git-annex-shell: proxy nodes located beyond remote cluster gateways

Walking a tightrope between security and convenience here, because
git-annex-shell needs to only proxy for things when there has been
an explicit, local action to configure them.

In this case, the user has to have run `git-annex extendcluster`,
which now sets annex-cluster-gateway on the remote.

Note that any repositories that the gateway is recorded to
proxy for will be proxied onward. This is not limited to cluster nodes,
because checking the node log would not add any security; someone could
add any uuid to it. The gateway of course then does its own
checking to determine if it will allow proxying for the remote.

CmdLine/GitAnnexShell.hs

@@ -206,18 +206,30 @@ checkProxy remoteuuid ouruuid = M.lookup ouruuid <$> getProxies >>= \case
 		rs <- concat . byCost <$> remoteList
 		myclusters <- annexClusters <$> Annex.getGitConfig
 		let sameuuid r = uuid r == remoteuuid
-		-- Only proxy for a remote when the git configuration
-		-- allows it.
-		let proxyconfigured r = remoteAnnexProxy (R.gitconfig r)
-			|| (any (`M.member` myclusters) $ fromMaybe [] $ remoteAnnexClusterNode $ R.gitconfig r)
 		let samename r p = name r == proxyRemoteName p
-		case headMaybe (filter (\r -> sameuuid r && proxyconfigured r && any (samename r) ps) rs) of
+		case headMaybe (filter (\r -> sameuuid r && proxyisconfigured rs myclusters r && any (samename r) ps) rs) of
 			Nothing -> notconfigured
 			Just r -> do
 				Annex.changeState $ \st ->
 					st { Annex.proxyremote = Just (Right r) }
 				return True
 	
+	-- Only proxy for a remote when the git configuration
+	-- allows it. This is important to prevent changes to 
+	-- the git-annex branch making git-annex-shell unexpectedly
+	-- proxy for remotes.
+	proxyisconfigured rs myclusters r
+		| remoteAnnexProxy (R.gitconfig r) = True
+		-- Proxy for remotes that are configured as cluster nodes.
+		| any (`M.member` myclusters) (fromMaybe [] $ remoteAnnexClusterNode $ R.gitconfig r) = True
+		-- Proxy for a remote when it is proxied by another remote
+		-- which is itself configured as a cluster gateway.
+		| otherwise = case remoteAnnexProxiedBy (R.gitconfig r) of
+			Just proxyuuid -> not $ null $ 
+				concatMap (remoteAnnexClusterGateway . R.gitconfig) $
+					filter (\p -> R.uuid p == proxyuuid) rs
+			Nothing -> False
+
 	proxyforcluster cu = do
 		clusters <- getClusters
 		if M.member cu (clusterUUIDs clusters)

Command/ExtendCluster.hs

@@ -13,6 +13,7 @@ import Command
 import qualified Annex
 import Types.Cluster
 import Config
+import Types.GitConfig
 import qualified Remote
 
 import qualified Data.Map as M
@@ -23,11 +24,13 @@ cmd = command "extendcluster" SectionSetup "add an gateway to a cluster"
 
 seek :: CmdParams -> CommandSeek
 seek (remotename:clustername:[]) = Remote.byName (Just clusterremotename) >>= \case
-	Just clusterremote -> 
+	Just clusterremote -> Remote.byName (Just remotename) >>= \case
+		Just gatewayremote -> 
 			case mkClusterUUID (Remote.uuid clusterremote) of
-			Just cu -> commandAction $ start cu clustername
+				Just cu -> commandAction $ start cu clustername gatewayremote
 				Nothing -> giveup $ clusterremotename 
 					++ " is not a cluster remote."
+		Nothing -> giveup $ "No remote named " ++ remotename ++ " exists."
 	Nothing -> giveup $ "Expected to find a cluster remote named " 
 		++ clusterremotename
 		++ " that is accessed via " ++ remotename
@@ -38,12 +41,14 @@ seek (remotename:clustername:[]) = Remote.byName (Just clusterremotename) >>= \c
 	clusterremotename = remotename ++ "-" ++ clustername
 seek _ = giveup "Expected two parameters, gateway and clustername."
 
