btrfs regression debugging

bisect-oracle.sh

#!/bin/bash
set -eEuo pipefail

cd "$(dirname "$0")"

pwd=$PWD

cd ~/work/extern/linux

if true
then
	~/libexec/git-clean -fdx

	~/libexec/ccache-run make ARCH=x86_64 defconfig || exit 125

	cat >> .config <<EOF
CONFIG_NET_9P=y
CONFIG_NET_9P_VIRTIO=y
CONFIG_NET_9P_DEBUG=y
CONFIG_9P_FS=y
CONFIG_9P_FS_POSIX_ACL=y
CONFIG_PCI=y
CONFIG_VIRTIO_PCI=y
CONFIG_PCI_HOST_GENERIC=y

CONFIG_BTRFS_FS=y

# CONFIG_PVPANIC=y
EOF

	~/libexec/ccache-run make ARCH=x86_64 olddefconfig || exit 125

	~/libexec/ccache-run make -j"$(nproc)" || exit 125
fi

fs=/tmp/2023-05-01/fs

rm -f "$fs"/result.txt

# shellcheck disable=SC2054
args=(
	env -i
	DISPLAY="$DISPLAY"
	HOME="$HOME"
	PATH=/usr/bin

	# ~/libexec/t

	~/libexec/bwrap-unshare
	--unshare-user
	--dev-bind / /
	--bind "$pwd"/init "$(realpath "$fs"/bin/init)"
	--dir "$fs"/btdu
	--ro-bind ~/work/btdu "$fs"/btdu

	qemu-system-x86_64
	-enable-kvm
	-kernel arch/x86/boot/bzImage
	# -drive file=fat:rw:"$fs"
	-virtfs local,path="$fs",mount_tag=root,security_model=passthrough,mount_tag=fsRoot	
	-drive file=/tmp/2023-05-01/badimage,format=raw
	# -device pvpanic
	# -fsdev local,security_model=passthrough,multidevs=remap,id=fsdev-fsRoot,path="$fs"
	# -device virtio-9p-pci,id=fsRoot,fsdev=fsdev-fsRoot,mount_tag=fsRoot
	-append "root=fsRoot rw rootfstype=9p rootflags=trans=virtio,version=9p2000.L,msize=5000000,cache=mmap,posixacl panic=-1 console=ttyS0"
	-nographic
	-no-reboot
	# -net nic -net user
) ; "${args[@]}"

cat "$fs"/result.txt || exit 125

if grep -qF '<UNREACHABLE>' "$fs"/result.txt ; then
	exit 0 # good/old
else
	exit 1 # bad/new
fi

dustmite-oracle.sh

#!/bin/bash
set -eEuo pipefail

patch="$PWD"/patch-rev.diff

cd "$(dirname "$0")"

pwd=$PWD

cd ~/work/extern/linux

if true
then
# 	~/libexec/git-clean -fdx

# 	~/libexec/ccache-run make ARCH=x86_64 defconfig || exit 125

# 	cat >> .config <<EOF
# CONFIG_NET_9P=y
# CONFIG_NET_9P_VIRTIO=y
# CONFIG_NET_9P_DEBUG=y
# CONFIG_9P_FS=y
# CONFIG_9P_FS_POSIX_ACL=y
# CONFIG_PCI=y
# CONFIG_VIRTIO_PCI=y
# CONFIG_PCI_HOST_GENERIC=y

# CONFIG_BTRFS_FS=y

# # CONFIG_PVPANIC=y
# EOF

# 	~/libexec/ccache-run make ARCH=x86_64 olddefconfig || exit 125

	git reset --hard
	git checkout 6ce6ba534418132f4c727d5707fe2794c797299c
	patch -p1 < "$patch"

	~/libexec/ccache-run make -j"$(nproc)" # || exit 125
fi

fs=/tmp/2023-05-01/fs

rm -f "$fs"/result.txt

# shellcheck disable=SC2054
args=(
	env -i
	DISPLAY="$DISPLAY"
	HOME="$HOME"
	PATH=/usr/bin

	# ~/libexec/t

	~/libexec/bwrap-unshare
	--unshare-user
	--dev-bind / /
	--bind "$pwd"/init "$(realpath "$fs"/bin/init)"
	--dir "$fs"/btdu
	--ro-bind ~/work/btdu "$fs"/btdu

