Keep dist suffix

Add shrinker counters for wkic and quota_info.

ReleaseNotes.md

@@ -1,6 +1,36 @@
 Versity ScoutFS Release Notes
 =============================
 
+---
+v1.21
+\
+*Jul 1, 2024*
+
+This release adds features that rely on incompatible changes to
+structure the file system.  The process of advancing the format version
+to enable these features is described in scoutfs(5).
+
+Added the ".indx." extended attribute tag which can be used to determine
+the sorting of files in a global index.
+
+Added ScoutFS quotas which let rules define file size and count limits
+in terms of ".totl." extended attribute totals.
+
+Added the project ID file attribute which is inherited from parent
+directories on creation.  ScoutFS quota rules can reference project IDs.
+
+Add a retention attribute for files which prevents modification once
+enabled.
+
+---
+v1.20
+\
+*Apr 22, 2024*
+
+Minor changes to packaging to better support "weak" module linking of
+the kernel module, and to including git hashes in the built package.  No
+changes in runtime behaviour.
+
 ---
 v1.19
 \

kmod/Makefile

@@ -12,17 +12,22 @@ else
 SP = @:
 endif
 
-SCOUTFS_GIT_DESCRIBE := \
+SCOUTFS_GIT_DESCRIBE ?= \
 	$(shell git describe --all --abbrev=6 --long 2>/dev/null || \
 		echo no-git)
 
+ESCAPED_GIT_DESCRIBE := \
+	$(shell echo $(SCOUTFS_GIT_DESCRIBE) |sed -e 's/\//\\\//g')
+
+RPM_GITHASH ?= $(shell git rev-parse --short HEAD)
+
 SCOUTFS_ARGS := SCOUTFS_GIT_DESCRIBE=$(SCOUTFS_GIT_DESCRIBE) \
+		RPM_GITHASH=$(RPM_GITHASH) \
 		CONFIG_SCOUTFS_FS=m -C $(SK_KSRC) M=$(CURDIR)/src \
 		EXTRA_CFLAGS="-Werror"
 
 # - We use the git describe from tags to set up the RPM versioning
 RPM_VERSION := $(shell git describe --long --tags | awk -F '-' '{gsub(/^v/,""); print $$1}')
-RPM_GITHASH := $(shell git rev-parse --short HEAD)
 TARFILE = scoutfs-kmod-$(RPM_VERSION).tar
 
 
@@ -41,7 +46,8 @@ modules_install:
 
 %.spec: %.spec.in .FORCE
 	sed -e 's/@@VERSION@@/$(RPM_VERSION)/g' \
-	    -e 's/@@GITHASH@@/$(RPM_GITHASH)/g' < $< > $@+
+	    -e 's/@@GITHASH@@/$(RPM_GITHASH)/g' \
+	    -e 's/@@GITDESCRIBE@@/$(ESCAPED_GIT_DESCRIBE)/g' < $< > $@+
 	mv $@+ $@
 
 

kmod/scoutfs-kmod.spec.in

@@ -1,6 +1,7 @@
 %define kmod_name scoutfs
 %define kmod_version @@VERSION@@
 %define kmod_git_hash @@GITHASH@@
+%define kmod_git_describe @@GITDESCRIBE@@
 %define pkg_date %(date +%%Y%%m%%d)
 
 # Disable the building of the debug package(s).
@@ -9,6 +10,7 @@
 # take kernel version or default to uname -r
 %{!?kversion: %global kversion %(uname -r)}
 %global kernel_version %{kversion}
+%global sanitized_kversion %(echo %{kernel_version} | sed -e 's/-/_/g' -e 's/\.el.*//')
 
 %if 0%{?el7}
 %global kernel_source() /usr/src/kernels/%{kernel_version}.$(arch)
@@ -27,7 +29,7 @@ Name:           kmod-%{kmod_name}
 %endif
 Summary:        %{kmod_name} kernel module
 Version:        %{kmod_version}
-Release:        %{_release}%{?dist}
+Release:        %{_release}+%{sanitized_kversion}%{?dist}
 License:        GPLv2
 Group:          System/Kernel
 URL:            http://scoutfs.org/
@@ -75,7 +77,7 @@ echo "Building for kernel: %{kernel_version} flavors: '%{flavors_to_build}'"
 for flavor in %flavors_to_build; do
     rm -rf obj/$flavor
     cp -r source obj/$flavor
-    make SK_KSRC=%{kernel_source $flavor} -C obj/$flavor module
+    make RPM_GITHASH=%{kmod_git_hash} SCOUTFS_GIT_DESCRIBE=%{kmod_git_describe} SK_KSRC=%{kernel_source $flavor} -C obj/$flavor module
 done
 
 %install
@@ -97,10 +99,21 @@ find %{buildroot} -type f -name \*.ko -exec %{__chmod} u+x \{\} \;
 /lib/modules
 
 %post
-weak-modules --add-kernel --no-initramfs
+echo /lib/modules/%{kversion}/%{install_mod_dir}/scoutfs.ko | weak-modules --add-modules --no-initramfs
 depmod -a
 %endif
 
 %clean
 rm -rf %{buildroot}
 
+%preun
+# stash our modules for postun cleanup
+SCOUTFS_RPM_NAME=$(rpm -q %{name} | grep "%{version}-%{release}")
+rpm -ql $SCOUTFS_RPM_NAME | grep '\.ko$' > /var/run/%{name}-modules-%{version}-%{release} || true
+
+%postun
+if [ -x /sbin/weak-modules ]; then
+    cat /var/run/%{name}-modules-%{version}-%{release} | /sbin/weak-modules --remove-modules --no-initramfs
+fi
+
+rm /var/run/%{name}-modules-%{version}-%{release} || true

kmod/src/Makefile

@@ -9,6 +9,7 @@ CFLAGS_scoutfs_trace.o = -I$(src) # define_trace.h double include
  
 scoutfs-y +=			\
 	acl.o			\
+	attr_x.o		\
 	avl.o			\
 	alloc.o			\
 	block.o			\
@@ -34,6 +35,7 @@ scoutfs-y +=			\
 	options.o		\
 	per_task.o		\
 	quorum.o		\
+	quota.o			\
 	recov.o			\
 	scoutfs_trace.o		\
 	server.o		\
@@ -42,10 +44,12 @@ scoutfs-y +=			\
 	srch.o			\
 	super.o			\
 	sysfs.o			\
+	totl.o			\
 	trans.o			\
 	triggers.o		\
 	tseq.o			\
 	volopt.o		\
+	wkic.o			\
 	xattr.o
 
 #

kmod/src/acl.c

@@ -98,11 +98,9 @@ struct posix_acl *scoutfs_get_acl_locked(struct inode *inode, int type, struct s
 		acl = ERR_PTR(ret);
 	}
 
-#ifndef KC___POSIX_ACL_CREATE
 	/* can set null negative cache */
 	if (!IS_ERR(acl))
 		set_cached_acl(inode, type, acl);
-#endif
 
 	kfree(value);
 
@@ -194,10 +192,8 @@ int scoutfs_set_acl_locked(struct inode *inode, struct posix_acl *acl, int type,
 	}
 
 out:
-#ifndef KC___POSIX_ACL_CREATE
 	if (!ret)
 		set_cached_acl(inode, type, acl);
-#endif
 
 	kfree(value);
 

kmod/src/attr_x.c

@@ -0,0 +1,252 @@
+/*
+ * Copyright (C) 2024 Versity Software, Inc.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+#include <linux/kernel.h>
+#include <linux/fs.h>
+
+#include "format.h"
+#include "super.h"
+#include "inode.h"
+#include "ioctl.h"
+#include "lock.h"
+#include "trans.h"
+#include "attr_x.h"
+
+static int validate_attr_x_input(struct super_block *sb, struct scoutfs_ioctl_inode_attr_x *iax)
+{
+	int ret;
+
+	if ((iax->x_mask & SCOUTFS_IOC_IAX__UNKNOWN) ||
+	    (iax->x_flags & SCOUTFS_IOC_IAX_F__UNKNOWN))
+		return -EINVAL;
+
+	if ((iax->x_mask & SCOUTFS_IOC_IAX_RETENTION) &&
+	    (ret = scoutfs_fmt_vers_unsupported(sb, SCOUTFS_FORMAT_VERSION_FEAT_RETENTION)))
+		    return ret;
+
+	if ((iax->x_mask & SCOUTFS_IOC_IAX_PROJECT_ID) &&
+	    (ret = scoutfs_fmt_vers_unsupported(sb, SCOUTFS_FORMAT_VERSION_FEAT_PROJECT_ID)))
+		    return ret;
+
+	return 0;
+}
+
+/*
+ * If the mask indicates interest in the given attr then set the field
+ * to the caller's value and return the new size if it didn't already
+ * include the attr field.
+ */
+#define fill_attr(size, iax, bit, field, val)							\
+({												\
+	__typeof__(iax) _iax = (iax);								\
+	__typeof__(size) _size = (size);							\
+												\
+	if (_iax->x_mask & (bit)) {								\
+		_iax->field = (val);								\
+		_size = max(_size, offsetof(struct scoutfs_ioctl_inode_attr_x, field) +		\
+				   sizeof_field(struct scoutfs_ioctl_inode_attr_x, field));	\
+	}											\
+												\
+	_size;											\
+})
+
+/*
+ * Returns -errno on error, or >= number of bytes filled by the
+ * response.  0 can be returned if no attributes are requested in the
+ * input x_mask.
+ */
+int scoutfs_get_attr_x(struct inode *inode, struct scoutfs_ioctl_inode_attr_x *iax)
+{
+	struct super_block *sb = inode->i_sb;
+	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
+	struct scoutfs_lock *lock = NULL;
+	size_t size = 0;
+	u64 offline;
+	u64 online;
+	u64 bits;
+	int ret;
+
+	if (iax->x_mask == 0) {
+		ret = 0;
+		goto out;
+	}
+
+	ret = validate_attr_x_input(sb, iax);
+	if (ret < 0)
+		goto out;
+
+	inode_lock(inode);
+
+	ret = scoutfs_lock_inode(sb, SCOUTFS_LOCK_READ, SCOUTFS_LKF_REFRESH_INODE, inode, &lock);
+	if (ret)
+		goto unlock;
+
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_META_SEQ,
+			 meta_seq, scoutfs_inode_meta_seq(inode));
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_DATA_SEQ,
+			 data_seq, scoutfs_inode_data_seq(inode));
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_DATA_VERSION,
+			 data_version, scoutfs_inode_data_version(inode));
+	if (iax->x_mask & (SCOUTFS_IOC_IAX_ONLINE_BLOCKS | SCOUTFS_IOC_IAX_OFFLINE_BLOCKS)) {
+		scoutfs_inode_get_onoff(inode, &online, &offline);
+		size = fill_attr(size, iax, SCOUTFS_IOC_IAX_ONLINE_BLOCKS,
+				 online_blocks, online);
+		size = fill_attr(size, iax, SCOUTFS_IOC_IAX_OFFLINE_BLOCKS,
+				 offline_blocks, offline);
+	}
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_CTIME, ctime_sec, inode->i_ctime.tv_sec);
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_CTIME, ctime_nsec, inode->i_ctime.tv_nsec);
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_CRTIME, crtime_sec, si->crtime.tv_sec);
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_CRTIME, crtime_nsec, si->crtime.tv_nsec);
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_SIZE, size, i_size_read(inode));
+	if (iax->x_mask & SCOUTFS_IOC_IAX__BITS) {
+		bits = 0;
+		if ((iax->x_mask & SCOUTFS_IOC_IAX_RETENTION) &&
+		    (scoutfs_inode_get_flags(inode) & SCOUTFS_INO_FLAG_RETENTION))
+			bits |= SCOUTFS_IOC_IAX_B_RETENTION;
+		size = fill_attr(size, iax, SCOUTFS_IOC_IAX__BITS, bits, bits);
+	}
+	size = fill_attr(size, iax, SCOUTFS_IOC_IAX_PROJECT_ID,
+			 project_id, scoutfs_inode_get_proj(inode));
+
+	ret = size;
+unlock:
+	scoutfs_unlock(sb, lock, SCOUTFS_LOCK_READ);
+	inode_unlock(inode);
+out:
+	return ret;
+}
+
+static bool valid_attr_changes(struct inode *inode, struct scoutfs_ioctl_inode_attr_x *iax)
+{
+	/* provided data_version must be non-zero */
+	if ((iax->x_mask & SCOUTFS_IOC_IAX_DATA_VERSION) && (iax->data_version == 0))
+		return false;
+
+	/* can only set size or data version in new regular files */
+	if (((iax->x_mask & SCOUTFS_IOC_IAX_SIZE) ||
+	     (iax->x_mask & SCOUTFS_IOC_IAX_DATA_VERSION)) &&
+	    (!S_ISREG(inode->i_mode) || scoutfs_inode_data_version(inode) != 0))
+		return false;
+
+	/* must provide non-zero data_version with non-zero size */
+	if (((iax->x_mask & SCOUTFS_IOC_IAX_SIZE) && (iax->size > 0)) &&
+	    (!(iax->x_mask & SCOUTFS_IOC_IAX_DATA_VERSION) || (iax->data_version == 0)))
+		return false;
+
+	/* must provide non-zero size when setting offline extents to that size */
+	if ((iax->x_flags & SCOUTFS_IOC_IAX_F_SIZE_OFFLINE) &&
+	    (!(iax->x_mask & SCOUTFS_IOC_IAX_SIZE) || (iax->size == 0)))
+		return false;
+
+	/* the retention bit only applies to regular files */
+	if ((iax->x_mask & SCOUTFS_IOC_IAX_RETENTION) && !S_ISREG(inode->i_mode))
+		return false;
+
+	return true;
+}
+
+int scoutfs_set_attr_x(struct inode *inode, struct scoutfs_ioctl_inode_attr_x *iax)
+{
+	struct super_block *sb = inode->i_sb;
+	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
+	struct scoutfs_lock *lock = NULL;
+	LIST_HEAD(ind_locks);
+	bool set_data_seq;
+	int ret;
+
+	/* initially all setting is root only, could loosen with finer grained checks */
+	if (!capable(CAP_SYS_ADMIN)) {
+		ret = -EPERM;
+		goto out;
+	}
+
+	if (iax->x_mask == 0) {
+		ret = 0;
+		goto out;
+	}
+
+	ret = validate_attr_x_input(sb, iax);
+	if (ret < 0)
+		goto out;
+
+	inode_lock(inode);
+
+	ret = scoutfs_lock_inode(sb, SCOUTFS_LOCK_WRITE, SCOUTFS_LKF_REFRESH_INODE, inode, &lock);
+	if (ret)
+		goto unlock;
+
+	/* check for errors before making any changes */
+	if (!valid_attr_changes(inode, iax)) {
+		ret = -EINVAL;
+		goto unlock;
+	}
+
+	/* retention prevents modification unless also clearing retention */
+	ret = scoutfs_inode_check_retention(inode);
+	if (ret < 0 && !((iax->x_mask & SCOUTFS_IOC_IAX_RETENTION) &&
+			 !(iax->bits & SCOUTFS_IOC_IAX_B_RETENTION)))
+		goto unlock;
+
+	/* setting only so we don't see 0 data seq with nonzero data_version */
+	if ((iax->x_mask & SCOUTFS_IOC_IAX_DATA_VERSION) && (iax->data_version > 0))
+		set_data_seq = true;
+	else
+		set_data_seq = false;
+
+	ret = scoutfs_inode_index_lock_hold(inode, &ind_locks, set_data_seq, true);
+	if (ret)
+		goto unlock;
+
+	ret = scoutfs_dirty_inode_item(inode, lock);
+	if (ret < 0)
+		goto release;
+
+	/* creating offline extent first, it might fail */
+	if (iax->x_flags & SCOUTFS_IOC_IAX_F_SIZE_OFFLINE) {
+		ret = scoutfs_data_init_offline_extent(inode, iax->size, lock);
+		if (ret)
+			goto release;
+	}
+
+	/* make all changes once they're all checked and will succeed */
+	if (iax->x_mask & SCOUTFS_IOC_IAX_DATA_VERSION)
+		scoutfs_inode_set_data_version(inode, iax->data_version);
+	if (iax->x_mask & SCOUTFS_IOC_IAX_SIZE)
+		i_size_write(inode, iax->size);
+	if (iax->x_mask & SCOUTFS_IOC_IAX_CTIME) {
+		inode->i_ctime.tv_sec = iax->ctime_sec;
+		inode->i_ctime.tv_nsec = iax->ctime_nsec;
+	}
+	if (iax->x_mask & SCOUTFS_IOC_IAX_CRTIME) {
+		si->crtime.tv_sec = iax->crtime_sec;
+		si->crtime.tv_nsec = iax->crtime_nsec;
+	}
+	if (iax->x_mask & SCOUTFS_IOC_IAX_RETENTION) {
+		scoutfs_inode_set_flags(inode, ~SCOUTFS_INO_FLAG_RETENTION,
+					(iax->bits & SCOUTFS_IOC_IAX_B_RETENTION) ?
+					SCOUTFS_INO_FLAG_RETENTION : 0);
+	}
+	if (iax->x_mask & SCOUTFS_IOC_IAX_PROJECT_ID)
+		scoutfs_inode_set_proj(inode, iax->project_id);
+
+	scoutfs_update_inode_item(inode, lock, &ind_locks);
+	ret = 0;
+release:
+	scoutfs_release_trans(sb);
+unlock:
+	scoutfs_inode_index_unlock(sb, &ind_locks);
+	scoutfs_unlock(sb, lock, SCOUTFS_LOCK_WRITE);
+	inode_unlock(inode);
+out:
+	return ret;
+}

