | | | | Test: atest resolv_integration_test Change-Id: Ibd31b4ed231c66a0cb7559cf16958621951e9287
* Add more traces in error casesKen Chen2022-06-271-1/+4
| | | | | | | | | Current logs are unable to tell why mdns fallback doesn't work. Add a few logs on error paths. Bug: 226325994 Test: adb locat Change-Id: Ifd61bbbd61b8866296d7b4583a2d9e78ba43c78d
* Enable some DNS logs in userdebug buildKen Chen2022-05-041-13/+13
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 1. Set DNS default logging level to INFO. 2. Adjust logging level on some traces. Output: [Success] resolv : GetAddrInfoHandler::run: {100 983140 100 983140 1021 0} resolv : res_nmkquery: (QUERY, IN, A) resolv : res_nsend: used send_dg 97 terrno: 0 resolv : doQuery: rcode=0, ancount=4 [Cache hit] resolv : GetAddrInfoHandler::run: {100 786532 100 983140 0 0} resolv : res_nmkquery: (QUERY, IN, AAAA) resolv : resolv_cache_lookup: FOUND IN CACHE entry=0xb4000075f51973d0 resolv : doQuery: rcode=0, ancount=1 [Error - NXDOMAIN] resolv : GetAddrInfoHandler::run: {100 786532 100 983140 0 0} resolv : res_nmkquery: (QUERY, IN, AAAA) resolv : res_nsend: used send_dg 104 terrno: 0 resolv : doQuery: rcode=3, ancount=0 [Timeout] resolv : GetAddrInfoHandler::run: {30 983070 30 983070 0 0} resolv : res_nmkquery: (QUERY, IN, A) resolv : res_nsend: used send_dg 0 terrno: 110 resolv : res_nsend: used send_dg 0 terrno: 110 resolv : doQuery: rcode=255, ancount=0 ... Bug: N/A Test: m; adb logcat Change-Id: I2a0ea2b416264d59382ef121d6d7755f8c7ffdf7
* Remove unicast-response bit from .local resolutionKen Chen2022-04-081-3/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | RFC 6762 section 6.7 says that the Multicast DNS responder MUST send a UDP response directly back to the querier, via unicast, to the query packet's source IP address and port if the source UDP port in a received Multicast DNS query is not 5353. Section 5.4 also mentions "...New questions created by local clients afterwards should be treated as normal 'QM' questions and SHOULD NOT have the unicast-response bit set on the first question of the series." DNS resolver works as a one-shot Multicast DNS querier, which send queries from random ports. The unicast-response bit is not necessary to be set. Setting the unicast-response bit may also cause malfunction on .local fallback queries. Currently, the bit is not cleared while .local resolution is fallbacked from Multicast to Unicast DNS queries. DNS server may send a no error response without Answer RR, or a failure. One way to fix this is clearing the unicast-response bit before fallback. However, it needs to parse packet bytes in res_nsend() because the query packet is made before the res_nsend() but the fallback decision is made in res_nsend(). Besides, it can cause problems in cache. The query class is counted in cache key hashing. The answer obtained by fallback queries is hashed (without unicast-response bit) and stored in cache. But subsequent .local queries cannot match the record because their unicast-response bit is initially set. To avoid adding more complexity, the unnecessary unicast-response bit should be removed. Bug: 227147672 Test: atest Test: Ping test.local on openWRT with fallback and without fallback. Change-Id: Ib703a7537f638669fdc1d9c6927800e5c901786a
* Exclude VPN and mobile data from .local resolutionKen Chen2022-03-041-0/+16
| | | | | | | | | There is no point in sending multicast DNS over VPN or mobile data. Bug: 209492138 Test: atest Test: adb root; adb shell dumpsys dnsresolver Change-Id: I8987d3c9f219c42247e9fd8f85880a1d49fcedad
* Merge "Replace manual buffer handling with std::span"Bruce Chen2021-09-031-24/+23
|\
| * Replace manual buffer handling with std::spanchenbruce2021-08-311-24/+23
