#include <linux/timer.h>
#include <linux/mm.h>
#include <linux/highmem.h>
+#include <linux/hrtimer.h>
static void __journal_temp_unlink_buffer(struct journal_head *jh);
{
transaction->t_journal = journal;
transaction->t_state = T_RUNNING;
+ transaction->t_start_time = ktime_get();
transaction->t_tid = journal->j_transaction_sequence++;
transaction->t_expires = jiffies + journal->j_commit_interval;
spin_lock_init(&transaction->t_handle_lock);
/* Set up the commit timer for the new transaction. */
- journal->j_commit_timer.expires = round_jiffies(transaction->t_expires);
+ journal->j_commit_timer.expires =
+ round_jiffies_up(transaction->t_expires);
add_timer(&journal->j_commit_timer);
J_ASSERT(journal->j_running_transaction == NULL);
__log_space_left(journal));
spin_unlock(&transaction->t_handle_lock);
spin_unlock(&journal->j_state_lock);
+
+ lock_map_acquire(&handle->h_lockdep_map);
out:
if (unlikely(new_transaction)) /* It's usually NULL */
kfree(new_transaction);
handle = ERR_PTR(err);
goto out;
}
-
- lock_acquire(&handle->h_lockdep_map, 0, 0, 0, 2, _THIS_IP_);
-
out:
return handle;
}
__log_start_commit(journal, transaction->t_tid);
spin_unlock(&journal->j_state_lock);
+ lock_map_release(&handle->h_lockdep_map);
handle->h_buffer_credits = nblocks;
ret = start_this_handle(journal, handle);
return ret;
wake_up(&journal->j_wait_transaction_locked);
}
-/*
- * Report any unexpected dirty buffers which turn up. Normally those
- * indicate an error, but they can occur if the user is running (say)
- * tune2fs to modify the live filesystem, so we need the option of
- * continuing as gracefully as possible. #
- *
- * The caller should already hold the journal lock and
- * j_list_lock spinlock: most callers will need those anyway
- * in order to probe the buffer's journaling state safely.
- */
-static void jbd_unexpected_dirty_buffer(struct journal_head *jh)
+static void warn_dirty_buffer(struct buffer_head *bh)
{
- int jlist;
-
- /* If this buffer is one which might reasonably be dirty
- * --- ie. data, or not part of this journal --- then
- * we're OK to leave it alone, but otherwise we need to
- * move the dirty bit to the journal's own internal
- * JBDDirty bit. */
- jlist = jh->b_jlist;
-
- if (jlist == BJ_Metadata || jlist == BJ_Reserved ||
- jlist == BJ_Shadow || jlist == BJ_Forget) {
- struct buffer_head *bh = jh2bh(jh);
+ char b[BDEVNAME_SIZE];
- if (test_clear_buffer_dirty(bh))
- set_buffer_jbddirty(bh);
- }
+ printk(KERN_WARNING
+ "JBD: Spotted dirty metadata buffer (dev = %s, blocknr = %llu). "
+ "There's a risk of filesystem corruption in case of system "
+ "crash.\n",
+ bdevname(bh->b_bdev, b), (unsigned long long)bh->b_blocknr);
}
/*
if (jh->b_next_transaction)
J_ASSERT_JH(jh, jh->b_next_transaction ==
transaction);
+ warn_dirty_buffer(bh);
}
/*
* In any case we need to clean the dirty flag and we must
* do it under the buffer lock to be sure we don't race
* with running write-out.
*/
- JBUFFER_TRACE(jh, "Unexpected dirty buffer");
- jbd_unexpected_dirty_buffer(jh);
+ JBUFFER_TRACE(jh, "Journalling dirty buffer");
+ clear_buffer_dirty(bh);
+ set_buffer_jbddirty(bh);
}
unlock_buffer(bh);
goto done;
/*
+ * this is the first time this transaction is touching this buffer,
+ * reset the modified flag
+ */
+ jh->b_modified = 0;
+
+ /*
* If there is already a copy-out version of this buffer, then we don't
* need to make another one
*/
if (!frozen_buffer) {
printk(KERN_EMERG
"%s: OOM for frozen_buffer\n",
- __FUNCTION__);
+ __func__);
JBUFFER_TRACE(jh, "oom!");
error = -ENOMEM;
jbd_lock_bh_state(bh);
* int journal_get_write_access() - notify intent to modify a buffer for metadata (not data) update.
* @handle: transaction to add buffer modifications to
* @bh: bh to be used for metadata writes
- * @credits: variable that will receive credits for the buffer
*
* Returns an error code or 0 on success.