-start :: ClusterUUID -> String -> CommandStart
+start :: ClusterUUID -> String -> Remote -> CommandStart
-start cu clustername = starting "extendcluster" ai si $ do
+start cu clustername gatewayremote = starting "extendcluster" ai si $ do
 	myclusters <- annexClusters <$> Annex.getGitConfig
+	let setcus f = setConfig f (fromUUID (fromClusterUUID cu))
 	unless (M.member clustername myclusters) $ do
-		setConfig (annexConfig ("cluster." <> encodeBS clustername))
+		setcus $ annexConfig ("cluster." <> encodeBS clustername)
-			(fromUUID (fromClusterUUID cu))
+	setcus $ remoteAnnexConfig gatewayremote $ 
+		remoteGitConfigKey ClusterGatewayField
 	next $ return True
   where
 	ai = ActionItemOther (Just (UnquotedString clustername))

Command/UpdateCluster.hs

@@ -80,5 +80,5 @@ findProxiedClusterNodes recordednodes =
   where
 	isproxynode r = 
 		asclusternode r `S.member` recordednodes
-			&& remoteAnnexProxied (R.gitconfig r)
+			&& isJust (remoteAnnexProxiedBy (R.gitconfig r))
 	asclusternode = ClusterNodeUUID . R.uuid

Command/UpdateProxy.hs

@@ -83,7 +83,7 @@ findRemoteProxiedClusterNodes = do
 		<$> Annex.getGitConfig
 	clusternodes <- clusterNodeUUIDs <$> getClusters
 	let isproxiedclusternode r
-		| remoteAnnexProxied (R.gitconfig r) =
+		| isJust (remoteAnnexProxiedBy (R.gitconfig r)) =
 			case M.lookup (ClusterNodeUUID (R.uuid r)) clusternodes of
 				Nothing -> False
 				Just s -> not $ S.null $ 

Remote.hs

@@ -455,7 +455,7 @@ gitSyncableRemote :: Remote -> Bool
 gitSyncableRemote r
 	| gitSyncableRemoteType (remotetype r) 
 		&& isJust (remoteUrl (gitconfig r)) =
-			not (remoteAnnexProxied (gitconfig r))
+			not (isJust (remoteAnnexProxiedBy (gitconfig r)))
 	| otherwise = case remoteUrl (gitconfig r) of
 		Just u | "annex::" `isPrefixOf` u -> True
 		_ -> False

Remote/Git.hs

@@ -794,21 +794,22 @@ listProxied proxies rs = concat <$> mapM go rs
 			then pure []
 			else case M.lookup cu proxies of
 				Nothing -> pure []
-				Just s -> catMaybes
+				Just proxied -> catMaybes
-					<$> mapM (mkproxied g r s) (S.toList s)
+					<$> mapM (mkproxied g r gc proxied)
+						(S.toList proxied)
 	
 	proxiedremotename r p = do
 		n <- Git.remoteName r
 		pure $ n ++ "-" ++ proxyRemoteName p
 
-	mkproxied g r proxied p = case proxiedremotename r p of
+	mkproxied g r gc proxied p = case proxiedremotename r p of
 		Nothing -> pure Nothing
-		Just proxyname -> mkproxied' g r proxied p proxyname
+		Just proxyname -> mkproxied' g r gc proxied p proxyname
 	
 	-- The proxied remote is constructed by renaming the proxy remote,
 	-- changing its uuid, and setting the proxied remote's inherited
 	-- configs and uuid in Annex state.
-	mkproxied' g r proxied p proxyname
+	mkproxied' g r gc proxied p proxyname
 		| any isconfig (M.keys (Git.config g)) = pure Nothing
 		| otherwise = do
 			clusters <- getClustersWith id
@@ -830,7 +831,7 @@ listProxied proxies rs = concat <$> mapM go rs
 		annexconfigadjuster clusters r' = 
 			let c = adduuid (configRepoUUID renamedr) $
 				addurl $
-				addproxied $
+				addproxiedby $
 				adjustclusternode clusters $
 				inheritconfigs $ Git.fullconfig r'
 			in r'
@@ -844,7 +845,10 @@ listProxied proxies rs = concat <$> mapM go rs
 		addurl = M.insert (remoteConfig renamedr (remoteGitConfigKey UrlField))
 			[Git.ConfigValue $ encodeBS $ Git.repoLocation r]
 		
-		addproxied = addremoteannexfield ProxiedField True
+		addproxiedby = case remoteAnnexUUID gc of
+			Just u -> addremoteannexfield ProxiedByField
+				[Git.ConfigValue $ fromUUID u]
+			Nothing -> id
 		