kmod/src/attr_x.h

@@ -0,0 +1,11 @@
+#ifndef _SCOUTFS_ATTR_X_H_
+#define _SCOUTFS_ATTR_X_H_
+
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include "ioctl.h"
+
+int scoutfs_get_attr_x(struct inode *inode, struct scoutfs_ioctl_inode_attr_x *iax);
+int scoutfs_set_attr_x(struct inode *inode, struct scoutfs_ioctl_inode_attr_x *iax);
+
+#endif

kmod/src/block.c

@@ -683,6 +683,7 @@ int scoutfs_block_read_ref(struct super_block *sb, struct scoutfs_block_ref *ref
 	struct scoutfs_block_header *hdr;
 	struct block_private *bp = NULL;
 	bool retried = false;
+	__le32 crc = 0;
 	int ret;
 
 retry:
@@ -695,7 +696,9 @@ retry:
 
 	/* corrupted writes might be a sign of a stale reference */
 	if (!test_bit(BLOCK_BIT_CRC_VALID, &bp->bits)) {
-		if (hdr->crc != block_calc_crc(hdr, SCOUTFS_BLOCK_LG_SIZE)) {
+		crc = block_calc_crc(hdr, SCOUTFS_BLOCK_LG_SIZE);
+		if (hdr->crc != crc) {
+			trace_scoutfs_block_stale(sb, ref, hdr, magic, le32_to_cpu(crc));
 			ret = -ESTALE;
 			goto out;
 		}
@@ -705,6 +708,7 @@ retry:
 
 	if (hdr->magic != cpu_to_le32(magic) || hdr->fsid != cpu_to_le64(sbi->fsid) ||
 	    hdr->seq != ref->seq || hdr->blkno != ref->blkno) {
+		trace_scoutfs_block_stale(sb, ref, hdr, magic, 0);
 		ret = -ESTALE;
 		goto out;
 	}

kmod/src/counters.h

@@ -162,6 +162,8 @@
 	EXPAND_COUNTER(orphan_scan_error)			\
 	EXPAND_COUNTER(orphan_scan_item)			\
 	EXPAND_COUNTER(orphan_scan_omap_set)			\
+	EXPAND_COUNTER(quota_info_count_objects)		\
+	EXPAND_COUNTER(quota_info_scan_objects)			\
 	EXPAND_COUNTER(quorum_candidate_server_stopping)	\
 	EXPAND_COUNTER(quorum_elected)				\
 	EXPAND_COUNTER(quorum_fence_error)			\
@@ -199,20 +201,19 @@
 	EXPAND_COUNTER(srch_read_stale)				\
 	EXPAND_COUNTER(statfs)					\
 	EXPAND_COUNTER(totl_read_copied)			\
-	EXPAND_COUNTER(totl_read_finalized)			\
-	EXPAND_COUNTER(totl_read_fs)				\
 	EXPAND_COUNTER(totl_read_item)				\
-	EXPAND_COUNTER(totl_read_logged)			\
 	EXPAND_COUNTER(trans_commit_data_alloc_low)		\
 	EXPAND_COUNTER(trans_commit_dirty_meta_full)		\
 	EXPAND_COUNTER(trans_commit_fsync)			\
 	EXPAND_COUNTER(trans_commit_meta_alloc_low)		\
 	EXPAND_COUNTER(trans_commit_sync_fs)			\
 	EXPAND_COUNTER(trans_commit_timer)			\
-	EXPAND_COUNTER(trans_commit_written)
+	EXPAND_COUNTER(trans_commit_written)			\
+	EXPAND_COUNTER(wkic_count_objects)			\
+	EXPAND_COUNTER(wkic_scan_objects)
 
 #define FIRST_COUNTER	alloc_alloc_data
-#define LAST_COUNTER	trans_commit_written
+#define LAST_COUNTER	wkic_scan_objects
 
 #undef EXPAND_COUNTER
 #define EXPAND_COUNTER(which) struct percpu_counter which;

kmod/src/data.c

@@ -586,6 +586,12 @@ static int scoutfs_get_block(struct inode *inode, sector_t iblock,
 		goto out;
 	}
 
+	if (create && !si->staging) {
+		ret = scoutfs_inode_check_retention(inode);
+		if (ret < 0)
+			goto out;
+	}
+
 	/* convert unwritten to written, could be staging */
 	if (create && ext.map && (ext.flags & SEF_UNWRITTEN)) {
 		un.start = iblock;
@@ -1104,6 +1110,10 @@ long scoutfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 
 	while(iblock <= last) {
 
+		ret = scoutfs_quota_check_data(sb, inode);
+		if (ret)
+			goto out_extent;
+
 		ret = scoutfs_inode_index_lock_hold(inode, &ind_locks, false, true);
 		if (ret)
 			goto out_extent;
@@ -1155,9 +1165,9 @@ out:
  * on regular files with no data extents.  It's used to restore a file
  * with an offline extent which can then trigger staging.
  *
- * The caller has taken care of locking the inode.  We're updating the
- * inode offline count as we create the offline extent so we take care
- * of the index locking, updating, and transaction.
+ * The caller must take care of cluster locking, transactions, inode
+ * updates, and index updates (so that they can atomically make this
+ * change along with other metadata changes).
  */
 int scoutfs_data_init_offline_extent(struct inode *inode, u64 size,
 				     struct scoutfs_lock *lock)
@@ -1171,7 +1181,6 @@ int scoutfs_data_init_offline_extent(struct inode *inode, u64 size,
 		.lock = lock,
 	};
 	const u64 count = DIV_ROUND_UP(size, SCOUTFS_BLOCK_SM_SIZE);
-	LIST_HEAD(ind_locks);
 	u64 on;
 	u64 off;
 	int ret;
@@ -1184,28 +1193,10 @@ int scoutfs_data_init_offline_extent(struct inode *inode, u64 size,
 		goto out;
 	}
 
-	/* we're updating meta_seq with offline block count */
-	ret = scoutfs_inode_index_lock_hold(inode, &ind_locks, false, true);
-	if (ret < 0)
-		goto out;
-
-	ret = scoutfs_dirty_inode_item(inode, lock);
-	if (ret < 0)
-		goto unlock;
-
 	down_write(&si->extent_sem);
 	ret = scoutfs_ext_insert(sb, &data_ext_ops, &args,
 				 0, count, 0, SEF_OFFLINE);
 	up_write(&si->extent_sem);
-	if (ret < 0)
-		goto unlock;
-
-	scoutfs_update_inode_item(inode, lock, &ind_locks);
-
-unlock:
-	scoutfs_release_trans(sb);
-	scoutfs_inode_index_unlock(sb, &ind_locks);
-	ret = 0;
 out:
 	return ret;
 }
@@ -1273,6 +1264,9 @@ int scoutfs_data_move_blocks(struct inode *from, u64 from_off,
 	if (ret)
 		goto out;
 
+	if (!is_stage && (ret = scoutfs_inode_check_retention(to)))
+		goto out;
+
 	if ((from_off & SCOUTFS_BLOCK_SM_MASK) ||
 	    (to_off & SCOUTFS_BLOCK_SM_MASK) ||
 	    ((byte_len & SCOUTFS_BLOCK_SM_MASK) &&
@@ -1807,37 +1801,6 @@ int scoutfs_data_wait_check_iov(struct inode *inode, const struct iovec *iov,
 	return ret;
 }
 
-int scoutfs_data_wait_check_iter(struct inode *inode, loff_t pos, struct iov_iter *iter,
-				 u8 sef, u8 op, struct scoutfs_data_wait *dw,
-				 struct scoutfs_lock *lock)
-{
-	size_t count = iov_iter_count(iter);
-	size_t off = iter->iov_offset;
-	const struct iovec *iov;
-	size_t len;
-	int ret = 0;
-
-	for (iov = iter->iov; count > 0; iov++) {
-		len = iov->iov_len - off;
-		if (len == 0)
-			continue;
-
-		/* aren't we waiting on too much data here ? */
-		ret = scoutfs_data_wait_check(inode, pos, len,
-					      sef, op, dw, lock);
-
-		if (ret != 0)
-			break;
-
-
-		pos += len;
-		count -= len;
-		off = 0;
-	}
-
-	return ret;
-}
-
 int scoutfs_data_wait(struct inode *inode, struct scoutfs_data_wait *dw)
 {
 	DECLARE_DATA_WAIT_ROOT(inode->i_sb, rt);

kmod/src/data.h

@@ -65,9 +65,6 @@ int scoutfs_data_wait_check_iov(struct inode *inode, const struct iovec *iov,
 				unsigned long nr_segs, loff_t pos, u8 sef,
 				u8 op, struct scoutfs_data_wait *ow,
 				struct scoutfs_lock *lock);
-int scoutfs_data_wait_check_iter(struct inode *inode, loff_t pos, struct iov_iter *iter,
-				 u8 sef, u8 op, struct scoutfs_data_wait *ow,
-				 struct scoutfs_lock *lock);
 bool scoutfs_data_wait_found(struct scoutfs_data_wait *ow);
 int scoutfs_data_wait(struct inode *inode,
 			      struct scoutfs_data_wait *ow);

kmod/src/dir.c

@@ -34,6 +34,7 @@
 #include "forest.h"
 #include "acl.h"
 #include "counters.h"
+#include "quota.h"
 #include "scoutfs_trace.h"
 
 /*
@@ -651,6 +652,10 @@ static struct inode *lock_hold_create(struct inode *dir, struct dentry *dentry,
 	if (ret)
 		goto out_unlock;
 
+	ret = scoutfs_quota_check_inode(sb, dir);
+	if (ret)
+		goto out_unlock;
+
 	if (orph_lock) {
 		ret = scoutfs_lock_orphan(sb, SCOUTFS_LOCK_WRITE_ONLY, 0, ino, orph_lock);
 		if (ret < 0)
@@ -672,6 +677,8 @@ retry:
 	if (ret < 0)
 		goto out;
 
+	scoutfs_inode_set_proj(inode, scoutfs_inode_get_proj(dir));
+
 	ret = scoutfs_dirty_inode_item(dir, *dir_lock);
 out:
 	if (ret)
@@ -926,12 +933,16 @@ static int scoutfs_unlink(struct inode *dir, struct dentry *dentry)
 		goto unlock;
 	}
 
+	ret = scoutfs_inode_check_retention(inode);
+	if (ret < 0)
+		goto unlock;
+
 	hash = dirent_name_hash(dentry->d_name.name, dentry->d_name.len);
 
 	ret = lookup_dirent(sb, scoutfs_ino(dir), dentry->d_name.name, dentry->d_name.len, hash,
 			    &dent, dir_lock);
 	if (ret < 0)
-		goto out;
+		goto unlock;
 
 	if (should_orphan(inode)) {
 		ret = scoutfs_lock_orphan(sb, SCOUTFS_LOCK_WRITE_ONLY, 0, scoutfs_ino(inode),
@@ -1632,6 +1643,10 @@ static int scoutfs_rename_common(struct inode *old_dir,
 		goto out_unlock;
 	}
 
+	if ((old_inode && (ret = scoutfs_inode_check_retention(old_inode))) ||
+	    (new_inode && (ret = scoutfs_inode_check_retention(new_inode))))
+		goto out_unlock;
+
 	if (should_orphan(new_inode)) {
 		ret = scoutfs_lock_orphan(sb, SCOUTFS_LOCK_WRITE_ONLY, 0, scoutfs_ino(new_inode),
 					  &orph_lock);

kmod/src/file.c

@@ -28,6 +28,7 @@
 #include "inode.h"
 #include "per_task.h"
 #include "omap.h"
+#include "quota.h"
 
 #ifdef KC_LINUX_HAVE_FOP_AIO_READ
 /*
@@ -108,6 +109,10 @@ retry:
 	if (ret)
 		goto out;
 