	qemu-system-x86_64
	-enable-kvm
	-kernel arch/x86/boot/bzImage
	# -drive file=fat:rw:"$fs"
	-virtfs local,path="$fs",mount_tag=root,security_model=passthrough,mount_tag=fsRoot	
	-drive file=/tmp/2023-05-01/badimage,format=raw
	# -device pvpanic
	# -fsdev local,security_model=passthrough,multidevs=remap,id=fsdev-fsRoot,path="$fs"
	# -device virtio-9p-pci,id=fsRoot,fsdev=fsdev-fsRoot,mount_tag=fsRoot
	-append "root=fsRoot rw rootfstype=9p rootflags=trans=virtio,version=9p2000.L,msize=5000000,cache=mmap,posixacl panic=-1 console=ttyS0"
	-nographic
	-no-reboot
	# -net nic -net user
) ; "${args[@]}"

cat "$fs"/result.txt || exit 125

if grep -qF '<UNREACHABLE>' "$fs"/result.txt ; then
	exit 0 # good/old
else
	exit 1 # bad/new
fi

init

#!/bin/bash

set -xeEuo pipefail

mount -t proc proc /proc
mount -t sysfs sysfs /sys

mount -o ro /dev/sda /mnt

#/bin/bash

/btdu/btdu --headless --du --max-samples=100000 /mnt > /result.txt

mkfs

#!/bin/bash
set -eEuo pipefail

cd /tmp/2023-05-01
exec pacstrap -N -K -c "$PWD"/fs base

mkimage

#!/bin/bash
set -eEuo pipefail

umount mnt || true

image=/tmp/2023-05-01/badimage

rm -f "$image"
rm -rf mnt
dd if=/dev/null of="$image" bs=2G count=0 seek=1
mkfs.btrfs "$image"
mkdir -p mnt
{
	trap 'rmdir mnt' EXIT
	sudo true
	{
		trap 'sudo umount mnt' EXIT
		sudo mount "$image" mnt

		sudo chown "$UID" mnt/.
		dd if=/dev/urandom of=mnt/file bs=1G count=1

		for _ in $(seq $((4*1024))) ; do
			sync mnt
			dd if=/dev/urandom of=mnt/file bs=$((1024*1024)) count=1 seek=$((RANDOM*RANDOM*RANDOM%(1024))) conv=notrunc
		done
	}
}

patch-rev.diff

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 432064ee788e..9be11a342de5 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -35,13 +35,15 @@ static int check_extent_in_eb(const struct btrfs_key *key,
 			      const struct extent_buffer *eb,
 			      const struct btrfs_file_extent_item *fi,
 			      u64 extent_item_pos,
-			      struct extent_inode_elem **eie)
+			      struct extent_inode_elem **eie,
+			      bool ignore_offset)
 {
 	const u64 data_len = btrfs_file_extent_num_bytes(eb, fi);
 	u64 offset = 0;
 	struct extent_inode_elem *e;
 
-	if (!btrfs_file_extent_compression(eb, fi) &&
+	if (!ignore_offset &&
+	    !btrfs_file_extent_compression(eb, fi) &&
 	    !btrfs_file_extent_encryption(eb, fi) &&
 	    !btrfs_file_extent_other_encoding(eb, fi)) {
 		u64 data_offset;
@@ -79,7 +81,8 @@ static void free_inode_elem_list(struct extent_inode_elem *eie)
 
 static int find_extent_in_eb(const struct extent_buffer *eb,
 			     u64 wanted_disk_byte, u64 extent_item_pos,
-			     struct extent_inode_elem **eie)
+			     struct extent_inode_elem **eie,
+			     bool ignore_offset)
 {
 	u64 disk_byte;
 	struct btrfs_key key;
@@ -108,7 +111,7 @@ static int find_extent_in_eb(const struct extent_buffer *eb,
 		if (disk_byte != wanted_disk_byte)
 			continue;
 