| | | | | | | | | | | | Bug: 196298216 Test: cd packages/modules/DnsResolver && atest Change-Id: I8b0b8b0cc0c1185eb9a9e99ed10a2fc109bfa1d7
* | Merge "Replace StringAppendF() with fmt::format_to"Treehugger Robot2021-08-271-5/+2
|\ \ | |/ |/|
| * Replace StringAppendF() with fmt::format_tochenbruce2021-08-231-5/+2
| | | | | | | | | | | | Bug: 196298216 Test: cd packages/modules/DnsResolver && atest Change-Id: Iddcd0e69f46911ced852cf3994e6e243e9637820
* | Subsampling DNS event for mDNSchenbruce2021-08-251-13/+21
|/ | | | | | | | | | | | | | | | | | | | Currently, a device generates some mDNS queries when the user uses mDNS .local resolution. Using the query info by subsampling events based on how interesting they are. Because the number of mDNS query is much less than DNS query, the mDNS subsampling rate is higer than DNS query. - if return_code == 0,2,7 -> log 1 in 1 event - if return_code == default -> log 1 in 1 event Also allow to use experiment flag to update sub-sampling denom. Example for dumpsys dnsresolver: NetId: 100 DnsEvent subsampling map for MDNS: default:1 Bug: 197092658 Test: cd packages/modules/DnsResolver && atest m statsd_testdrive and run "statsd_testdrive 116" Change-Id: I76073aa9a1cea43bda2675334592ed22e96a238e
* Replace StringPrintf() with fmt::format()chenbruce2021-08-201-7/+5
| | | | | | Bug: 196298216 Test: cd packages/modules/DnsResolver && atest Change-Id: I581c282bba1527afd2be9daf925e9ec5bd2e5cd6
* use std::span for res_pquerychenbruce2021-08-181-2/+2
| | | | | | | | | Manual buffer handling may make mistakes easily. Replacing it by std::span. Bug: 196298216 Test: cd packages/modules/DnsResolver && atest Change-Id: I2a9ebb5c9f7eb9d09fa7bdcafe4748eb2042ca6b
* Implement a new IPC setResolverOptions in DnsResolverLuke Huang2021-08-061-17/+25
| | | | | | | | | | 1. Implement the new method added in V9, setResolverOptions 2. Add its relevant binder test and update integration test to use this new method if the remote version is new enough. Test: atest Bug: 194048056 Change-Id: I9a18f11769fd154e7cc6f35090e6c8fe61853d2a
* Make resolverOptions optional in DnsResolver.Lorenzo Colitti2021-08-061-12/+17
| | | | | | | | The only change is currently to make resolverOptions optional. Test: m Bug: 194048056 Change-Id: I42a07d2bad1b3ee0e66e3f4e2a0f068686e1d1a3
* Set UNICAST-RESPONSE bit for mDNSchenbruce2021-07-121-1/+3
| | | | | | | | The field is used to minimize unnecessary broadcasts on the network. Bug: 140857615 Test: cd packages/modules/DnsResolver && atest Change-Id: I54bc6f7cd41ff9687d4de8c9e87780ce805df050
* Add Mdns statistics logchenbruce2021-07-021-7/+30
| | | | | | | | | | When developers run "dumpsys dnsresolver", Mdns statistics will show the information in the log if the Mdns queries happened. Bug: 140857615 Test: cd packages/modules/DnsResolver && atest Change-Id: Iec0657d417ddf329590f1fc0a599497aaf45639b
* Rename definitions and funtions of stats.chenbruce2021-06-291-3/+3
| | | | | | | | | | StatsRecords only include informations of dns servers. When adding the information of Mdns statistics, we have to modifiy the names of server's definitions and funtions to include mdns in stats files. Bug: 140857615 Test: cd packages/modules/DnsResolver && atest Change-Id: I38befe452d5d37b4172ee5dd16822083a08e4c88
* Revert "Support evaluating private DNS by latency"Mike Yu2021-06-131-12/+0
| | | | | | | | | | | This reverts commit 1aede8135e8a227e127f826f38073eba7447c382. This is no longer needed since we will implement a simpler mechanics for DoT validation. Bug: 188153519 Test: cd packages/modules/DnsResolver && atest Change-Id: I4a2f8b926f27fd38c58aea3a993a311a74fc4950
* Support evaluating private DNS by latencyMike Yu2021-06-051-0/+12
