[sheepdog] [PATCH v3 3/3] journal: new journal flag JF_DELETE_OBJ for deleting object

[Hitoshi Mitake]

Current journal implementation doesn't consider such a case:

1. make new journal for new object creation
2. actual creation of the object
3. delete object
4. crash

After such a sequence, the journaling mechanism recovers the deleted
object. This behavior isn't valid. This patch implements the new
journal flag: JF_DELETE_OBJ. If the journaling mechanism finds an
entry with this flag during recovery process, it deletes an
corresponding object.

Signed-off-by: Hitoshi Mitake <mitake.hitoshi at lab.ntt.co.jp>
---
 sheep/journal.c     |   29 +++++++++++++++++++++++++++--
 sheep/plain_store.c |    3 +++
 sheep/sheep_priv.h  |    1 +
 3 files changed, 31 insertions(+), 2 deletions(-)

diff --git a/sheep/journal.c b/sheep/journal.c
index be1a6b7..aea0115 100644
--- a/sheep/journal.c
+++ b/sheep/journal.c
@@ -53,6 +53,7 @@ struct journal_descriptor {
 #define JF_STORE 0
 #define JF_EPOCH 1
 #define JF_CONFIG 2
+#define JF_DELETE_OBJ 3
 
 static const char *jfile_name[2] = { "journal_file0", "journal_file1", };
 static int jfile_fds[2];
@@ -140,10 +141,14 @@ static void journal_get_path(struct journal_descriptor *jd, char *path)
 {
 	switch (jd->flag) {
 	case JF_STORE:
+	case JF_DELETE_OBJ:
 		snprintf(path, PATH_MAX, "%s/%016"PRIx64,
 			 get_object_path(jd->oid), jd->oid);
-		sd_iprintf("%s, size %"PRIu64", off %"PRIu64", %d",
-			   path, jd->size, jd->offset, jd->create);
+		if (jd->flag == JF_STORE)
+			sd_iprintf("%s, size %"PRIu64", off %"PRIu64", %d",
+				path, jd->size, jd->offset, jd->create);
+		else
+			sd_iprintf("%s (delete)", path);
 		break;
 	case JF_EPOCH:
 		snprintf(path, PATH_MAX, "%s/%08"PRIu32, epoch_path, jd->epoch);
@@ -154,6 +159,9 @@ static void journal_get_path(struct journal_descriptor *jd, char *path)
 		snprintf(path, PATH_MAX, "%s", config_path);
 		sd_iprintf("%s, size %"PRIu64, path, jd->size);
 		break;
+	default:
+		panic("unknown type of journal flag: %d", jd->flag);
+		break;
 	}
 }
 
@@ -165,6 +173,12 @@ static int replay_journal_entry(struct journal_descriptor *jd)
 	void *buf = NULL;
 	char *p = (char *)jd;
 
+	if (jd->flag == JF_DELETE_OBJ) {
+		journal_get_path(jd, path);
+		unlink(path);
+		return;
+	}
+
 	if (jd->create)
 		flags |= O_CREAT;
 
@@ -453,3 +467,14 @@ int journal_write_config(const char *buf, size_t size)
 	};
 	return journal_file_write(&jd, buf);
 }
+
+int journal_delete_object(uint64_t oid)
+{
+	struct journal_descriptor jd = {
+		.magic = JOURNAL_DESC_MAGIC,
+		.flag = JF_DELETE_OBJ,
+		.size = 0,
+	};
+	jd.oid = oid;
+	return journal_file_write(&jd, NULL);
+}
diff --git a/sheep/plain_store.c b/sheep/plain_store.c
index b5bb395..38c5c62 100644
--- a/sheep/plain_store.c
+++ b/sheep/plain_store.c
@@ -452,6 +452,9 @@ int default_remove_object(uint64_t oid)
 {
 	char path[PATH_MAX];
 
+	if (uatomic_is_true(&sys->use_journal))
+		journal_delete_object(oid);
+
 	get_obj_path(oid, path);
 
 	if (unlink(path) < 0) {
diff --git a/sheep/sheep_priv.h b/sheep/sheep_priv.h
index a447387..9b3a7ac 100644
--- a/sheep/sheep_priv.h
+++ b/sheep/sheep_priv.h
@@ -407,6 +407,7 @@ int
 journal_write_store(uint64_t oid, const char *buf, size_t size, off_t, bool);
 int journal_write_epoch(const char *buf, size_t size, uint32_t epoch);
 int journal_write_config(const char *buf, size_t size);
+int journal_delete_object(uint64_t oid);
 
 /* md.c */
 void md_add_disk(char *path);
-- 
1.7.5.1