git-annex/Command/P2P.hs

328 lines
11 KiB
Haskell
Raw Normal View History

2016-11-30 18:35:24 +00:00
{- git-annex command
-
- Copyright 2016 Joey Hess <id@joeyh.name>
-
- Licensed under the GNU AGPL version 3 or higher.
2016-11-30 18:35:24 +00:00
-}
{-# LANGUAGE OverloadedStrings #-}
2016-11-30 18:35:24 +00:00
module Command.P2P where
import Command
import P2P.Address
import P2P.Auth
import P2P.IO
import qualified P2P.Protocol as P2P
import Git.Types
import qualified Git.Remote
import qualified Git.Command
import qualified Annex
import Annex.UUID
import Config
import Utility.AuthToken
import Utility.Tmp.Dir
import Utility.FileMode
import Utility.ThreadScheduler
import qualified Utility.MagicWormhole as Wormhole
import Control.Concurrent.Async
import qualified Data.Text as T
import Data.Time.Clock.POSIX
2016-11-30 18:35:24 +00:00
cmd :: Command
cmd = command "p2p" SectionSetup
"configure peer-2-peer links between repositories"
paramNothing (seek <$$> optParser)
data P2POpts
= GenAddresses
| LinkRemote
| Pair
2016-11-30 18:35:24 +00:00
optParser :: CmdParamsDesc -> Parser (P2POpts, Maybe RemoteName)
optParser _ = (,)
<$> (pair <|> linkremote <|> genaddresses)
<*> optional name
2016-11-30 18:35:24 +00:00
where
genaddresses = flag' GenAddresses
( long "gen-addresses"
<> help "generate addresses that allow accessing this repository over P2P networks"
)
linkremote = flag' LinkRemote
( long "link"
<> help "set up a P2P link to a git remote"
)
pair = flag' Pair
( long "pair"
<> help "pair with another repository"
)
name = Git.Remote.makeLegalName <$> strOption
( long "name"
<> metavar paramName
<> help "name of remote"
)
2016-11-30 18:35:24 +00:00
seek :: (P2POpts, Maybe RemoteName) -> CommandSeek
seek (GenAddresses, _) = genAddresses =<< loadP2PAddresses
seek (LinkRemote, Just name) = commandAction $
linkRemote name
seek (LinkRemote, Nothing) = commandAction $
linkRemote =<< unusedPeerRemoteName
seek (Pair, Just name) = commandAction $
2016-12-18 21:01:15 +00:00
startPairing name =<< loadP2PAddresses
seek (Pair, Nothing) = commandAction $ do
name <- unusedPeerRemoteName
2016-12-18 21:01:15 +00:00
startPairing name =<< loadP2PAddresses
unusedPeerRemoteName :: Annex RemoteName
unusedPeerRemoteName = go (1 :: Integer) =<< usednames
where
usednames = mapMaybe remoteName <$> Annex.getGitRemotes
go n names = do
let name = "peer" ++ show n
if name `elem` names
then go (n+1) names
else return name
-- Only addresses are output to stdout, to allow scripting.
genAddresses :: [P2PAddress] -> Annex ()
genAddresses [] = giveup "No P2P networks are currrently available."
genAddresses addrs = do
authtoken <- liftIO $ genAuthToken 128
storeP2PAuthToken authtoken
earlyWarning "These addresses allow access to this git-annex repository. Only share them with people you trust with that access, using trusted communication channels!"
liftIO $ putStr $ unlines $
map formatP2PAddress $
map (`P2PAddressAuth` authtoken) addrs
-- Address is read from stdin, to avoid leaking it in shell history.
