bcachefs: Print cycle on unrecoverable deadlock

Some lock operations can't fail; a cycle of nofail locks is impossible
to recover from. So we want to get rid of these nofail locking
operations, but as this is tricky it'll be done incrementally.

If such a cycle happens, this patch prints out which codepaths are
involved so we know what to work on next.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

4 files changed, 45 insertions, 22 deletions

diff --git a/fs/bcachefs/btree_locking.c b/fs/bcachefs/btree_locking.c
index e83690d2ddea..458f870c7682 100644
--- a/fs/bcachefs/btree_locking.c
+++ b/fs/bcachefs/btree_locking.c
@@ -138,7 +138,29 @@ static noinline int break_cycle(struct lock_graph *g)
 		return abort_lock(g, i);
 	}
 
-	BUG();
+	{
+		struct bch_fs *c = g->g->trans->c;
+		struct printbuf buf = PRINTBUF;
+
+		bch_err(c, "cycle of nofail locks");
+
+		for (i = g->g; i < g->g + g->nr; i++) {
+			struct btree_trans *trans = i->trans;
+
+			bch2_btree_trans_to_text(&buf, trans);
+
+			prt_printf(&buf, "backtrace:");
+			prt_newline(&buf);
+			printbuf_indent_add(&buf, 2);
+			bch2_prt_backtrace(&buf, trans->locking_wait.task);
+			printbuf_indent_sub(&buf, 2);
+			prt_newline(&buf);
+		}
+
+		bch2_print_string_as_lines(KERN_ERR, buf.buf);
+		printbuf_exit(&buf);
+		BUG();
+	}
 }
 
 static void lock_graph_pop(struct lock_graph *g)
diff --git a/fs/bcachefs/debug.c b/fs/bcachefs/debug.c
index c7d558381388..7abc707d2f38 100644
--- a/fs/bcachefs/debug.c
+++ b/fs/bcachefs/debug.c
@@ -501,26 +501,6 @@ static const struct file_operations cached_btree_nodes_ops = {
 };
 
 #ifdef CONFIG_BCACHEFS_DEBUG_TRANSACTIONS
-static int prt_backtrace(struct printbuf *out, struct task_struct *task)
-{
-	unsigned long entries[32];
-	unsigned i, nr_entries;
-	int ret;
-
-	ret = down_read_killable(&task->signal->exec_update_lock);
-	if (ret)
-		return ret;
-
-	nr_entries = stack_trace_save_tsk(task, entries, ARRAY_SIZE(entries), 0);
-	for (i = 0; i < nr_entries; i++) {
-		prt_printf(out, "[<0>] %pB", (void *)entries[i]);
-		prt_newline(out);
-	}
-
-	up_read(&task->signal->exec_update_lock);
-	return 0;
-}
-
 static ssize_t bch2_btree_transactions_read(struct file *file, char __user *buf,
 					    size_t size, loff_t *ppos)
 {
@@ -547,7 +527,7 @@ static ssize_t bch2_btree_transactions_read(struct file *file, char __user *buf,
 		prt_printf(&i->buf, "backtrace:");
 		prt_newline(&i->buf);
 		printbuf_indent_add(&i->buf, 2);
-		prt_backtrace(&i->buf, trans->locking_wait.task);
+		bch2_prt_backtrace(&i->buf, trans->locking_wait.task);
 		printbuf_indent_sub(&i->buf, 2);
 		prt_newline(&i->buf);
 
diff --git a/fs/bcachefs/util.c b/fs/bcachefs/util.c
index ae580bccf650..eb8633133829 100644
--- a/fs/bcachefs/util.c
+++ b/fs/bcachefs/util.c
@@ -295,6 +295,26 @@ void bch2_print_string_as_lines(const char *prefix, const char *lines)
 	console_unlock();
 }
 
+int bch2_prt_backtrace(struct printbuf *out, struct task_struct *task)
+{
+	unsigned long entries[32];
+	unsigned i, nr_entries;
+	int ret;
+
+	ret = down_read_killable(&task->signal->exec_update_lock);
+	if (ret)
+		return ret;
+
+	nr_entries = stack_trace_save_tsk(task, entries, ARRAY_SIZE(entries), 0);
+	for (i = 0; i < nr_entries; i++) {
+		prt_printf(out, "[<0>] %pB", (void *)entries[i]);
+		prt_newline(out);
+	}
+
+	up_read(&task->signal->exec_update_lock);
+	return 0;
+}
+
 /* time stats: */
 
 #ifndef CONFIG_BCACHEFS_NO_LATENCY_ACCT
diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
index a16f8bb9d415..3b0090faef4d 100644
--- a/fs/bcachefs/util.h
+++ b/fs/bcachefs/util.h
@@ -383,6 +383,7 @@ u64 bch2_read_flag_list(char *, const char * const[]);
 void bch2_prt_u64_binary(struct printbuf *, u64, unsigned);
 
 void bch2_print_string_as_lines(const char *prefix, const char *lines);
+int bch2_prt_backtrace(struct printbuf *, struct task_struct *);
 
 #define NR_QUANTILES	15
 #define QUANTILE_IDX(i)	inorder_to_eytzinger0(i, NR_QUANTILES)