Skip to contentSkip to conversation

btrfs: move transaction abort message to __btrfs_abort_transaction()

PATCH Superseded David Sterba 4 participants 6 messages Apr 28 – Apr 29 Btrfs file system

lore
Share & tools
lore

David Sterba

author maintainer
#0 lore raw

PATCH 1/1 btrfs: move transaction abort message to __btrfs_abort_transaction()

The btrfs_abort_transaction() is called at the location where we want to report the abort. It must be a macro so we get the correct line and stack trace. This inlines the necessary code and the rest is pushed to __btrfs_abort_transaction().

There's a possibility to reduce the inlined code if we move the message to the helper function as well, without loss of information. The difference is only that the WARN will not print it inside the stack report but after:

--[ cut here ]--

  WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
  ...
  --[ end trace ] --
  BTRFS error (device dm-0 state A): Transaction aborted (error -28)

While previously there would be one more line like:

--[ cut here ]--
BTRFS: Transaction aborted (error -28)

  WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
  ...
  --[ end trace ] --

This removes about 22KiB of btrfs.ko on a release config.

---
 fs/btrfs/transaction.c |  3 +++
 fs/btrfs/transaction.h | 11 +----------
 2 files changed, 4 insertions(+), 10 deletions(-)
fs/btrfs/transaction.c +3 -0
index 194f581b36f3..5125cb20ef85 100644
@@ -2729,6 +2729,9 @@ void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_fs_info *fs_info = trans->fs_info;
 
+	btrfs_err(trans->fs_info, "Transaction %llu aborted (error %d)",
+		  trans->transid, error);
+
 	WRITE_ONCE(trans->aborted, error);
 	WRITE_ONCE(trans->transaction->aborted, error);
 	if (first_hit && error == -ENOSPC)
fs/btrfs/transaction.h +1 -10
index 264dcd4b3788..f1cb05460cec 100644
@@ -253,16 +253,7 @@ do {								\
 	if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,	\
 			&((trans)->fs_info->fs_state))) {	\
 		__first = true;					\
-		if (WARN(btrfs_abort_should_print_stack(error),	\
-			KERN_ERR				\
-			"BTRFS: Transaction %llu aborted (error %d)\n",	\
-			(trans)->transid, (error))) {			\
-			/* Stack trace printed. */			\
-		} else {						\
-			btrfs_err((trans)->fs_info,			\
-			"Transaction %llu aborted (error %d)",	\
-				  (trans)->transid, (error));	\
-		}						\
+		WARN_ON(btrfs_abort_should_print_stack(error));	\
 	}							\
 	__btrfs_abort_transaction((trans), __func__,		\
 				  __LINE__, (error), __first);	\
Signature · 1 line
2.51.0
#1 lore raw
Quoted 37 lines · David Sterba

The btrfs_abort_transaction() is called at the location where we want to report the abort. It must be a macro so we get the correct line and stack trace. This inlines the necessary code and the rest is pushed to __btrfs_abort_transaction().

There's a possibility to reduce the inlined code if we move the message to the helper function as well, without loss of information. The difference is only that the WARN will not print it inside the stack report but after:

   --[ cut here ]--
   WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
   ...
   --[ end trace ] --
   BTRFS error (device dm-0 state A): Transaction aborted (error -28)

While previously there would be one more line like:

   --[ cut here ]--
   BTRFS: Transaction aborted (error -28)
   WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
   ...
   --[ end trace ] --

This removes about 22KiB of btrfs.ko on a release config.

---
  fs/btrfs/transaction.c |  3 +++
  fs/btrfs/transaction.h | 11 +----------
  2 files changed, 4 insertions(+), 10 deletions(-)
fs/btrfs/transaction.c +0 -0
index 194f581b36f3..5125cb20ef85 100644

fs/btrfs/transaction.c void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans, · line 2734

Quoted 6 lines
fs/btrfs/transaction.c +3 -0
@@ -2729,6 +2729,9 @@ void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
  {
  	struct btrfs_fs_info *fs_info = trans->fs_info;
 
+	btrfs_err(trans->fs_info, "Transaction %llu aborted (error %d)",
+		  trans->transid, error);
+

This message will be outputted for every btrfs_abort_transaction() call.

Shouldn't we move it behind first_hit check?

Thanks,
Qu

#2 lore raw
Quoted 43 lines · Qu Wenruo
Quoted 40 lines · David Sterba

The btrfs_abort_transaction() is called at the location where we want to report the abort. It must be a macro so we get the correct line and stack trace. This inlines the necessary code and the rest is pushed to __btrfs_abort_transaction().