 		-- A node of a cluster that is being proxied along with
 		-- that cluster does not need to be synced with
@@ -854,14 +858,14 @@ listProxied proxies rs = concat <$> mapM go rs
 			case M.lookup (ClusterNodeUUID (proxyRemoteUUID p)) (clusterNodeUUIDs clusters) of
 				Just cs
 					| any (\c -> S.member (fromClusterUUID c) proxieduuids) (S.toList cs) ->
-						addremoteannexfield SyncField False
+						addremoteannexfield SyncField
+							[Git.ConfigValue $ Git.Config.boolConfig' False]
 				_ -> id
 
 		proxieduuids = S.map proxyRemoteUUID proxied
 
-		addremoteannexfield f b = M.insert
+		addremoteannexfield f = M.insert
 			(remoteAnnexConfig renamedr (remoteGitConfigKey f))
-			[Git.ConfigValue $ Git.Config.boolConfig' b]
 
 		inheritconfigs c = foldl' inheritconfig c proxyInheritedFields
 		

Remote/Helper/Git.hs

@@ -65,6 +65,6 @@ gitRepoInfo r = do
 	return
 		[ ("repository location", Git.repoLocation repo)
 		, ("proxied", Git.Config.boolConfig 
-			(remoteAnnexProxied (Remote.gitconfig r)))
+			(isJust (remoteAnnexProxiedBy (Remote.gitconfig r))))
 		, ("last synced", lastsynctime)
 		]

Types/GitConfig.hs

@@ -386,12 +386,14 @@ data RemoteGitConfig = RemoteGitConfig
 	, remoteAnnexBwLimitUpload :: Maybe BwRate
 	, remoteAnnexBwLimitDownload :: Maybe BwRate
 	, remoteAnnexAllowUnverifiedDownloads :: Bool
+	, remoteAnnexUUID :: Maybe UUID
 	, remoteAnnexConfigUUID :: Maybe UUID
 	, remoteAnnexMaxGitBundles :: Int
 	, remoteAnnexAllowEncryptedGitRepo :: Bool
 	, remoteAnnexProxy :: Bool
-	, remoteAnnexProxied :: Bool
+	, remoteAnnexProxiedBy :: Maybe UUID
 	, remoteAnnexClusterNode :: Maybe [RemoteName]
+	, remoteAnnexClusterGateway :: [ClusterUUID]
 	, remoteUrl :: Maybe String
 
 	{- These settings are specific to particular types of remotes
@@ -471,16 +473,20 @@ extractRemoteGitConfig r remotename = do
 			readBwRatePerSecond =<< getmaybe BWLimitDownloadField
 		, remoteAnnexAllowUnverifiedDownloads = (== Just "ACKTHPPT") $
 			getmaybe SecurityAllowUnverifiedDownloadsField
+		, remoteAnnexUUID = toUUID <$> getmaybe UUIDField
 		, remoteAnnexConfigUUID = toUUID <$> getmaybe ConfigUUIDField
 		, remoteAnnexMaxGitBundles =
 			fromMaybe 100 (getmayberead MaxGitBundlesField)
 		, remoteAnnexAllowEncryptedGitRepo = 
 			getbool AllowEncryptedGitRepoField False
 		, remoteAnnexProxy = getbool ProxyField False
-		, remoteAnnexProxied = getbool ProxiedField False
+		, remoteAnnexProxiedBy = toUUID <$> getmaybe ProxiedByField
 		, remoteAnnexClusterNode = 
 			(filter isLegalName . words)
 				<$> getmaybe ClusterNodeField
+		, remoteAnnexClusterGateway = fromMaybe [] $
+			(mapMaybe (mkClusterUUID . toUUID) . words)
+				<$> getmaybe ClusterGatewayField
 		, remoteUrl = 
 			case Git.Config.getMaybe (remoteConfig remotename (remoteGitConfigKey UrlField)) r of
 				Just (ConfigValue b)
@@ -553,13 +559,15 @@ data RemoteGitConfigField
 	| BWLimitField
 	| BWLimitUploadField
 	| BWLimitDownloadField
+	| UUIDField
 	| ConfigUUIDField
 	| SecurityAllowUnverifiedDownloadsField
 	| MaxGitBundlesField
 	| AllowEncryptedGitRepoField
 	| ProxyField
-	| ProxiedField
+	| ProxiedByField
 	| ClusterNodeField
+	| ClusterGatewayField
 	| UrlField
 	| ShellField
 	| SshOptionsField
@@ -618,14 +626,16 @@ remoteGitConfigField = \case
 	BWLimitField -> inherited "bwlimit"
 	BWLimitUploadField -> inherited "bwlimit-upload"
 	BWLimitDownloadField -> inherited "bwlimit-upload"
+	UUIDField -> uninherited "uuid"
 	ConfigUUIDField -> uninherited "config-uuid"
 	SecurityAllowUnverifiedDownloadsField -> inherited "security-allow-unverified-downloads"
 	MaxGitBundlesField -> inherited "max-git-bundles"
 	AllowEncryptedGitRepoField -> inherited "allow-encrypted-gitrepo"
 	-- Allow proxy chains.
 	ProxyField -> inherited "proxy"
-	ProxiedField -> uninherited "proxied"
+	ProxiedByField -> uninherited "proxied-by"
 	ClusterNodeField -> uninherited "cluster-node"
+	ClusterGatewayField -> uninherited "cluster-gateway"
 	UrlField -> uninherited "url"
 	ShellField -> inherited "shell"
 	SshOptionsField -> inherited "ssh-options"