linkRemote :: RemoteName -> CommandStart
make CommandStart return a StartMessage The goal is to be able to run CommandStart in the main thread when -J is used, rather than unncessarily passing it off to a worker thread, which incurs overhead that is signficant when the CommandStart is going to quickly decide to stop. To do that, the message it displays needs to be displayed in the worker thread, after the CommandStart has run. Also, the change will mean that CommandStart will no longer necessarily run with the same Annex state as CommandPerform. While its docs already said it should avoid modifying Annex state, I audited all the CommandStart code as part of the conversion. (Note that CommandSeek already sometimes runs with a different Annex state, and that has not been a source of any problems, so I am not too worried that this change will lead to breakage going forward.) The only modification of Annex state I found was it calling allowMessages in some Commands that default to noMessages. Dealt with that by adding a startCustomOutput and a startingUsualMessages. This lets a command start with noMessages and then select the output it wants for each CommandStart. One bit of breakage: onlyActionOn has been removed from commands that used it. The plan is that, since a StartMessage contains an ActionItem, when a Key can be extracted from that, the parallel job runner can run onlyActionOn' automatically. Then commands won't need to worry about this detail. Future work. Otherwise, this was a fairly straightforward process of making each CommandStart compile again. Hopefully other behavior changes were mostly avoided. In a few cases, a command had a CommandStart that called a CommandPerform that then called showStart multiple times. I have collapsed those down to a single start action. The main command to perhaps suffer from it is Command.Direct, which used to show a start for each file, and no longer does. Another minor behavior change is that some commands used showStart before, but had an associated file and a Key available, so were changed to ShowStart with an ActionItemAssociatedFile. That will not change the normal output or behavior, but --json output will now include the key. This should not break it for anyone using a real json parser.
2019-06-06 19:42:30 +00:00
linkRemote remotename = starting "p2p link" (ActionItemOther (Just remotename)) $
next promptaddr
where
2017-05-11 22:31:14 +00:00
promptaddr = do
liftIO $ putStrLn ""
liftIO $ putStr "Enter peer address: "
liftIO $ hFlush stdout
s <- liftIO getLine
if null s
then do
liftIO $ hPutStrLn stderr "Nothing entered, giving up."
return False
else case unformatP2PAddress s of
Nothing -> do
liftIO $ hPutStrLn stderr "Unable to parse that address, please check its format and try again."
2017-05-11 22:31:14 +00:00
promptaddr
Just addr -> do
r <- setupLink remotename addr
case r of
LinkSuccess -> return True
ConnectionError e -> giveup e
AuthenticationError e -> giveup e
2016-12-18 21:01:15 +00:00
startPairing :: RemoteName -> [P2PAddress] -> CommandStart
startPairing _ [] = giveup "No P2P networks are currrently available."
make CommandStart return a StartMessage The goal is to be able to run CommandStart in the main thread when -J is used, rather than unncessarily passing it off to a worker thread, which incurs overhead that is signficant when the CommandStart is going to quickly decide to stop. To do that, the message it displays needs to be displayed in the worker thread, after the CommandStart has run. Also, the change will mean that CommandStart will no longer necessarily run with the same Annex state as CommandPerform. While its docs already said it should avoid modifying Annex state, I audited all the CommandStart code as part of the conversion. (Note that CommandSeek already sometimes runs with a different Annex state, and that has not been a source of any problems, so I am not too worried that this change will lead to breakage going forward.) The only modification of Annex state I found was it calling allowMessages in some Commands that default to noMessages. Dealt with that by adding a startCustomOutput and a startingUsualMessages. This lets a command start with noMessages and then select the output it wants for each CommandStart. One bit of breakage: onlyActionOn has been removed from commands that used it. The plan is that, since a StartMessage contains an ActionItem, when a Key can be extracted from that, the parallel job runner can run onlyActionOn' automatically. Then commands won't need to worry about this detail. Future work. Otherwise, this was a fairly straightforward process of making each CommandStart compile again. Hopefully other behavior changes were mostly avoided. In a few cases, a command had a CommandStart that called a CommandPerform that then called showStart multiple times. I have collapsed those down to a single start action. The main command to perhaps suffer from it is Command.Direct, which used to show a start for each file, and no longer does. Another minor behavior change is that some commands used showStart before, but had an associated file and a Key available, so were changed to ShowStart with an ActionItemAssociatedFile. That will not change the normal output or behavior, but --json output will now include the key. This should not break it for anyone using a real json parser.
