{-# LANGUAGE CPP #-} {-# LANGUAGE ScopedTypeVariables #-} {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE NamedFieldPuns, BangPatterns #-} {-| Thread-safe Logging with bonus controlled-schedule debugging capabilities. This module supports logging to memory, serializing messages and deferring the work of actually printing them. Another thread can flush the logged messages at its leisure. The second capability of this infrastructure is to use the debugging print messages as points at which to gate the execution of the program. That is, each `logStrLn_` call becomes a place where the program blocks and checks in with a central coordinator, which only allows one thread to unblock at a time. Thus, if there are sufficient debug logging messages in the program, this can enable a form of deterministic replay (and quickcheck-style testing of different interleavings). -} module Control.LVish.Logging ( -- * Global variables dbgLvl, -- * New logger interface newLogger, logOn, Logger(closeIt, flushLogs), WaitMode(..), LogMsg(..), OutDest(..), -- * General utilities forkWithExceptions ) where import Control.Monad import qualified Control.Exception as E import qualified Control.Concurrent.Async as A import Data.IORef import qualified Data.Sequence as Seq import Data.List (sortBy) import GHC.Conc hiding (yield) import Control.Concurrent import System.IO.Unsafe (unsafePerformIO) import System.IO (stderr, stdout, hFlush, hPutStrLn, Handle) import System.Environment(getEnvironment) import System.Random import Text.Printf (printf, hPrintf) import Debug.Trace (trace, traceEventIO) import Control.LVish.Types -- import qualified Control.LVish.SchedIdempotentInternal as Sched ---------------------------------------------------------------------------------------------------- -- | A Logger coordinates a set of threads that print debug logging messages. -- -- This are abstract objects supporting only the operations provided by this module -- and the non-hidden fields of the Logger. data Logger = Logger { coordinator :: A.Async () -- ThreadId -- ^ (private) The thread that chooses which action to unblock next -- and handles printing to the screen as well. , minLvl :: Int -- ^ The minimum level of messages accepted by this logger (usually 0). , maxLvl :: Int -- ^ The maximum level of messages accepted by this logger. , checkPoint :: SmplChan Writer -- ^ The serialized queue of writers attempting to log dbg messages. , closeIt :: IO () -- ^ (public) A method to complete flushing, close down the helper thread, -- and generally wrap up. , loutDests :: [OutDest] -- ^ Where to send output. If empty, messages dropped entirely. , logged :: IORef [String] -- ^ (private) In-memory buffer of messages, if OutputInMemory is selected. -- This is stored in reverse-temporal order during execution. , flushLogs :: IO [String] -- ^ Clear buffered log messages and return in the order they occurred. , waitWorkers :: WaitMode } -- | A single thread attempting to log a message. It only unblocks when the attached -- MVar is filled. data Writer = Writer { who :: String , continue :: MVar () , msg :: LogMsg -- TODO: Indicate whether this writer has useful work to do or -- is about to block... this provides a simple notion of -- priority. } -- | Several different ways we know to wait for quiescence in the concurrent mutator -- before proceeding. data WaitMode = WaitTids [ThreadId] (IO Bool) -- ^ Wait until a certain set of threads is blocked before proceeding. -- If that conditional holds ALSO make sure the provided polling action -- returns True as well. | WaitDynamic -- ^ UNFINISHED: Dynamically track tasks/workers. The -- num workers starts at 1 and then is modified -- with `incrTasks` and `decrTasks`. | WaitNum { numThreads :: Int, -- ^ How many threads total must check in? downThreads :: IO Int -- ^ Poll how many threads won't