The patch titled
Subject: ocfs2: fix mounting crash if journal is not alloced
has been added to the -mm tree. Its filename is
ocfs2-fix-mounting-crash-if-journal-is-not-alloced.patch
This patch should soon appear at
https://ozlabs.org/~akpm/mmots/broken-out/ocfs2-fix-mounting-crash-if-journal-is-not-alloced.patch
and later at
https://ozlabs.org/~akpm/mmotm/broken-out/ocfs2-fix-mounting-crash-if-journal-is-not-alloced.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Heming Zhao via Ocfs2-devel <ocfs2-devel@xxxxxxxxxxxxxx>
Subject: ocfs2: fix mounting crash if journal is not alloced
Patch series "rewrite error handling during mounting stage".
This patch (of 5):
After commit da5e7c87827e8 ("ocfs2: cleanup journal init and shutdown"),
journal init later than before, it makes NULL pointer access in free
routine.
Crash flow:
ocfs2_fill_super
+ ocfs2_mount_volume
| + ocfs2_dlm_init //fail & return, osb->journal is NULL.
| + ...
| + ocfs2_check_volume //no chance to init osb->journal
|
+ ...
+ ocfs2_dismount_volume
ocfs2_release_system_inodes
...
evict
...
ocfs2_clear_inode
ocfs2_checkpoint_inode
ocfs2_ci_fully_checkpointed
time_after(journal->j_trans_id, ci->ci_last_trans)
+ journal is empty, crash!
For fixing, there are three solutions:
1> Partly revert commit da5e7c87827e8
For avoiding kernel crash, this make sense for us. We only
concerned whether there has any non-system inode access before dlm
init. The answer is NO. And all journal replay/recovery handling
happen after dlm & journal init done. So this method is not graceful
but workable.
2> Add osb->journal check in free inode routine (eg ocfs2_clear_inode)
The fix code is special for mounting phase, but it will continue
working after mounting stage. In another word, this method adds
useless code in normal inode free flow.
3> Do directly free inode in mounting phase
This method is brutal/complex and may introduce unsafe code,
currently maintainer didn't like.
At last, we chose method <1> and did partly reverted job. We reverted
journal init codes, and kept cleanup codes flow.
Link: https://lkml.kernel.org/r/20220424130952.2436-1-heming.zhao@xxxxxxxx
Link: https://lkml.kernel.org/r/20220424130952.2436-2-heming.zhao@xxxxxxxx
Fixes: da5e7c87827e8 ("ocfs2: cleanup journal init and shutdown")
Signed-off-by: Heming Zhao <heming.zhao@xxxxxxxx>
Reviewed-by: Joseph Qi <joseph.qi@xxxxxxxxxxxxxxxxx>
Cc: Mark Fasheh <mark@xxxxxxxxxx>
Cc: Joel Becker <jlbec@xxxxxxxxxxxx>
Cc: Junxiao Bi <junxiao.bi@xxxxxxxxxx>
Cc: Changwei Ge <gechangwei@xxxxxxx>
Cc: Gang He <ghe@xxxxxxxx>
Cc: Jun Piao <piaojun@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
fs/ocfs2/inode.c | 4 ++--
fs/ocfs2/journal.c | 33 +++++++++++++++++++++++----------
fs/ocfs2/journal.h | 2 ++
fs/ocfs2/super.c | 15 +++++++++++++++
4 files changed, 42 insertions(+), 12 deletions(-)
--- a/fs/ocfs2/inode.c~ocfs2-fix-mounting-crash-if-journal-is-not-alloced
+++ a/fs/ocfs2/inode.c
@@ -125,6 +125,7 @@ struct inode *ocfs2_iget(struct ocfs2_su
struct inode *inode = NULL;
struct super_block *sb = osb->sb;
struct ocfs2_find_inode_args args;
+ journal_t *journal = osb->journal->j_journal;
trace_ocfs2_iget_begin((unsigned long long)blkno, flags,
sysfile_type);
@@ -171,11 +172,10 @@ struct inode *ocfs2_iget(struct ocfs2_su
* part of the transaction - the inode could have been reclaimed and
* now it is reread from disk.
*/
- if (osb->journal) {
+ if (journal) {
transaction_t *transaction;
tid_t tid;
struct ocfs2_inode_info *oi = OCFS2_I(inode);
- journal_t *journal = osb->journal->j_journal;
read_lock(&journal->j_state_lock);
if (journal->j_running_transaction)
--- a/fs/ocfs2/journal.c~ocfs2-fix-mounting-crash-if-journal-is-not-alloced
+++ a/fs/ocfs2/journal.c
@@ -810,22 +810,20 @@ void ocfs2_set_journal_params(struct ocf
write_unlock(&journal->j_state_lock);
}
-int ocfs2_journal_init(struct ocfs2_super *osb, int *dirty)
+/*
+ * alloc & initialize skeleton for journal structure.