+	ret = scoutfs_inode_check_retention(inode);
+	if (ret < 0)
+		goto out;
+
 	ret = scoutfs_complete_truncate(inode, scoutfs_inode_lock);
 	if (ret)
 		goto out;
@@ -122,6 +127,10 @@ retry:
 			goto out;
 	}
 
+	ret = scoutfs_quota_check_data(sb, inode);
+	if (ret)
+		goto out;
+
 	/* XXX: remove SUID bit */
 
 	ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos);
@@ -171,10 +180,8 @@ retry:
 		goto out;
 
 	if (scoutfs_per_task_add_excl(&si->pt_data_lock, &pt_ent, scoutfs_inode_lock)) {
-		ret = scoutfs_data_wait_check_iter(inode, iocb->ki_pos, to,
-						   SEF_OFFLINE,
-						   SCOUTFS_IOC_DWO_READ,
-						   &dw, scoutfs_inode_lock);
+		ret = scoutfs_data_wait_check(inode, iocb->ki_pos, iov_iter_count(to), SEF_OFFLINE,
+					      SCOUTFS_IOC_DWO_READ, &dw, scoutfs_inode_lock);
 		if (ret != 0)
 			goto out;
 	} else {
@@ -205,8 +212,7 @@ ssize_t scoutfs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 	struct scoutfs_lock *scoutfs_inode_lock = NULL;
 	SCOUTFS_DECLARE_PER_TASK_ENTRY(pt_ent);
 	DECLARE_DATA_WAIT(dw);
-	int ret;
-	int written;
+	ssize_t ret;
 
 retry:
 	inode_lock(inode);
@@ -219,23 +225,29 @@ retry:
 	if (ret <= 0)
 		goto out;
 
+	ret = scoutfs_inode_check_retention(inode);
+	if (ret < 0)
+		goto out;
+
 	ret = scoutfs_complete_truncate(inode, scoutfs_inode_lock);
 	if (ret)
 		goto out;
 
+	ret = scoutfs_quota_check_data(sb, inode);
+	if (ret)
+		goto out;
+
 	if (scoutfs_per_task_add_excl(&si->pt_data_lock, &pt_ent, scoutfs_inode_lock)) {
 		/* data_version is per inode, whole file must be online */
-		ret = scoutfs_data_wait_check_iter(inode, iocb->ki_pos, from,
-						   SEF_OFFLINE,
-						   SCOUTFS_IOC_DWO_WRITE,
-						   &dw, scoutfs_inode_lock);
+		ret = scoutfs_data_wait_check(inode, 0, i_size_read(inode), SEF_OFFLINE,
+					      SCOUTFS_IOC_DWO_WRITE, &dw, scoutfs_inode_lock);
 		if (ret != 0)
 			goto out;
 	}
 
 	/* XXX: remove SUID bit */
 
-	written = __generic_file_write_iter(iocb, from);
+	ret = __generic_file_write_iter(iocb, from);
 
 out:
 	scoutfs_per_task_del(&si->pt_data_lock, &pt_ent);
@@ -248,10 +260,10 @@ out:
 			goto retry;
 	}
 
-	if (ret > 0 || ret == -EIOCBQUEUED)
-		ret = generic_write_sync(iocb, written);
+	if (ret > 0)
+		ret = generic_write_sync(iocb, ret);
 
-	return written ? written : ret;
+	return ret;
 }
 #endif
 

kmod/src/forest.c

@@ -238,19 +238,16 @@ static int forest_read_items(struct super_block *sb, struct scoutfs_key *key, u6
  * We return -ESTALE if we hit stale blocks to give the caller a chance
  * to reset their state and retry with a newer version of the btrees.
  */
-int scoutfs_forest_read_items(struct super_block *sb,
-			      struct scoutfs_key *key,
-			      struct scoutfs_key *bloom_key,
-			      struct scoutfs_key *start,
-			      struct scoutfs_key *end,
-			      scoutfs_forest_item_cb cb, void *arg)
+int scoutfs_forest_read_items_roots(struct super_block *sb, struct scoutfs_net_roots *roots,
+				    struct scoutfs_key *key, struct scoutfs_key *bloom_key,
+				    struct scoutfs_key *start, struct scoutfs_key *end,
+				    scoutfs_forest_item_cb cb, void *arg)
 {
 	struct forest_read_items_data rid = {
 		.cb = cb,
 		.cb_arg = arg,
 	};
 	struct scoutfs_log_trees lt;
-	struct scoutfs_net_roots roots;
 	struct scoutfs_bloom_block *bb;
 	struct forest_bloom_nrs bloom;
 	SCOUTFS_BTREE_ITEM_REF(iref);
@@ -264,18 +261,14 @@ int scoutfs_forest_read_items(struct super_block *sb,
 	scoutfs_inc_counter(sb, forest_read_items);
 	calc_bloom_nrs(&bloom, bloom_key);
 
-	ret = scoutfs_client_get_roots(sb, &roots);
-	if (ret)
-		goto out;
-
-	trace_scoutfs_forest_using_roots(sb, &roots.fs_root, &roots.logs_root);
+	trace_scoutfs_forest_using_roots(sb, &roots->fs_root, &roots->logs_root);
 
 	*start = orig_start;
 	*end = orig_end;
 
 	/* start with fs root items */
 	rid.fic |= FIC_FS_ROOT;
-	ret = scoutfs_btree_read_items(sb, &roots.fs_root, key, start, end,
+	ret = scoutfs_btree_read_items(sb, &roots->fs_root, key, start, end,
 				       forest_read_items, &rid);
 	if (ret < 0)
 		goto out;
@@ -283,7 +276,7 @@ int scoutfs_forest_read_items(struct super_block *sb,
 
 	scoutfs_key_init_log_trees(&ltk, 0, 0);
 	for (;; scoutfs_key_inc(&ltk)) {
-		ret = scoutfs_btree_next(sb, &roots.logs_root, &ltk, &iref);
+		ret = scoutfs_btree_next(sb, &roots->logs_root, &ltk, &iref);
 		if (ret == 0) {
 			if (iref.val_len == sizeof(lt)) {
 				ltk = *iref.key;
@@ -340,6 +333,23 @@ out:
 	return ret;
 }
 
+int scoutfs_forest_read_items(struct super_block *sb,
+			      struct scoutfs_key *key,
+			      struct scoutfs_key *bloom_key,
+			      struct scoutfs_key *start,
+			      struct scoutfs_key *end,
+			      scoutfs_forest_item_cb cb, void *arg)
+{
+	struct scoutfs_net_roots roots;
+	int ret;
+
+	ret = scoutfs_client_get_roots(sb, &roots);
+	if (ret == 0)
+		ret = scoutfs_forest_read_items_roots(sb, &roots, key, bloom_key, start, end,
+						      cb, arg);
+	return ret;
+}
+
 /*
  * If the items are deltas then combine the src with the destination
  * value and store the result in the destination.

kmod/src/forest.h

@@ -4,6 +4,7 @@
 struct scoutfs_alloc;
 struct scoutfs_block_writer;
 struct scoutfs_block;
+struct scoutfs_lock;
 
 #include "btree.h"
 
@@ -23,6 +24,10 @@ int scoutfs_forest_read_items(struct super_block *sb,
 			      struct scoutfs_key *start,
 			      struct scoutfs_key *end,
 			      scoutfs_forest_item_cb cb, void *arg);
+int scoutfs_forest_read_items_roots(struct super_block *sb, struct scoutfs_net_roots *roots,
+				    struct scoutfs_key *key, struct scoutfs_key *bloom_key,
+				    struct scoutfs_key *start, struct scoutfs_key *end,
+				    scoutfs_forest_item_cb cb, void *arg);
 int scoutfs_forest_set_bloom_bits(struct super_block *sb,
 				  struct scoutfs_lock *lock);
 void scoutfs_forest_set_max_seq(struct super_block *sb, u64 max_seq);

kmod/src/format.h

@@ -8,9 +8,14 @@
  */
 #define SCOUTFS_FORMAT_VERSION_MIN		1
 #define SCOUTFS_FORMAT_VERSION_MIN_STR	__stringify(SCOUTFS_FORMAT_VERSION_MIN)
-#define SCOUTFS_FORMAT_VERSION_MAX		1
+#define SCOUTFS_FORMAT_VERSION_MAX		2
 #define SCOUTFS_FORMAT_VERSION_MAX_STR	__stringify(SCOUTFS_FORMAT_VERSION_MAX)
 
+#define SCOUTFS_FORMAT_VERSION_FEAT_RETENTION	2
+#define SCOUTFS_FORMAT_VERSION_FEAT_PROJECT_ID	2
+#define SCOUTFS_FORMAT_VERSION_FEAT_QUOTA	2
+#define SCOUTFS_FORMAT_VERSION_FEAT_INDX_TAG	2
+
 /* statfs(2) f_type */
 #define SCOUTFS_SUPER_MAGIC	0x554f4353		/* "SCOU" */
 
@@ -175,6 +180,10 @@ struct scoutfs_key {
 #define sko_rid		_sk_first
 #define sko_ino		_sk_second
 
+/* quota rules */
+#define skqr_hash	_sk_second
+#define skqr_coll_nr	_sk_third
+
 /* xattr totl */
 #define skxt_a		_sk_first
 #define skxt_b		_sk_second
@@ -585,7 +594,9 @@ struct scoutfs_log_merge_freeing {
  */
 #define SCOUTFS_INODE_INDEX_ZONE		4
 #define SCOUTFS_ORPHAN_ZONE			8
+#define SCOUTFS_QUOTA_ZONE			10
 #define SCOUTFS_XATTR_TOTL_ZONE			12
+#define SCOUTFS_XATTR_INDX_ZONE			14
 #define SCOUTFS_FS_ZONE				16
 #define SCOUTFS_LOCK_ZONE			20
 /* Items only stored in server btrees */
@@ -608,6 +619,9 @@ struct scoutfs_log_merge_freeing {
 /* orphan zone, redundant type used for clarity */
 #define SCOUTFS_ORPHAN_TYPE			4
 
+/* quota zone */
+#define SCOUTFS_QUOTA_RULE_TYPE			4
+
 /* fs zone */
 #define SCOUTFS_INODE_TYPE			4
 #define SCOUTFS_XATTR_TYPE			8
@@ -661,6 +675,34 @@ struct scoutfs_xattr_totl_val {
 	__le64 count;
 };
 
+#define SQ_RF_TOTL_COUNT	(1 << 0)
+#define SQ_RF__UNKNOWN	(~((1 << 1) - 1))
+
+#define SQ_NS_LITERAL		0
+#define SQ_NS_PROJ		1
+#define SQ_NS_UID		2
+#define SQ_NS_GID		3
+#define SQ_NS__NR		4
+#define SQ_NS__NR_SELECT	(SQ_NS__NR - 1) /* !literal */
+
+#define SQ_NF_SELECT	(1 << 0)
+#define SQ_NF__UNKNOWN	(~((1 << 1) - 1))
+
+#define SQ_OP_INODE	0
+#define SQ_OP_DATA	1
+#define SQ_OP__NR	2
+
+struct scoutfs_quota_rule_val {
+	__le64 name_val[3];
+	__le64 limit;
+	__u8 prio;
+	__u8 op;
+	__u8 rule_flags;
+	__u8 name_source[3];
+	__u8 name_flags[3];
+	__u8 _pad[7];
+};
+
 /* XXX does this exist upstream somewhere? */
 #define member_sizeof(TYPE, MEMBER) (sizeof(((TYPE *)0)->MEMBER))
 
@@ -859,9 +901,38 @@ struct scoutfs_inode {
 	struct scoutfs_timespec ctime;
 	struct scoutfs_timespec mtime;
 	struct scoutfs_timespec crtime;
+	__le64 proj;
 };
 
-#define SCOUTFS_INO_FLAG_TRUNCATE 0x1
+#define SCOUTFS_INODE_FMT_V1_BYTES offsetof(struct scoutfs_inode, proj)
+
+/*
+ * There are so few versions that we don't mind doing this work inline
+ * so that both utils and kernel can share these.  Mounting has already
+ * checked that the format version is within the supported min and max,
+ * so these functions only deal with size variance within that band.
+ */
+/* Returns the native written inode size for the given format version, 0 for bad version */
+static inline int scoutfs_inode_vers_bytes(__u64 fmt_vers)
+{
+	if (fmt_vers == 1)
+		return SCOUTFS_INODE_FMT_V1_BYTES;
+	else
+		return sizeof(struct scoutfs_inode);
+}
+/*
+ * Returns true if bytes is a valid inode size to read from the given
+ * version.  The given version must be greater than the version that
+ * introduced the size.
+ */
+static inline int scoutfs_inode_valid_vers_bytes(__u64 fmt_vers, int bytes)
+{
+	return (bytes == sizeof(struct scoutfs_inode) && fmt_vers == SCOUTFS_FORMAT_VERSION_MAX) ||
+	       (bytes == SCOUTFS_INODE_FMT_V1_BYTES);
+}
+
+#define SCOUTFS_INO_FLAG_TRUNCATE	0x1
+#define SCOUTFS_INO_FLAG_RETENTION	0x2
 
 #define SCOUTFS_ROOT_INO 1
 

kmod/src/inode.c

@@ -91,7 +91,7 @@ static void scoutfs_inode_ctor(void *obj)
 
 	init_rwsem(&si->extent_sem);
 	mutex_init(&si->item_mutex);
-	seqcount_init(&si->seqcount);
+	seqlock_init(&si->seqlock);
 	si->staging = false;
 	scoutfs_per_task_init(&si->pt_data_lock);
 	atomic64_set(&si->data_waitq.changed, 0);
@@ -250,7 +250,7 @@ static void set_item_info(struct scoutfs_inode_info *si,
 	set_item_major(si, SCOUTFS_INODE_INDEX_DATA_SEQ_TYPE, sinode->data_seq);
 }
 
-static void load_inode(struct inode *inode, struct scoutfs_inode *cinode)
+static void load_inode(struct inode *inode, struct scoutfs_inode *cinode, int inode_bytes)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
 
@@ -278,6 +278,7 @@ static void load_inode(struct inode *inode, struct scoutfs_inode *cinode)
 	si->flags = le32_to_cpu(cinode->flags);
 	si->crtime.tv_sec = le64_to_cpu(cinode->crtime.sec);
 	si->crtime.tv_nsec = le32_to_cpu(cinode->crtime.nsec);
+	si->proj = le64_to_cpu(cinode->proj);
 
 	/*
 	 * i_blocks is initialized from online and offline and is then
@@ -298,6 +299,24 @@ void scoutfs_inode_init_key(struct scoutfs_key *key, u64 ino)
 	};
 }
 