doc/git-annex.mdwn

@@ -1398,11 +1398,12 @@ repository, using [[git-annex-config]]. See its man page for a list.)
 
 * `annex.cluster.<name>`
 
-  [[git-annex-updatecluster]] sets this to the UUID of a cluster
+  This is set to make the repository be a gateway to a cluster.
-  based on `remote.<name>.annex-cluster-node` configuration.
+  The value is the cluster UUID. Note that cluster UUIDs are not
+  the same as repository UUIDs, and a repository UUID cannot be used here.
 
-  Note that cluster UUIDs are not the same as repository UUIDs,
+  Usually this is set up by running [[git-annex-initcluster]] or
-  and a repository UUID cannot be used here.
+  [[git-annex-extendcluster]].
 
 # CONFIGURATION OF REMOTES
 
@@ -1681,14 +1682,11 @@ Remotes are configured using these settings in `.git/config`.
   After configuring this, run [[git-annex-updateproxy](1) to store
   the new configuration in the git-annex branch.
 
-* `remote.<name>.annex-proxied`
+* `remote.<name>.annex-proxied-by`
-
-  Setting this to "true" indicates that a remote is proxied via the
-  git-annex repository that its remote points to. That prevents commands
-  like `git-annex sync` from pulling and pushing the remote. 
 
   Usually this is used internally, when git-annex sets up proxied remotes,
-  and will not need to be set.
+  and will not need to be configured. The value is the UUID of the
+  git-annex repository that proxies access to this remote.
 
 * `remote.<name>.annex-cluster-node`
 
@@ -1699,6 +1697,15 @@ Remotes are configured using these settings in `.git/config`.
   After configuring this, run [[git-annex-updatecluster](1) to store
   the new configuration in the git-annex branch.
 
+* `remote.<name>.annex-cluster-gateway`
+
+  Set to the UUID of a cluster that this remote serves as a gateway for.
+  Multiple UUIDs can be listed, separated by whitespace. When the local
+  repository is also a gateway for that cluster, it will proxy for the
+  nodes of the remote gateway.
+  
+  Usually this is set up by running [[git-annex-extendcluster]].
+
 * `remote.<name>.annex-private`
 
   When this is set to true, no information about the remote will be

doc/todo/git-annex_proxies.mdwn

@@ -33,8 +33,11 @@ For June's work on [[design/passthrough_proxy]], remaining todos:
   protocol messages on to any remotes that have the same UUID as
   the cluster. Needs VIA extension to P2P protocol to avoid cycles.
 
-* `git-annex updatecluster` needs changes to support a distributed cluster.
+  Current status: Distributed cluster nodes are visible,
-  Currently it will remove nodes that are behind another gateway.
+  and can be accessed directly, but trying to GET from a cluster
+  fails when the content is located behind a remote gateway.
+  And PUT only sends to the immediate nodes
+  of the cluster, not on to other gateways.
 
 * Getting a key from a cluster currently always selects the lowest cost
   remote, and always the same remote if cost is the same. Should
@@ -106,3 +109,8 @@ For June's work on [[design/passthrough_proxy]], remaining todos:
 
 * Support annex.jobs for clusters. (done)
 
+* Add `git-annex extendcluster` command and extend `git-annex updatecluster` 
+  to support clusters with multiple gateways. (done)
+
+* Support proxying for a remote that is proxied by another gateway of
+  a cluster. (done)