bcachefs: Improved errcodes

Instead of overloading standard error codes (EINTR/EAGAIN), and defining
short lists of error codes in multiple places that potentially end up
overlapping & conflicting, we're now going to have one master list of
error codes.

Error codes are defined with an x-macro: thus we also have
bch2_err_str() now.

Also, error codes have a class field. Now, instead of checking for
errors with ==, code should use bch2_err_matches(), which returns true
if the error is equal to or a sub-error of the error class.

This means we can define unique errors for every source location where
an error is generated, which will help improve our error messages.

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

fs/bcachefs/Kconfig

@@ -21,6 +21,7 @@ config BCACHEFS_FS
 	select XOR_BLOCKS
 	select XXHASH
 	select SRCU
+	select SYMBOLIC_ERRNAME
 	help
 	The bcachefs filesystem - a modern, copy on write filesystem, with
 	support for multiple devices, compression, checksumming, etc.

fs/bcachefs/Makefile

@@ -27,6 +27,7 @@ bcachefs-y		:=	\
 	disk_groups.o		\
 	data_update.o		\
 	ec.o			\
+	errcode.o		\
 	error.o			\
 	extents.o		\
 	extent_update.o		\

fs/bcachefs/alloc_background.c

@@ -1051,7 +1051,8 @@ static void bch2_do_discards_work(struct work_struct *work)
 
 	percpu_ref_put(&c->writes);
 
-	trace_discard_buckets(c, seen, open, need_journal_commit, discarded, ret);
+	trace_discard_buckets(c, seen, open, need_journal_commit, discarded,
+			      bch2_err_str(ret));
 }
 
 void bch2_do_discards(struct bch_fs *c)

fs/bcachefs/alloc_foreground.c

@@ -238,7 +238,7 @@ static struct open_bucket *__try_alloc_bucket(struct bch_fs *c, struct bch_dev *
 			c->blocked_allocate_open_bucket = local_clock();
 
 		spin_unlock(&c->freelist_lock);
-		return ERR_PTR(-OPEN_BUCKETS_EMPTY);
+		return ERR_PTR(-BCH_ERR_open_buckets_empty);
 	}
 
 	/* Recheck under lock: */
@@ -440,7 +440,7 @@ again:
 		goto again;
 	}
 
-	return ob ?: ERR_PTR(ret ?: -FREELIST_EMPTY);
+	return ob ?: ERR_PTR(ret ?: -BCH_ERR_no_buckets_found);
 }
 
 static struct open_bucket *bch2_bucket_alloc_freelist(struct btree_trans *trans,
@@ -548,7 +548,7 @@ again:
 		if (!c->blocked_allocate)
 			c->blocked_allocate = local_clock();
 
-		ob = ERR_PTR(-FREELIST_EMPTY);
+		ob = ERR_PTR(-BCH_ERR_freelist_empty);
 		goto err;
 	}
 
@@ -579,7 +579,7 @@ again:
 		bch2_journal_flush_async(&c->journal, NULL);
 err:
 	if (!ob)
-		ob = ERR_PTR(-FREELIST_EMPTY);
+		ob = ERR_PTR(-BCH_ERR_no_buckets_found);
 
 	if (!IS_ERR(ob)) {
 		trace_bucket_alloc(ca, bch2_alloc_reserves[reserve],
@@ -591,7 +591,8 @@ err:
 				   skipped_open,
 				   skipped_need_journal_commit,
 				   skipped_nouse,
-				   cl == NULL, PTR_ERR_OR_ZERO(ob));
+				   cl == NULL,
+				   "");
 	} else {
 		trace_bucket_alloc_fail(ca, bch2_alloc_reserves[reserve],
 				   usage.d[BCH_DATA_free].buckets,
@@ -602,7 +603,8 @@ err:
 				   skipped_open,
 				   skipped_need_journal_commit,
 				   skipped_nouse,
-				   cl == NULL, PTR_ERR_OR_ZERO(ob));
+				   cl == NULL,
+				   bch2_err_str(PTR_ERR(ob)));
 		atomic_long_inc(&c->bucket_alloc_fail);
 	}
 