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The evaluation is limited to opportunistic mode and is implemented as a flag-off feature. It is introduced to avoid from using high latency private DNS servers. The latency of a server is considered high if it's higher than a latency threshold which is calculated based on the average latency of cleartext DNS server: latency threshold = std::clamp(3 * mean_do53_latency_ms, min_private_dns_latency_threshold_ms, max_private_dns_latency_threshold_ms) , where min_private_dns_latency_threshold_ms is 500 ms by default and max_private_dns_latency_threshold_ms is 2000 ms by default. If there's no Do53 average latency for reference, the latency threshold is min_private_dns_latency_threshold_ms. The evaluation of a private DNS server works in two phases. Phase 1: In this phase, Private DNS Validation is being performed, and the server is not considered validated. The server latency is evaluated by sending a probe. If the latency is lower than a the latency threshold, the server state is transitioned to Validation::success. The evaluation goes to phase 2. Phase 2: In this phase, the server is considered validated and DnsResolver can send DNS queries to the server. The server latency is evaluated by the query response time, and the same latency threshold is used. If there are several, 10 by default, query response time failed to meet the time threshold in a row, the server state is transitioned to Validation::in_process. The evaluation goes to phase 1. Bug: 188153519 Test: run atest with all the flags off/on avoid_bad_private_dns: 0 / 1 sort_nameservers: 0 / 1 dot_xport_unusable_threshold: -1 / 20 dot_query_timeout_ms: -1 / 10000 min_private_dns_latency_threshold_ms: -1 / 500 keep_listening_udp: 0 / 1 parallel_lookup_sleep_time: 2 / 2 dot_revalidation_threshold: -1 / 10 max_private_dns_latency_threshold_ms: -1 / 2000 dot_async_handshake: 0 / 1 dot_maxtries: 3 / 1 dot_connect_timeout_ms: 127000 / 10000 parallel_lookup_release: UNSET / UNSET Change-Id: Ib681b1ea1417eadac9c013f19549a9fa7c408696
* Update the default sampling flag of DNS resolver metrics eventslifr2021-04-211-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | The sample rate of DNS resolver metrics events depends on Experiment flag and default sampling flag. Currently, the sampling experiment has been completed. Therefore update the new default sampling flag to reduce the metrics event size. 1. Current default sampling rate: DEFAULT_SUBSAMPLING_MAP[] = "default:1 0:100 7:10 metrics data size: 170k byte per device 2. Old Experiment: DEFAULT_SUBSAMPLING_MAP[] = "default:5 0:300 2:80 7:80" metrics data size: 11k byte per device 3. New default sampling rate: DEFAULT_SUBSAMPLING_MAP[] = "default:8 0:400 2:110 7:110" metrics data size: 7.9k byte per device 4. New Experiment:(including the temporary fix of b/185097438) DEFAULT_SUBSAMPLING_MAP[] = "default:8 0:400 2:110 4:110 7:110" metrics data size: 7.9k byte per device Because there is already a solution for b/185097438 in the code. Therefore, what this commit will do is item3 to reduce the size of the event. Bug: 184830442 Test: atest resolv_unit_test Change-Id: I3c0ab4a4504dfd5f8c789348587de33e01951920
* Move definitions of HostMapping next to its one and only referenceBernie Innocenti2020-12-031-0/+4
| | | | | | | This also brings back a function comment next to the definition it was related to. Change-Id: I41778a3df0911f6673b1f4bcee2b10eece7bc0c8
* DnsResolver: Support case-insensitive domain name hashPabba Prakash2020-09-071-10/+25
| | | | | | | | | | | | | | | | - Support case-insensitive hashing i.e., convert domain names into lowercase before hashing/comparison so that there is no additional query when domain names are same. Test: as follows - build, and boot - Perform DNS queries, and ensure they are not case sensitive - resolv_integration_test - resolv_unit_test Bug: 111586865 Change-Id: I75f2ef8d09314e2f89a4764ccfd0247675340271
