Make the garbage collector more concurrent

Make the garbage collector more concurrent by deleting valid paths
outside the region where we're holding the global GC lock.  This
should greatly reduce the time during which new builds are blocked,
since the deletion accounts for the vast majority of the time spent in
the GC.

To ensure that this is safe, the valid paths are invalidated and
renamed to some arbitrary path while we're holding the lock.  This
ensures that we when we finally delete the path, it's not a (newly)
valid or locked path.

3 files changed, 48 insertions, 21 deletions

diff --git a/src/libstore/gc.cc b/src/libstore/gc.cc
index d81bf40c73da..65c8dc210bd2 100644
--- a/src/libstore/gc.cc
+++ b/src/libstore/gc.cc
@@ -1,6 +1,7 @@
 #include "globals.hh"
 #include "misc.hh"
 #include "local-store.hh"
+#include "immutable.hh"
 
 #include <boost/shared_ptr.hpp>
 
@@ -396,11 +397,11 @@ struct LocalStore::GCState
     PathSet deleted;
     PathSet live;
     PathSet busy;
+    PathSet invalidated;
     bool gcKeepOutputs;
     bool gcKeepDerivations;
-    GCState(GCResults & results_) : results(results_)
-    {
-    }
+    unsigned long long bytesInvalidated;
+    GCState(GCResults & results_) : results(results_), bytesInvalidated(0) { }
 };
 
 
@@ -419,6 +420,16 @@ bool LocalStore::isActiveTempFile(const GCState & state,
 }
 
 
+void LocalStore::deleteGarbage(GCState & state, const Path & path)
+{
+    printMsg(lvlInfo, format("deleting `%1%'") % path);
+    unsigned long long bytesFreed, blocksFreed;
+    deletePathWrapped(path, bytesFreed, blocksFreed);
+    state.results.bytesFreed += bytesFreed;
+    state.results.blocksFreed += blocksFreed;
+}
+
+
 bool LocalStore::tryToDelete(GCState & state, const Path & path)
 {
     checkInterrupt();
@@ -502,15 +513,27 @@ bool LocalStore::tryToDelete(GCState & state, const Path & path)
 
     /* The path is garbage, so delete it. */
     if (shouldDelete(state.options.action)) {
-        printMsg(lvlInfo, format("deleting `%1%'") % path);
-
-        unsigned long long bytesFreed, blocksFreed;
-        deleteFromStore(path, bytesFreed, blocksFreed);
-        state.results.bytesFreed += bytesFreed;
-        state.results.blocksFreed += blocksFreed;
 
-        if (state.options.maxFreed && state.results.bytesFreed > state.options.maxFreed) {
-            printMsg(lvlInfo, format("deleted more than %1% bytes; stopping") % state.options.maxFreed);
+        if (isValidPath(path)) {
+            /* If it's a valid path, invalidate it, rename it, and
+               schedule it for deletion.  The renaming is to ensure
+               that later (when we're not holding the global GC lock)
+               we can delete the path without being afraid that the
+               path has become alive again. */
+            printMsg(lvlInfo, format("invalidating `%1%'") % path);
+            // Estimate the amount freed using the narSize field.
+            state.bytesInvalidated += queryPathInfo(path).narSize;
+            invalidatePathChecked(path);
+            Path tmp = (format("%1%-gc-%2%") % path % getpid()).str();
+            makeMutable(path.c_str());
+            if (rename(path.c_str(), tmp.c_str()))
+                throw SysError(format("unable to rename `%1%' to `%2%'") % path % tmp);
+            state.invalidated.insert(tmp);
+        } else
+            deleteGarbage(state, path);
+
+        if (state.options.maxFreed && state.results.bytesFreed + state.bytesInvalidated > state.options.maxFreed) {
+            printMsg(lvlInfo, format("deleted or invalidated more than %1% bytes; stopping") % state.options.maxFreed);
             throw GCLimitReached();
         }
 
@@ -635,6 +658,14 @@ void LocalStore::collectGarbage(const GCOptions & options, GCResults & results)
         } catch (GCLimitReached & e) {
         }
     }
+
+    /* Allow other processes to add to the store from here on. */
+    fdGCLock.close();
+
+    /* Delete the invalidated paths now that the lock has been
+       released. */
+    foreach (PathSet::iterator, i, state.invalidated)
+        deleteGarbage(state, *i);
 }
 
 
diff --git a/src/libstore/local-store.cc b/src/libstore/local-store.cc
index f3b779dd05f3..1ab9d15eb66b 100644
--- a/src/libstore/local-store.cc
+++ b/src/libstore/local-store.cc
@@ -1380,11 +1380,8 @@ Paths LocalStore::importPaths(bool requireSignature, Source & source)
 }
 
 
-void LocalStore::deleteFromStore(const Path & path, unsigned long long & bytesFreed,
-    unsigned long long & blocksFreed)
+void LocalStore::invalidatePathChecked(const Path & path)
 {
-    bytesFreed = 0;
-
     assertStorePath(path);
 
     while (1) {
@@ -1404,8 +1401,6 @@ void LocalStore::deleteFromStore(const Path & path, unsigned long long & bytesFr
             break;
         } catch (SQLiteBusy & e) { };
     }
-    
-    deletePathWrapped(path, bytesFreed, blocksFreed);
 }
 
 
diff --git a/src/libstore/local-store.hh b/src/libstore/local-store.hh
index 2739c4eea69d..8e3cbe5ce11b 100644
--- a/src/libstore/local-store.hh
+++ b/src/libstore/local-store.hh
@@ -164,10 +164,6 @@ public:
 
     void collectGarbage(const GCOptions & options, GCResults & results);
 
-    /* Delete a path from the Nix store. */
-    void deleteFromStore(const Path & path, unsigned long long & bytesFreed,
-        unsigned long long & blocksFreed);
-    
     /* Optimise the disk space usage of the Nix store by hard-linking
        files with the same contents. */
     void optimiseStore(bool dryRun, OptimiseStats & stats);
@@ -238,6 +234,9 @@ private:
 
     void invalidatePath(const Path & path);
 
+    /* Delete a path from the Nix store. */
+    void invalidatePathChecked(const Path & path);
+    
     void verifyPath(const Path & path, const PathSet & store,
         PathSet & done, PathSet & validPaths);
 
@@ -249,6 +248,8 @@ private:
 
     struct GCState;
 
+    void deleteGarbage(GCState & state, const Path & path);
+    
     bool tryToDelete(GCState & state, const Path & path);
     
     bool isActiveTempFile(const GCState & state,