@@ -750,7 +752,7 @@ static int bch2_bucket_alloc_set_trans(struct btree_trans *trans,
 	if (*nr_effective >= nr_replicas)
 		ret = 0;
 	else if (!ret)
-		ret = -INSUFFICIENT_DEVICES;
+		ret = -BCH_ERR_insufficient_devices;
 
 	return ret;
 }
@@ -923,8 +925,8 @@ static int open_bucket_add_buckets(struct btree_trans *trans,
 						 nr_replicas, nr_effective,
 						 have_cache, flags, _cl);
 			if (ret == -EINTR ||
-			    ret == -FREELIST_EMPTY ||
-			    ret == -OPEN_BUCKETS_EMPTY)
+			    bch2_err_matches(ret, BCH_ERR_freelist_empty) ||
+			    bch2_err_matches(ret, BCH_ERR_open_buckets_empty))
 				return ret;
 			if (*nr_effective >= nr_replicas)
 				return 0;
@@ -947,7 +949,7 @@ retry_blocking:
 				reserve, flags, cl);
 	if (ret &&
 	    ret != -EINTR &&
-	    ret != -INSUFFICIENT_DEVICES &&
+	    !bch2_err_matches(ret, BCH_ERR_insufficient_devices) &&
 	    !cl && _cl) {
 		cl = _cl;
 		goto retry_blocking;
@@ -1203,7 +1205,7 @@ alloc_done:
 	if (erasure_code && !ec_open_bucket(c, &ptrs))
 		pr_debug("failed to get ec bucket: ret %u", ret);
 
-	if (ret == -INSUFFICIENT_DEVICES &&
+	if (ret == -BCH_ERR_insufficient_devices &&
 	    nr_effective >= nr_replicas_required)
 		ret = 0;
 
@@ -1234,19 +1236,18 @@ err:
 
 	mutex_unlock(&wp->lock);
 
-	if (ret == -FREELIST_EMPTY &&
+	if (bch2_err_matches(ret, BCH_ERR_freelist_empty) &&
 	    try_decrease_writepoints(c, write_points_nr))
 		goto retry;
 
-	switch (ret) {
-	case -OPEN_BUCKETS_EMPTY:
-	case -FREELIST_EMPTY:
+	if (bch2_err_matches(ret, BCH_ERR_open_buckets_empty) ||
+	    bch2_err_matches(ret, BCH_ERR_freelist_empty))
 		return cl ? -EAGAIN : -ENOSPC;
-	case -INSUFFICIENT_DEVICES:
+
+	if (bch2_err_matches(ret, BCH_ERR_insufficient_devices))
 		return -EROFS;
-	default:
-		return ret;
-	}
+
+	return ret;
 }
 
 int bch2_alloc_sectors_start(struct bch_fs *c,

fs/bcachefs/errcode.c

@@ -0,0 +1,51 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include "bcachefs.h"
+#include "errcode.h"
+
+#include <linux/errname.h>
+
+static const char * const bch2_errcode_strs[] = {
+#define x(class, err) [BCH_ERR_##err - BCH_ERR_START] = #err,
+	BCH_ERRCODES()
+#undef x
+	NULL
+};
+
+#define BCH_ERR_0	0
+
+static unsigned bch2_errcode_parents[] = {
+#define x(class, err) [BCH_ERR_##err - BCH_ERR_START] = BCH_ERR_##class,
+	BCH_ERRCODES()
+#undef x
+};
+
+const char *bch2_err_str(int err)
+{
+	const char *errstr;
+	err = abs(err);
+
+	BUG_ON(err >= BCH_ERR_MAX);
+
+	if (err >= BCH_ERR_START)
+		errstr = bch2_errcode_strs[err - BCH_ERR_START];
+	else if (err)
+		errstr = errname(err);
+	else
+		errstr = "(No error)";
+	return errstr ?: "(Invalid error)";
+}
+
+bool __bch2_err_matches(int err, int class)
+{
+	err	= abs(err);
+	class	= abs(class);
+
+	BUG_ON(err	>= BCH_ERR_MAX);
+	BUG_ON(class	>= BCH_ERR_MAX);
+
+	while (err >= BCH_ERR_START && err != class)
+		err = bch2_errcode_parents[err - BCH_ERR_START];
+
+	return err == class;
+}