* Support prioritizing DNS serversMike Yu2020-07-141-1/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The change introduces a way to prioritize DNS servers on the basis of DNS query response time, which aims to replace the current design that is biased towards using the first DNS server assigned from networks. The quality is evaluated based on the heuristics: - The more latency it is, the less likely it is used. - The longer time it is not used, the more likely it is used. Compared to the current design, the proposed method detects bad DNS servers more quickly. For instance, a server which is unreachable or times out can be detected and deprioritized with few trials by backoff penalty and abnormal latency. Similar to the current design, a server which has been regarded as bad quality can be used again, but it depends on how much worse it is. A counter is used to count how many times a DNS server not being used, which avoids from constantly using the same DNS server. This change comprises: [1] Allow the resolver to sort DNS servers on the basis of DNS query response time. [2] Add an experiment flag to enable/disable the sorting. [3] Show the result of the quantified quality of DNS servers in dumpsys dnsresolver. [4] Add unit tests for DnsStats::getSortedServers(). [5] Revise the integration tests which are sensitive to the nameserver sorting, including two big changes in SkipBadServersDueToInternalError and SkipBadServersDueToTimeout and some minor changes. Bug: 137169582 Test: ran resolv_unit_test ran resolv_integration_test with the sorting enabled ran resolv_integration_test with the sorting disabled Change-Id: I24b6a317f135a942ce0ea310c81dfe658bada6a7
* Add a resolver option to enforce AID_DNS on queryKen Chen2020-06-201-0/+3
| | | | | | | | | | The default behavior is that plaintext DNS queries are sent by the application's UID using fchown(). DoT are sent with an UID of AID_DNS This option control the plaintext uid of DNS query. Bug: 154910763 Test: atest Change-Id: Iada5d850d8bb9d7b0ad46f5c28a1fff22c7d11a6
* Fix the side channel attack by using aysnc DnsResolver API with ↵Luke Huang2020-05-281-6/+4
| | | | | | | | | | | | | | FLAG_NO_CACHE_STORE Before this CL, FLAG_NO_CACHE_STORE could be used to perform side channel attack. Because this flag ensures the result is never added to the cache, but will return a cached response if one exists. So make FLAG_NO_CACHE_STORE imply FLAG_NO_CACHE_LOOKUP to block the possibility of side channel attacking. Bug: 150371903 Test: atest Change-Id: I37391ffe315b90c0cdfd86888c6bf68b2b89f601
* Add Experiments to handle experiment flags logicLuke Huang2020-04-231-1/+0
| | | | | | | | | | | update experiment flags everytime when a resolver network is created or destroyed. Bug: 135717624 Bug: 151698212 Test: atest Change-Id: I5375e78831c5994af74b9ecaca991f18db03eca6
* Add a new NetworkType to resolverchenbruce2020-04-161-2/+15
| | | | | | | | The new NetworkType is WIFI + CELLULAR + VPN. Bug: 143732914 Test: atest Change-Id: I44c3ff1ec27417a23652815edd94389a14bbd3b2
* Rename ResolverExperimentalOptions to ResolverOptionsKen Chen2020-04-071-10/+9
| | | | | | | | Remove 'Experimental' from parcel name and variable name for better naming. Bug: 139646101 Change-Id: Iaa75e8028e98998eca4c9b821aa07effab19dac3
* Refine AIDL constant definition for TC_MODE_*Luke Huang2020-04-061-2/+3
| | | | | | | | | Address comments in AIDL API review before AIDL freezing. Bug: 151074535 Bug: 133526962 Test: atest Change-Id: Iaf5e233666b9a05a86d4951a751145a8e94cb94b
