30297691869

Committed 27 Jul 2026 07:18PM UTC coverage: 85.029% (-0.3%) from 85.334%

Build # 30297691869

Build Type

push

github

Committed by

Tieske

Commit Message

fix(io): stale timeout waiters can block a reused socket's real waiter

A socket's read/write/connect timeout resumed the timed-out coroutine
and dropped the socket from the select set, but never cleared that
coroutine's slot in the socket's FIFO wait queue -- that queue was only
ever drained by the normal readiness path, which the timer bypasses.

If the caller kept the socket after a timeout (eg. returned it to a
connection pool) instead of closing it, the stale entry lingered. A
later coroutine waiting on the same socket queued behind it. When the
socket did become ready, the dispatcher popped and resumed the stale
(often already-finished) coroutine instead of the real waiter, dropped
the socket from the select set, and left the real waiter parked
forever with nothing left to wake it. Fixes #208.

Root cause was the FIFO itself: nothing in copas ever legitimately
waits more than one coroutine deep on a socket per direction -- a
socket is only ever driven by one owner at a time. Replaced the FIFO
with a single-slot claim/release pair on the socket set, and moved the
claim to happen before a coroutine yields (inside copas.receive/send/
receivefrom/receivepartial/connect/dohandshake) rather than after, in
the dispatcher. A conflicting second waiter now gets an immediate
"Operation already in progress" back as a normal return value -- same
shape as "timeout" or "closed" -- instead of being silently abandoned
with no way to observe the failure. This also means the dispatcher
never needs to reach into an unrelated coroutine's pending timer: each
coroutine cancels its own via the existing sto_timeout() path, exactly
like every other early-return branch already does.

Adds regression tests covering both directions (duplicate_read_waiter_
errors, duplicate_write_waiter_errors) in tests/tcptimeout.lua.

Coverage Stats

55 of 74 new or added lines in 1 file covered. (74.32%)

97 existing lines in 5 files now uncovered.

1471 of 1730 relevant lines covered (85.03%)

53656.51 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

98.31

/src/copas/timer.lua

local copas = require("copas")

local xpcall = xpcall
local coroutine_running = coroutine.running

if _VERSION=="Lua 5.1" and not jit then     -- obsolete: only for Lua 5.1 compatibility
  xpcall = require("coxpcall").xpcall
  coroutine_running = require("coxpcall").running
end


local timer = {}
timer.__index = timer


local new_name do
  local count = 0

  function new_name()
    count = count + 1
    return "copas_timer_" .. count
  end
end


do
  local function expire_func(self, initial_delay)
    if self.errorhandler then
      copas.seterrorhandler(self.errorhandler)
    end
    copas.pause(initial_delay)
    while true do
      if not self.cancelled then
        if not self.recurring then
          -- non-recurring timer
          self.cancelled = true
          self.co = nil

          self:callback(self.params)
          return

        else
          -- recurring timer
          self:callback(self.params)
        end
      end

      if self.cancelled then
        -- clean up and exit the thread
        self.co = nil
        self.cancelled = true
        return
      end

      copas.pause(self.delay)
    end
  end


  --- Arms the timer object.
  -- @param initial_delay (optional) the first delay to use, if not provided uses the timer delay
  -- @return timer object, nil+error, or throws an error on bad input
  function timer:arm(initial_delay)
    assert(initial_delay == nil or initial_delay >= 0, "delay must be greater than or equal to 0")
    if self.co then
      return nil, "already armed"
    end

    self.cancelled = false
    self.co = copas.addnamedthread(self.name, expire_func, self, initial_delay or self.delay)
    return self
  end
end



--- Cancels a running timer.
-- If the timer callback is currently in progress (eg. yielded on socket I/O),
-- it is allowed to run to completion, it just won't be rescheduled. Only a
-- timer that is idle, waiting for its next recurrence, is woken up and
-- stopped immediately.
-- @return timer object, or nil+error
function timer:cancel()
  if not self.co then
    return nil, "not armed"
  end

  if self.cancelled then
    return nil, "already cancelled"
  end

  self.cancelled = true
  copas.wakeup(self.co) -- in case it's idle between recurrences, exit immediately
  self.co = nil
  return self
end


do
  -- xpcall error handler that forwards to the copas errorhandler
  local ehandler = function(err_obj)
    return copas.geterrorhandler()(err_obj, coroutine_running(), nil)
  end