fs/bcachefs/errcode.h

@@ -2,12 +2,33 @@
 #ifndef _BCACHEFS_ERRCODE_H
 #define _BCACHEFS_ERRCODE_H
 
-enum {
-	/* Bucket allocator: */
-	OPEN_BUCKETS_EMPTY =	2048,
-	FREELIST_EMPTY,		/* Allocator thread not keeping up */
-	INSUFFICIENT_DEVICES,
-	NEED_SNAPSHOT_CLEANUP,
+#define BCH_ERRCODES()							\
+	x(0,			open_buckets_empty)			\
+	x(0,			freelist_empty)				\
+	x(freelist_empty,	no_buckets_found)			\
+	x(0,			insufficient_devices)			\
+	x(0,			need_snapshot_cleanup)
+
+enum bch_errcode {
+	BCH_ERR_START		= 2048,
+#define x(class, err) BCH_ERR_##err,
+	BCH_ERRCODES()
+#undef x
+	BCH_ERR_MAX
 };
 
+const char *bch2_err_str(int);
+bool __bch2_err_matches(int, int);
+
+static inline bool _bch2_err_matches(int err, int class)
+{
+	return err && __bch2_err_matches(err, class);
+}
+
+#define bch2_err_matches(_err, _class)			\
+({							\
+	BUILD_BUG_ON(!__builtin_constant_p(_class));	\
+	_bch2_err_matches(_err, _class);		\
+})
+
 #endif /* _BCACHFES_ERRCODE_H */

fs/bcachefs/fsck.c

@@ -534,7 +534,7 @@ static int snapshots_seen_update(struct bch_fs *c, struct snapshots_seen *s,
 					bch2_btree_ids[btree_id],
 					pos.inode, pos.offset,
 					i->id, n.id, n.equiv);
-				return -NEED_SNAPSHOT_CLEANUP;
+				return -BCH_ERR_need_snapshot_cleanup;
 			}
 
 			return 0;
@@ -2371,7 +2371,7 @@ again:
 		check_nlinks(c) ?:
 		fix_reflink_p(c);
 
-	if (ret == -NEED_SNAPSHOT_CLEANUP) {
+	if (bch2_err_matches(ret, BCH_ERR_need_snapshot_cleanup)) {
 		set_bit(BCH_FS_HAVE_DELETED_SNAPSHOTS, &c->flags);
 		goto again;
 	}

fs/bcachefs/trace.h

@@ -449,9 +449,9 @@ DECLARE_EVENT_CLASS(bucket_alloc,
 		 u64 need_journal_commit,
 		 u64 nouse,
 		 bool nonblocking,
-		 int ret),
+		 const char *err),
 	TP_ARGS(ca, alloc_reserve, free, avail, copygc_wait_amount, copygc_waiting_for,
-		seen, open, need_journal_commit, nouse, nonblocking, ret),
+		seen, open, need_journal_commit, nouse, nonblocking, err),
 
 	TP_STRUCT__entry(
 		__field(dev_t,			dev			)
@@ -465,7 +465,7 @@ DECLARE_EVENT_CLASS(bucket_alloc,
 		__field(u64,			need_journal_commit	)
 		__field(u64,			nouse			)
 		__field(bool,			nonblocking		)
-		__field(int,			ret			)
+		__array(char,			err,	16		)
 	),
 
 	TP_fast_assign(
@@ -480,10 +480,10 @@ DECLARE_EVENT_CLASS(bucket_alloc,
 		__entry->need_journal_commit = need_journal_commit;
 		__entry->nouse		= nouse;
 		__entry->nonblocking	= nonblocking;
-		__entry->ret		= ret;
+		strlcpy(__entry->err, err, sizeof(__entry->err));
 	),
 
-	TP_printk("%d,%d reserve %s free %llu avail %llu copygc_wait %llu/%lli seen %llu open %llu need_journal_commit %llu nouse %llu nonblocking %u ret %i",
+	TP_printk("%d,%d reserve %s free %llu avail %llu copygc_wait %llu/%lli seen %llu open %llu need_journal_commit %llu nouse %llu nonblocking %u err %s",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __entry->reserve,
 		  __entry->free,
@@ -495,7 +495,7 @@ DECLARE_EVENT_CLASS(bucket_alloc,
 		  __entry->need_journal_commit,
 		  __entry->nouse,
 		  __entry->nonblocking,
-		  __entry->ret)
+		  __entry->err)
 );
 
 DEFINE_EVENT(bucket_alloc, bucket_alloc,
@@ -509,9 +509,9 @@ DEFINE_EVENT(bucket_alloc, bucket_alloc,
 		 u64 need_journal_commit,
 		 u64 nouse,
 		 bool nonblocking,
-		 int ret),
+		 const char *err),
 	TP_ARGS(ca, alloc_reserve, free, avail, copygc_wait_amount, copygc_waiting_for,
-		seen, open, need_journal_commit, nouse, nonblocking, ret)
+		seen, open, need_journal_commit, nouse, nonblocking, err)
 );
 
 DEFINE_EVENT(bucket_alloc, bucket_alloc_fail,
@@ -525,15 +525,15 @@ DEFINE_EVENT(bucket_alloc, bucket_alloc_fail,
 		 u64 need_journal_commit,
 		 u64 nouse,
 		 bool nonblocking,
-		 int ret),
+		 const char *err),
 	TP_ARGS(ca, alloc_reserve, free, avail, copygc_wait_amount, copygc_waiting_for,
-		seen, open, need_journal_commit, nouse, nonblocking, ret)
+		seen, open, need_journal_commit, nouse, nonblocking, err)
 );
 
 TRACE_EVENT(discard_buckets,
 	TP_PROTO(struct bch_fs *c, u64 seen, u64 open,
-		 u64 need_journal_commit, u64 discarded, int ret),
-	TP_ARGS(c, seen, open, need_journal_commit, discarded, ret),
+		 u64 need_journal_commit, u64 discarded, const char *err),
+	TP_ARGS(c, seen, open, need_journal_commit, discarded, err),
 
 	TP_STRUCT__entry(
 		__field(dev_t,		dev			)
@@ -541,7 +541,7 @@ TRACE_EVENT(discard_buckets,
 		__field(u64,		open			)
 		__field(u64,		need_journal_commit	)
 		__field(u64,		discarded		)
-		__field(int,		ret			)
+		__array(char,		err,	16		)
 	),
 
 	TP_fast_assign(
@@ -550,16 +550,16 @@ TRACE_EVENT(discard_buckets,
 		__entry->open			= open;
 		__entry->need_journal_commit	= need_journal_commit;
 		__entry->discarded		= discarded;
-		__entry->ret			= ret;
+		strlcpy(__entry->err, err, sizeof(__entry->err));
 	),
 
-	TP_printk("%d%d seen %llu open %llu need_journal_commit %llu discarded %llu ret %i",
+	TP_printk("%d%d seen %llu open %llu need_journal_commit %llu discarded %llu err %s",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __entry->seen,
 		  __entry->open,
 		  __entry->need_journal_commit,
 		  __entry->discarded,
-		  __entry->ret)
+		  __entry->err)
 );
 
 TRACE_EVENT(invalidate_bucket,