* Add transport types to a given networkchenbruce2020-04-011-17/+94
| | | | | | | | | | - Resolver doesn't know what the transport types are when having a given network. Set tranport types by SetResolverConfiguration. Bug: 143732914 Test: atest Change-Id: I337d4204afd9fbb790fd42cc191759b22dd66bca
* Remove nscount in ResState and NetConfigMike Yu2020-02-201-35/+23
| | | | | | | | The number of nameservers can be derived from std::vector::size(). It doesn't need a variable to store the value. Test: cd packages/modules/DnsResolver && atest Change-Id: Ifc2e1401b1d081595a00b60fbaf6dd96c3d44a71
* Retrieve the res_stats based on a given list of nameserver addressesMike Yu2020-02-201-6/+23
| | | | | | | | | | | | | | Similar to aosp/1136436, each independent lookup thread should get the stats based on its own nameserver address list rather than the list stored in NetConfig. With this change, plus aosp/1136436, it's able to populate nameserver addresses in random order to lookup threads without the worry of breaking the legacy stats stored in NetConfig. Bug: 137169582 Test: cd packages/modules/DnsResolver && atest Change-Id: I8d5c639326d2931bf7e5f5ae668e7aba830e00fb
* Refactor ResState to store nameserver addresses by IPSockAddrMike Yu2020-02-201-19/+6
| | | | | | | | | | | IPSockAddr is more safer and convenient to store socket addresses, to compare two socket addresses, and to make the code more readable. The change also removes get_nsaddr(), a static function in res_send.cpp. Bug: 137169582 Test: cd packages/modules/DnsResolver && atest Change-Id: I694c293139b01a39c40cc50ba8c4f067a2ac4b07
* Fix async DNS flag NO_CACHE_STORE doesn't work as expectedLuke Huang2020-02-111-1/+1
| | | | | | | | | | | The stale cache case isn't handled correctly while performing cahce_lookup with flag NO_CACHE_STORE, which caused this problem. Fix it and add a test to ensure it won't happen again. Test: atest Bug: 148842821 Change-Id: I72a2211a636cadc72009a5542f7c755c30329c43
* Delete dead code in res_comp.cpp and misc cleanupsBernie Innocenti2020-01-241-0/+1
| | | | | | | | | | | | | | No functionality changes, this is mostly meant to improve our code coverage and cleanup our headers a bit. - delete dead functions res_ownok() and res_mailok() - Move prototypes of the remaining functions to the new header res_comp.h - Replace resolv_private.h with res_comp.h as needed - Drop the dependency on Bionic's <resolv.h> wherever possible - Eliminate the horrible maybe_XXX() macros - Make res_hnok() and res_dnok() return a proper bool instead of 0 and 1 Change-Id: Ic2e27753355d873925044a7561ccb78bd2c2d162
* Merge "Fix harmless cache key bug in res_cache"Treehugger Robot2020-01-231-0/+1
|\
| * Fix harmless cache key bug in res_cacheBernie Innocenti2020-01-221-0/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The resolver's cache is keyed by a hash of the query, excluding certain sections such as the query ID and a few others. However, a loop bug caused the entire packet to be hashed to the end after the first invocation of _dnsPacket_hashBytes() in _dnsPacket_hashQuery(). The bug predates the inclusion of this code in Bionic in 2009, and was likely harmless: since _dnsPacket_hashBytes() adjusts the cursor to the end of the packet, the subsequent calls to _dnsPacket_readInt16() fail, returning return -1, which causes the loops to execute 0 times. This whole oldschool code could use a good rewrite into safer C++ and, while we're at it, also switch to a proper hash function. For now, let's start with a conservative bugfix, so at least it behaves the way it reads. Test: cd packages/modules/DnsResolver && atest . Change-Id: I9368115bc16373fd64989f2a37f1dcd8f3e71b94
generated by cgit v1.2.3 (git 2.25.1) at 2025-12-20 02:43:36 +0000