git-annex/Annex/Ssh.hs

140 lines
4.3 KiB
Haskell
Raw Normal View History

{- git-annex ssh interface, with connection caching
-
- Copyright 2012 Joey Hess <joey@kitenet.net>
-
- Licensed under the GNU GPL version 3 or higher.
-}
module Annex.Ssh (
sshParams,
sshCleanup,
) where
import qualified Data.Map as M
import Common.Annex
import Annex.LockPool
Clean up handling of git directory and git worktree. Baked into the code was an assumption that a repository's git directory could be determined by adding ".git" to its work tree (or nothing for bare repos). That fails when core.worktree, or GIT_DIR and GIT_WORK_TREE are used to separate the two. This was attacked at the type level, by storing the gitdir and worktree separately, so Nothing for the worktree means a bare repo. A complication arose because we don't learn where a repository is bare until its configuration is read. So another Location type handles repositories that have not had their config read yet. I am not entirely happy with this being a Location type, rather than representing them entirely separate from the Git type. The new code is not worse than the old, but better types could enforce more safety. Added support for core.worktree. Overriding it with -c isn't supported because it's not really clear what to do if a git repo's config is read, is not bare, and is then overridden to bare. What is the right git directory in this case? I will worry about this if/when someone has a use case for overriding core.worktree with -c. (See Git.Config.updateLocation) Also removed and renamed some functions like gitDir and workTree that misused git's terminology. One minor regression is known: git annex add in a bare repository does not print a nice error message, but runs git ls-files in a way that fails earlier with a less nice error message. This is because before --work-tree was always passed to git commands, even in a bare repo, while now it's not.
2012-05-18 20:38:26 +00:00
import qualified Git.Config
2012-03-22 04:23:15 +00:00
import Config
import qualified Build.SysConfig as SysConfig
import Annex.Perms
{- Generates parameters to ssh to a given host (or user@host) on a given
- port, with connection caching. -}
sshParams :: (String, Maybe Integer) -> [CommandParam] -> Annex [CommandParam]
sshParams (host, port) opts = go =<< sshInfo (host, port)
where
go (Nothing, params) = ret params
2012-01-20 21:13:36 +00:00
go (Just socketfile, params) = do
cleanstale
liftIO $ createDirectoryIfMissing True $ parentDir socketfile
lockFile $ socket2lock socketfile
ret params
ret ps = return $ ps ++ opts ++ portParams port ++ [Param host]
-- If the lock pool is empty, this is the first ssh of this
-- run. There could be stale ssh connections hanging around
-- from a previous git-annex run that was interrupted.
2012-02-16 04:41:30 +00:00
cleanstale = whenM (not . any isLock . M.keys <$> getPool) $
sshCleanup
2012-01-20 21:13:36 +00:00
sshInfo :: (String, Maybe Integer) -> Annex (Maybe FilePath, [CommandParam])
sshInfo (host, port) = ifM caching
( do
dir <- fromRepo gitAnnexSshDir
let socketfile = dir </> hostport2socket host port
if valid_unix_socket_path socketfile
then return (Just socketfile, cacheParams socketfile)
else do
socketfile' <- liftIO $ relPathCwdToFile socketfile
if valid_unix_socket_path socketfile'
then return (Just socketfile', cacheParams socketfile')
else return (Nothing, [])
, return (Nothing, [])
)
where
caching = fromMaybe SysConfig.sshconnectioncaching
Clean up handling of git directory and git worktree. Baked into the code was an assumption that a repository's git directory could be determined by adding ".git" to its work tree (or nothing for bare repos). That fails when core.worktree, or GIT_DIR and GIT_WORK_TREE are used to separate the two. This was attacked at the type level, by storing the gitdir and worktree separately, so Nothing for the worktree means a bare repo. A complication arose because we don't learn where a repository is bare until its configuration is read. So another Location type handles repositories that have not had their config read yet. I am not entirely happy with this being a Location type, rather than representing them entirely separate from the Git type. The new code is not worse than the old, but better types could enforce more safety. Added support for core.worktree. Overriding it with -c isn't supported because it's not really clear what to do if a git repo's config is read, is not bare, and is then overridden to bare. What is the right git directory in this case? I will worry about this if/when someone has a use case for overriding core.worktree with -c. (See Git.Config.updateLocation) Also removed and renamed some functions like gitDir and workTree that misused git's terminology. One minor regression is known: git annex add in a bare repository does not print a nice error message, but runs git ls-files in a way that fails earlier with a less nice error message. This is because before --work-tree was always passed to git commands, even in a bare repo, while now it's not.
