aboutsummaryrefslogtreecommitdiffstats
path: root/org.eclipse.jgit
diff options
context:
space:
mode:
authorShawn O. Pearce <spearce@spearce.org>2010-12-07 14:05:40 -0800
committerChris Aniszczyk <caniszczyk@gmail.com>2010-12-08 11:30:11 -0600
commitc181e1ab8ac52ff8ce02f95affc8f79cfbfff323 (patch)
treecb2f2eb7594d00b15c181b93c0a06cf8225f93e7 /org.eclipse.jgit
parentbc1130c6aac83a5db0da47fc2cbd25f1a61e3601 (diff)
downloadjgit-c181e1ab8ac52ff8ce02f95affc8f79cfbfff323.tar.gz
jgit-c181e1ab8ac52ff8ce02f95affc8f79cfbfff323.zip
IndexPack: Use streaming for large whole blobs
When indexing large blobs that are stored whole (non-delta form), avoid allocating the entire blob in memory and instead stream it through the SHA-1 checksum computation. This reduces the size of memory required by IndexPack when processing very big blobs, such as a 500 MiB uncompressable binary. If the large blob already exists in the local repository, its contents needs to be compared byte-for-byte after the entire pack has been indexed, to ensure there isn't an unexpected SHA-1 collision which may result in later data corruption. This compare is performed as a streaming compare, again avoiding the large object allocation. This change doesn't improve on memory utilization for large objects stored as deltas. The change also doesn't improve handling for any large commits, trees or annotated tags. There isn't much to be done here for those objects, because they need to be passed down to the ObjectChecker as a byte[]. Fortunately it isn't common for these object types to be that large, Bug: 312868 Change-Id: I862afd4cb78013ee033d4ec68c067b1774a05be8 Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Chris Aniszczyk <caniszczyk@gmail.com> CC: Roberto Tyley <roberto.tyley@guardian.co.uk>
Diffstat (limited to 'org.eclipse.jgit')
-rw-r--r--org.eclipse.jgit/src/org/eclipse/jgit/transport/IndexPack.java93
1 files changed, 87 insertions, 6 deletions
diff --git a/org.eclipse.jgit/src/org/eclipse/jgit/transport/IndexPack.java b/org.eclipse.jgit/src/org/eclipse/jgit/transport/IndexPack.java
index d7d1373c8c..8feb483fe2 100644
--- a/org.eclipse.jgit/src/org/eclipse/jgit/transport/IndexPack.java
+++ b/org.eclipse.jgit/src/org/eclipse/jgit/transport/IndexPack.java
@@ -75,6 +75,7 @@ import org.eclipse.jgit.lib.ObjectId;
import org.eclipse.jgit.lib.ObjectIdSubclassMap;
import org.eclipse.jgit.lib.ObjectLoader;
import org.eclipse.jgit.lib.ObjectReader;
+import org.eclipse.jgit.lib.ObjectStream;
import org.eclipse.jgit.lib.ProgressMonitor;
import org.eclipse.jgit.lib.Repository;
import org.eclipse.jgit.storage.file.PackIndexWriter;
@@ -144,6 +145,8 @@ public class IndexPack {
private final Repository repo;
+ private int streamFileThreshold;
+
/**
* Object database used for loading existing objects
*/
@@ -151,6 +154,8 @@ public class IndexPack {
private InflaterStream inflater;
+ private byte[] readBuffer;
+
private final MessageDigest objectDigest;
private final MutableObjectId tempObjectId;
@@ -211,6 +216,9 @@ public class IndexPack {
private LongMap<UnresolvedDelta> baseByPos;
+ /** Blobs whose contents need to be double-checked after indexing. */
+ private List<PackedObjectInfo> deferredCheckBlobs;
+
private MessageDigest packDigest;
private RandomAccessFile packOut;
@@ -236,11 +244,13 @@ public class IndexPack {
public IndexPack(final Repository db, final InputStream src,
final File dstBase) throws IOException {
repo = db;
+ streamFileThreshold = 5 * (1 << 20); // A reasonable default for now.
objectDatabase = db.getObjectDatabase().newCachedDatabase();
in = src;
inflater = new InflaterStream();
readCurs = objectDatabase.newReader();
buf = new byte[BUFFER_SIZE];
+ readBuffer = new byte[BUFFER_SIZE];
objectDigest = Constants.newMessageDigest();
tempObjectId = new MutableObjectId();
packDigest = Constants.newMessageDigest();
@@ -258,6 +268,10 @@ public class IndexPack {
}
}
+ void setStreamFileThreshold(int sz) {
+ streamFileThreshold = sz;
+ }
+
/**
* Set the pack index file format version this instance will create.
