Specify how to disable timeouts #4069
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Of course I had to forget something, and it was So I think we could safely make zero no longer special, and in that case have a 1ms timed wait, or less. |
dridi
force-pushed
the
timeout_never
branch
2 times, most recently
from
9e3283f to
0a2f97d
Compare
From now on, INFINITY disables the timeout and NAN is no longer allowed. Refs varnishcache#3045
For all intents and purposes, it currently is exactly the same as the timeout tweak. The duration parameters are either not really timeouts, or timeouts that cannot be disabled. In other words, the timeout tweak will grow the ability to formally disable a timeout. Refs varnishcache#3045
The timeout type is the authority from which both the tweak and flag are derived.
They were already relying on NAN to fall back to parameters.
We can't use NAN in VGC code today, so in order to convey the lack of timeout setting in a backend definition, only a negative value makes sense since zero will eventually mean zero instead of undefined. This is an implicit breakage of the VRT ABI for the meaning of struct vrt_backend.
Unlike session timeouts, they were not falling back to their parameter counterparts from NAN. Since NAN is not allowed in VCL, they now behave like session timeouts, returning their value or falling back to the parameter when unset.
The vtim_real in Lck_CondWaitUntil() is not meant to be INFINITY, and it should only apply to the vtim_dur passed to Lck_CondWaitTimeout(). In practice it does since the latter calls the former. It could be enforced by extracting a static function from Lck_CondWaitUntil(). This accommodates a few timeout parameters that are used for cond waits.
The timeout_* parameters are used in duration arithmetics and the simplest is to not grant them the ability to be disabled. Both timeout_idle and idle_send_timeout appear in a CondWait context. For this reason both need to be capped to 1h, matching the implicit limit in Lck_CondWaitTimeout().
walid-git
added a commit
to walid-git/varnish-cache
that referenced
this pull request
Jul 15, 2024
walid-git
added a commit
to walid-git/varnish-cache
that referenced
this pull request
Jul 15, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Following consensus, this patch series formalizes several things:
NANmeans that a timeout is not set (only possible in VCL, parameters must have a default value)unsettimeout variables in VCLsess.*timeout*)INFINITYdisables a timeoutVRT_DECIMAL_MAXin VCLpoll(2)case it translates to -1 (akaINFTIM)setsockopt(2)case it translates to a zero timevalsetsockopt(2)translates to a 1us timeout (should it be 1ms?)bereqtimeouts have two levels of fallbackINFINITY(nomath.h)I tried to be as thorough as possible, following the trail of relevant timeout parameters.
Fixes #3045