participate this round. } -- ^ A fixed set of threads must check-in each round before proceeding. | DontWait -- ^ In this mode, logging calls are non-blocking and return -- immediately, rather than waiting on a central coordinator. -- This is what we want if we're simply printing debugging output, -- not controlling the schedule for stress testing. deriving Show instance Show (IO Bool) where show _ = "<IO Bool>" instance Show (IO Int) where show _ = "<IO Int>" -- | We allow logging in O(1) time in String or ByteString format. In practice the -- distinction is not that important, because only *thunks* should be logged; the -- thread printing the logs should deal with forcing those thunks. data LogMsg = StrMsg { lvl::Int, body::String } -- | ByteStrMsg { lvl::Int, } toString x@(StrMsg{}) = body x maxWait :: Int maxWait = 10*1000 -- 10ms andM :: [IO Bool] -> IO a -> IO a -> IO a andM [] t _f = t andM (hd:tl) t f = do b <- hd if b then andM tl t f else f catchAll :: ThreadId -> E.SomeException -> IO () catchAll parent exn = case E.fromException exn of Just E.ThreadKilled -> return () _ -> do hPutStrLn stderr ("! Exception on Logger thread: "++show exn) hFlush stderr E.throwTo parent exn E.throwIO exn -------------------------------------------------------------------------------- -- | Create a new logger, which includes forking a coordinator thread. -- Takes as argument the number of worker threads participating in the computation. newLogger :: (Int,Int) -- ^ What inclusive range of messages do we accept? Defaults to `(0,dbgLvl)`. -> [OutDest] -> WaitMode -> IO Logger newLogger (minLvl, maxLvl) loutDests waitWorkers = do logged <- newIORef [] checkPoint <- newSmplChan parent <- myThreadId let flushLogs = atomicModifyIORef' logged $ \ ls -> ([],reverse ls) let -- When all threads are quiescent, we can flush the remaining messagers from -- the channel to get the whole set of waiting tasks. Return in chronological order. flushChan !acc = do x <- tryReadSmplChan checkPoint case x of Just h -> flushChan (h:acc) Nothing -> return $ reverse acc -- This is the format we use for debugging messages formatMessage extra Writer{msg} = "|"++show (lvl msg)++ "| "++extra++ toString msg -- One of these message reports how many tasks are in parallel with it: messageInContext pos len wr = formatMessage ("#"++show (1+pos)++" of "++show len ++": ") wr printOne str (OutputTo h) = hPrintf h "%s\n" str printOne str OutputEvents = traceEventIO str printOne str OutputInMemory = -- This needs to be atomic because other messages might be calling "flush" -- at the same time. atomicModifyIORef' logged $ \ ls -> (str:ls,()) printAll str = mapM_ (printOne str) loutDests shutdownFlag <- newIORef False -- When true, time to shutdown. shutdownComplete <- newEmptyMVar -- Here's the new thread that corresponds to this logger: coordinator <- A.async $ E.handle (catchAll parent) $ -- BEGIN defs for the async task: -------------------------------------------------------------------------------- -- Proceed in rounds, gather the set of actions that may happen in parallel, then -- pick one. We log the series of decisions we make for reproducability. let schedloop :: Int -> Int -- ^ length of list `waiting` -> [Writer] -> Backoff -> IO () schedloop !iters !num !waiting !bkoff = do when (iters > 0 && iters `mod` 500 == 0) $ putStrLn $ "Warning: logger has spun for "++show iters++" iterations, "++show num++" are waiting." hFlush stdout fl <- readIORef shutdownFlag if fl then flushLoop else do let keepWaiting = do b <- backoff bkoff schedloop (iters+1) num waiting b waitMore = do w <- readSmplChan checkPoint -- Blocking! (or spinning) b <- newBackoff maxWait -- We got something, reset this. schedloop (iters+1) (num+1) (w:waiting) b case waitWorkers of DontWait -> error "newLogger: internal invariant broken." WaitNum target extra -> do n <- extra -- Atomically check how many extra workers are blocked. if (num + n >= target) then pickAndProceed waiting else waitMore WaitTids tids poll -> do -- FIXME: This is not watertight... it will work with high probability but can't be trusted: andM [checkTids tids, poll, checkTids tids, poll] (do ls <- flushChan waiting case ls of [] -> do chatter " [Logger] Warning: No active tasks?" bk2 <- backoff bkoff schedloop (iters+1) 0 [] bk2 _ -> pickAndProceed ls) keepWaiting -- | Keep printing messages until there is (transiently) nothing left. flushLoop = do x <- tryReadSmplChan checkPoint case x of Just wr -> do printAll (formatMessage "" wr) flushLoop Nothing -> return () -- | A simpler alternative schedloop that only does printing (e.g. for DontWait mode). printLoop = do fl <- readIORef shutdownFlag if fl then flushLoop else do wr <- readSmplChan checkPoint printAll (formatMessage "" wr) printLoop -- Take the set of logically-in-parallel tasks, choose one, execute it, and -- then return to the main scheduler loop. pickAndProceed [] = error "pickAndProceed: this should only be called on a non-empty list" pickAndProceed waiting = do let order a b = let s1 = toString (msg a) s2 = toString (msg b) in case compare s1 s2 of GT -> GT LT -> LT EQ -> error $" [Logger] Need in-parallel log messages to have an ordering, got two equal:\n "++s1 sorted = sortBy order waiting len = length waiting -- For now let's randomly pick an action: pos <- randomRIO (0,len-1) let pick = sorted !! pos (pref,suf) = splitAt pos sorted rst = pref ++ tail suf unblockTask pos len pick -- The task will asynchronously run when it can. yield -- If running on one thread, give it a chance to run. -- Return to the scheduler to wait for the next quiescent point: bnew <- newBackoff maxWait schedloop 0 (length rst) rst bnew unblockTask pos len wr@Writer{continue} = do printAll (messageInContext pos len wr) putMVar continue () -- Signal that the thread may continue. -- Check whether the worker threads are all quiesced checkTids [] = return True checkTids (tid:rst) = do st <- threadStatus tid case st of ThreadRunning -> return False ThreadFinished -> checkTids rst -- WARNING: this design is flawed because it is possible when compiled -- with -threaded that IO will spuriously showed up as BlockedOnMVar: ThreadBlocked BlockedOnMVar -> checkTids rst ThreadBlocked _ -> return False ThreadDied -> checkTids rst -- Should this be an error condition!? in -- Main body of async task: do case waitWorkers of DontWait -> printLoop _ -> schedloop (0::Int) (0::Int) [] =<< newBackoff maxWait -- Kick things off. putMVar shutdownComplete () return () -- End: async thread -- END async task. -------------------------------------------------------------------------------- let closeIt = do atomicModifyIORef' shutdownFlag (\_ -> (True,())) readMVar shutdownComplete A.cancel coordinator -- Just to make sure its completely done. return $! Logger { coordinator, checkPoint, closeIt, loutDests, logged, flushLogs, waitWorkers, minLvl, maxLvl } chatter :: String -> IO () -- chatter = hPrintf stderr -- chatter = printf "%s\n" chatter _ = return () printNTrace s = do putStrLn s; traceEventIO s; hFlush stdout -- UNFINISHED: incrTasks = undefined decrTasks = undefined -- | Write a log message from the current thread, IF the level of the -- message falls into the range accepted by the given `Logger`, -- otherwise, the message is ignored. logOn :: Logger -> LogMsg -> IO () logOn Logger{checkPoint,minLvl,maxLvl,waitWorkers} msg | (minLvl <= lvl msg) && (lvl msg <= maxLvl) = do case waitWorkers of -- In this mode we are non-blocking: DontWait -> writeSmplChan checkPoint Writer{who="",continue=dummyMVar,msg} _ -> do continue <- newEmptyMVar writeSmplChan checkPoint Writer{who="",continue,msg} takeMVar continue -- Block until we're given permission to proceed. | otherwise = return () {-# NOINLINE dummyMVar #-} dummyMVar :: MVar () dummyMVar = unsafePerformIO newEmptyMVar ---------------------------------------------------------------------------------------------------- -- Simple back-off strategy. -- | The state for an exponential backoff. data Backoff = Backoff { current :: !Int , cap :: !Int -- ^ Maximum nanoseconds to wait. } deriving Show newBackoff :: Int -> IO Backoff newBackoff cap = return Backoff{cap,current=0} backoff :: Backoff -> IO Backoff -- backoff b = do yield; return b backoff Backoff{current,cap} = case current of -- Yield once before we start delaying: 0 -> do yield return Backoff{cap,current=1} n -> do let next = min cap (2*n) threadDelay n return Backoff{cap,current=next} ---------------------------------------------------------------------------------------------------- -- Simple channels: we need non-blocking reads so we can't use -- Control.Concurrent.Chan. We could use TChan, but I don't want to bring STM into -- it right now. -- type MyChan a = Chan a -- -- | A simple channel. Take-before-put is the protocol. -- type SmplChan a = MVar [a] -- | Simple channels that don't support real blocking. type SmplChan a = IORef (Seq.Seq a) -- New elements pushed on right. newSmplChan :: IO (SmplChan a) newSmplChan = newIORef Seq.empty -- | Non-blocking read. tryReadSmplChan :: SmplChan a -> IO (Maybe a) tryReadSmplChan ch = do x <- atomicModifyIORef' ch $ \ sq -> case Seq.viewl sq of Seq.EmptyL -> (Seq.empty, Nothing) h Seq.:< t -> (t, Just h) return x -- | A synchronous read that must block or busy-wait until a value is available. readSmplChan :: SmplChan a -> IO a readSmplChan ch = loop =<< newBackoff maxWait where loop bk = do x <- tryReadSmplChan ch case x of Nothing -> do b2 <- backoff bk loop b2 Just h -> return h -- | Always succeeds. Asynchronous write to channel. writeSmplChan :: SmplChan a -> a -> IO () writeSmplChan ch x = do atomicModifyIORef' ch $ \ s -> (s Seq.|> x,()) ---------------------------------------------------------------------------------------------------- {-# NOINLINE theEnv #-} theEnv :: [(String, String)] theEnv = unsafePerformIO getEnvironment -- | Debugging flag shared by several modules. -- This is activated by setting the environment variable @DEBUG=1..5@. -- -- By convention @DEBUG=100@ turns on full sequentialization of the program and -- control over the interleavings in concurrent code, enabling systematic debugging -- of concurrency problems. dbgLvl :: Int #ifdef DEBUG_LVAR {-# NOINLINE dbgLvl #-} dbgLvl = case lookup "DEBUG" theEnv of Nothing -> defaultDbg Just "" -> defaultDbg Just "0" -> defaultDbg Just s -> case reads s of ((n,_):_) -> trace (" [!] LVish responding to env Var: DEBUG="++show n) n [] -> error$"Attempt to parse DEBUG env var as Int failed: "++show s #else {-# INLINE dbgLvl #-} dbgLvl = 0 #endif defaultDbg :: Int defaultDbg = 0 replayDbg :: Int replayDbg = 100 -- | Exceptions that walk up the fork-tree of threads. -- -- WARNING: By holding onto the ThreadId we keep the parent thread from being -- garbage collected (at least as of GHC 7.6). This means that even if it was -- complete, it will still be hanging around to accept the exception below. forkWithExceptions :: (IO () -> IO ThreadId) -> String -> IO () -> IO ThreadId forkWithExceptions forkit descr action = do parent <- myThreadId forkit $ do tid <- myThreadId E.catch action (\ e -> case E.fromException e of Just E.ThreadKilled -> do -- Killing worker threads is normal now when exception handling, so this chatter is restricted to debug mode: #ifdef DEBUG_LVAR printf "\nThreadKilled exception inside child thread, %s (not propagating!): %s\n" (show tid) (show descr) #endif return () _ -> do #ifdef DEBUG_LVAR printf "\nException inside child thread %s, %s: %s\n" (show descr) (show tid) (show e) #endif E.throwTo parent (e :: E.SomeException) )