2012-05-18 20:38:26 +00:00
. Git.Config.isTrue
<$> getConfig (annexConfig "sshcaching") ""
cacheParams :: FilePath -> [CommandParam]
cacheParams socketfile =
[ Param "-S", Param socketfile
, Params "-o ControlMaster=auto -o ControlPersist=yes"
]
portParams :: Maybe Integer -> [CommandParam]
portParams Nothing = []
portParams (Just port) = [Param "-p", Param $ show port]
{- Stop any unused ssh processes. -}
sshCleanup :: Annex ()
sshCleanup = do
2012-02-16 04:41:30 +00:00
dir <- fromRepo gitAnnexSshDir
2012-03-06 17:56:20 +00:00
sockets <- filter (not . isLock) <$>
liftIO (catchDefaultIO (dirContents dir) [])
forM_ sockets cleanup
where
cleanup socketfile = do
-- Drop any shared lock we have, and take an
-- exclusive lock, without blocking. If the lock
-- succeeds, nothing is using this ssh, and it can
-- be stopped.
let lockfile = socket2lock socketfile
unlockFile lockfile
mode <- annexFileMode
fd <- liftIO $ noUmask mode $
openFd lockfile ReadWrite (Just mode) defaultFileFlags
v <- liftIO $ tryIO $
setLock fd (WriteLock, AbsoluteSeek, 0, 0)
case v of
2012-04-22 03:32:33 +00:00
Left _ -> noop
Right _ -> stopssh socketfile
liftIO $ closeFd fd
stopssh socketfile = do
let (host, port) = socket2hostport socketfile
(_, params) <- sshInfo (host, port)
void $ liftIO $ do
-- "ssh -O stop" is noisy on stderr even with -q
let cmd = unwords $ toCommand $
[ Params "-O stop"
] ++ params ++ [Param host]
boolSystem "sh"
[ Param "-c"
, Param $ "ssh " ++ cmd ++ " >/dev/null 2>/dev/null"
]
-- Cannot remove the lock file; other processes may
-- be waiting on our exclusive lock to use it.
hostport2socket :: String -> Maybe Integer -> FilePath
hostport2socket host Nothing = host
hostport2socket host (Just port) = host ++ "!" ++ show port
socket2hostport :: FilePath -> (String, Maybe Integer)
socket2hostport socket
| null p = (h, Nothing)
| otherwise = (h, readish p)
where
(h, p) = separate (== '!') $ takeFileName socket
socket2lock :: FilePath -> FilePath
socket2lock socket = socket ++ lockExt
isLock :: FilePath -> Bool
isLock f = lockExt `isSuffixOf` f
lockExt :: String
lockExt = ".lock"
{- This is the size of the sun_path component of sockaddr_un, which
- is the limit to the total length of the filename of a unix socket.
-
- On Linux, this is 108. On OSX, 104. TODO: Probe
-}
sizeof_sockaddr_un_sun_path :: Int
sizeof_sockaddr_un_sun_path = 100
{- Note that this looks at the true length of the path in bytes, as it will
- appear on disk. -}
valid_unix_socket_path :: FilePath -> Bool
valid_unix_socket_path f = length (decodeW8 f) < sizeof_sockaddr_un_sun_path