+/*
+ * Read an inode item into the caller's buffer and return the size that
+ * we read.   Returns errors if the inode size is unsupported or doesn't
+ * make sense for the format version.
+ */
+static int lookup_inode_item(struct super_block *sb, struct scoutfs_key *key,
+			     struct scoutfs_inode *sinode, struct scoutfs_lock *lock)
+{
+	struct scoutfs_sb_info *sbi = SCOUTFS_SB(sb);
+	int ret;
+
+	ret = scoutfs_item_lookup_smaller_zero(sb, key, sinode, sizeof(struct scoutfs_inode), lock);
+	if (ret >= 0 && !scoutfs_inode_valid_vers_bytes(sbi->fmt_vers, ret))
+		return -EIO;
+
+	return ret;
+}
+
 /*
  * Refresh the vfs inode fields if the lock indicates that the current
  * contents could be stale.
@@ -333,12 +352,12 @@ int scoutfs_inode_refresh(struct inode *inode, struct scoutfs_lock *lock)
 
 	mutex_lock(&si->item_mutex);
 	if (atomic64_read(&si->last_refreshed) < refresh_gen) {
-		ret = scoutfs_item_lookup_exact(sb, &key, &sinode,
-						sizeof(sinode), lock);
-		if (ret == 0) {
-			load_inode(inode, &sinode);
+		ret = lookup_inode_item(sb, &key, &sinode, lock);
+		if (ret > 0) {
+			load_inode(inode, &sinode, ret);
 			atomic64_set(&si->last_refreshed, refresh_gen);
 			scoutfs_lock_add_coverage(sb, lock, &si->ino_lock_cov);
+			ret = 0;
 		}
 	} else {
 		ret = 0;
@@ -486,6 +505,10 @@ retry:
 	if (ret)
 		goto out;
 
+	ret = scoutfs_inode_check_retention(inode);
+	if (ret < 0)
+		goto out;
+
 	attr_size = (attr->ia_valid & ATTR_SIZE) ? attr->ia_size :
 		i_size_read(inode);
 
@@ -566,11 +589,9 @@ static void set_trans_seq(struct inode *inode, u64 *seq)
 	struct scoutfs_sb_info *sbi = SCOUTFS_SB(sb);
 
 	if (*seq != sbi->trans_seq) {
-		preempt_disable();
-		write_seqcount_begin(&si->seqcount);
+		write_seqlock(&si->seqlock);
 		*seq = sbi->trans_seq;
-		write_seqcount_end(&si->seqcount);
-		preempt_enable();
+		write_sequnlock(&si->seqlock);
 	}
 }
 
@@ -592,22 +613,18 @@ void scoutfs_inode_inc_data_version(struct inode *inode)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
 
-	preempt_disable();
-	write_seqcount_begin(&si->seqcount);
+	write_seqlock(&si->seqlock);
 	si->data_version++;
-	write_seqcount_end(&si->seqcount);
-	preempt_enable();
+	write_sequnlock(&si->seqlock);
 }
 
 void scoutfs_inode_set_data_version(struct inode *inode, u64 data_version)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
 
-	preempt_disable();
-	write_seqcount_begin(&si->seqcount);
+	write_seqlock(&si->seqlock);
 	si->data_version = data_version;
-	write_seqcount_end(&si->seqcount);
-	preempt_enable();
+	write_sequnlock(&si->seqlock);
 }
 
 void scoutfs_inode_add_onoff(struct inode *inode, s64 on, s64 off)
@@ -616,8 +633,7 @@ void scoutfs_inode_add_onoff(struct inode *inode, s64 on, s64 off)
 
 	if (inode && (on || off)) {
 		si = SCOUTFS_I(inode);
-		preempt_disable();
-		write_seqcount_begin(&si->seqcount);
+		write_seqlock(&si->seqlock);
 
 		/* inode and extents out of sync, bad callers */
 		if (((s64)si->online_blocks + on < 0) ||
@@ -638,8 +654,7 @@ void scoutfs_inode_add_onoff(struct inode *inode, s64 on, s64 off)
 						    si->online_blocks,
 						    si->offline_blocks);
 
-		write_seqcount_end(&si->seqcount);
-		preempt_enable();
+		write_sequnlock(&si->seqlock);
 	}
 
 	/* any time offline extents decreased we try and wake waiters */
@@ -647,16 +662,16 @@ void scoutfs_inode_add_onoff(struct inode *inode, s64 on, s64 off)
 		scoutfs_data_wait_changed(inode);
 }
 
-static u64 read_seqcount_u64(struct inode *inode, u64 *val)
+static u64 read_seqlock_u64(struct inode *inode, u64 *val)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
-	unsigned int seq;
+	unsigned seq;
 	u64 v;
 
 	do {
-		seq = read_seqcount_begin(&si->seqcount);
+		seq = read_seqbegin(&si->seqlock);
 		v = *val;
-	} while (read_seqcount_retry(&si->seqcount, seq));
+	} while (read_seqretry(&si->seqlock, seq));
 
 	return v;
 }
@@ -665,33 +680,82 @@ u64 scoutfs_inode_meta_seq(struct inode *inode)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
 
-	return read_seqcount_u64(inode, &si->meta_seq);
+	return read_seqlock_u64(inode, &si->meta_seq);
 }
 
 u64 scoutfs_inode_data_seq(struct inode *inode)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
 
-	return read_seqcount_u64(inode, &si->data_seq);
+	return read_seqlock_u64(inode, &si->data_seq);
 }
 
 u64 scoutfs_inode_data_version(struct inode *inode)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
 
-	return read_seqcount_u64(inode, &si->data_version);
+	return read_seqlock_u64(inode, &si->data_version);
 }
 
 void scoutfs_inode_get_onoff(struct inode *inode, s64 *on, s64 *off)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
-	unsigned int seq;
+	unsigned seq;
 
 	do {
-		seq = read_seqcount_begin(&si->seqcount);
+		seq = read_seqbegin(&si->seqlock);
 		*on = SCOUTFS_I(inode)->online_blocks;
 		*off = SCOUTFS_I(inode)->offline_blocks;
-	} while (read_seqcount_retry(&si->seqcount, seq));
+	} while (read_seqretry(&si->seqlock, seq));
+}
+
+/*
+ * Get our private scoutfs inode flags, not the vfs i_flags.
+ */
+u32 scoutfs_inode_get_flags(struct inode *inode)
+{
+	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
+	unsigned seq;
+	u32 flags;
+
+	do {
+		seq = read_seqbegin(&si->seqlock);
+		flags = si->flags;
+	} while (read_seqretry(&si->seqlock, seq));
+
+	return flags;
+}
+
+void scoutfs_inode_set_flags(struct inode *inode, u32 and, u32 or)
+{
+	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
+
+	write_seqlock(&si->seqlock);
+	si->flags = (si->flags & and) | or;
+	write_sequnlock(&si->seqlock);
+}
+
+u64 scoutfs_inode_get_proj(struct inode *inode)
+{
+	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
+	unsigned seq;
+	u64 proj;
+
+	do {
+		seq = read_seqbegin(&si->seqlock);
+		proj = si->proj;
+	} while (read_seqretry(&si->seqlock, seq));
+
+	return proj;
+}
+
+void scoutfs_inode_set_proj(struct inode *inode, u64 proj)
+{
+	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
+
+	write_seqlock(&si->seqlock);
+	si->proj = proj;
+	write_sequnlock(&si->seqlock);
 }
 
 static int scoutfs_iget_test(struct inode *inode, void *arg)
@@ -803,7 +867,7 @@ out:
 	return inode;
 }
 
-static void store_inode(struct scoutfs_inode *cinode, struct inode *inode)
+static void store_inode(struct scoutfs_inode *cinode, struct inode *inode, int inode_bytes)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
 	u64 online_blocks;
@@ -839,6 +903,7 @@ static void store_inode(struct scoutfs_inode *cinode, struct inode *inode)
 	cinode->crtime.sec = cpu_to_le64(si->crtime.tv_sec);
 	cinode->crtime.nsec = cpu_to_le32(si->crtime.tv_nsec);
 	memset(cinode->crtime.__pad, 0, sizeof(cinode->crtime.__pad));
+	cinode->proj = cpu_to_le64(si->proj);
 }
 
 /*
@@ -862,15 +927,18 @@ static void store_inode(struct scoutfs_inode *cinode, struct inode *inode)
 int scoutfs_dirty_inode_item(struct inode *inode, struct scoutfs_lock *lock)
 {
 	struct super_block *sb = inode->i_sb;
+	struct scoutfs_sb_info *sbi = SCOUTFS_SB(sb);
 	struct scoutfs_inode sinode;
 	struct scoutfs_key key;
+	int inode_bytes;
 	int ret;
 
-	store_inode(&sinode, inode);
+	inode_bytes = scoutfs_inode_vers_bytes(sbi->fmt_vers);
+	store_inode(&sinode, inode, inode_bytes);
 
 	scoutfs_inode_init_key(&key, scoutfs_ino(inode));
 
-	ret = scoutfs_item_update(sb, &key, &sinode, sizeof(sinode), lock);
+	ret = scoutfs_item_update(sb, &key, &sinode, inode_bytes, lock);
 	if (!ret)
 		trace_scoutfs_dirty_inode(inode);
 	return ret;
@@ -1072,9 +1140,11 @@ void scoutfs_update_inode_item(struct inode *inode, struct scoutfs_lock *lock,
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
 	struct super_block *sb = inode->i_sb;
+	struct scoutfs_sb_info *sbi = SCOUTFS_SB(sb);
 	const u64 ino = scoutfs_ino(inode);
-	struct scoutfs_key key;
 	struct scoutfs_inode sinode;
+	struct scoutfs_key key;
+	int inode_bytes;
 	int ret;
 	int err;
 
@@ -1083,15 +1153,17 @@ void scoutfs_update_inode_item(struct inode *inode, struct scoutfs_lock *lock,
 	/* set the meta version once per trans for any inode updates */
 	scoutfs_inode_set_meta_seq(inode);
 
+	inode_bytes = scoutfs_inode_vers_bytes(sbi->fmt_vers);
+
 	/* only race with other inode field stores once */
-	store_inode(&sinode, inode);
+	store_inode(&sinode, inode, inode_bytes);
 
 	ret = update_indices(sb, si, ino, inode->i_mode, &sinode, lock_list, lock);
 	BUG_ON(ret);
 
 	scoutfs_inode_init_key(&key, ino);
 
-	err = scoutfs_item_update(sb, &key, &sinode, sizeof(sinode), lock);
+	err = scoutfs_item_update(sb, &key, &sinode, inode_bytes, lock);
 	if (err) {
 		scoutfs_err(sb, "inode %llu update err %d", ino, err);
 		BUG_ON(err);
@@ -1459,10 +1531,12 @@ out:
 int scoutfs_new_inode(struct super_block *sb, struct inode *dir, umode_t mode, dev_t rdev,
 		      u64 ino, struct scoutfs_lock *lock, struct inode **inode_ret)
 {
+	struct scoutfs_sb_info *sbi = SCOUTFS_SB(sb);
 	struct scoutfs_inode_info *si;
-	struct scoutfs_key key;
 	struct scoutfs_inode sinode;
+	struct scoutfs_key key;
 	struct inode *inode;
+	int inode_bytes;
 	int ret;
 
 	inode = new_inode(sb);
@@ -1478,6 +1552,7 @@ int scoutfs_new_inode(struct super_block *sb, struct inode *dir, umode_t mode, d
 	si->offline_blocks = 0;
 	si->next_readdir_pos = SCOUTFS_DIRENT_FIRST_POS;
 	si->next_xattr_id = 0;
+	si->proj = 0;
 	si->have_item = false;
 	atomic64_set(&si->last_refreshed, lock->refresh_gen);
 	scoutfs_lock_add_coverage(sb, lock, &si->ino_lock_cov);
@@ -1493,14 +1568,16 @@ int scoutfs_new_inode(struct super_block *sb, struct inode *dir, umode_t mode, d
 	inode->i_rdev = rdev;
 	set_inode_ops(inode);
 
-	store_inode(&sinode, inode);
+	inode_bytes = scoutfs_inode_vers_bytes(sbi->fmt_vers);
+
+	store_inode(&sinode, inode, inode_bytes);
 	scoutfs_inode_init_key(&key, scoutfs_ino(inode));
 
 	ret = scoutfs_omap_set(sb, ino);
 	if (ret < 0)
 		goto out;
 
-	ret = scoutfs_item_create(sb, &key, &sinode, sizeof(sinode), lock);
+	ret = scoutfs_item_create(sb, &key, &sinode, inode_bytes, lock);
 	if (ret < 0)
 		scoutfs_omap_clear(sb, ino);
 out:
@@ -1754,7 +1831,7 @@ static int try_delete_inode_items(struct super_block *sb, u64 ino)
 	}
 
 	scoutfs_inode_init_key(&key, ino);
-	ret = scoutfs_item_lookup_exact(sb, &key, &sinode, sizeof(sinode), lock);
+	ret = lookup_inode_item(sb, &key, &sinode, lock);
 	if (ret < 0) {
 		if (ret == -ENOENT)
 			ret = 0;
@@ -2143,6 +2220,17 @@ out:
 	return ret;
 }
 
+/*
+ * Return an error if the inode has the retention flag set and can not
+ * be modified.  This mimics the errno returned by the vfs whan an
+ * inode's immutable flag is set.  The flag won't be set on older format
+ * versions so we don't check the mounted format version here.
+ */
+int scoutfs_inode_check_retention(struct inode *inode)
+{
+	return (scoutfs_inode_get_flags(inode) & SCOUTFS_INO_FLAG_RETENTION) ? -EPERM : 0;
+}
+
 int scoutfs_inode_setup(struct super_block *sb)
 {
 	struct scoutfs_sb_info *sbi = SCOUTFS_SB(sb);

kmod/src/inode.h

@@ -21,6 +21,7 @@ struct scoutfs_inode_info {
 	u64 data_version;
 	u64 online_blocks;
 	u64 offline_blocks;
+	u64 proj;
 	u32 flags;
 	struct kc_timespec crtime;
 
@@ -47,7 +48,7 @@ struct scoutfs_inode_info {
 	atomic64_t last_refreshed;
 
 	/* initialized once for slab object */
-	seqcount_t seqcount;
+	seqlock_t seqlock;
 	bool staging;			/* holder of i_mutex is staging */
 	struct scoutfs_per_task pt_data_lock;
 	struct scoutfs_data_waitq data_waitq;
@@ -120,8 +121,15 @@ u64 scoutfs_inode_meta_seq(struct inode *inode);
 u64 scoutfs_inode_data_seq(struct inode *inode);
 u64 scoutfs_inode_data_version(struct inode *inode);
 void scoutfs_inode_get_onoff(struct inode *inode, s64 *on, s64 *off);
+u32 scoutfs_inode_get_flags(struct inode *inode);
+void scoutfs_inode_set_flags(struct inode *inode, u32 and, u32 or);
+u64 scoutfs_inode_get_proj(struct inode *inode);
+void scoutfs_inode_set_proj(struct inode *inode, u64 proj);
+
 int scoutfs_complete_truncate(struct inode *inode, struct scoutfs_lock *lock);
 