*
@@ -396,6 +410,7 @@ public class IndexPack {
entries = new PackedObjectInfo[(int) objectCount];
baseById = new ObjectIdSubclassMap<DeltaChain>();
baseByPos = new LongMap<UnresolvedDelta>();
+ deferredCheckBlobs = new ArrayList<PackedObjectInfo>();
progress.beginTask(JGitText.get().receivingObjects,
(int) objectCount);
@@ -407,6 +422,8 @@ public class IndexPack {
}
readPackFooter();
endInput();
+ if (!deferredCheckBlobs.isEmpty())
+ doDeferredCheckBlobs();
progress.endTask();
if (deltaCount > 0) {
if (packOut == null)
@@ -837,17 +854,38 @@ public class IndexPack {
private void whole(final int type, final long pos, final long sz)
throws IOException {
- final byte[] data = inflateAndReturn(Source.INPUT, sz);
objectDigest.update(Constants.encodedTypeString(type));
objectDigest.update((byte) ' ');
objectDigest.update(Constants.encodeASCII(sz));
objectDigest.update((byte) 0);
- objectDigest.update(data);
- tempObjectId.fromRaw(objectDigest.digest(), 0);
- verifySafeObject(tempObjectId, type, data);
+ boolean checkContentLater = false;
+ if (type == Constants.OBJ_BLOB && sz >= streamFileThreshold) {
+ InputStream inf = inflate(Source.INPUT, sz);
+ long cnt = 0;
+ while (cnt < sz) {
+ int r = inf.read(readBuffer);
+ if (r <= 0)
+ break;
+ objectDigest.update(readBuffer, 0, r);
+ cnt += r;
+ }
+ inf.close();
+ tempObjectId.fromRaw(objectDigest.digest(), 0);
+ checkContentLater = readCurs.has(tempObjectId);
+
+ } else {
+ final byte[] data = inflateAndReturn(Source.INPUT, sz);
+ objectDigest.update(data);
+ tempObjectId.fromRaw(objectDigest.digest(), 0);
+ verifySafeObject(tempObjectId, type, data);
+ }
+
final int crc32 = (int) crc.getValue();
- addObjectAndTrack(new PackedObjectInfo(pos, crc32, tempObjectId));
+ PackedObjectInfo obj = new PackedObjectInfo(pos, crc32, tempObjectId);
+ addObjectAndTrack(obj);
+ if (checkContentLater)
+ deferredCheckBlobs.add(obj);
}
private void verifySafeObject(final AnyObjectId id, final int type,
@@ -863,7 +901,7 @@ public class IndexPack {
try {
final ObjectLoader ldr = readCurs.open(id, type);
- final byte[] existingData = ldr.getCachedBytes(Integer.MAX_VALUE);
+ final byte[] existingData = ldr.getCachedBytes(data.length);
if (!Arrays.equals(data, existingData)) {
throw new IOException(MessageFormat.format(JGitText.get().collisionOn, id.name()));
}
@@ -874,6 +912,49 @@ public class IndexPack {
}
}
+ private void doDeferredCheckBlobs() throws IOException {
+ final byte[] curBuffer = new byte[readBuffer.length];
+ for (PackedObjectInfo obj : deferredCheckBlobs) {
+ position(obj.getOffset());
+
+ int c = readFrom(Source.FILE);
+ final int type = (c >> 4) & 7;
+ long sz = c & 15;
+ int shift = 4;
+ while ((c & 0x80) != 0) {
+ c = readFrom(Source.FILE);
+ sz += (c & 0x7f) << shift;
+ shift += 7;
+ }
+
+ if (type != Constants.OBJ_BLOB)
+ throw new IOException(MessageFormat.format(
+ JGitText.get().unknownObjectType, type));
+
+ ObjectStream cur = readCurs.open(obj, type).openStream();
+ try {
+ if (cur.getSize() != sz)
+ throw new IOException(MessageFormat.format(
+ JGitText.get().collisionOn, obj.name()));
+ InputStream pck = inflate(Source.FILE, sz);
+ while (0 < sz) {
+ int n = (int) Math.min(readBuffer.length, sz);
+ IO.readFully(cur, curBuffer, 0, n);
+ IO.readFully(pck, readBuffer, 0, n);
+ for (int i = 0; i < n; i++) {
+ if (curBuffer[i] != readBuffer[i])
+ throw new IOException(MessageFormat.format(JGitText
+ .get().collisionOn, obj.name()));
+ }
+ sz -= n;
+ }
+ pck.close();
+ } finally {
+ cur.close();
+ }
+ }
+ }
+
// Current position of {@link #bOffset} within the entire file.
private long position() {
return bBase + bOffset;