2019-06-06 19:42:30 +00:00
startPairing remotename addrs = ifM (liftIO Wormhole.isInstalled)
( starting "p2p pair" (ActionItemOther (Just remotename)) $
performPairing remotename addrs
, giveup "Magic Wormhole is not installed, and is needed for pairing. Install it from your distribution or from https://github.com/warner/magic-wormhole/"
)
2016-12-18 21:01:15 +00:00
performPairing :: RemoteName -> [P2PAddress] -> CommandPerform
performPairing remotename addrs = do
-- This note is displayed mainly so when magic wormhole
-- complains about possible protocol mismatches or other problems,
-- it's clear what's doing the complaining.
2016-12-18 21:31:02 +00:00
showNote "using Magic Wormhole"
2016-12-18 21:01:15 +00:00
next $ do
showOutput
r <- wormholePairing remotename addrs ui
case r of
PairSuccess -> return True
SendFailed -> do
warning "Failed sending data to pair."
return False
ReceiveFailed -> do
warning "Failed receiving data from pair."
return False
LinkFailed e -> do
warning $ "Failed linking to pair: " ++ e
return False
where
ui observer producer = do
ourcode <- Wormhole.waitCode observer
putStrLn ""
putStrLn $ "This repository's pairing code is: " ++
Wormhole.fromCode ourcode
putStrLn ""
theircode <- getcode ourcode
Wormhole.sendCode producer theircode
getcode ourcode = do
putStr "Enter the other repository's pairing code: "
hFlush stdout
l <- getLine
case Wormhole.toCode l of
Just code
2016-12-18 21:01:15 +00:00
| code /= ourcode -> do
2016-12-18 21:13:06 +00:00
putStrLn "Exchanging pairing data..."
2016-12-18 21:01:15 +00:00
return code
| otherwise -> do
putStrLn "Oops -- You entered this repository's pairing code. Enter the pairing code of the *other* repository."
getcode ourcode
Nothing -> do
2016-12-24 20:56:56 +00:00
putStrLn "That does not look like a valiad pairing code. Try again..."
getcode ourcode
-- We generate half of the authtoken; the pair will provide
-- the other half.
newtype HalfAuthToken = HalfAuthToken T.Text
deriving (Show)
data PairData = PairData HalfAuthToken [P2PAddress]
deriving (Show)
serializePairData :: PairData -> String
serializePairData (PairData (HalfAuthToken ha) addrs) = unlines $
T.unpack ha : map formatP2PAddress addrs
deserializePairData :: String -> Maybe PairData
deserializePairData s = case lines s of
[] -> Nothing
(ha:l) -> do
addrs <- mapM unformatP2PAddress l
return (PairData (HalfAuthToken (T.pack ha)) addrs)
data PairingResult
= PairSuccess
| SendFailed
| ReceiveFailed
| LinkFailed String
wormholePairing
:: RemoteName
-> [P2PAddress]
-> (Wormhole.CodeObserver -> Wormhole.CodeProducer -> IO ())
-> Annex PairingResult
wormholePairing remotename ouraddrs ui = do
ourhalf <- liftIO $ HalfAuthToken . fromAuthToken
<$> genAuthToken 64
let ourpairdata = PairData ourhalf ouraddrs
-- The magic wormhole interface only supports exchanging
-- files. Permissions of received files may allow others
-- to read them. So, set up a temp directory that only
-- we can read.
withTmpDir "pair" $ \tmp -> do
liftIO $ void $ tryIO $ modifyFileMode tmp $
removeModes otherGroupModes
let sendf = tmp </> "send"
let recvf = tmp </> "recv"
liftIO $ writeFileProtected sendf $
serializePairData ourpairdata
observer <- liftIO Wormhole.mkCodeObserver
producer <- liftIO Wormhole.mkCodeProducer
void $ liftIO $ async $ ui observer producer
-- Provide an appid to magic wormhole, to avoid using
-- the same channels that other wormhole users use.
--
-- Since a version of git-annex that did not provide an
-- appid is shipping in Debian 9, and having one side
-- provide an appid while the other does not will make
-- wormhole fail, this is deferred until 2021-12-31.