There's a possibility to reduce the inlined code if we move the message to the helper function as well, without loss of information. The difference is only that the WARN will not print it inside the stack report but after:

   --[ cut here ]--
   WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
   ...
   --[ end trace ] --
   BTRFS error (device dm-0 state A): Transaction aborted (error -28)

While previously there would be one more line like:

   --[ cut here ]--
   BTRFS: Transaction aborted (error -28)
   WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
   ...
   --[ end trace ] --

This removes about 22KiB of btrfs.ko on a release config.

---
  fs/btrfs/transaction.c |  3 +++
  fs/btrfs/transaction.h | 11 +----------
  2 files changed, 4 insertions(+), 10 deletions(-)
fs/btrfs/transaction.c +0 -0
index 194f581b36f3..5125cb20ef85 100644
@@ -2729,6 +2729,9 @@ void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
  {
  	struct btrfs_fs_info *fs_info = trans->fs_info;
 
Quoted 7 lines

fs/btrfs/transaction.c void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans, · line 2734

Quoted 3 lines
fs/btrfs/transaction.c +3 -0
+	btrfs_err(trans->fs_info, "Transaction %llu aborted (error %d)",
+		  trans->transid, error);
+

This message will be outputted for every btrfs_abort_transaction() call.

Shouldn't we move it behind first_hit check?

Right, we should, to keep the previous behaviour. I'll update it, thanks.

author maintainer
#3 lore raw

PATCH v2 1/1 btrfs: move transaction abort message to __btrfs_abort_transaction()

In reply to David Sterba

The btrfs_abort_transaction() is called at the location where we want to report the abort. It must be a macro so we get the correct line and stack trace. This inlines the necessary code and the rest is pushed to __btrfs_abort_transaction().

There's a possibility to reduce the inlined code if we move the message to the helper function as well, without loss of information. The difference is only that the WARN will not print it inside the stack report but after:

--[ cut here ]--

  WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
  ...
  --[ end trace ] --
  BTRFS error (device dm-0 state A): Transaction aborted (error -28)

While previously there would be one more line like:

--[ cut here ]--
BTRFS: Transaction aborted (error -28)

  WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
  ...
  --[ end trace ] --

This removes about 20KiB of btrfs.ko on a release config.

---

v2:
- move message in __btrfs_abort_transaction() under 'first_hit', also to make the WRITE_ONCE() assignments immediate
- savings updated to 20K, it depends on the config and compiler, use an average of a few setups

 fs/btrfs/transaction.c |  8 ++++++--
 fs/btrfs/transaction.h | 11 +----------
 2 files changed, 7 insertions(+), 12 deletions(-)
fs/btrfs/transaction.c +6 -2
index 194f581b36f3..184fe976450e 100644
@@ -2731,8 +2731,12 @@ void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 
 	WRITE_ONCE(trans->aborted, error);
 	WRITE_ONCE(trans->transaction->aborted, error);
-	if (first_hit && error == -ENOSPC)
-		btrfs_dump_space_info_for_trans_abort(fs_info);
+	if (first_hit) {
+		btrfs_err(trans->fs_info, "Transaction %llu aborted (error %d)",
+			  trans->transid, error);
+		if (error == -ENOSPC)
+			btrfs_dump_space_info_for_trans_abort(fs_info);
+	}
 	/* Wake up anybody who may be waiting on this transaction */
 	wake_up(&fs_info->transaction_wait);
 	wake_up(&fs_info->transaction_blocked_wait);
fs/btrfs/transaction.h +1 -10
index 7d70fe486758..f1cb05460cec 100644
@@ -253,16 +253,7 @@ do {								\
 	if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,	\
 			&((trans)->fs_info->fs_state))) {	\
 		__first = true;					\
-		if (WARN(btrfs_abort_should_print_stack(error),	\
-			KERN_ERR				\
-			"BTRFS: Transaction aborted (error %d)\n",	\
-			(error))) {					\
-			/* Stack trace printed. */			\
-		} else {						\
-			btrfs_err((trans)->fs_info,			\
-				  "Transaction aborted (error %d)",	\
-				  (error));			\
-		}						\
+		WARN_ON(btrfs_abort_should_print_stack(error));	\
 	}							\
 	__btrfs_abort_transaction((trans), __func__,		\
 				  __LINE__, (error), __first);	\
Signature · 1 line
2.53.0
#4 lore raw
Quoted 45 lines · David Sterba

The btrfs_abort_transaction() is called at the location where we want to report the abort. It must be a macro so we get the correct line and stack trace. This inlines the necessary code and the rest is pushed to __btrfs_abort_transaction().

There's a possibility to reduce the inlined code if we move the message to the helper function as well, without loss of information. The difference is only that the WARN will not print it inside the stack report but after:

--[ cut here ]--

  WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
  ...
  --[ end trace ] --
  BTRFS error (device dm-0 state A): Transaction aborted (error -28)

While previously there would be one more line like:

--[ cut here ]--
BTRFS: Transaction aborted (error -28)

  WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
  ...
  --[ end trace ] --

This removes about 20KiB of btrfs.ko on a release config.