-		ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie);
+		ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie, ignore_offset);
 		if (ret < 0)
 			return ret;
 	}
@@ -447,9 +450,9 @@ static int is_shared_data_backref(struct preftrees *preftrees, u64 bytenr)
 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 			   struct ulist *parents,
 			   struct preftrees *preftrees, struct prelim_ref *ref,
-			   int level, u64 time_seq, u64 extent_item_pos)
+			   int level, u64 time_seq, const u64 *extent_item_pos,
+			   bool ignore_offset)
 {
-	const bool ignore_offset = (extent_item_pos == BTRFS_IGNORE_EXTENT_OFFSET);
 	int ret = 0;
 	int slot;
 	struct extent_buffer *eb;
@@ -525,9 +528,10 @@ static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 				count++;
 			else
 				goto next;
-			if (!ignore_offset) {
+			if (extent_item_pos) {
 				ret = check_extent_in_eb(&key, eb, fi,
-							 extent_item_pos, &eie);
+						*extent_item_pos,
+						&eie, ignore_offset);
 				if (ret < 0)
 					break;
 			}
@@ -537,7 +541,7 @@ static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 						  eie, (void **)&old, GFP_NOFS);
 			if (ret < 0)
 				break;
-			if (!ret && !ignore_offset) {
+			if (!ret && extent_item_pos) {
 				while (old->next)
 					old = old->next;
 				old->next = eie;
@@ -566,7 +570,7 @@ static int resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 				struct btrfs_path *path, u64 time_seq,
 				struct preftrees *preftrees,
 				struct prelim_ref *ref, struct ulist *parents,
-				u64 extent_item_pos)
+				const u64 *extent_item_pos, bool ignore_offset)
 {
 	struct btrfs_root *root;
 	struct extent_buffer *eb;
@@ -660,7 +664,7 @@ static int resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 	}
 
 	ret = add_all_parents(root, path, parents, preftrees, ref, level,
-			      time_seq, extent_item_pos);
+			      time_seq, extent_item_pos, ignore_offset);
 out:
 	btrfs_put_root(root);
 out_free:
@@ -708,8 +712,8 @@ static void free_leaf_list(struct ulist *ulist)
 static int resolve_indirect_refs(struct btrfs_fs_info *fs_info,
 				 struct btrfs_path *path, u64 time_seq,
 				 struct preftrees *preftrees,
-				 u64 extent_item_pos,
-				 struct share_check *sc)
+				 const u64 *extent_item_pos,
+				 struct share_check *sc, bool ignore_offset)
 {
 	int err;
 	int ret = 0;
@@ -752,7 +756,8 @@ static int resolve_indirect_refs(struct btrfs_fs_info *fs_info,
 			goto out;
 		}
 		err = resolve_indirect_ref(fs_info, path, time_seq, preftrees,
-					   ref, parents, extent_item_pos);
+					   ref, parents, extent_item_pos,
+					   ignore_offset);
 		/*
 		 * we can only tolerate ENOENT,otherwise,we should catch error
 		 * and return directly.
@@ -1335,21 +1340,18 @@ static void store_backref_shared_cache(struct btrfs_backref_share_check_ctx *ctx
  *
  * Otherwise this returns 0 for success and <0 for an error.
  *
- * @extent_item_pos is meaningful only if we are dealing with a data extent.
- * If its value is not BTRFS_IGNORE_EXTENT_OFFSET, then only collect references
- * from file extent items that refer to a section of the data extent that
- * contains @extent_item_pos. If its value is BTRFS_IGNORE_EXTENT_OFFSET then
- * collect references for every file extent item that points to the data extent.
+ * If ignore_offset is set to false, only extent refs whose offsets match
+ * extent_item_pos are returned.  If true, every extent ref is returned
+ * and extent_item_pos is ignored.
  *
  * FIXME some caching might speed things up
  */
 static int find_parent_nodes(struct btrfs_trans_handle *trans,
 			     struct btrfs_fs_info *fs_info, u64 bytenr,
 			     u64 time_seq, struct ulist *refs,
-			     struct ulist *roots, u64 extent_item_pos,
-			     struct share_check *sc)
+			     struct ulist *roots, const u64 *extent_item_pos,
+			     struct share_check *sc, bool ignore_offset)
 {
-	const bool ignore_offset = (extent_item_pos == BTRFS_IGNORE_EXTENT_OFFSET);
 	struct btrfs_root *root = btrfs_extent_root(fs_info, bytenr);
 	struct btrfs_key key;
 	struct btrfs_path *path;
@@ -1537,7 +1539,7 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 	WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect_missing_keys.root.rb_root));
 
 	ret = resolve_indirect_refs(fs_info, path, time_seq, &preftrees,
-				    extent_item_pos, sc);
+				    extent_item_pos, sc, ignore_offset);
 	if (ret)
 		goto out;
 
