[Sheepdog] [PATCH v4 05/12] farm: remove stale object in backend store

[Liu Yuan]

From: Liu Yuan <tailai.ly at taobao.com>

For storing one object into backend store, either

a) no content change, then point to the same old sha1_file (no stale object)
or
b) content updated, then will point to a new object with a new sha1.

So we need to remove stale object in case b), only in the assumption it is the
object generated by recovery code. [*]

When we try store new snapshot object into the backend store, it is safe and
good timing for us to remove the old object with the same object ID.

[*] Here I assume we don't need to restore to 'sys epoch' state.

Signed-off-by: Liu Yuan <tailai.ly at taobao.com>
---
 sheep/farm/trunk.c |  136 ++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 files changed, 133 insertions(+), 3 deletions(-)

diff --git a/sheep/farm/trunk.c b/sheep/farm/trunk.c
index ec36081..fd0a525 100644
--- a/sheep/farm/trunk.c
+++ b/sheep/farm/trunk.c
@@ -1,5 +1,7 @@
 #include <pthread.h>
 #include <dirent.h>
+#include <sys/stat.h>
+#include <unistd.h>
 
 #include "farm.h"
 #include "strbuf.h"
@@ -19,6 +21,11 @@ static struct hlist_head trunk_hashtable[HASH_SIZE];
 static pthread_mutex_t hashtable_lock[HASH_SIZE] = { [0 ... HASH_SIZE - 1] = PTHREAD_MUTEX_INITIALIZER };
 static unsigned int trunk_entry_active_nr;
 
+struct omap_entry {
+	uint64_t oid;
+	unsigned char sha1[SHA1_LEN];
+};
+
 static inline int trunk_entry_is_dirty(struct __trunk_entry *entry)
 {
 	return entry->flags & TRUNK_ENTRY_DIRTY;
@@ -78,6 +85,27 @@ out:
 	return entry;
 }
 
+static int create_files(void)
+{
+	int fd, ret = 0;
+	struct strbuf buf = STRBUF_INIT;
+
+	strbuf_addstr(&buf, farm_dir);
+	strbuf_addf(&buf, "/%s", "obj_map");
+
+	fd = open(buf.buf, O_CREAT | O_EXCL, 0666);
+	if (fd < 0) {
+		if (errno != EEXIST) {
+			ret = -1;
+			goto out;
+		}
+	}
+	close(fd);
+out:
+	strbuf_release(&buf);
+	return ret;
+}
+
 int trunk_init(void)
 {
 	DIR *dir;
@@ -96,6 +124,10 @@ int trunk_init(void)
 			continue;
 		lookup_trunk_entry(oid, 1);
 	}
+
+	if (create_files() < 0)
+		return -1;
+
 	closedir(dir);
 	return 0;
 }
@@ -148,9 +180,92 @@ static void put_entry(struct __trunk_entry *entry)
 	trunk_entry_active_nr--;
 }
 