-- After that point, all git-annex's should have been
-- upgraded to include this code, and they will start
-- providing an appid.
--
-- This assumes reasonably good client clocks. If the clock
-- is completely wrong, it won't use the appid at that
-- point, and pairing will fail. On 2021-12-31, minor clock
-- skew may also cause transient problems.
--
-- After 2021-12-31, this can be changed to simply
-- always provide the appid.
now <- liftIO getPOSIXTime
let wormholeparams = if now < 1640950000
then []
else Wormhole.appId "git-annex.branchable.com/p2p-setup"
(sendres, recvres) <- liftIO $
Wormhole.sendFile sendf observer wormholeparams
`concurrently`
Wormhole.receiveFile recvf producer wormholeparams
liftIO $ nukeFile sendf
if sendres /= True
then return SendFailed
else if recvres /= True
then return ReceiveFailed
else do
r <- liftIO $ tryIO $
readFileStrict recvf
case r of
Left _e -> return ReceiveFailed
Right s -> maybe
(return ReceiveFailed)
(finishPairing 100 remotename ourhalf)
(deserializePairData s)
-- | Allow the peer we're pairing with to authenticate to us,
-- using an authtoken constructed from the two HalfAuthTokens.
-- Connect to the peer we're pairing with, and try to link to them.
--
-- Multiple addresses may have been received for the peer. This only
-- makes a link to one address.
--
-- Since we're racing the peer as they do the same, the first try is likely
-- to fail to authenticate. Can retry any number of times, to avoid the
-- users needing to redo the whole process.
finishPairing :: Int -> RemoteName -> HalfAuthToken -> PairData -> Annex PairingResult
finishPairing retries remotename (HalfAuthToken ourhalf) (PairData (HalfAuthToken theirhalf) theiraddrs) = do
case (toAuthToken (ourhalf <> theirhalf), toAuthToken (theirhalf <> ourhalf)) of
(Just ourauthtoken, Just theirauthtoken) -> do
2016-12-18 21:13:06 +00:00
liftIO $ putStrLn $ "Successfully exchanged pairing data. Connecting to " ++ remotename ++ "..."
storeP2PAuthToken ourauthtoken
go retries theiraddrs theirauthtoken
_ -> return ReceiveFailed
where
go 0 [] _ = return $ LinkFailed $ "Unable to connect to " ++ remotename ++ "."
go n [] theirauthtoken = do
liftIO $ threadDelaySeconds (Seconds 2)
liftIO $ putStrLn $ "Unable to connect to " ++ remotename ++ ". Retrying..."
go (n-1) theiraddrs theirauthtoken
go n (addr:rest) theirauthtoken = do
r <- setupLink remotename (P2PAddressAuth addr theirauthtoken)
case r of
LinkSuccess -> return PairSuccess
_ -> go n rest theirauthtoken
data LinkResult
= LinkSuccess
| ConnectionError String
| AuthenticationError String
setupLink :: RemoteName -> P2PAddressAuth -> Annex LinkResult
setupLink remotename (P2PAddressAuth addr authtoken) = do
g <- Annex.gitRepo
cv <- liftIO $ tryNonAsync $ connectPeer g addr
case cv of
Left e -> return $ ConnectionError $ "Unable to connect with peer. Please check that the peer is connected to the network, and try again. (" ++ show e ++ ")"
Right conn -> do
u <- getUUID
let proto = P2P.auth u authtoken noop
runst <- liftIO $ mkRunState Client
go =<< liftIO (runNetProto runst conn proto)
where
go (Right (Just theiruuid)) = do
ok <- inRepo $ Git.Command.runBool
[ Param "remote", Param "add"
, Param remotename
, Param (formatP2PAddress addr)
]
when ok $ do
storeUUIDIn (remoteConfig remotename "uuid") theiruuid
storeP2PRemoteAuthToken addr authtoken
return LinkSuccess
go (Right Nothing) = return $ AuthenticationError "Unable to authenticate with peer. Please check the address and try again."
go (Left e) = return $ AuthenticationError $ "Unable to authenticate with peer: " ++ describeProtoFailure e