@@ -1571,7 +1573,8 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 				goto out;
 		}
 		if (ref->count && ref->parent) {
-			if (!ignore_offset && !ref->inode_list && ref->level == 0) {
+			if (extent_item_pos && !ref->inode_list &&
+			    ref->level == 0) {
 				struct extent_buffer *eb;
 
 				eb = read_tree_block(fs_info, ref->parent, 0,
@@ -1589,7 +1592,7 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 				if (!path->skip_locking)
 					btrfs_tree_read_lock(eb);
 				ret = find_extent_in_eb(eb, bytenr,
-							extent_item_pos, &eie);
+							*extent_item_pos, &eie, ignore_offset);
 				if (!path->skip_locking)
 					btrfs_tree_read_unlock(eb);
 				free_extent_buffer(eb);
@@ -1608,7 +1611,7 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 						  (void **)&eie, GFP_NOFS);
 			if (ret < 0)
 				goto out;
-			if (!ret && !ignore_offset) {
+			if (!ret && extent_item_pos) {
 				/*
 				 * We've recorded that parent, so we must extend
 				 * its inode list here.
@@ -1662,7 +1665,7 @@ static int find_parent_nodes(struct btrfs_trans_handle *trans,
 int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
 			 struct btrfs_fs_info *fs_info, u64 bytenr,
 			 u64 time_seq, struct ulist **leafs,
-			 u64 extent_item_pos)
+			 const u64 *extent_item_pos, bool ignore_offset)
 {
 	int ret;
 
@@ -1671,7 +1674,7 @@ int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
 		return -ENOMEM;
 
 	ret = find_parent_nodes(trans, fs_info, bytenr, time_seq,
-				*leafs, NULL, extent_item_pos, NULL);
+				*leafs, NULL, extent_item_pos, NULL, ignore_offset);
 	if (ret < 0 && ret != -ENOENT) {
 		free_leaf_list(*leafs);
 		return ret;
@@ -1695,7 +1698,8 @@ int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
  */
 static int btrfs_find_all_roots_safe(struct btrfs_trans_handle *trans,
 				     struct btrfs_fs_info *fs_info, u64 bytenr,
-				     u64 time_seq, struct ulist **roots)
+				     u64 time_seq, struct ulist **roots,
+				     bool ignore_offset)
 {
 	struct ulist *tmp;
 	struct ulist_node *node = NULL;
@@ -1714,8 +1718,7 @@ static int btrfs_find_all_roots_safe(struct btrfs_trans_handle *trans,
 	ULIST_ITER_INIT(&uiter);
 	while (1) {
 		ret = find_parent_nodes(trans, fs_info, bytenr, time_seq,
-					tmp, *roots, BTRFS_IGNORE_EXTENT_OFFSET,
-					NULL);
+					tmp, *roots, NULL, NULL, ignore_offset);
 		if (ret < 0 && ret != -ENOENT) {
 			ulist_free(tmp);
 			ulist_free(*roots);
@@ -1742,7 +1745,8 @@ int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
 
 	if (!trans && !skip_commit_root_sem)
 		down_read(&fs_info->commit_root_sem);
-	ret = btrfs_find_all_roots_safe(trans, fs_info, bytenr, time_seq, roots);
+	ret = btrfs_find_all_roots_safe(trans, fs_info, bytenr,
+					time_seq, roots, false);
 	if (!trans && !skip_commit_root_sem)
 		up_read(&fs_info->commit_root_sem);
 	return ret;
@@ -1841,7 +1845,7 @@ int btrfs_is_data_extent_shared(struct btrfs_inode *inode, u64 bytenr,
 		bool cached;
 
 		ret = find_parent_nodes(trans, fs_info, bytenr, elem.seq, &ctx->refs,
-					NULL, BTRFS_IGNORE_EXTENT_OFFSET, &shared);
+					NULL, NULL, &shared, false);
 		if (ret == BACKREF_FOUND_SHARED ||
 		    ret == BACKREF_FOUND_NOT_SHARED) {
 			/* If shared must return 1, otherwise return 0. */
@@ -2282,7 +2286,8 @@ static int iterate_leaf_refs(struct btrfs_fs_info *fs_info,
 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
 				u64 extent_item_objectid, u64 extent_item_pos,
 				int search_commit_root,
-				iterate_extent_inodes_t *iterate, void *ctx)
+				iterate_extent_inodes_t *iterate, void *ctx,
+				bool ignore_offset)
 {
 	int ret;
 	struct btrfs_trans_handle *trans = NULL;
@@ -2313,14 +2318,16 @@ int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
 		down_read(&fs_info->commit_root_sem);
 
 	ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
-				   seq_elem.seq, &refs, extent_item_pos);
+				   seq_elem.seq, &refs,
+				   &extent_item_pos, ignore_offset);
 	if (ret)
 		goto out;
 