+
+static int omap_file_init(struct strbuf *outbuf)
+{
+	int fd;
+	int len = -1;
+	void *buffer = NULL;
+	struct strbuf buf = STRBUF_INIT;
+	struct stat st;
+
+ 	strbuf_addf(&buf, "%s/%s", farm_dir, "obj_map");
+	fd = open(buf.buf, O_RDONLY);
+	if (fd < 0)
+		goto out;
+
+        if (fstat(fd, &st) < 0) {
+                dprintf("%m\n");
+                goto out_close;
+        }
+
+        len = st.st_size;
+	if (len == 0)
+		goto out_close;
+
+        buffer = xmalloc(len);
+        len = xread(fd, buffer, len);
+        if (len != st.st_size) {
+		len = -1;
+                goto out_close;
+        }
+	strbuf_add(outbuf, buffer, len); 
+out_close:
+	close(fd);
+out:
+	strbuf_release(&buf);
+        free(buffer);
+	return len;
+}
+
+static int omap_file_final(struct strbuf *omap_buf)
+{
+        int fd, ret = -1;
+        struct strbuf buf = STRBUF_INIT;
+
+	if (omap_buf->len == 0)
+		return 0;
+
+ 	strbuf_addf(&buf, "%s/%s", farm_dir, "obj_map");
+        fd = open(buf.buf, O_WRONLY);
+        if (fd < 0) {
+                dprintf("%m\n");
+                goto out;
+        }
+
+        ret = xwrite(fd, omap_buf->buf, omap_buf->len);
+        if (ret != omap_buf->len)
+                ret = -1;
+
+        close(fd);
+out:
+        strbuf_release(&buf);
+        return ret;
+}
+
+static struct omap_entry *omap_file_insert(struct strbuf *buf, struct omap_entry *new)
+{
+	unsigned i, nr = buf->len / sizeof(struct omap_entry);
+	struct omap_entry *omap_buf = (struct omap_entry *)buf->buf;
+	for (i = 0; i < nr; i++, omap_buf++)
+		if (omap_buf->oid == new->oid) {
+			if (memcmp(omap_buf->sha1, new->sha1, SHA1_LEN) == 0)
+				return NULL;
+			else {
+				struct omap_entry *old = xmalloc(sizeof(*old));
+				memcpy(old, omap_buf, sizeof(*old));
+				memcpy(omap_buf->sha1, new->sha1, SHA1_LEN);
+				return old;
+			}
+		}
+	dprintf("oid, %"PRIx64"\n", new->oid);
+	strbuf_add(buf, new, sizeof(*new));
+	return NULL;
+}
+
 int trunk_file_write(unsigned char *outsha1, int user)
 {
-	struct strbuf buf;
+	struct strbuf buf, omap_buf = STRBUF_INIT;
 	uint64_t data_size = sizeof(struct trunk_entry) * trunk_entry_active_nr;
 	struct sha1_file_hdr hdr = { .size = data_size,
 				     .priv = trunk_entry_active_nr };
@@ -159,18 +274,32 @@ int trunk_file_write(unsigned char *outsha1, int user)
 
 	memcpy(hdr.tag, TAG_TRUNK, TAG_LEN);
 	strbuf_init(&buf, sizeof(hdr) + data_size);
+	omap_file_init(&omap_buf);
 
 	strbuf_add(&buf, &hdr, sizeof(hdr));
 	list_for_each_entry_safe(entry, t, &trunk_active_list, active_list) {
-		if (trunk_entry_no_sha1(entry) || trunk_entry_is_dirty(entry)) {
+		if (trunk_entry_no_sha1(entry) || trunk_entry_is_dirty(entry))
 			if (fill_entry_new_sha1(entry) < 0) {
 				ret = -1;
 				goto out;
 			}
-		}
 		strbuf_add(&buf, &entry->raw, sizeof(struct trunk_entry));
+		if (!user) {
+			struct omap_entry new = { .oid = entry->raw.oid }, *old;
+			memcpy(new.sha1, entry->raw.sha1, SHA1_LEN);
+			old = omap_file_insert(&omap_buf, &new);
+			if (old) {
+				dprintf("try delete stale %s, oid %"PRIx64"\n",
+					sha1_to_hex(old->sha1), old->oid);
+				sha1_file_delete(old->sha1);
+				free(old);
+			}
+		}
 		put_entry(entry);
 	}
+	if (omap_file_final(&omap_buf) < 0)
+		eprintf("omap_file_final failed\n");
+
 	if (sha1_file_write((void *)buf.buf, buf.len, outsha1) < 0) {
 		ret = -1;
 		goto out;
@@ -178,6 +307,7 @@ int trunk_file_write(unsigned char *outsha1, int user)
 	dprintf("trunk sha1: %s\n", sha1_to_hex(outsha1));
 out:
 	strbuf_release(&buf);
+	strbuf_release(&omap_buf);
 	return ret;
 }
 
-- 
1.7.8.rc3