+int scoutfs_inode_check_retention(struct inode *inode);
+
 int scoutfs_inode_refresh(struct inode *inode, struct scoutfs_lock *lock);
 #ifdef KC_LINUX_HAVE_RHEL_IOPS_WRAPPER
 int scoutfs_getattr(struct vfsmount *mnt, struct dentry *dentry,

kmod/src/ioctl.c

@@ -42,6 +42,10 @@
 #include "alloc.h"
 #include "server.h"
 #include "counters.h"
+#include "attr_x.h"
+#include "totl.h"
+#include "wkic.h"
+#include "quota.h"
 #include "scoutfs_trace.h"
 
 /*
@@ -545,20 +549,41 @@ out:
 static long scoutfs_ioc_stat_more(struct file *file, unsigned long arg)
 {
 	struct inode *inode = file_inode(file);
-	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
-	struct scoutfs_ioctl_stat_more stm;
+	struct scoutfs_ioctl_inode_attr_x *iax = NULL;
+	struct scoutfs_ioctl_stat_more *stm = NULL;
+	int ret;
 
-	stm.meta_seq = scoutfs_inode_meta_seq(inode);
-	stm.data_seq = scoutfs_inode_data_seq(inode);
-	stm.data_version = scoutfs_inode_data_version(inode);
-	scoutfs_inode_get_onoff(inode, &stm.online_blocks, &stm.offline_blocks);
-	stm.crtime_sec = si->crtime.tv_sec;
-	stm.crtime_nsec = si->crtime.tv_nsec;
+	iax = kmalloc(sizeof(struct scoutfs_ioctl_inode_attr_x), GFP_KERNEL);
+	stm = kmalloc(sizeof(struct scoutfs_ioctl_stat_more), GFP_KERNEL);
+	if (!iax || !stm) {
+		ret = -ENOMEM;
+		goto out;
+	}
 
-	if (copy_to_user((void __user *)arg, &stm, sizeof(stm)))
-		return -EFAULT;
+	iax->x_mask = SCOUTFS_IOC_IAX_META_SEQ | SCOUTFS_IOC_IAX_DATA_SEQ |
+		      SCOUTFS_IOC_IAX_DATA_VERSION | SCOUTFS_IOC_IAX_ONLINE_BLOCKS |
+		      SCOUTFS_IOC_IAX_OFFLINE_BLOCKS | SCOUTFS_IOC_IAX_CRTIME;
+	iax->x_flags = 0;
+	ret = scoutfs_get_attr_x(inode, iax);
+	if (ret < 0)
+		goto out;
 
-	return 0;
+	stm->meta_seq = iax->meta_seq;
+	stm->data_seq = iax->data_seq;
+	stm->data_version = iax->data_version;
+	stm->online_blocks = iax->online_blocks;
+	stm->offline_blocks = iax->offline_blocks;
+	stm->crtime_sec = iax->crtime_sec;
+	stm->crtime_nsec = iax->crtime_nsec;
+
+	if (copy_to_user((void __user *)arg, stm, sizeof(struct scoutfs_ioctl_stat_more)))
+		ret = -EFAULT;
+	else
+		ret = 0;
+out:
+	kfree(iax);
+	kfree(stm);
+	return ret;
 }
 
 static bool inc_wrapped(u64 *ino, u64 *iblock)
@@ -615,24 +640,19 @@ static long scoutfs_ioc_data_waiting(struct file *file, unsigned long arg)
  * This is used when restoring files, it lets the caller set all the
  * inode attributes which are otherwise unreachable.  Changing the file
  * size can only be done for regular files with a data_version of 0.
+ *
+ * We unconditionally fill the iax attributes from the sm set and let
+ * set_attr_x check them.
  */
 static long scoutfs_ioc_setattr_more(struct file *file, unsigned long arg)
 {
-	struct inode *inode = file->f_inode;
-	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
-	struct super_block *sb = inode->i_sb;
+	struct inode *inode = file_inode(file);
 	struct scoutfs_ioctl_setattr_more __user *usm = (void __user *)arg;
+	struct scoutfs_ioctl_inode_attr_x *iax = NULL;
 	struct scoutfs_ioctl_setattr_more sm;
-	struct scoutfs_lock *lock = NULL;
 	LIST_HEAD(ind_locks);
-	bool set_data_seq;
 	int ret;
 
-	if (!capable(CAP_SYS_ADMIN)) {
-		ret = -EPERM;
-		goto out;
-	}
-
 	if (!(file->f_mode & FMODE_WRITE)) {
 		ret = -EBADF;
 		goto out;
@@ -643,65 +663,38 @@ static long scoutfs_ioc_setattr_more(struct file *file, unsigned long arg)
 		goto out;
 	}
 
-	if ((sm.i_size > 0 && sm.data_version == 0) ||
-	    ((sm.flags & SCOUTFS_IOC_SETATTR_MORE_OFFLINE) && !sm.i_size) ||
-	    (sm.flags & SCOUTFS_IOC_SETATTR_MORE_UNKNOWN)) {
+	if (sm.flags & SCOUTFS_IOC_SETATTR_MORE_UNKNOWN) {
 		ret = -EINVAL;
 		goto out;
 	}
 
+	iax = kzalloc(sizeof(struct scoutfs_ioctl_inode_attr_x), GFP_KERNEL);
+	if (!iax) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	iax->x_mask = SCOUTFS_IOC_IAX_DATA_VERSION | SCOUTFS_IOC_IAX_CTIME |
+		      SCOUTFS_IOC_IAX_CRTIME | SCOUTFS_IOC_IAX_SIZE;
+	iax->data_version = sm.data_version;
+	iax->ctime_sec = sm.ctime_sec;
+	iax->ctime_nsec = sm.ctime_nsec;
+	iax->crtime_sec = sm.crtime_sec;
+	iax->crtime_nsec = sm.crtime_nsec;
+	iax->size = sm.i_size;
+
+	if (sm.flags & SCOUTFS_IOC_SETATTR_MORE_OFFLINE)
+		iax->x_flags |= SCOUTFS_IOC_IAX_F_SIZE_OFFLINE;
+
 	ret = mnt_want_write_file(file);
-	if (ret)
+	if (ret < 0)
 		goto out;
 
-	inode_lock(inode);
+	ret = scoutfs_set_attr_x(inode, iax);
 
-	ret = scoutfs_lock_inode(sb, SCOUTFS_LOCK_WRITE,
-				 SCOUTFS_LKF_REFRESH_INODE, inode, &lock);
-	if (ret)
-		goto unlock;
-
-	/* can only change size/dv on untouched regular files */
-	if ((sm.i_size != 0 || sm.data_version != 0) &&
-	    ((!S_ISREG(inode->i_mode) ||
-	      scoutfs_inode_data_version(inode) != 0))) {
-		ret = -EINVAL;
-		goto unlock;
-	}
-
-	/* create offline extents in potentially many transactions */
-	if (sm.flags & SCOUTFS_IOC_SETATTR_MORE_OFFLINE) {
-		ret = scoutfs_data_init_offline_extent(inode, sm.i_size, lock);
-		if (ret)
-			goto unlock;
-	}
-
-	/* setting only so we don't see 0 data seq with nonzero data_version */
-	set_data_seq = sm.data_version != 0 ? true : false;
-	ret = scoutfs_inode_index_lock_hold(inode, &ind_locks, set_data_seq, false);
-	if (ret)
-		goto unlock;
-
-	if (sm.data_version)
-		scoutfs_inode_set_data_version(inode, sm.data_version);
-	if (sm.i_size)
-		i_size_write(inode, sm.i_size);
-	inode->i_ctime.tv_sec = sm.ctime_sec;
-	inode->i_ctime.tv_nsec = sm.ctime_nsec;
-	si->crtime.tv_sec = sm.crtime_sec;
-	si->crtime.tv_nsec = sm.crtime_nsec;
-
-	scoutfs_update_inode_item(inode, lock, &ind_locks);
-	ret = 0;
-
-	scoutfs_release_trans(sb);
-unlock:
-	scoutfs_inode_index_unlock(sb, &ind_locks);
-	scoutfs_unlock(sb, lock, SCOUTFS_LOCK_WRITE);
-	inode_unlock(inode);
 	mnt_drop_write_file(file);
 out:
-
+	kfree(iax);
 	return ret;
 }
 
@@ -1035,124 +1028,32 @@ out:
 	return ret;
 }
 
-struct xattr_total_entry {
-	struct rb_node node;
-	struct scoutfs_ioctl_xattr_total xt;
-	u64 fs_seq;
-	u64 fs_total;
-	u64 fs_count;
-	u64 fin_seq;
-	u64 fin_total;
-	s64 fin_count;
-	u64 log_seq;
-	u64 log_total;
-	s64 log_count;
+struct read_xattr_total_iter_cb_args {
+	struct scoutfs_ioctl_xattr_total *xt;
+	unsigned int copied;
+	unsigned int total;
 };
 
-static int cmp_xt_entry_name(const struct xattr_total_entry *a,
-			     const struct xattr_total_entry *b)
-
-{
-	return scoutfs_cmp_u64s(a->xt.name[0], b->xt.name[0]) ?:
-	       scoutfs_cmp_u64s(a->xt.name[1], b->xt.name[1]) ?:
-	       scoutfs_cmp_u64s(a->xt.name[2], b->xt.name[2]);
-}
-
 /*
- * Record the contribution of the three classes of logged items we can
- * see: the item in the fs_root, items from finalized log btrees, and
- * items from active log btrees.  Once we have the full set the caller
- * can decide which of the items contribute to the total it sends to the
- * user.
+ * This is called under an RCU read lock so it can't copy to userspace.
  */
-static int read_xattr_total_item(struct super_block *sb, struct scoutfs_key *key,
-				 u64 seq, u8 flags, void *val, int val_len, int fic, void *arg)
+static int read_xattr_total_iter_cb(struct scoutfs_key *key, void *val, unsigned int val_len,
+				    void *cb_arg)
 {
+	struct read_xattr_total_iter_cb_args *cba = cb_arg;
 	struct scoutfs_xattr_totl_val *tval = val;
-	struct xattr_total_entry *ent;
-	struct xattr_total_entry rd;
-	struct rb_root *root = arg;
-	struct rb_node *parent;
-	struct rb_node **node;
-	int cmp;
+	struct scoutfs_ioctl_xattr_total *xt = &cba->xt[cba->copied];
 
-	rd.xt.name[0] = le64_to_cpu(key->skxt_a);
-	rd.xt.name[1] = le64_to_cpu(key->skxt_b);
-	rd.xt.name[2] = le64_to_cpu(key->skxt_c);
+	xt->name[0] = le64_to_cpu(key->skxt_a);
+	xt->name[1] = le64_to_cpu(key->skxt_b);
+	xt->name[2] = le64_to_cpu(key->skxt_c);
+	xt->total = le64_to_cpu(tval->total);
+	xt->count = le64_to_cpu(tval->count);
 
-	/* find entry matching name */
-	node = &root->rb_node;
-	parent = NULL;
-	cmp = -1;
-	while (*node) {
-		parent = *node;
-		ent = container_of(*node, struct xattr_total_entry, node);
-
-		/* sort merge items by key then newest to oldest */
-		cmp = cmp_xt_entry_name(&rd, ent);
-		if (cmp < 0)
-			node = &(*node)->rb_left;
-		else if (cmp > 0)
-			node = &(*node)->rb_right;
-		else
-			break;
-	}
-
-	/* allocate and insert new node if we need to */
-	if (cmp != 0) {
-		ent = kzalloc(sizeof(*ent), GFP_KERNEL);
-		if (!ent)
-			return -ENOMEM;
-
-		memcpy(&ent->xt.name, &rd.xt.name, sizeof(ent->xt.name));
-
-		rb_link_node(&ent->node, parent, node);
-		rb_insert_color(&ent->node, root);
-	}
-
-	if (fic & FIC_FS_ROOT) {
-		ent->fs_seq = seq;
-		ent->fs_total = le64_to_cpu(tval->total);
-		ent->fs_count = le64_to_cpu(tval->count);
-	} else if (fic & FIC_FINALIZED) {
-		ent->fin_seq = seq;
-		ent->fin_total += le64_to_cpu(tval->total);
-		ent->fin_count += le64_to_cpu(tval->count);
-	} else {
-		ent->log_seq = seq;
-		ent->log_total += le64_to_cpu(tval->total);
-		ent->log_count += le64_to_cpu(tval->count);
-	}
-
-	scoutfs_inc_counter(sb, totl_read_item);
-
-	return 0;
-}
-
-/* these are always _safe, node stores next */
-#define for_each_xt_ent(ent, node, root)					\
-	for (node = rb_first(root);						\
-	     node && (ent = rb_entry(node, struct xattr_total_entry, node),	\
-		      node = rb_next(node), 1); )
-
-#define for_each_xt_ent_reverse(ent, node, root)				\
-	for (node = rb_last(root);						\
-	     node && (ent = rb_entry(node, struct xattr_total_entry, node),	\
-		      node = rb_prev(node), 1); )
-
-static void free_xt_ent(struct rb_root *root, struct xattr_total_entry *ent)
-{
-	rb_erase(&ent->node, root);
-	kfree(ent);
-}
-
-static void free_all_xt_ents(struct rb_root *root)
-{
-	struct xattr_total_entry *ent;
-	struct rb_node *node;
-
-	for_each_xt_ent(ent, node, root)
-		free_xt_ent(root, ent);
+	if (++cba->copied < cba->total)
+		return -EAGAIN;
+	else
+		return 0;
 }
 
 /*
@@ -1162,30 +1063,6 @@ static void free_all_xt_ents(struct rb_root *root)
  * have been committed.  It doesn't use locking to force commits and
  * block writers so it can be a little bit out of date with respect to
  * dirty xattrs in memory across the system.
- *
- * Our reader has to be careful because the log btree merging code can
- * write partial results to the fs_root.  This means that a reader can
- * see both cases where new finalized logs should be applied to the old
- * fs items and where old finalized logs have already been applied to
- * the partially merged fs items.  Currently active logged items are
- * always applied on top of all cases.
- *
- * These cases are differentiated with a combination of sequence numbers
- * in items, the count of contributing xattrs, and a flag
- * differentiating finalized and active logged items.  This lets us
- * recognize all cases, including when finalized logs were merged and
- * deleted the fs item.
- *
- * We're allocating a tracking struct for each totl name we see while
- * traversing the item btrees.  The forest reader is providing the items
- * it finds in leaf blocks that contain the search key.  In the worst
- * case all of these blocks are full and none of the items overlap.  At
- * most, figure order a thousand names per mount.  But in practice many
- * of these factors fall away: leaf blocks aren't fill, leaf items
- * overlap, there aren't finalized log btrees, and not all mounts are
- * actively changing totals.   We're much more likely to only read a
- * leaf block's worth of totals that have been long since merged into
- * the fs_root.
  */
 static long scoutfs_ioc_read_xattr_totals(struct file *file, unsigned long arg)
 {
@@ -1193,14 +1070,13 @@ static long scoutfs_ioc_read_xattr_totals(struct file *file, unsigned long arg)
 	struct scoutfs_ioctl_read_xattr_totals __user *urxt = (void __user *)arg;
 	struct scoutfs_ioctl_read_xattr_totals rxt;
 	struct scoutfs_ioctl_xattr_total __user *uxt;
-	struct xattr_total_entry *ent;
+	struct read_xattr_total_iter_cb_args cba = {NULL, };
+	struct scoutfs_key range_start;
+	struct scoutfs_key range_end;
 	struct scoutfs_key key;
-	struct scoutfs_key bloom_key;
-	struct scoutfs_key start;
-	struct scoutfs_key end;
-	struct rb_root root = RB_ROOT;
-	struct rb_node *node;
-	int count = 0;
+	unsigned int copied = 0;
+	unsigned int total;
+	unsigned int ready;
 	int ret;
 
 	if (!(file->f_mode & FMODE_READ)) {
@@ -1213,6 +1089,13 @@ static long scoutfs_ioc_read_xattr_totals(struct file *file, unsigned long arg)
 		goto out;
 	}
 
+	cba.xt = (void *)__get_free_page(GFP_KERNEL);
+	if (!cba.xt) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	cba.total = PAGE_SIZE / sizeof(struct scoutfs_ioctl_xattr_total);
+
 	if (copy_from_user(&rxt, urxt, sizeof(rxt))) {
 		ret = -EFAULT;
 		goto out;
@@ -1225,101 +1108,40 @@ static long scoutfs_ioc_read_xattr_totals(struct file *file, unsigned long arg)
 		goto out;
 	}
 
-	scoutfs_key_set_zeros(&bloom_key);
-	bloom_key.sk_zone = SCOUTFS_XATTR_TOTL_ZONE;
-	scoutfs_xattr_init_totl_key(&start, rxt.pos_name);
+	total = div_u64(min_t(u64, rxt.totals_bytes, INT_MAX),
+			sizeof(struct scoutfs_ioctl_xattr_total));
 
-	while (rxt.totals_bytes >= sizeof(struct scoutfs_ioctl_xattr_total)) {
+	scoutfs_totl_set_range(&range_start, &range_end);
+	scoutfs_xattr_init_totl_key(&key, rxt.pos_name);
 
-		scoutfs_key_set_ones(&end);
-		end.sk_zone = SCOUTFS_XATTR_TOTL_ZONE;
-		if (scoutfs_key_compare(&start, &end) > 0)
+	while (copied < total) {
+		cba.copied = 0;
+		ret = scoutfs_wkic_iterate(sb, &key, &range_end, &range_start, &range_end,
+					   read_xattr_total_iter_cb, &cba);
+		if (ret < 0)
+			goto out;
+
+		if (cba.copied == 0)
 			break;
 
-		key = start;
-		ret = scoutfs_forest_read_items(sb, &key, &bloom_key, &start, &end,
-						read_xattr_total_item, &root);
-		if (ret < 0) {
-			if (ret == -ESTALE) {
-				free_all_xt_ents(&root);
-				continue;
-			}
+		ready = min(total - copied, cba.copied);
+
+		if (copy_to_user(&uxt[copied], cba.xt, ready * sizeof(cba.xt[0]))) {
+			ret = -EFAULT;
 			goto out;
 		}
 
-		if (RB_EMPTY_ROOT(&root))
-			break;
-
-		/* trim totals that fall outside of the consistent range */
-		for_each_xt_ent(ent, node, &root) {
-			scoutfs_xattr_init_totl_key(&key, ent->xt.name);
-			if (scoutfs_key_compare(&key, &start) < 0) {
-				free_xt_ent(&root, ent);
-			} else {
-				break;
-			}
-		}
-		for_each_xt_ent_reverse(ent, node, &root) {
-			scoutfs_xattr_init_totl_key(&key, ent->xt.name);
-			if (scoutfs_key_compare(&key, &end) > 0) {
-				free_xt_ent(&root, ent);
-			} else {
-				break;
-			}
-		}
-
-		/* copy resulting unique non-zero totals to userspace */
-		for_each_xt_ent(ent, node, &root) {
-			if (rxt.totals_bytes < sizeof(ent->xt))
-				break;
-
-			/* start with the fs item if we have it */
-			if (ent->fs_seq != 0) {
-				ent->xt.total = ent->fs_total;
-				ent->xt.count = ent->fs_count;
-				scoutfs_inc_counter(sb, totl_read_fs);
-			}
-
-			/* apply finalized logs if they're newer or creating */
-			if (((ent->fs_seq != 0) && (ent->fin_seq > ent->fs_seq)) ||
-			    ((ent->fs_seq == 0) && (ent->fin_count > 0))) {
-				ent->xt.total += ent->fin_total;
-				ent->xt.count += ent->fin_count;
-				scoutfs_inc_counter(sb, totl_read_finalized);
-			}
-
-			/* always apply active logs which must be newer than fs and finalized */
-			if (ent->log_seq > 0) {
-				ent->xt.total += ent->log_total;
-				ent->xt.count += ent->log_count;
-				scoutfs_inc_counter(sb, totl_read_logged);
-			}
-
-			if (ent->xt.total != 0 || ent->xt.count != 0) {
-				if (copy_to_user(uxt, &ent->xt, sizeof(ent->xt))) {
-					ret = -EFAULT;
-					goto out;
-				}
-
-				uxt++;
-				rxt.totals_bytes -= sizeof(ent->xt);
-				count++;
-				scoutfs_inc_counter(sb, totl_read_copied);
-			}
-
-			free_xt_ent(&root, ent);
-		}
-
-		/* continue after the last possible key read */
-		start = end;
-		scoutfs_key_inc(&start);
+		scoutfs_xattr_init_totl_key(&key, cba.xt[ready - 1].name);
+		scoutfs_key_inc(&key);
+		copied += ready;
 	}
 
 	ret = 0;
 out:
-	free_all_xt_ents(&root);
+	if (cba.xt)
+		free_page((long)cba.xt);
 
-	return ret ?: count;
+	return ret ?: copied;
 }
 
 static long scoutfs_ioc_get_allocated_inos(struct file *file, unsigned long arg)
@@ -1504,6 +1326,265 @@ out:
 	return nr ?: ret;
 }
 
+static long scoutfs_ioc_get_attr_x(struct file *file, unsigned long arg)
+{
+	struct inode *inode = file_inode(file);
+	struct scoutfs_ioctl_inode_attr_x __user *uiax = (void __user *)arg;
+	struct scoutfs_ioctl_inode_attr_x *iax = NULL;
+	int ret;
+
+	iax = kmalloc(sizeof(struct scoutfs_ioctl_inode_attr_x), GFP_KERNEL);
+	if (!iax) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ret = get_user(iax->x_mask, &uiax->x_mask) ?:
+	      get_user(iax->x_flags, &uiax->x_flags);
+	if (ret < 0)
+		goto out;
+
+	ret = scoutfs_get_attr_x(inode, iax);
+	if (ret < 0)
+		goto out;
+
+	/* only copy results after dropping cluster locks (could fault) */
+	if (ret > 0 && copy_to_user(uiax, iax, ret) != 0)
+		ret = -EFAULT;
+	else
+		ret = 0;
+out:
+	kfree(iax);
+	return ret;
+}
+
+static long scoutfs_ioc_set_attr_x(struct file *file, unsigned long arg)
+{
+	struct inode *inode = file_inode(file);
+	struct scoutfs_ioctl_inode_attr_x __user *uiax = (void __user *)arg;
+	struct scoutfs_ioctl_inode_attr_x *iax = NULL;
+	int ret;
+
+	iax = kmalloc(sizeof(struct scoutfs_ioctl_inode_attr_x), GFP_KERNEL);
+	if (!iax) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	if (copy_from_user(iax, uiax, sizeof(struct scoutfs_ioctl_inode_attr_x))) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	ret = mnt_want_write_file(file);
+	if (ret < 0)
+		goto out;
+
+	ret = scoutfs_set_attr_x(inode, iax);
+
+	mnt_drop_write_file(file);
+out:
+	kfree(iax);
+	return ret;
+}
+
+static long scoutfs_ioc_get_quota_rules(struct file *file, unsigned long arg)
+{
+	struct super_block *sb = file_inode(file)->i_sb;
+	struct scoutfs_ioctl_get_quota_rules __user *ugqr = (void __user *)arg;
+	struct scoutfs_ioctl_get_quota_rules gqr;
+	struct scoutfs_ioctl_quota_rule __user *uirules;
+	struct scoutfs_ioctl_quota_rule *irules;
+	struct page *page = NULL;
+	int copied = 0;
+	int nr;
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	if (copy_from_user(&gqr, ugqr, sizeof(gqr)))
+		return -EFAULT;
+
+	if (gqr.rules_nr == 0)
+		return 0;
+
+	uirules = (void __user *)gqr.rules_ptr;
+	/* limit rules copied per call */
+	gqr.rules_nr = min_t(u64, gqr.rules_nr, INT_MAX);
+
+	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
+	if (!page) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	irules = page_address(page);
+
+	while (copied < gqr.rules_nr) {
+		nr = min_t(u64, gqr.rules_nr - copied,
+				PAGE_SIZE / sizeof(struct scoutfs_ioctl_quota_rule));
+		ret = scoutfs_quota_get_rules(sb, gqr.iterator, page_address(page), nr);
+		if (ret <= 0)
+			goto out;
+
+		if (copy_to_user(&uirules[copied], irules, ret * sizeof(irules[0]))) {
+			ret = -EFAULT;
+			goto out;
+		}
+
+		copied += ret;
+	}
+
+	ret = 0;
+out:
+	if (page)
+		__free_page(page);
+
+	if (ret == 0 && copy_to_user(ugqr->iterator, gqr.iterator, sizeof(gqr.iterator)))
+		ret = -EFAULT;
+
+	return ret ?: copied;
+}
+
+static long scoutfs_ioc_mod_quota_rule(struct file *file, unsigned long arg, bool is_add)
+{
+	struct super_block *sb = file_inode(file)->i_sb;
+	struct scoutfs_ioctl_quota_rule __user *uirule = (void __user *)arg;
+	struct scoutfs_ioctl_quota_rule irule;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	if (copy_from_user(&irule, uirule, sizeof(irule)))
+		return -EFAULT;
+
+	return scoutfs_quota_mod_rule(sb, is_add, &irule);
+}
+
+struct read_index_buf {
+	int nr;
+	int size;
+	struct scoutfs_ioctl_xattr_index_entry ents[0];
+};
+
+#define READ_INDEX_BUF_MAX_ENTS \
+	((PAGE_SIZE - sizeof(struct read_index_buf)) / \
+		sizeof(struct scoutfs_ioctl_xattr_index_entry))
+
+/*
+ * This doesn't filter out duplicates, the caller filters them out to
+ * catch duplicates between iteration calls.
+ */
+static int read_index_cb(struct scoutfs_key *key, void *val, unsigned int val_len, void *cb_arg)
+{
+	struct read_index_buf *rib = cb_arg;
+	struct scoutfs_ioctl_xattr_index_entry *ent = &rib->ents[rib->nr];
+	u64 xid;
+
+	if (val_len != 0)
+		return -EIO;
+
+	/* discard the xid, they're not exposed to ioctl callers */
+	scoutfs_xattr_get_indx_key(key, &ent->major, &ent->minor, &ent->ino, &xid);
+
+	if (++rib->nr == rib->size)
+		return rib->nr;
+
+	return -EAGAIN;
+}
+
+static long scoutfs_ioc_read_xattr_index(struct file *file, unsigned long arg)
+{
+	struct super_block *sb = file_inode(file)->i_sb;
+	struct scoutfs_ioctl_read_xattr_index __user *urxi = (void __user *)arg;
+	struct scoutfs_ioctl_xattr_index_entry __user *uents;
+	struct scoutfs_ioctl_xattr_index_entry *ent;
+	struct scoutfs_ioctl_xattr_index_entry prev;
+	struct scoutfs_ioctl_read_xattr_index rxi;
+	struct read_index_buf *rib;
+	struct page *page = NULL;
+	struct scoutfs_key first;
+	struct scoutfs_key last;
+	struct scoutfs_key start;
+	struct scoutfs_key end;
+	int copied = 0;
+	int ret;
+	int i;
+
+	if (!capable(CAP_SYS_ADMIN)) {
+		ret = -EPERM;
+		goto out;
+	}
+
+	if (copy_from_user(&rxi, urxi, sizeof(rxi))) {
+		ret = -EFAULT;
+		goto out;
+	}
+	uents = (void __user *)rxi.entries_ptr;
+	rxi.entries_nr = min_t(u64, rxi.entries_nr, INT_MAX);
+
+	page = alloc_page(GFP_KERNEL);
+	if (!page) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	rib = page_address(page);
+
+	scoutfs_xattr_init_indx_key(&first, rxi.first.major, rxi.first.minor, rxi.first.ino, 0);
+	scoutfs_xattr_init_indx_key(&last, rxi.last.major, rxi.last.minor, rxi.last.ino, U64_MAX);
+	scoutfs_xattr_indx_get_range(&start, &end);
+
+	if (scoutfs_key_compare(&first, &last) > 0) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* 0 ino doesn't exist, can't ever match entry to return */
+	memset(&prev, 0, sizeof(prev));
+
+	while (copied < rxi.entries_nr) {
+		rib->nr = 0;
+		rib->size = min_t(u64, rxi.entries_nr - copied, READ_INDEX_BUF_MAX_ENTS);
+		ret = scoutfs_wkic_iterate(sb, &first, &last, &start, &end,
+					   read_index_cb, rib);
+		if (ret < 0)
+			goto out;
+		if (rib->nr == 0)
+			break;
+
+		/*
+		 * Copy entries to userspace, skipping duplicate entries
+		 * that can result from multiple xattrs indexing an
+		 * inode at the same position and which can span
+		 * multiple cache iterations.  (Comparing in order of
+		 * most likely to change to fail fast.)
+		 */
+		for (i = 0, ent = rib->ents; i < rib->nr; i++, ent++) {
+			if (ent->ino == prev.ino && ent->minor == prev.minor &&
+			    ent->major == prev.major)
+				continue;
+
+			if (copy_to_user(&uents[copied], ent, sizeof(*ent))) {
+				ret = -EFAULT;
+				goto out;
+			}
+
+			prev = *ent;
+			copied++;
+		}
+
+		scoutfs_xattr_init_indx_key(&first, prev.major, prev.minor, prev.ino, U64_MAX);
+		scoutfs_key_inc(&first);
+	}
+
+	ret = copied;
+out:
+	if (page)
+		__free_page(page);
+
+	return ret;
+}
+
 long scoutfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
 	switch (cmd) {
@@ -1541,6 +1622,18 @@ long scoutfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 		return scoutfs_ioc_get_allocated_inos(file, arg);
 	case SCOUTFS_IOC_GET_REFERRING_ENTRIES:
 		return scoutfs_ioc_get_referring_entries(file, arg);
+	case SCOUTFS_IOC_GET_ATTR_X:
+		return scoutfs_ioc_get_attr_x(file, arg);
+	case SCOUTFS_IOC_SET_ATTR_X:
+		return scoutfs_ioc_set_attr_x(file, arg);
+	case SCOUTFS_IOC_GET_QUOTA_RULES:
+		return scoutfs_ioc_get_quota_rules(file, arg);
+	case SCOUTFS_IOC_ADD_QUOTA_RULE:
+		return scoutfs_ioc_mod_quota_rule(file, arg, true);
+	case SCOUTFS_IOC_DEL_QUOTA_RULE:
+		return scoutfs_ioc_mod_quota_rule(file, arg, false);
+	case SCOUTFS_IOC_READ_XATTR_INDEX:
+		return scoutfs_ioc_read_xattr_index(file, arg);
 	}
 
 	return -ENOTTY;

kmod/src/ioctl.h

@@ -673,4 +673,174 @@ struct scoutfs_ioctl_dirent {
 #define SCOUTFS_IOC_GET_REFERRING_ENTRIES \
 	_IOW(SCOUTFS_IOCTL_MAGIC, 17, struct scoutfs_ioctl_get_referring_entries)
 
+struct scoutfs_ioctl_inode_attr_x {
+	__u64 x_mask;
+	__u64 x_flags;
+	__u64 meta_seq;
+	__u64 data_seq;
+	__u64 data_version;
+	__u64 online_blocks;
+	__u64 offline_blocks;
+	__u64 ctime_sec;
+	__u32 ctime_nsec;
+	__u32 crtime_nsec;
+	__u64 crtime_sec;
+	__u64 size;
+	__u64 bits;
+	__u64 project_id;
+};
+
+/*
+ * Behavioral flags set in the x_flags field.  These flags don't
+ * necessarily correspond to specific attributes, but instead change the
+ * behaviour of a _get_ or _set_ operation.
+ *
+ * @SCOUTFS_IOC_IAX_F_SIZE_OFFLINE: When setting i_size, also create
+ * extents which are marked offline for the region of the file from
+ * offset 0 to the new set size.  This can only be set when setting the
+ * size and has no effect if setting the size fails.
+ */
+#define SCOUTFS_IOC_IAX_F_SIZE_OFFLINE	(1ULL << 0)
+#define SCOUTFS_IOC_IAX_F__UNKNOWN	(U64_MAX << 1)
+
+/*
+ * Single-bit values stored in the @bits field.  These indicate whether
+ * the bit is set, or not.  The main _IAX_ bits set in the mask indicate
+ * whether this value bit is populated by _get or stored by _set. 
+ */
+#define SCOUTFS_IOC_IAX_B_RETENTION	(1ULL << 0)
+
+/*
+ * x_mask bits which indicate which attributes of the inode to populate
+ * on return for _get or to set on the inode for _set.  Each mask bit
+ * corresponds to the matching named field in the attr_x struct passed
+ * to the _get_ and _set_ calls.
+ *
+ * Each field can have different permissions or other attribute
+ * requirements which can cause calls to fail.  If _set_ fails then no
+ * other attribute changes will have been made by the same call.
+ *
+ * @SCOUTFS_IOC_IAX_RETENTION: Mark a file for retention.  When marked,
+ * no modification can be made to the file other than changing extended
+ * attributes outside the "user." prefix and clearing the retention
+ * mark.  This can only be set on regular files and requires root (the
+ * CAP_SYS_ADMIN capability).  Other attributes can be set with a
+ * set_attr_x call on a retention inode as long as that call also
+ * successfully clears the retention mark.
+ */
+#define SCOUTFS_IOC_IAX_META_SEQ	(1ULL << 0)
+#define SCOUTFS_IOC_IAX_DATA_SEQ	(1ULL << 1)
+#define SCOUTFS_IOC_IAX_DATA_VERSION	(1ULL << 2)
+#define SCOUTFS_IOC_IAX_ONLINE_BLOCKS	(1ULL << 3)
+#define SCOUTFS_IOC_IAX_OFFLINE_BLOCKS	(1ULL << 4)
+#define SCOUTFS_IOC_IAX_CTIME		(1ULL << 5)
+#define SCOUTFS_IOC_IAX_CRTIME		(1ULL << 6)
+#define SCOUTFS_IOC_IAX_SIZE		(1ULL << 7)
+#define SCOUTFS_IOC_IAX_RETENTION	(1ULL << 8)
+#define SCOUTFS_IOC_IAX_PROJECT_ID	(1ULL << 9)
+
+/* single bit attributes that are packed in the bits field as _B_ */
+#define SCOUTFS_IOC_IAX__BITS		(SCOUTFS_IOC_IAX_RETENTION)
+/* inverse of all the bits we understand */
+#define SCOUTFS_IOC_IAX__UNKNOWN	(U64_MAX << 10)
+
+#define SCOUTFS_IOC_GET_ATTR_X \
+	_IOW(SCOUTFS_IOCTL_MAGIC, 18, struct scoutfs_ioctl_inode_attr_x)
+
+#define SCOUTFS_IOC_SET_ATTR_X \
+	_IOW(SCOUTFS_IOCTL_MAGIC, 19, struct scoutfs_ioctl_inode_attr_x)
+
+/*
+ * (These fields are documented in the order that they're displayed by
+ * the scoutfs cli utility which matches the sort order of the rules.)
+ *
+ * @prio: The priority of the rule.  Rules are sorted by their fields
+ * with prio at the highest magnitude.  When multiple rules match the
+ * rule with the highest sort order is enforced.  The priority field
+ * lets rules override the default field sort order.
+ *
+ * @name_val[3]: The three 64bit values that make up the name of the
+ * totl xattr whose total will be checked against the rule's limit to
+ * see if the quota rule has been exceeded.  The behavior of the values
+ * can be changed by their corresponding name_source and name_flags.
+ *
+ * @name_source[3]: The SQ_NS_ enums that control where the value comes
+ * from.  _LITERAL uses the value from name_val.  Inode attribute
+ * sources (_PROJ, _UID, _GID) are taken from the inode of the operation
+ * that is being checked against the rule.
+ *
+ * @name_flags[3]: The SQ_NF_ enums that alter the name values.  _SELECT
+ * makes the rule only match if the inode attribute of the operation
+ * matches the attribute value stored in name_val.  This lets rules
+ * match a specific value of an attribute rather than mapping all
+ * attribute values of to totl names.
+ *
+ * @op: The SQ_OP_ enums which specify the operation that can't exceed
+ * the rule's limit.  _INODE checks inode creation and the inode
+ * attributes are taken from the inode that would be created.  _DATA
+ * checks file data block allocation and the inode fields come from the
+ * inode that is allocating the blocks.
+ *
+ * @limit: The 64bit value that is checked against the totl value
+ * described by the rule.  If the totl value is greater than or equal to
+ * this value of the matching rule then the operation will return
+ * -EDQUOT.
+ *
+ * @rule_flags: SQ_RF_TOTL_COUNT indicates that the rule's limit should
+ * be checked against the number of xattrs contributing to a totl value
+ * instead of the sum of the xattrs.
+ */
+struct scoutfs_ioctl_quota_rule {
+	__u64 name_val[3];
+	__u64 limit;
+	__u8 prio;
+	__u8 op;
+	__u8 rule_flags;
+	__u8 name_source[3];
+	__u8 name_flags[3];
+	__u8 _pad[7];
+};
+
+struct scoutfs_ioctl_get_quota_rules {
+	__u64 iterator[2];
+	__u64 rules_ptr;
+	__u64 rules_nr;
+};
+
+/*
+ * Rules are uniquely identified by their non-padded fields.  Addition will fail
+ * with -EEXIST if the specified rule already exists and deletion must find a rule
+ * with all matching fields to delete.
+ */
+#define SCOUTFS_IOC_GET_QUOTA_RULES \
+	_IOR(SCOUTFS_IOCTL_MAGIC, 20, struct scoutfs_ioctl_get_quota_rules)
+#define SCOUTFS_IOC_ADD_QUOTA_RULE \
+	_IOW(SCOUTFS_IOCTL_MAGIC, 21, struct scoutfs_ioctl_quota_rule)
+#define SCOUTFS_IOC_DEL_QUOTA_RULE \
+	_IOW(SCOUTFS_IOCTL_MAGIC, 22, struct scoutfs_ioctl_quota_rule)
+
+/*
+ * Inodes can be indexed in a global key space at a position determined
+ * by a .indx. tagged xattr.  The xattr name specifies the two index
+ * position values, with major having the more significant comparison
+ * order.
+ */
+struct scoutfs_ioctl_xattr_index_entry {
+	__u64 minor;
+	__u64 ino;
+	__u8 major;
+	__u8 _pad[7];
+};
+
+struct scoutfs_ioctl_read_xattr_index {
+	__u64 flags;
+	struct scoutfs_ioctl_xattr_index_entry first;
+	struct scoutfs_ioctl_xattr_index_entry last;
+	__u64 entries_ptr;
+	__u64 entries_nr;
+};
+
+#define SCOUTFS_IOC_READ_XATTR_INDEX \
+	_IOR(SCOUTFS_IOCTL_MAGIC, 23, struct scoutfs_ioctl_read_xattr_index)
+
 #endif

kmod/src/item.c

@@ -24,6 +24,7 @@
 #include "item.h"
 #include "forest.h"
 #include "block.h"
+#include "msg.h"
 #include "trans.h"
 #include "counters.h"
 #include "scoutfs_trace.h"
@@ -1670,13 +1671,24 @@ out:
 	return ret;
 }
 
-static int lock_safe(struct scoutfs_lock *lock, struct scoutfs_key *key,
+static int lock_safe(struct super_block *sb, struct scoutfs_lock *lock, struct scoutfs_key *key,
 		     int mode)
 {
-	if (WARN_ON_ONCE(!scoutfs_lock_protected(lock, key, mode)))
+	bool prot = scoutfs_lock_protected(lock, key, mode);
+
+	if (!prot) {
+		static bool once = false;
+		if (!once) {
+			scoutfs_err(sb, "lock (start "SK_FMT" end "SK_FMT" mode 0x%x) does not protect operation (key "SK_FMT" mode 0x%x)",
+				    SK_ARG(&lock->start), SK_ARG(&lock->end), lock->mode,
+				    SK_ARG(key), mode);
+			dump_stack();
+			once = true;
+		}
 		return -EINVAL;
-	else
-		return 0;
+	}
+
+	return 0;
 }
 
 static int optional_lock_mode_match(struct scoutfs_lock *lock, int mode)
@@ -1708,8 +1720,8 @@ static int copy_val(void *dst, int dst_len, void *src, int src_len)
  * The amount of bytes copied is returned which can be 0 or truncated if
  * the caller's buffer isn't big enough.
  */
-int scoutfs_item_lookup(struct super_block *sb, struct scoutfs_key *key,
-			void *val, int val_len, struct scoutfs_lock *lock)
+static int item_lookup(struct super_block *sb, struct scoutfs_key *key,
+		       void *val, int val_len, int len_limit, struct scoutfs_lock *lock)
 {
 	DECLARE_ITEM_CACHE_INFO(sb, cinf);
 	struct cached_item *item;
@@ -1718,7 +1730,7 @@ int scoutfs_item_lookup(struct super_block *sb, struct scoutfs_key *key,
 
 	scoutfs_inc_counter(sb, item_lookup);
 
-	if ((ret = lock_safe(lock, key, SCOUTFS_LOCK_READ)))
+	if ((ret = lock_safe(sb, lock, key, SCOUTFS_LOCK_READ)))
 		goto out;
 
 	ret = get_cached_page(sb, cinf, lock, key, false, false, 0, &pg);
@@ -1729,6 +1741,8 @@ int scoutfs_item_lookup(struct super_block *sb, struct scoutfs_key *key,
 	item = item_rbtree_walk(&pg->item_root, key, NULL, NULL, NULL);
 	if (!item || item->deletion)
 		ret = -ENOENT;
+	else if (len_limit > 0 && item->val_len > len_limit)
+		ret = -EIO;
 	else
 		ret = copy_val(val, val_len, item->val, item->val_len);
 
@@ -1737,13 +1751,38 @@ out:
 	return ret;
 }
 
+int scoutfs_item_lookup(struct super_block *sb, struct scoutfs_key *key,
+			void *val, int val_len, struct scoutfs_lock *lock)
+{
+	return item_lookup(sb, key, val, val_len, 0, lock);
+}
+
+/*
+ * Copy an item's value into the caller's buffer.  If the item's value
+ * is larger than the caller's buffer then -EIO is returned.  If the
+ * item is smaller then the bytes from the end of the copied value to
+ * the end of the buffer are zeroed.  The number of value bytes copied
+ * is returned, and 0 can be returned for an item with no value.
+ */
+int scoutfs_item_lookup_smaller_zero(struct super_block *sb, struct scoutfs_key *key,
+				     void *val, int val_len, struct scoutfs_lock *lock)
+{
+	int ret;
+
+	ret = item_lookup(sb, key, val, val_len, val_len, lock);
+	if (ret >= 0 && ret < val_len)
+		memset(val + ret, 0, val_len - ret);
+
+	return ret;
+}
+
 int scoutfs_item_lookup_exact(struct super_block *sb, struct scoutfs_key *key,
 			      void *val, int val_len,
 			      struct scoutfs_lock *lock)
 {
 	int ret;
 
-	ret = scoutfs_item_lookup(sb, key, val, val_len, lock);
+	ret = item_lookup(sb, key, val, val_len, 0, lock);
 	if (ret == val_len)
 		ret = 0;
 	else if (ret >= 0)
@@ -1793,7 +1832,7 @@ int scoutfs_item_next(struct super_block *sb, struct scoutfs_key *key,
 		goto out;
 	}
 
