bcachefs: Make allocator stuck timeout configurable, ratelimit messages

Limit these messages to once every 2 minutes to avoid spamming logs;
with multiple devices the output can be quite significant.

Also, up the default timeout to 30 seconds from 10 seconds.

Signed-off-by: Kent Overstreet <[email protected]>

Author: Kent Overstreet

fs/bcachefs/alloc_foreground.c

@@ -1758,11 +1758,12 @@ void bch2_dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 	prt_printf(out, "buckets to invalidate\t%llu\r\n",	should_invalidate_buckets(ca, stats));
 }
 
-void bch2_print_allocator_stuck(struct bch_fs *c)
+static noinline void bch2_print_allocator_stuck(struct bch_fs *c)
 {
 	struct printbuf buf = PRINTBUF;
 
-	prt_printf(&buf, "Allocator stuck? Waited for 10 seconds\n");
+	prt_printf(&buf, "Allocator stuck? Waited for %u seconds\n",
+		   c->opts.allocator_stuck_timeout);
 
 	prt_printf(&buf, "Allocator debug:\n");
 	printbuf_indent_add(&buf, 2);
@@ -1792,3 +1793,24 @@ void bch2_print_allocator_stuck(struct bch_fs *c)
 	bch2_print_string_as_lines(KERN_ERR, buf.buf);
 	printbuf_exit(&buf);
 }
+
+static inline unsigned allocator_wait_timeout(struct bch_fs *c)
+{
+	if (c->allocator_last_stuck &&
+	    time_after(c->allocator_last_stuck + HZ * 60 * 2, jiffies))
+		return 0;
+
+	return c->opts.allocator_stuck_timeout * HZ;
+}
+
+void __bch2_wait_on_allocator(struct bch_fs *c, struct closure *cl)
+{
+	unsigned t = allocator_wait_timeout(c);
+
+	if (t && closure_sync_timeout(cl, t)) {
+		c->allocator_last_stuck = jiffies;
+		bch2_print_allocator_stuck(c);
+	}
+
+	closure_sync(cl);
+}

fs/bcachefs/alloc_foreground.h

@@ -231,6 +231,11 @@ void bch2_write_points_to_text(struct printbuf *, struct bch_fs *);
 void bch2_fs_alloc_debug_to_text(struct printbuf *, struct bch_fs *);
 void bch2_dev_alloc_debug_to_text(struct printbuf *, struct bch_dev *);
 
-void bch2_print_allocator_stuck(struct bch_fs *);
+void __bch2_wait_on_allocator(struct bch_fs *, struct closure *);
+static inline void bch2_wait_on_allocator(struct bch_fs *c, struct closure *cl)
+{
+	if (cl->closure_get_happened)
+		__bch2_wait_on_allocator(c, cl);
+}
 
 #endif /* _BCACHEFS_ALLOC_FOREGROUND_H */

fs/bcachefs/bcachefs.h

@@ -893,6 +893,8 @@ struct bch_fs {
 	struct bch_fs_usage_base __percpu *usage;
 	u64 __percpu		*online_reserved;
 
+	unsigned long		allocator_last_stuck;
+
 	struct io_clock		io_clock[2];
 
 	/* JOURNAL SEQ BLACKLIST */

fs/bcachefs/bcachefs_format.h

@@ -836,6 +836,8 @@ LE64_BITMASK(BCH_SB_BACKGROUND_COMPRESSION_TYPE_HI,
 
 LE64_BITMASK(BCH_SB_VERSION_UPGRADE_COMPLETE,
 					struct bch_sb, flags[5],  0, 16);
+LE64_BITMASK(BCH_SB_ALLOCATOR_STUCK_TIMEOUT,
+					struct bch_sb, flags[5], 16, 32);
 
 static inline __u64 BCH_SB_COMPRESSION_TYPE(const struct bch_sb *sb)
 {

fs/bcachefs/io_misc.c

@@ -126,11 +126,7 @@ int bch2_extent_fallocate(struct btree_trans *trans,
 
 	if (closure_nr_remaining(&cl) != 1) {
 		bch2_trans_unlock_long(trans);
-
-		if (closure_sync_timeout(&cl, HZ * 10)) {
-			bch2_print_allocator_stuck(c);
-			closure_sync(&cl);
-		}
+		bch2_wait_on_allocator(c, &cl);
 	}
 
 	return ret;

fs/bcachefs/io_write.c

@@ -1503,10 +1503,7 @@ static void __bch2_write(struct bch_write_op *op)
 	if ((op->flags & BCH_WRITE_SYNC) ||
 	    (!(op->flags & BCH_WRITE_SUBMITTED) &&
 	     !(op->flags & BCH_WRITE_IN_WORKER))) {
-		if (closure_sync_timeout(&op->cl, HZ * 10)) {
-			bch2_print_allocator_stuck(c);
-			closure_sync(&op->cl);
-		}
+		bch2_wait_on_allocator(c, &op->cl);
 
 		__bch2_write_index(op);
 

fs/bcachefs/opts.h

@@ -391,6 +391,11 @@ enum fsck_err_opts {
 	  OPT_BOOL(),							\
 	  BCH_SB_JOURNAL_TRANSACTION_NAMES, true,			\
 	  NULL,		"Log transaction function names in journal")	\
+	x(allocator_stuck_timeout,	u16,				\
+	  OPT_FS|OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME,			\
+	  OPT_UINT(0, U16_MAX),						\
+	  BCH_SB_ALLOCATOR_STUCK_TIMEOUT, 30,				\
+	  NULL,		"Default timeout in seconds for stuck allocator messages")\
 	x(noexcl,			u8,				\
 	  OPT_FS|OPT_MOUNT,						\
 	  OPT_BOOL(),							\

fs/bcachefs/super-io.c

@@ -414,6 +414,10 @@ static int bch2_sb_validate(struct bch_sb_handle *disk_sb,
 
 		if (!BCH_SB_VERSION_UPGRADE_COMPLETE(sb))
 			SET_BCH_SB_VERSION_UPGRADE_COMPLETE(sb, le16_to_cpu(sb->version));
+
+		if (le16_to_cpu(sb->version) <= bcachefs_metadata_version_disk_accounting_v2 &&
+		    !BCH_SB_ALLOCATOR_STUCK_TIMEOUT(sb))
+			SET_BCH_SB_ALLOCATOR_STUCK_TIMEOUT(sb, 30);
 	}
 
 	for (opt_id = 0; opt_id < bch2_opts_nr; opt_id++) {