This patch attempts to record the reason why we fall back to DNS.
I considered using categorical probes for this, but they have a max limit of
20 categories, so we have to use a linear probe. I chose 50 buckets to allow
us to add more failure reasons in the future.
The recorded values are defined in nsHostRecord::TRRSkippedReason.
nsHostRecord::RecordReason is called whenever we encounter a condition that
will cause us to skip TRR in nsHostResolver.
For failures that occur inside TRR.cpp, each TRR object holds its own reason
that is recorded in a similar way. When all TRR requests are complete we
report the one that failed (or if both failed we report the one for the A
request).
Due to the fact that we might also follow CNAME requests, and the final
TRR request might not be the one that was issued at first, TRR requests
must pass back the reason as an argument to CompleteLookup.
Finally, this patch records the reason in two probes:
TRR_SKIP_REASON_TRR_FIRST - only reported in TRR-first mode
TRR_SKIP_REASON_DNS_WORKED - only reported in TRR-first mode when the
fallback DNS request succeeded. This allows us to filter for complete
network failures.
Differential Revision: https://phabricator.services.mozilla.com/D82168
mTRRUsed is a variable that we check to gate several telemetry probes, and to
decide if TRR really failed and we should add a domain to the TRR blocklist.
The problem with setting this too early is that when this is true but we
don't actually send the TRR request, then we will report that we fell back
to Do53 and potentially skip next TRR requests in the future.
The solution here is to only set mTRRUsed if TRRServiceChannel::AsyncOpen
succeeds.
Differential Revision: https://phabricator.services.mozilla.com/D81517
mTRRUsed is a variable that we check to gate several telemetry probes, and to
decide if TRR really failed and we should add a domain to the TRR blocklist.
The problem with setting this too early is that when this is true but we
don't actually send the TRR request, then we will report that we fell back
to Do53 and potentially skip next TRR requests in the future.
The solution here is to only set mTRRUsed if TRRServiceChannel::AsyncOpen
succeeds.
Differential Revision: https://phabricator.services.mozilla.com/D81517
mTRRUsed is a variable that we check to gate several telemetry probes, and to
decide if TRR really failed and we should add a domain to the TRR blocklist.
The problem with setting this too early is that when this is true but we
don't actually send the TRR request, then we will report that we fell back
to Do53 and potentially skip next TRR requests in the future.
The solution here is to only set mTRRUsed if TRRServiceChannel::AsyncOpen
succeeds.
Differential Revision: https://phabricator.services.mozilla.com/D81517
* Makes it possible to selectively enable TRR for pbmode/container/window/etc
Differential Revision: https://phabricator.services.mozilla.com/D48363
--HG--
extra : moz-landing-system : lando
* Makes it possible to selectively enable TRR for pbmode/container/window/etc
Differential Revision: https://phabricator.services.mozilla.com/D48363
--HG--
extra : moz-landing-system : lando
* Makes it possible to selectively enable TRR for pbmode/container/window/etc
Differential Revision: https://phabricator.services.mozilla.com/D48363
--HG--
extra : moz-landing-system : lando
* Makes it possible to selectively enable TRR for pbmode/container/window/etc
Differential Revision: https://phabricator.services.mozilla.com/D48363
--HG--
extra : moz-landing-system : lando
This patch calls NetworkConnectivityService::GetSingleton() on the main thread
and keeps a ref to the service until shutdown.
Even though calling ncs->GetIPv6() off-main-thread is technically a data-race
in practice that's OK because only the simple decision whether to send
AAAA requests is made based on that value, which in itself is an optimization.
I filed bug 1556967 for making the connectivity service thread safe.
Differential Revision: https://phabricator.services.mozilla.com/D33765
--HG--
extra : moz-landing-system : lando
- modify line wrap up to 80 chars; (tw=80)
- modify size of tab to 2 chars everywhere; (sts=2, sw=2)
--HG--
extra : rebase_source : 7eedce0311b340c9a5a1265dc42d3121cc0f32a0
extra : amend_source : 9cb4ffdd5005f5c4c14172390dd00b04b2066cd7
This is a best effort attempt at ensuring that the adverse impact of
reformatting the entire tree over the comments would be minimal. I've used a
combination of strategies including disabling of formatting, some manual
formatting and some changes to formatting to work around some clang-format
limitations.
Differential Revision: https://phabricator.services.mozilla.com/D13371
--HG--
extra : moz-landing-system : lando
... when that NS check is used to check the "parent" domain of a
blacklisted host.
Previously, additional TRRblacklist entries due to this would always be
added with the originSuffix "" which was incorrect for all uses of other
suffxes.
MozReview-Commit-ID: EeorQuuRCRX
Differential Revision: https://phabricator.services.mozilla.com/D10192
--HG--
extra : moz-landing-system : lando
Valentin Gosu
Kershaw Chang
Butkovits Atila
Narcis Beleuzu
Andrea Marchesini
Mihai Alexandru Michis
Razvan Maries
Dorel Luca
Dragana Damjanovic
Bogdan Tara
Noemi Erli
Sylvestre Ledru
Benjamin Bouvier
Ehsan Akhgari
Daniel Stenberg