+ * ocfs2_journal_init() will make fs have journal ability.
+ */
+int ocfs2_journal_alloc(struct ocfs2_super *osb)
{
- int status = -1;
- struct inode *inode = NULL; /* the journal inode */
- journal_t *j_journal = NULL;
- struct ocfs2_journal *journal = NULL;
- struct ocfs2_dinode *di = NULL;
- struct buffer_head *bh = NULL;
- int inode_lock = 0;
+ int status = 0;
+ struct ocfs2_journal *journal;
- /* initialize our journal structure */
journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
if (!journal) {
mlog(ML_ERROR, "unable to alloc journal\n");
status = -ENOMEM;
- goto done;
+ goto bail;
}
osb->journal = journal;
journal->j_osb = osb;
@@ -839,6 +837,21 @@ int ocfs2_journal_init(struct ocfs2_supe
INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
journal->j_state = OCFS2_JOURNAL_FREE;
+bail:
+ return status;
+}
+
+int ocfs2_journal_init(struct ocfs2_super *osb, int *dirty)
+{
+ int status = -1;
+ struct inode *inode = NULL; /* the journal inode */
+ journal_t *j_journal = NULL;
+ struct ocfs2_journal *journal = osb->journal;
+ struct ocfs2_dinode *di = NULL;
+ struct buffer_head *bh = NULL;
+ int inode_lock = 0;
+
+ BUG_ON(!journal);
/* already have the inode for our journal */
inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE,
osb->slot_num);
--- a/fs/ocfs2/journal.h~ocfs2-fix-mounting-crash-if-journal-is-not-alloced
+++ a/fs/ocfs2/journal.h
@@ -154,6 +154,7 @@ int ocfs2_compute_replay_slots(struct oc
* Journal Control:
* Initialize, Load, Shutdown, Wipe a journal.
*
+ * ocfs2_journal_alloc - Initialize skeleton for journal structure.
* ocfs2_journal_init - Initialize journal structures in the OSB.
* ocfs2_journal_load - Load the given journal off disk. Replay it if
* there's transactions still in there.
@@ -167,6 +168,7 @@ int ocfs2_compute_replay_slots(struct oc
* ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint.
*/
void ocfs2_set_journal_params(struct ocfs2_super *osb);
+int ocfs2_journal_alloc(struct ocfs2_super *osb);
int ocfs2_journal_init(struct ocfs2_super *osb, int *dirty);
void ocfs2_journal_shutdown(struct ocfs2_super *osb);
int ocfs2_journal_wipe(struct ocfs2_journal *journal,
--- a/fs/ocfs2/super.c~ocfs2-fix-mounting-crash-if-journal-is-not-alloced
+++ a/fs/ocfs2/super.c
@@ -2195,6 +2195,15 @@ static int ocfs2_initialize_super(struct
get_random_bytes(&osb->s_next_generation, sizeof(u32));
+ /*
+ * FIXME
+ * This should be done in ocfs2_journal_init(), but any inode
+ * writes back operation will cause the filesystem to crash.
+ */
+ status = ocfs2_journal_alloc(osb);
+ if (status < 0)
+ goto bail;
+
INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
init_llist_head(&osb->dquot_drop_list);
@@ -2483,6 +2492,12 @@ static void ocfs2_delete_osb(struct ocfs
kfree(osb->osb_orphan_wipes);
kfree(osb->slot_recovery_generations);
+ /* FIXME
+ * This belongs in journal shutdown, but because we have to
+ * allocate osb->journal at the middle of ocfs2_initialize_super(),
+ * we free it here.
+ */
+ kfree(osb->journal);
kfree(osb->local_alloc_copy);
kfree(osb->uuid_str);
kfree(osb->vol_label);
_
Patches currently in -mm which might be from ocfs2-devel@xxxxxxxxxxxxxx are
ocfs2-fix-mounting-crash-if-journal-is-not-alloced.patch
ocfs2-change-return-type-of-ocfs2_resmap_init.patch
ocfs2-ocfs2_initialize_super-does-cleanup-job-before-return-error.patch
ocfs2-ocfs2_mount_volume-does-cleanup-job-before-return-error.patch
ocfs2-rewrite-error-handling-of-ocfs2_fill_super.patch