 	ULIST_ITER_INIT(&ref_uiter);
 	while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
 		ret = btrfs_find_all_roots_safe(trans, fs_info, ref_node->val,
-						seq_elem.seq, &roots);
+						seq_elem.seq, &roots,
+						ignore_offset);
 		if (ret)
 			break;
 		ULIST_ITER_INIT(&root_uiter);
@@ -2388,14 +2395,10 @@ int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
 	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
 		return -EINVAL;
 
-	if (ignore_offset)
-		extent_item_pos = BTRFS_IGNORE_EXTENT_OFFSET;
-	else
-		extent_item_pos = logical - found_key.objectid;
-
+	extent_item_pos = logical - found_key.objectid;
 	ret = iterate_extent_inodes(fs_info, found_key.objectid,
 					extent_item_pos, search_commit_root,
-					build_ino_list, ctx);
+					build_ino_list, ctx, ignore_offset);
 
 	return ret;
 }
diff --git a/fs/btrfs/backref.h b/fs/btrfs/backref.h
index 2eb99f23cc8f..8d3598155f3b 100644
--- a/fs/btrfs/backref.h
+++ b/fs/btrfs/backref.h
@@ -12,13 +12,6 @@
 #include "disk-io.h"
 #include "extent_io.h"
 
-/*
- * Pass to backref walking functions to tell them to include references from
- * all file extent items that point to the target data extent, regardless if
- * they refer to the whole extent or just sections of it (bookend extents).
- */
-#define BTRFS_IGNORE_EXTENT_OFFSET   ((u64)-1)
-
 struct inode_fs_paths {
 	struct btrfs_path		*btrfs_path;
 	struct btrfs_root		*fs_root;
@@ -99,7 +92,8 @@ int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
 				u64 extent_item_objectid,
 				u64 extent_offset, int search_commit_root,
-				iterate_extent_inodes_t *iterate, void *ctx);
+				iterate_extent_inodes_t *iterate, void *ctx,
+				bool ignore_offset);
 
 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
 				struct btrfs_path *path, void *ctx,
@@ -110,7 +104,7 @@ int paths_from_inode(u64 inum, struct inode_fs_paths *ipath);
 int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
 			 struct btrfs_fs_info *fs_info, u64 bytenr,
 			 u64 time_seq, struct ulist **leafs,
-			 u64 extent_item_pos);
+			 const u64 *extent_item_pos, bool ignore_offset);
 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
 			 struct btrfs_fs_info *fs_info, u64 bytenr,
 			 u64 time_seq, struct ulist **roots,
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 45690f7b5900..d119986d1599 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -3417,7 +3417,7 @@ int add_data_references(struct reloc_control *rc,
 
 	btrfs_release_path(path);
 	ret = btrfs_find_all_leafs(NULL, fs_info, extent_key->objectid,
-				   0, &leaves, BTRFS_IGNORE_EXTENT_OFFSET);
+				   0, &leaves, NULL, true);
 	if (ret < 0)
 		return ret;
 
diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c
index 6c7dc89709fc..3c22573bfe0d 100644
--- a/fs/btrfs/scrub.c
+++ b/fs/btrfs/scrub.c
@@ -969,7 +969,7 @@ static void scrub_print_warning(const char *errstr, struct scrub_block *sblock)
 		swarn.dev = dev;
 		iterate_extent_inodes(fs_info, found_key.objectid,
 					extent_item_pos, 1,
-					scrub_print_warning_inode, &swarn);
+					scrub_print_warning_inode, &swarn, false);
 	}
 
 out:
diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c
index 6bf06939f891..49759cd9eecb 100644
--- a/fs/btrfs/send.c
+++ b/fs/btrfs/send.c
@@ -1467,7 +1467,7 @@ static int find_extent_clone(struct send_ctx *sctx,
 		extent_item_pos = 0;
 	ret = iterate_extent_inodes(fs_info, found_key.objectid,
 				    extent_item_pos, 1, __iterate_backrefs,
-				    &backref_ctx);
+				    &backref_ctx, false);
 
 	if (ret < 0)
 		goto out;

run-bisect

#!/bin/bash
set -eEuo pipefail

git -C ~/work/extern/linux bisect reset || true
git -C ~/work/extern/linux bisect start v6.2 v6.0
git -C ~/work/extern/linux bisect run "$PWD"/bisect-oracle.sh