UBIFS: do not start the commit if there is nothing to commit

This patch fixes suboptimal UBIFS 'sync_fs()' implementation which causes
flash I/O even if the file-system is synchronized. E.g., a 'printk()'
in the MTD erasure function (e.g., 'nand_erase_nand()') can show that
for every 'sync' shell command UBIFS erases at least one eraseblock.

So '$ while true; do sync; done' will cause huge amount of flash I/O.

The reason for this is that UBIFS commits in 'sync_fs()', and starts the
commit even if there is nothing to commit, e.g., it anyway changes the
log. This patch adds a check in the 'do_commit()' UBIFS functions which
prevents the commit if there is nothing to commit.

Reported-by: Hans J. Koch <hjk@linutronix.de>
Tested-by: John Ogness <john.ogness@linutronix.de>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>

diff --git a/fs/ubifs/commit.c b/fs/ubifs/commit.c
index 02429d81ca3374fef11d9646391d3ee7f1878139..b148fbc80f8d1eb71ccc3a9b0f6a52b761fedab2 100644 (file)
--- a/fs/ubifs/commit.c
+++ b/fs/ubifs/commit.c
@@ -48,6 +48,56 @@
 #include <linux/slab.h>
 #include "ubifs.h"
 
+/*
+ * nothing_to_commit - check if there is nothing to commit.
+ * @c: UBIFS file-system description object
+ *
+ * This is a helper function which checks if there is anything to commit. It is
+ * used as an optimization to avoid starting the commit if it is not really
+ * necessary. Indeed, the commit operation always assumes flash I/O (e.g.,
+ * writing the commit start node to the log), and it is better to avoid doing
+ * this unnecessarily. E.g., 'ubifs_sync_fs()' runs the commit, but if there is
+ * nothing to commit, it is more optimal to avoid any flash I/O.
+ *
+ * This function has to be called with @c->commit_sem locked for writing -
+ * this function does not take LPT/TNC locks because the @c->commit_sem
+ * guarantees that we have exclusive access to the TNC and LPT data structures.
+ *
+ * This function returns %1 if there is nothing to commit and %0 otherwise.
+ */
+static int nothing_to_commit(struct ubifs_info *c)
+{
+       /*
+        * During mounting or remounting from R/O mode to R/W mode we may
+        * commit for various recovery-related reasons.
+        */
+       if (c->mounting || c->remounting_rw)
+               return 0;
+
+       /*
+        * If the root TNC node is dirty, we definitely have something to
+        * commit.
+        */
+       if (c->zroot.znode && test_bit(DIRTY_ZNODE, &c->zroot.znode->flags))
+               return 0;
+
+       /*
+        * Even though the TNC is clean, the LPT tree may have dirty nodes. For
+        * example, this may happen if the budgeting subsystem invoked GC to
+        * make some free space, and the GC found an LEB with only dirty and
+        * free space. In this case GC would just change the lprops of this
+        * LEB (by turning all space into free space) and unmap it.
+        */
+       if (c->nroot && test_bit(DIRTY_CNODE, &c->nroot->flags))
+               return 0;
+
+       ubifs_assert(atomic_long_read(&c->dirty_zn_cnt) == 0);
+       ubifs_assert(c->dirty_pn_cnt == 0);
+       ubifs_assert(c->dirty_nn_cnt == 0);
+
+       return 1;
+}
+
 /**
  * do_commit - commit the journal.
  * @c: UBIFS file-system description object
@@ -70,6 +120,12 @@ static int do_commit(struct ubifs_info *c)
                goto out_up;
        }
 
+       if (nothing_to_commit(c)) {
+               up_write(&c->commit_sem);
+               err = 0;
+               goto out_cancel;
+       }
+
        /* Sync all write buffers (necessary for recovery) */
        for (i = 0; i < c->jhead_cnt; i++) {
                err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
@@ -162,12 +218,12 @@ static int do_commit(struct ubifs_info *c)
        if (err)
                goto out;
 
+out_cancel:
        spin_lock(&c->cs_lock);
        c->cmt_state = COMMIT_RESTING;
        wake_up(&c->cmt_wq);
        dbg_cmt("commit end");
        spin_unlock(&c->cs_lock);
-
        return 0;
 
 out_up: