On Mon, Apr 02, 2012 at 09:47:22PM -0400, Dave Jones wrote:
> 49b25e0540904be0bf558b84475c69d72e4de66e is the first bad commit
> btrfs: enhance transaction abort infrastructure
Attached patch adds several debugging printks to help to track down
where the EIOs come from. As there are no messages in syslog, it happens
on a regular path and not after a transaction abort.
I was not able to trigger the problem with either fsx or full xfstests
suite (3.4-rc).
Patch is based on top of the bisected commit, please retest, thanks.
david
---------8<--------
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 0517bd7..161baa4 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -216,12 +216,12 @@ void __btrfs_abort_transaction(struct btrfs_trans_handle
*trans,
struct btrfs_root *root, const char *function,
unsigned int line, int errno)
{
- WARN_ON_ONCE(1);
+ WARN_ON(1);
trans->aborted = errno;
/* Nothing used. The other threads that have joined this
* transaction may be able to continue. */
if (!trans->blocks_used) {
- btrfs_printk(root->fs_info, "Aborting unused transaction.\n");
+ btrfs_printk(root->fs_info, "Aborting unused transaction
(errno=%d).\n", errno);
return;
}
trans->transaction->aborted = errno;
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 5a4999a..b7ae8e2 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -73,8 +73,10 @@ loop:
cur_trans = root->fs_info->running_transaction;
if (cur_trans) {
- if (cur_trans->aborted)
+ if (cur_trans->aborted) {
+ spin_unlock(&root->fs_info->trans_lock);
return cur_trans->aborted;
+ }
atomic_inc(&cur_trans->use_count);
atomic_inc(&cur_trans->num_writers);
cur_trans->num_joined++;
@@ -466,8 +468,10 @@ int btrfs_should_end_transaction(struct btrfs_trans_handle
*trans,
trans->delayed_ref_updates = 0;
if (updates) {
err = btrfs_run_delayed_refs(trans, root, updates);
- if (err) /* Error code will also eval true */
+ if (err) { /* Error code will also eval true */
+ printk(KERN_DEBUG "btrfs: %s %d from
btrfs_run_delayed_refs\n", __func__, err);
return err;
+ }
}
trans->block_rsv = rsv;
@@ -491,11 +495,14 @@ static int __btrfs_end_transaction(struct
btrfs_trans_handle *trans,
trans->block_rsv = NULL;
while (count < 2) {
unsigned long cur = trans->delayed_ref_updates;
+ int ret;
+
trans->delayed_ref_updates = 0;
if (cur &&
trans->transaction->delayed_refs.num_heads_ready > 64) {
trans->delayed_ref_updates = 0;
- btrfs_run_delayed_refs(trans, root, cur);
+ ret = btrfs_run_delayed_refs(trans, root, cur);
+ if (ret) printk(KERN_DEBUG "btrfs: %s %d from
btrfs_run_delayed_refs\n", __func__, ret);
} else {
break;
}
@@ -541,6 +548,10 @@ static int __btrfs_end_transaction(struct
btrfs_trans_handle *trans,
if (trans->aborted ||
root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
+ printk(KERN_DEBUG "btrfs: %s -EIO abored=%d (%s)\n", __func__,
+ trans->aborted,
+ (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) ?
+ "super error" : "no super error");
return -EIO;
}
@@ -740,8 +751,10 @@ static noinline int commit_cowonly_roots(struct
btrfs_trans_handle *trans,
int ret;
ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
- if (ret)
+ if (ret) {
+ printk(KERN_DEBUG "btrfs: %s %d from btrfs_run_delayed_refs\n",
__func__, ret);
return ret;
+ }
eb = btrfs_lock_root_node(fs_info->tree_root);
ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL,
@@ -749,12 +762,16 @@ static noinline int commit_cowonly_roots(struct
btrfs_trans_handle *trans,
btrfs_tree_unlock(eb);
free_extent_buffer(eb);
- if (ret)
+ if (ret) {
+ printk(KERN_DEBUG "btrfs: %s %d from btrfs_cow_block\n",
__func__, ret);
return ret;
+ }
ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
- if (ret)
+ if (ret) {
+ printk(KERN_DEBUG "btrfs: %s %d from btrfs_run_delayed_refs\n",
__func__, ret);
return ret;
+ }
while (!list_empty(&fs_info->dirty_cowonly_roots)) {
next = fs_info->dirty_cowonly_roots.next;
@@ -762,8 +779,10 @@ static noinline int commit_cowonly_roots(struct
btrfs_trans_handle *trans,
root = list_entry(next, struct btrfs_root, dirty_list);
ret = update_cowonly_root(trans, root);
- if (ret)
+ if (ret) {
+ printk(KERN_DEBUG "btrfs: %s %d from
update_cowonly_root\n", __func__, ret);
return ret;
+ }
}
down_write(&fs_info->extent_commit_sem);
@@ -1234,8 +1253,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle
*trans,
* any runnings procs may add more while we are here
*/
ret = btrfs_run_delayed_refs(trans, root, 0);
- if (ret)
+ if (ret) {
+ printk(KERN_DEBUG "btrfs: %s %d from btrfs_run_delayed_refs\n",
__func__, ret);
goto cleanup_transaction;
+ }
cur_trans = trans->transaction;
@@ -1246,8 +1267,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle
*trans,
cur_trans->delayed_refs.flushing = 1;
ret = btrfs_run_delayed_refs(trans, root, 0);
- if (ret)
+ if (ret) {
+ printk(KERN_DEBUG "btrfs: %s %d from btrfs_run_delayed_refs\n",
__func__, ret);
goto cleanup_transaction;
+ }
spin_lock(&cur_trans->commit_lock);
if (cur_trans->in_commit) {
@@ -1360,6 +1383,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle
*trans,
ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
if (ret) {
mutex_unlock(&root->fs_info->reloc_mutex);
+ printk(KERN_DEBUG "btrfs: %s %d from btrfs_run_delayed_refs\n",
__func__, ret);
goto cleanup_transaction;
}
@@ -1390,6 +1414,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle
*trans,
ret = commit_fs_roots(trans, root);
if (ret) {
mutex_unlock(&root->fs_info->tree_log_mutex);
+ mutex_unlock(&root->fs_info->reloc_mutex);
goto cleanup_transaction;
}
@@ -1401,6 +1426,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle
*trans,
ret = commit_cowonly_roots(trans, root);
if (ret) {
mutex_unlock(&root->fs_info->tree_log_mutex);
+ mutex_unlock(&root->fs_info->reloc_mutex);
goto cleanup_transaction;
}
--
1.7.6.233.gd79bc
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