*
J_ASSERT_JH(jh, buffer_locked(jh2bh(jh)));
if (jh->b_transaction == NULL) {
+ /*
+ * Previous journal_forget() could have left the buffer
+ * with jbddirty bit set because it was being committed. When
+ * the commit finished, we've filed the buffer for
+ * checkpointing and marked it dirty. Now we are reallocating
+ * the buffer so the transaction freeing it must have
+ * committed and so it's safe to clear the dirty bit.
+ */
+ clear_buffer_dirty(jh2bh(jh));
jh->b_transaction = transaction;
+
+ /* first access by this transaction */
+ jh->b_modified = 0;
+
JBUFFER_TRACE(jh, "file as BJ_Reserved");
__journal_file_buffer(jh, transaction, BJ_Reserved);
} else if (jh->b_transaction == journal->j_committing_transaction) {
+ /* first access by this transaction */
+ jh->b_modified = 0;
+
JBUFFER_TRACE(jh, "set next transaction");
jh->b_next_transaction = transaction;
}
* int journal_get_undo_access() - Notify intent to modify metadata with non-rewindable consequences
* @handle: transaction
* @bh: buffer to undo
- * @credits: store the number of taken credits here (if not NULL)
*
* Sometimes there is a need to distinguish between metadata which has
* been committed to disk and that which has not. The ext3fs code uses
committed_data = jbd_alloc(jh2bh(jh)->b_size, GFP_NOFS);
if (!committed_data) {
printk(KERN_EMERG "%s: No memory for committed data\n",
- __FUNCTION__);
+ __func__);
err = -ENOMEM;
goto out;
}
journal_t *journal = handle->h_transaction->t_journal;
int need_brelse = 0;
struct journal_head *jh;
+ int ret = 0;
if (is_handle_aborted(handle))
- return 0;
+ return ret;
jh = journal_add_journal_head(bh);
JBUFFER_TRACE(jh, "entry");
time if it is redirtied */
}
- /* journal_clean_data_list() may have got there first */
+ /*
+ * We cannot remove the buffer with io error from the
+ * committing transaction, because otherwise it would
+ * miss the error and the commit would not abort.
+ */
+ if (unlikely(!buffer_uptodate(bh))) {
+ ret = -EIO;
+ goto no_journal;
+ }
+
if (jh->b_transaction != NULL) {
JBUFFER_TRACE(jh, "unfile from commit");
__journal_temp_unlink_buffer(jh);
}
JBUFFER_TRACE(jh, "exit");
journal_put_journal_head(jh);
- return 0;
+ return ret;
}
/**
struct journal_head *jh;
int drop_reserve = 0;
int err = 0;
+ int was_modified = 0;
BUFFER_TRACE(bh, "entry");
goto not_jbd;
}
+ /* keep track of wether or not this transaction modified us */
+ was_modified = jh->b_modified;
+
/*
* The buffer's going from the transaction, we must drop
* all references -bzzz
JBUFFER_TRACE(jh, "belongs to current transaction: unfile");
- drop_reserve = 1;
+ /*
+ * we only want to drop a reference if this transaction
+ * modified the buffer
+ */
+ if (was_modified)
+ drop_reserve = 1;
/*
* We are no longer going to journal this buffer.
if (jh->b_next_transaction) {
J_ASSERT(jh->b_next_transaction == transaction);
jh->b_next_transaction = NULL;
- drop_reserve = 1;
+
+ /*
+ * only drop a reference if this transaction modified
+ * the buffer
+ */
+ if (was_modified)
+ drop_reserve = 1;
}
}
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
- int old_handle_count, err;
+ int err;
pid_t pid;
J_ASSERT(journal_current_handle() == handle);
* on IO anyway. Speeds up many-threaded, many-dir operations
* by 30x or more...
*
+ * We try and optimize the sleep time against what the underlying disk
+ * can do, instead of having a static sleep time. This is usefull for
+ * the case where our storage is so fast that it is more optimal to go
+ * ahead and force a flush and wait for the transaction to be committed
+ * than it is to wait for an arbitrary amount of time for new writers to
+ * join the transaction. We acheive this by measuring how long it takes
+ * to commit a transaction, and compare it with how long this
+ * transaction has been running, and if run time < commit time then we
+ * sleep for the delta and commit. This greatly helps super fast disks
+ * that would see slowdowns as more threads started doing fsyncs.
+ *
* But don't do this if this process was the most recent one to
* perform a synchronous write. We do this to detect the case where a
* single process is doing a stream of sync writes. No point in waiting
*/
pid = current->pid;
if (handle->h_sync && journal->j_last_sync_writer != pid) {
+ u64 commit_time, trans_time;
+
journal->j_last_sync_writer = pid;
- do {
- old_handle_count = transaction->t_handle_count;
- schedule_timeout_uninterruptible(1);
- } while (old_handle_count != transaction->t_handle_count);
+
+ spin_lock(&journal->j_state_lock);
+ commit_time = journal->j_average_commit_time;
+ spin_unlock(&journal->j_state_lock);
+
+ trans_time = ktime_to_ns(ktime_sub(ktime_get(),
+ transaction->t_start_time));
+
+ commit_time = min_t(u64, commit_time,
+ 1000*jiffies_to_usecs(1));
+
+ if (trans_time < commit_time) {
+ ktime_t expires = ktime_add_ns(ktime_get(),
+ commit_time);
+ set_current_state(TASK_UNINTERRUPTIBLE);
+ schedule_hrtimeout(&expires, HRTIMER_MODE_ABS);
+ }
}
+ if (handle->h_sync)
+ transaction->t_synchronous_commit = 1;
current->journal_info = NULL;
spin_lock(&journal->j_state_lock);
spin_lock(&transaction->t_handle_lock);
spin_unlock(&journal->j_state_lock);
}
- lock_release(&handle->h_lockdep_map, 1, _THIS_IP_);
+ lock_map_release(&handle->h_lockdep_map);
jbd_free_handle(handle);
return err;
}
-/**int journal_force_commit() - force any uncommitted transactions
+/**
+ * int journal_force_commit() - force any uncommitted transactions
* @journal: journal to force
*
* For synchronous operations: force any uncommitted transactions
return;
}
-
/**
* int journal_try_to_free_buffers() - try to free page buffers.
* @journal: journal for operation
* @page: to try and free
- * @unused_gfp_mask: unused
+ * @gfp_mask: we use the mask to detect how hard should we try to release
+ * buffers. If __GFP_WAIT and __GFP_FS is set, we wait for commit code to
+ * release the buffers.
*
*
* For all the buffers on this page,
* journal_try_to_free_buffer() is changing its state. But that
* cannot happen because we never reallocate freed data as metadata
* while the data is part of a transaction. Yes?
+ *
+ * Return 0 on failure, 1 on success
*/
int journal_try_to_free_buffers(journal_t *journal,
- struct page *page, gfp_t unused_gfp_mask)
+ struct page *page, gfp_t gfp_mask)
{
struct buffer_head *head;
struct buffer_head *bh;
if (buffer_jbd(bh))
goto busy;
} while ((bh = bh->b_this_page) != head);
+
ret = try_to_free_buffers(page);
+
busy:
return ret;
}
if (jh->b_cp_transaction) {
JBUFFER_TRACE(jh, "on running+cp transaction");
+ /*
+ * We don't want to write the buffer anymore, clear the
+ * bit so that we don't confuse checks in
+ * __journal_file_buffer
+ */
+ clear_buffer_dirty(bh);
__journal_file_buffer(jh, transaction, BJ_Forget);
- clear_buffer_jbddirty(bh);
may_free = 0;
} else {
JBUFFER_TRACE(jh, "on running transaction");
}
/**
- * void journal_invalidatepage()
- * @journal: journal to use for flush...
+ * void journal_invalidatepage() - invalidate a journal page
+ * @journal: journal to use for flush
* @page: page to flush
* @offset: length of page to invalidate.
*
* Reap page buffers containing data after offset in page.
- *
*/
void journal_invalidatepage(journal_t *journal,
struct page *page,
if (jh->b_transaction && jh->b_jlist == jlist)
return;
- /* The following list of buffer states needs to be consistent
- * with __jbd_unexpected_dirty_buffer()'s handling of dirty
- * state. */
-
if (jlist == BJ_Metadata || jlist == BJ_Reserved ||
jlist == BJ_Shadow || jlist == BJ_Forget) {
+ /*
+ * For metadata buffers, we track dirty bit in buffer_jbddirty
+ * instead of buffer_dirty. We should not see a dirty bit set
+ * here because we clear it in do_get_write_access but e.g.
+ * tune2fs can modify the sb and set the dirty bit at any time
+ * so we try to gracefully handle that.
+ */
+ if (buffer_dirty(bh))
+ warn_dirty_buffer(bh);
if (test_clear_buffer_dirty(bh) ||
test_clear_buffer_jbddirty(bh))
was_dirty = 1;
jh->b_transaction = jh->b_next_transaction;
jh->b_next_transaction = NULL;
__journal_file_buffer(jh, jh->b_transaction,
- was_dirty ? BJ_Metadata : BJ_Reserved);
+ jh->b_modified ? BJ_Metadata : BJ_Reserved);
J_ASSERT_JH(jh, jh->b_transaction->t_state == T_RUNNING);
if (was_dirty)