---

v2:
- move message in __btrfs_abort_transaction() under 'first_hit', also to make the WRITE_ONCE() assignments immediate
- savings updated to 20K, it depends on the config and compiler, use an average of a few setups

 fs/btrfs/transaction.c |  8 ++++++--
 fs/btrfs/transaction.h | 11 +----------
 2 files changed, 7 insertions(+), 12 deletions(-)
fs/btrfs/transaction.c +0 -0
index 194f581b36f3..184fe976450e 100644

fs/btrfs/transaction.c void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans, · line 2735

Quoted 7 lines
fs/btrfs/transaction.c +2 -2
@@ -2731,8 +2731,12 @@ void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 
        WRITE_ONCE(trans->aborted, error);
        WRITE_ONCE(trans->transaction->aborted, error);
-       if (first_hit && error == -ENOSPC)
-               btrfs_dump_space_info_for_trans_abort(fs_info);
+       if (first_hit) {
+               btrfs_err(trans->fs_info, "Transaction %llu aborted (error %d)",

There's already a local fs_info variable, no need to trans->fs_info.

Otherwise:

Filipe Manana <[email protected]>

Quoted 33 lines
fs/btrfs/transaction.c +4 -0
+                         trans->transid, error);
+               if (error == -ENOSPC)
+                       btrfs_dump_space_info_for_trans_abort(fs_info);
+       }
        /* Wake up anybody who may be waiting on this transaction */
        wake_up(&fs_info->transaction_wait);
        wake_up(&fs_info->transaction_blocked_wait);
fs/btrfs/transaction.h +1 -11
index 7d70fe486758..f1cb05460cec 100644
@@ -253,16 +253,7 @@ do {                                                               \
        if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,     \
                        &((trans)->fs_info->fs_state))) {       \
                __first = true;                                 \
-               if (WARN(btrfs_abort_should_print_stack(error), \
-                       KERN_ERR                                \
-                       "BTRFS: Transaction aborted (error %d)\n",      \
-                       (error))) {                                     \
-                       /* Stack trace printed. */                      \
-               } else {                                                \
-                       btrfs_err((trans)->fs_info,                     \
-                                 "Transaction aborted (error %d)",     \
-                                 (error));                     \
-               }                                               \
+               WARN_ON(btrfs_abort_should_print_stack(error)); \
        }                                                       \
        __btrfs_abort_transaction((trans), __func__,            \
                                  __LINE__, (error), __first);  \
--

2.53.0

#5 lore raw
Quoted 47 lines · Filipe Manana
Quoted 46 lines · David Sterba

The btrfs_abort_transaction() is called at the location where we want to report the abort. It must be a macro so we get the correct line and stack trace. This inlines the necessary code and the rest is pushed to __btrfs_abort_transaction().

There's a possibility to reduce the inlined code if we move the message to the helper function as well, without loss of information. The difference is only that the WARN will not print it inside the stack report but after:

--[ cut here ]--

  WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
  ...
  --[ end trace ] --
  BTRFS error (device dm-0 state A): Transaction aborted (error -28)

While previously there would be one more line like:

--[ cut here ]--
BTRFS: Transaction aborted (error -28)

  WARNING: fs/btrfs/transaction.c:2045 at btrfs_commit_transaction+0xa21/0xd30 [btrfs], CPU#11: bonnie++/3377975
  ...
  --[ end trace ] --

This removes about 20KiB of btrfs.ko on a release config.

---

v2:
- move message in __btrfs_abort_transaction() under 'first_hit', also to make the WRITE_ONCE() assignments immediate
- savings updated to 20K, it depends on the config and compiler, use an average of a few setups

 fs/btrfs/transaction.c |  8 ++++++--
 fs/btrfs/transaction.h | 11 +----------
 2 files changed, 7 insertions(+), 12 deletions(-)
fs/btrfs/transaction.c +0 -0
index 194f581b36f3..184fe976450e 100644
@@ -2731,8 +2731,12 @@ void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 
Quoted 8 lines

fs/btrfs/transaction.c void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans, · line 2735

Quoted 6 lines
fs/btrfs/transaction.c +2 -2
        WRITE_ONCE(trans->aborted, error);
        WRITE_ONCE(trans->transaction->aborted, error);
-       if (first_hit && error == -ENOSPC)
-               btrfs_dump_space_info_for_trans_abort(fs_info);
+       if (first_hit) {
+               btrfs_err(trans->fs_info, "Transaction %llu aborted (error %d)",

There's already a local fs_info variable, no need to trans->fs_info.

Ah, copy&paste leftover, I'll fix it in for-next, thanks.