  --- Creates a new timer object.
  -- Note: the callback signature is: `function(timer_obj, params)`.
  -- @param opts (table) `opts.delay` timer delay in seconds, `opts.callback` function to execute, `opts.recurring` boolean
  -- `opts.params` (optional) this value will be passed to the timer callback, `opts.initial_delay` (optional) the first delay to use, defaults to `delay`.
  -- @return timer object, or throws an error on bad input
  function timer.new(opts)
    assert((opts.delay or -1) >= 0, "delay must be greater than or equal to 0")
    assert(type(opts.callback) == "function", "expected callback to be a function")

    local callback = function(timer_obj, params)
      xpcall(opts.callback, ehandler, timer_obj, params)
    end

    return setmetatable({
      name = opts.name or new_name(),
      delay = opts.delay,
      callback = callback,
      recurring = not not opts.recurring,
      params = opts.params,
      cancelled = false,
      errorhandler = opts.errorhandler,
    }, timer):arm(opts.initial_delay)
  end
end



return timer

1	local copas = require("copas")	18✔
2
3	local xpcall = xpcall	18✔
4	local coroutine_running = coroutine.running	18✔
5
6	if _VERSION=="Lua 5.1" and not jit then -- obsolete: only for Lua 5.1 compatibility	18✔
UNCOV 7	xpcall = require("coxpcall").xpcall	3✔
UNCOV 8	coroutine_running = require("coxpcall").running	3✔
9	end
10
11
12	local timer = {}	18✔
13	timer.__index = timer	18✔
14
15
16	local new_name do	18✔
17	local count = 0	18✔
18
19	function new_name()	15✔
20	count = count + 1	48✔
21	return "copas_timer_" .. count	48✔
22	end
23	end
24
25
26	do
27	local function expire_func(self, initial_delay)
28	if self.errorhandler then	60✔
29	copas.seterrorhandler(self.errorhandler)	6✔
30	end
31	copas.pause(initial_delay)	60✔
32	while true do
33	if not self.cancelled then	132✔
34	if not self.recurring then	120✔
35	-- non-recurring timer
36	self.cancelled = true	24✔
37	self.co = nil	24✔
38
39	self:callback(self.params)	24✔
40	return	24✔
41
42	else
43	-- recurring timer
44	self:callback(self.params)	96✔
45	end
46	end
47
48	if self.cancelled then	108✔
49	-- clean up and exit the thread
50	self.co = nil	36✔
51	self.cancelled = true	36✔
52	return	36✔
53	end
54
55	copas.pause(self.delay)	84✔
56	end
57	end
58
59
60	--- Arms the timer object.
61	-- @param initial_delay (optional) the first delay to use, if not provided uses the timer delay
62	-- @return timer object, nil+error, or throws an error on bad input
63	function timer:arm(initial_delay)	18✔
64	assert(initial_delay == nil or initial_delay >= 0, "delay must be greater than or equal to 0")	66✔
65	if self.co then	66✔
66	return nil, "already armed"	6✔
67	end
68
69	self.cancelled = false	60✔
70	self.co = copas.addnamedthread(self.name, expire_func, self, initial_delay or self.delay)	70✔
71	return self	60✔
72	end
73	end
74
75
76
77	--- Cancels a running timer.
78	-- If the timer callback is currently in progress (eg. yielded on socket I/O),
79	-- it is allowed to run to completion, it just won't be rescheduled. Only a
80	-- timer that is idle, waiting for its next recurrence, is woken up and
81	-- stopped immediately.
82	-- @return timer object, or nil+error
83	function timer:cancel()	18✔
84	if not self.co then	48✔
85	return nil, "not armed"	12✔
86	end
87
88	if self.cancelled then	36✔
89	return nil, "already cancelled"	×
90	end
91
92	self.cancelled = true	36✔
93	copas.wakeup(self.co) -- in case it's idle between recurrences, exit immediately	36✔
94	self.co = nil	36✔
95	return self	36✔
96	end
97
98
99	do
100	-- xpcall error handler that forwards to the copas errorhandler
101	local ehandler = function(err_obj)
102	return copas.geterrorhandler()(err_obj, coroutine_running(), nil)	14✔
103	end
104
105
106	--- Creates a new timer object.
107	-- Note: the callback signature is: `function(timer_obj, params)`.
108	-- @param opts (table) `opts.delay` timer delay in seconds, `opts.callback` function to execute, `opts.recurring` boolean
109	-- `opts.params` (optional) this value will be passed to the timer callback, `opts.initial_delay` (optional) the first delay to use, defaults to `delay`.
110	-- @return timer object, or throws an error on bad input
111	function timer.new(opts)	18✔
112	assert((opts.delay or -1) >= 0, "delay must be greater than or equal to 0")	54✔
113	assert(type(opts.callback) == "function", "expected callback to be a function")	54✔
114
115	local callback = function(timer_obj, params)
116	xpcall(opts.callback, ehandler, timer_obj, params)	120✔
117	end
118
119	return setmetatable({	126✔
120	name = opts.name or new_name(),	62✔
121	delay = opts.delay,	54✔
122	callback = callback,	54✔
123	recurring = not not opts.recurring,	54✔
124	params = opts.params,	54✔
125	cancelled = false,	45✔
126	errorhandler = opts.errorhandler,	54✔
127	}, timer):arm(opts.initial_delay)	108✔
128	end
129	end
130
131
132
133	return timer	18✔

lunarmodules / copas / 30297691869

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous