Merge tag 'bcachefs-2024-01-01' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs from Kent Overstreet:
 "More bcachefs bugfixes for 6.7, and forwards compatibility work:

   - fix for a nasty extents + snapshot interaction, reported when
     reflink of a snapshotted file wouldn't complete but turned out to
     be a more general bug

   - fix for an invalid free in dio write path when iov vector was
     longer than our inline vector

   - fix for a buffer overflow in the nocow write path -
     BCH_REPLICAS_MAX doesn't actually limit the number of pointers in
     an extent when cached pointers are included

   - RO snapshots are actually RO now

   - And, a new superblock section to avoid future breakage when the
     disk space acounting rewrite rolls out: the new superblock section
     describes versions that need work to downgrade, where the work
     required is a list of recovery passes and errors to silently fix"

* tag 'bcachefs-2024-01-01' of https://evilpiepirate.org/git/bcachefs:
  bcachefs: make RO snapshots actually RO
  bcachefs: bch_sb_field_downgrade
  bcachefs: bch_sb.recovery_passes_required
  bcachefs: Add persistent identifiers for recovery passes
  bcachefs: prt_bitflags_vector()
  bcachefs: move BCH_SB_ERRS() to sb-errors_types.h
  bcachefs: fix buffer overflow in nocow write path
  bcachefs: DARRAY_PREALLOCATED()
  bcachefs: Switch darray to kvmalloc()
  bcachefs: Factor out darray resize slowpath
  bcachefs: fix setting version_upgrade_complete
  bcachefs: fix invalid free in dio write path
  bcachefs: Fix extents iteration + snapshots interaction

Author: torvalds

fs/bcachefs/Makefile

@@ -28,6 +28,7 @@ bcachefs-y		:=	\
 	clock.o			\
 	compress.o		\
 	counters.o		\
+	darray.o		\
 	debug.o			\
 	dirent.o		\
 	disk_groups.o		\
@@ -70,6 +71,7 @@ bcachefs-y		:=	\
 	reflink.o		\
 	replicas.o		\
 	sb-clean.o		\
+	sb-downgrade.o		\
 	sb-errors.o		\
 	sb-members.o		\
 	siphash.o		\

fs/bcachefs/acl.c

@@ -366,7 +366,8 @@ int bch2_set_acl(struct mnt_idmap *idmap,
 	bch2_trans_begin(trans);
 	acl = _acl;
 
-	ret = bch2_inode_peek(trans, &inode_iter, &inode_u, inode_inum(inode),
+	ret   = bch2_subvol_is_ro_trans(trans, inode->ei_subvol) ?:
+		bch2_inode_peek(trans, &inode_iter, &inode_u, inode_inum(inode),
 			      BTREE_ITER_INTENT);
 	if (ret)
 		goto btree_err;

fs/bcachefs/bcachefs.h

@@ -737,6 +737,7 @@ struct bch_fs {
 		unsigned	nsec_per_time_unit;
 		u64		features;
 		u64		compat;
+		unsigned long	errors_silent[BITS_TO_LONGS(BCH_SB_ERR_MAX)];
 	}			sb;
 
 

fs/bcachefs/bcachefs_format.h

@@ -1207,19 +1207,21 @@ struct bch_sb_field {
 };
 
 #define BCH_SB_FIELDS()				\
-	x(journal,	0)			\
-	x(members_v1,	1)			\
-	x(crypt,	2)			\
-	x(replicas_v0,	3)			\
-	x(quota,	4)			\
-	x(disk_groups,	5)			\
-	x(clean,	6)			\
-	x(replicas,	7)			\
-	x(journal_seq_blacklist, 8)		\
-	x(journal_v2,	9)			\
-	x(counters,	10)			\
-	x(members_v2,	11)			\
-	x(errors,	12)
+	x(journal,			0)	\
+	x(members_v1,			1)	\
+	x(crypt,			2)	\
+	x(replicas_v0,			3)	\
+	x(quota,			4)	\
+	x(disk_groups,			5)	\
+	x(clean,			6)	\
+	x(replicas,			7)	\
+	x(journal_seq_blacklist,	8)	\
+	x(journal_v2,			9)	\
+	x(counters,			10)	\
+	x(members_v2,			11)	\
+	x(errors,			12)	\
+	x(ext,				13)	\
+	x(downgrade,			14)
 
 enum bch_sb_field_type {
 #define x(f, nr)	BCH_SB_FIELD_##f = nr,
@@ -1631,6 +1633,24 @@ struct bch_sb_field_errors {
 LE64_BITMASK(BCH_SB_ERROR_ENTRY_ID,	struct bch_sb_field_error_entry, v,  0, 16);
 LE64_BITMASK(BCH_SB_ERROR_ENTRY_NR,	struct bch_sb_field_error_entry, v, 16, 64);
 
+struct bch_sb_field_ext {
+	struct bch_sb_field	field;
+	__le64			recovery_passes_required[2];
+	__le64			errors_silent[8];
+};
+
+struct bch_sb_field_downgrade_entry {
+	__le16			version;
+	__le64			recovery_passes[2];
+	__le16			nr_errors;
+	__le16			errors[] __counted_by(nr_errors);
+} __packed __aligned(2);
+
+struct bch_sb_field_downgrade {
+	struct bch_sb_field	field;
+	struct bch_sb_field_downgrade_entry entries[];
+};
+
 /* Superblock: */
 
 /*
@@ -1644,6 +1664,11 @@ LE64_BITMASK(BCH_SB_ERROR_ENTRY_NR,	struct bch_sb_field_error_entry, v, 16, 64);
 
 #define RECOVERY_PASS_ALL_FSCK		(1ULL << 63)
 
+/*
+ * field 1:		version name
+ * field 2:		BCH_VERSION(major, minor)
+ * field 3:		recovery passess required on upgrade
+ */
 #define BCH_METADATA_VERSIONS()						\
 	x(bkey_renumber,		BCH_VERSION(0, 10),		\
 	  RECOVERY_PASS_ALL_FSCK)					\

fs/bcachefs/btree_iter.c

@@ -2085,18 +2085,16 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 			goto out_no_locked;
 
 		/*
-		 * iter->pos should be mononotically increasing, and always be
-		 * equal to the key we just returned - except extents can
-		 * straddle iter->pos:
+		 * We need to check against @end before FILTER_SNAPSHOTS because
+		 * if we get to a different inode that requested we might be
+		 * seeing keys for a different snapshot tree that will all be
+		 * filtered out.
+		 *
+		 * But we can't do the full check here, because bkey_start_pos()
+		 * isn't monotonically increasing before FILTER_SNAPSHOTS, and
+		 * that's what we check against in extents mode:
 		 */
-		if (!(iter->flags & BTREE_ITER_IS_EXTENTS))
-			iter_pos = k.k->p;
-		else
-			iter_pos = bkey_max(iter->pos, bkey_start_pos(k.k));
-
-		if (unlikely(!(iter->flags & BTREE_ITER_IS_EXTENTS)
-			     ? bkey_gt(iter_pos, end)
-			     : bkey_ge(iter_pos, end)))
+		if (k.k->p.inode > end.inode)
 			goto end;
 
 		if (iter->update_path &&
@@ -2155,6 +2153,21 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
 			continue;
 		}
 
+		/*
+		 * iter->pos should be mononotically increasing, and always be
+		 * equal to the key we just returned - except extents can
+		 * straddle iter->pos:
+		 */
+		if (!(iter->flags & BTREE_ITER_IS_EXTENTS))
+			iter_pos = k.k->p;
+		else
+			iter_pos = bkey_max(iter->pos, bkey_start_pos(k.k));
+
+		if (unlikely(!(iter->flags & BTREE_ITER_IS_EXTENTS)
+			     ? bkey_gt(iter_pos, end)
+			     : bkey_ge(iter_pos, end)))
+			goto end;
+
 		break;
 	}
 

fs/bcachefs/darray.c

@@ -0,0 +1,24 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/log2.h>
+#include <linux/slab.h>
+#include "darray.h"
+
+int __bch2_darray_resize(darray_char *d, size_t element_size, size_t new_size, gfp_t gfp)
+{
+	if (new_size > d->size) {
+		new_size = roundup_pow_of_two(new_size);
+
+		void *data = kvmalloc_array(new_size, element_size, gfp);
+		if (!data)
+			return -ENOMEM;
+
+		memcpy(data, d->data, d->size * element_size);
+		if (d->data != d->preallocated)
+			kvfree(d->data);
+		d->data	= data;
+		d->size = new_size;
+	}
+
+	return 0;
+}

fs/bcachefs/darray.h

@@ -8,39 +8,48 @@
  * Inspired by CCAN's darray
  */
 
-#include "util.h"
 #include <linux/slab.h>
 
-#define DARRAY(type)							\
+#define DARRAY_PREALLOCATED(_type, _nr)					\
 struct {								\
 	size_t nr, size;						\
-	type *data;							\
+	_type *data;							\
+	_type preallocated[_nr];					\
 }
 
-typedef DARRAY(void) darray_void;
+#define DARRAY(_type) DARRAY_PREALLOCATED(_type, 0)
 
-static inline int __darray_make_room(darray_void *d, size_t t_size, size_t more, gfp_t gfp)
+typedef DARRAY(char)	darray_char;
+
+int __bch2_darray_resize(darray_char *, size_t, size_t, gfp_t);
+
+static inline int __darray_resize(darray_char *d, size_t element_size,
+				  size_t new_size, gfp_t gfp)
 {
-	if (d->nr + more > d->size) {
-		size_t new_size = roundup_pow_of_two(d->nr + more);
-		void *data = krealloc_array(d->data, new_size, t_size, gfp);
+	return unlikely(new_size > d->size)
+		? __bch2_darray_resize(d, element_size, new_size, gfp)
+		: 0;
+}
 
-		if (!data)
-			return -ENOMEM;
+#define darray_resize_gfp(_d, _new_size, _gfp)				\
+	unlikely(__darray_resize((darray_char *) (_d), sizeof((_d)->data[0]), (_new_size), _gfp))
 
-		d->data	= data;
-		d->size = new_size;
-	}
+#define darray_resize(_d, _new_size)					\
+	darray_resize_gfp(_d, _new_size, GFP_KERNEL)
 
-	return 0;
+static inline int __darray_make_room(darray_char *d, size_t t_size, size_t more, gfp_t gfp)
+{
+	return __darray_resize(d, t_size, d->nr + more, gfp);
 }
 
 #define darray_make_room_gfp(_d, _more, _gfp)				\
-	__darray_make_room((darray_void *) (_d), sizeof((_d)->data[0]), (_more), _gfp)
+	__darray_make_room((darray_char *) (_d), sizeof((_d)->data[0]), (_more), _gfp)
 
 #define darray_make_room(_d, _more)					\
 	darray_make_room_gfp(_d, _more, GFP_KERNEL)
 
+#define darray_room(_d)		((_d).size - (_d).nr)
+
 #define darray_top(_d)		((_d).data[(_d).nr])
 
 #define darray_push_gfp(_d, _item, _gfp)				\
@@ -80,13 +89,16 @@ static inline int __darray_make_room(darray_void *d, size_t t_size, size_t more,
 
 #define darray_init(_d)							\
 do {									\
-	(_d)->data = NULL;						\
-	(_d)->nr = (_d)->size = 0;					\
+	(_d)->nr = 0;							\
+	(_d)->size = ARRAY_SIZE((_d)->preallocated);			\
+	(_d)->data = (_d)->size ? (_d)->preallocated : NULL;		\
 } while (0)
 
 #define darray_exit(_d)							\
 do {									\
-	kfree((_d)->data);						\
+	if (!ARRAY_SIZE((_d)->preallocated) ||				\
+	    (_d)->data != (_d)->preallocated)				\
+		kvfree((_d)->data);					\
 	darray_init(_d);						\
 } while (0)
 

fs/bcachefs/errcode.h

@@ -95,6 +95,7 @@
 	x(ENOSPC,			ENOSPC_sb_members)			\
 	x(ENOSPC,			ENOSPC_sb_members_v2)			\
 	x(ENOSPC,			ENOSPC_sb_crypt)			\
+	x(ENOSPC,			ENOSPC_sb_downgrade)			\
 	x(ENOSPC,			ENOSPC_btree_slot)			\
 	x(ENOSPC,			ENOSPC_snapshot_tree)			\
 	x(ENOENT,			ENOENT_bkey_type_mismatch)		\
@@ -218,6 +219,8 @@
 	x(BCH_ERR_invalid_sb,		invalid_sb_quota)			\
 	x(BCH_ERR_invalid_sb,		invalid_sb_errors)			\
 	x(BCH_ERR_invalid_sb,		invalid_sb_opt_compression)		\
+	x(BCH_ERR_invalid_sb,		invalid_sb_ext)				\
+	x(BCH_ERR_invalid_sb,		invalid_sb_downgrade)			\
 	x(BCH_ERR_invalid,		invalid_bkey)				\
 	x(BCH_ERR_operation_blocked,    nocow_lock_blocked)			\
 	x(EIO,				btree_node_read_err)			\

fs/bcachefs/error.c

@@ -152,6 +152,9 @@ int bch2_fsck_err(struct bch_fs *c,
 	struct printbuf buf = PRINTBUF, *out = &buf;
 	int ret = -BCH_ERR_fsck_ignore;
 
+	if (test_bit(err, c->sb.errors_silent))
+		return -BCH_ERR_fsck_fix;
+
 	bch2_sb_error_count(c, err);
 
 	va_start(args, fmt);

fs/bcachefs/fs-io-direct.c

@@ -216,11 +216,11 @@ struct dio_write {
 	struct address_space		*mapping;
 	struct bch_inode_info		*inode;
 	struct mm_struct		*mm;
+	const struct iovec		*iov;
 	unsigned			loop:1,
 					extending:1,
 					sync:1,
-					flush:1,
-					free_iov:1;
+					flush:1;
 	struct quota_res		quota_res;
 	u64				written;
 
@@ -312,12 +312,10 @@ static noinline int bch2_dio_write_copy_iov(struct dio_write *dio)
 		return -1;
 
 	if (dio->iter.nr_segs > ARRAY_SIZE(dio->inline_vecs)) {
-		iov = kmalloc_array(dio->iter.nr_segs, sizeof(*iov),
+		dio->iov = iov = kmalloc_array(dio->iter.nr_segs, sizeof(*iov),
 				    GFP_KERNEL);
 		if (unlikely(!iov))
 			return -ENOMEM;
-
-		dio->free_iov = true;
 	}
 
 	memcpy(iov, dio->iter.__iov, dio->iter.nr_segs * sizeof(*iov));
@@ -381,8 +379,7 @@ static __always_inline long bch2_dio_write_done(struct dio_write *dio)
 
 	bch2_pagecache_block_put(inode);
 
-	if (dio->free_iov)
-		kfree(dio->iter.__iov);
+	kfree(dio->iov);
 
 	ret = dio->op.error ?: ((long) dio->written << 9);
 	bio_put(&dio->op.wbio.bio);
@@ -626,11 +623,11 @@ ssize_t bch2_direct_write(struct kiocb *req, struct iov_iter *iter)
 	dio->mapping		= mapping;
 	dio->inode		= inode;
 	dio->mm			= current->mm;
+	dio->iov		= NULL;
 	dio->loop		= false;
 	dio->extending		= extending;
 	dio->sync		= is_sync_kiocb(req) || extending;
 	dio->flush		= iocb_is_dsync(req) && !c->opts.journal_flush_disabled;
-	dio->free_iov		= false;
 	dio->quota_res.sectors	= 0;
 	dio->written		= 0;
 	dio->iter		= *iter;