-	if ((ret = lock_safe(lock, key, SCOUTFS_LOCK_READ)))
+	if ((ret = lock_safe(sb, lock, key, SCOUTFS_LOCK_READ)))
 		goto out;
 
 	pos = *key;
@@ -1874,7 +1913,7 @@ int scoutfs_item_dirty(struct super_block *sb, struct scoutfs_key *key,
 
 	scoutfs_inc_counter(sb, item_dirty);
 
-	if ((ret = lock_safe(lock, key, SCOUTFS_LOCK_WRITE)))
+	if ((ret = lock_safe(sb, lock, key, SCOUTFS_LOCK_WRITE)))
 		goto out;
 
 	ret = scoutfs_forest_set_bloom_bits(sb, lock);
@@ -1920,7 +1959,7 @@ static int item_create(struct super_block *sb, struct scoutfs_key *key,
 
 	scoutfs_inc_counter(sb, item_create);
 
-	if ((ret = lock_safe(lock, key, mode)) ||
+	if ((ret = lock_safe(sb, lock, key, mode)) ||
 	    (ret = optional_lock_mode_match(primary, SCOUTFS_LOCK_WRITE)))
 		goto out;
 
@@ -1963,7 +2002,7 @@ int scoutfs_item_create(struct super_block *sb, struct scoutfs_key *key,
 			void *val, int val_len, struct scoutfs_lock *lock)
 {
 	return item_create(sb, key, val, val_len, lock, NULL,
-			   SCOUTFS_LOCK_READ, false);
+			   SCOUTFS_LOCK_WRITE, false);
 }
 
 int scoutfs_item_create_force(struct super_block *sb, struct scoutfs_key *key,
@@ -1994,7 +2033,7 @@ int scoutfs_item_update(struct super_block *sb, struct scoutfs_key *key,
 
 	scoutfs_inc_counter(sb, item_update);
 
-	if ((ret = lock_safe(lock, key, SCOUTFS_LOCK_WRITE)))
+	if ((ret = lock_safe(sb, lock, key, SCOUTFS_LOCK_WRITE)))
 		goto out;
 
 	ret = scoutfs_forest_set_bloom_bits(sb, lock);
@@ -2062,7 +2101,7 @@ int scoutfs_item_delta(struct super_block *sb, struct scoutfs_key *key,
 
 	scoutfs_inc_counter(sb, item_delta);
 
-	if ((ret = lock_safe(lock, key, SCOUTFS_LOCK_WRITE_ONLY)))
+	if ((ret = lock_safe(sb, lock, key, SCOUTFS_LOCK_WRITE_ONLY)))
 		goto out;
 
 	ret = scoutfs_forest_set_bloom_bits(sb, lock);
@@ -2135,7 +2174,7 @@ static int item_delete(struct super_block *sb, struct scoutfs_key *key,
 
 	scoutfs_inc_counter(sb, item_delete);
 
-	if ((ret = lock_safe(lock, key, mode)) ||
+	if ((ret = lock_safe(sb, lock, key, mode)) ||
 	    (ret = optional_lock_mode_match(primary, SCOUTFS_LOCK_WRITE)))
 		goto out;
 

kmod/src/item.h

@@ -3,6 +3,8 @@
 
 int scoutfs_item_lookup(struct super_block *sb, struct scoutfs_key *key,
 			void *val, int val_len, struct scoutfs_lock *lock);
+int scoutfs_item_lookup_smaller_zero(struct super_block *sb, struct scoutfs_key *key,
+				     void *val, int val_len, struct scoutfs_lock *lock);
 int scoutfs_item_lookup_exact(struct super_block *sb, struct scoutfs_key *key,
 			      void *val, int val_len,
 			      struct scoutfs_lock *lock);

kmod/src/lock.c

@@ -36,6 +36,8 @@
 #include "item.h"
 #include "omap.h"
 #include "util.h"
+#include "totl.h"
+#include "quota.h"
 
 /*
  * scoutfs uses a lock service to manage item cache consistency between
@@ -185,6 +187,9 @@ static int lock_invalidate(struct super_block *sb, struct scoutfs_lock *lock,
 			return ret;
 	}
 
+	if (lock->start.sk_zone == SCOUTFS_QUOTA_ZONE && !lock_mode_can_read(mode))
+		scoutfs_quota_invalidate(sb);
+
 	/* have to invalidate if we're not in the only usable case */
 	if (!(prev == SCOUTFS_LOCK_WRITE && mode == SCOUTFS_LOCK_READ)) {
 retry:
@@ -1244,10 +1249,29 @@ int scoutfs_lock_xattr_totl(struct super_block *sb, enum scoutfs_lock_mode mode,
 	struct scoutfs_key start;
 	struct scoutfs_key end;
 
-	scoutfs_key_set_zeros(&start);
-	start.sk_zone = SCOUTFS_XATTR_TOTL_ZONE;
-	scoutfs_key_set_ones(&end);
-	end.sk_zone = SCOUTFS_XATTR_TOTL_ZONE;
+	scoutfs_totl_set_range(&start, &end);
+
+	return lock_key_range(sb, mode, flags, &start, &end, lock);
+}
+
+int scoutfs_lock_xattr_indx(struct super_block *sb, enum scoutfs_lock_mode mode, int flags,
+			    struct scoutfs_lock **lock)
+{
+	struct scoutfs_key start;
+	struct scoutfs_key end;
+
+	scoutfs_xattr_indx_get_range(&start, &end);
+
+	return lock_key_range(sb, mode, flags, &start, &end, lock);
+}
+
+int scoutfs_lock_quota(struct super_block *sb, enum scoutfs_lock_mode mode, int flags,
+		       struct scoutfs_lock **lock)
+{
+	struct scoutfs_key start;
+	struct scoutfs_key end;
+
+	scoutfs_quota_get_lock_range(&start, &end);
 
 	return lock_key_range(sb, mode, flags, &start, &end, lock);
 }

kmod/src/lock.h

@@ -86,6 +86,10 @@ int scoutfs_lock_orphan(struct super_block *sb, enum scoutfs_lock_mode mode, int
 		        u64 ino, struct scoutfs_lock **lock);
 int scoutfs_lock_xattr_totl(struct super_block *sb, enum scoutfs_lock_mode mode, int flags,
 			    struct scoutfs_lock **lock);
+int scoutfs_lock_xattr_indx(struct super_block *sb, enum scoutfs_lock_mode mode, int flags,
+			    struct scoutfs_lock **lock);
+int scoutfs_lock_quota(struct super_block *sb, enum scoutfs_lock_mode mode, int flags,
+		       struct scoutfs_lock **lock);
 void scoutfs_unlock(struct super_block *sb, struct scoutfs_lock *lock,
 		    enum scoutfs_lock_mode mode);
 

kmod/src/quota.h

@@ -0,0 +1,48 @@
+#ifndef _SCOUTFS_QUOTA_H_
+#define _SCOUTFS_QUOTA_H_
+
+#include "ioctl.h"
+
+/*
+ * Each rule's name can be in the ruleset's rbtree associated with the
+ * source attr that it selects.  This lets checks only test rules that
+ * the inputs could match.  The 'i' field indicates which name is in the
+ * tree so we can find the containing rule.
+ *
+ * This is mostly private to quota.c but we expose it for tracing.
+ */
+struct squota_rule {
+	u64 limit;
+	u8 prio;
+	u8 op;
+	u8 rule_flags;
+	struct squota_rule_name {
+		struct rb_node node;
+		u64 val;
+		u8 source;
+		u8 flags;
+		u8 i;
+	} names[3];
+};
+
+/* private to quota.c, only here for tracing */
+struct squota_input {
+	u64 attrs[SQ_NS__NR_SELECT];
+	u8 op;
+};
+
+int scoutfs_quota_check_inode(struct super_block *sb, struct inode *dir);
+int scoutfs_quota_check_data(struct super_block *sb, struct inode *inode);
+
+int scoutfs_quota_get_rules(struct super_block *sb, u64 *iterator,
+			    struct scoutfs_ioctl_quota_rule *irules, int nr);
+int scoutfs_quota_mod_rule(struct super_block *sb, bool is_add,
+			   struct scoutfs_ioctl_quota_rule *irule);
+
+void scoutfs_quota_get_lock_range(struct scoutfs_key *start, struct scoutfs_key *end);
+void scoutfs_quota_invalidate(struct super_block *sb);
+
+int scoutfs_quota_setup(struct super_block *sb);
+void scoutfs_quota_destroy(struct super_block *sb);
+
+#endif

kmod/src/scoutfs_trace.h

@@ -37,6 +37,10 @@
 #include "net.h"
 #include "data.h"
 #include "ext.h"
+#include "quota.h"
+
+#include "trace/quota.h"
+#include "trace/wkic.h"
 
 struct lock_info;
 
@@ -2395,6 +2399,44 @@ TRACE_EVENT(scoutfs_block_dirty_ref,
 		  __entry->block_blkno, __entry->block_seq)
 );
 
+TRACE_EVENT(scoutfs_block_stale,
+	TP_PROTO(struct super_block *sb, struct scoutfs_block_ref *ref,
+		 struct scoutfs_block_header *hdr, u32 magic, u32 crc),
+
+	TP_ARGS(sb, ref, hdr, magic, crc),
+
+	TP_STRUCT__entry(
+		SCSB_TRACE_FIELDS
+		__field(__u64, ref_blkno)
+		__field(__u64, ref_seq)
+		__field(__u32, hdr_crc)
+		__field(__u32, hdr_magic)
+		__field(__u64, hdr_fsid)
+		__field(__u64, hdr_seq)
+		__field(__u64, hdr_blkno)
+		__field(__u32, magic)
+		__field(__u32, crc)
+	),
+
+	TP_fast_assign(
+		SCSB_TRACE_ASSIGN(sb);
+		__entry->ref_blkno = le64_to_cpu(ref->blkno);
+		__entry->ref_seq = le64_to_cpu(ref->seq);
+		__entry->hdr_crc = le32_to_cpu(hdr->crc);
+		__entry->hdr_magic = le32_to_cpu(hdr->magic);
+		__entry->hdr_fsid = le64_to_cpu(hdr->fsid);
+		__entry->hdr_seq = le64_to_cpu(hdr->seq);
+		__entry->hdr_blkno = le64_to_cpu(hdr->blkno);
+		__entry->magic = magic;
+		__entry->crc = crc;
+	),
+
+	TP_printk(SCSBF" ref_blkno %llu ref_seq %016llx hdr_crc %08x hdr_magic %08x hdr_fsid %016llx hdr_seq %016llx hdr_blkno %llu magic %08x crc %08x",
+		  SCSB_TRACE_ARGS, __entry->ref_blkno, __entry->ref_seq, __entry->hdr_crc,
+		  __entry->hdr_magic, __entry->hdr_fsid, __entry->hdr_seq, __entry->hdr_blkno,
+		  __entry->magic, __entry->crc)
+);
+
 DECLARE_EVENT_CLASS(scoutfs_block_class,
 	TP_PROTO(struct super_block *sb, void *bp, u64 blkno, int refcount, int io_count,
 		 unsigned long bits, __u64 accessed),

kmod/src/super.c

@@ -49,6 +49,8 @@
 #include "volopt.h"
 #include "fence.h"
 #include "xattr.h"
+#include "wkic.h"
+#include "quota.h"
 #include "scoutfs_trace.h"
 
 static struct dentry *scoutfs_debugfs_root;
@@ -194,7 +196,9 @@ static void scoutfs_put_super(struct super_block *sb)
 	scoutfs_shutdown_trans(sb);
 	scoutfs_volopt_destroy(sb);
 	scoutfs_client_destroy(sb);
+	scoutfs_quota_destroy(sb);
 	scoutfs_inode_destroy(sb);
+	scoutfs_wkic_destroy(sb);
 	scoutfs_item_destroy(sb);
 	scoutfs_forest_destroy(sb);
 	scoutfs_data_destroy(sb);
@@ -544,7 +548,9 @@ static int scoutfs_fill_super(struct super_block *sb, void *data, int silent)
 	      scoutfs_block_setup(sb) ?:
 	      scoutfs_forest_setup(sb) ?:
 	      scoutfs_item_setup(sb) ?:
+	      scoutfs_wkic_setup(sb) ?:
 	      scoutfs_inode_setup(sb) ?:
+	      scoutfs_quota_setup(sb) ?:
 	      scoutfs_data_setup(sb) ?:
 	      scoutfs_setup_trans(sb) ?:
 	      scoutfs_omap_setup(sb) ?:

kmod/src/super.h

@@ -30,6 +30,8 @@ struct recov_info;
 struct omap_info;
 struct volopt_info;
 struct fence_info;
+struct wkic_info;
+struct squota_info;
 
 struct scoutfs_sb_info {
 	struct super_block *sb;
@@ -55,6 +57,8 @@ struct scoutfs_sb_info {
 	struct omap_info *omap_info;
 	struct volopt_info *volopt_info;
 	struct item_cache_info *item_cache_info;
+	struct wkic_info *wkic_info;
+	struct squota_info *squota_info;
 	struct fence_info *fence_info;
 
 	/* tracks tasks waiting for data extents */
@@ -156,4 +160,17 @@ int scoutfs_write_super(struct super_block *sb,
 /* to keep this out of the ioctl.h public interface definition */
 long scoutfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 
+/*
+ * Returns 0 when supported, non-zero -errno when unsupported.
+ */
+static inline int scoutfs_fmt_vers_unsupported(struct super_block *sb, u64 vers)
+{
+	struct scoutfs_sb_info *sbi = SCOUTFS_SB(sb);
+
+	if (sbi && (sbi->fmt_vers < vers))
+		return -EOPNOTSUPP;
+	else
+		return 0;
+}
+
 #endif

kmod/src/totl.c

@@ -0,0 +1,90 @@
+/*
+ * Copyright (C) 2023 Versity Software, Inc.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+#include <linux/kernel.h>
+#include <linux/string.h>
+
+#include "format.h"
+#include "forest.h"
+#include "totl.h"
+
+void scoutfs_totl_set_range(struct scoutfs_key *start, struct scoutfs_key *end)
+{
+	scoutfs_key_set_zeros(start);
+	start->sk_zone = SCOUTFS_XATTR_TOTL_ZONE;
+	scoutfs_key_set_ones(end);
+	end->sk_zone = SCOUTFS_XATTR_TOTL_ZONE;
+}
+
+void scoutfs_totl_merge_init(struct scoutfs_totl_merging *merg)
+{
+	memset(merg, 0, sizeof(struct scoutfs_totl_merging));
+}
+
+void scoutfs_totl_merge_contribute(struct scoutfs_totl_merging *merg,
+				   u64 seq, u8 flags, void *val, int val_len, int fic)
+{
+	struct scoutfs_xattr_totl_val *tval = val;
+
+	if (fic & FIC_FS_ROOT) {
+		merg->fs_seq = seq;
+		merg->fs_total = le64_to_cpu(tval->total);
+		merg->fs_count = le64_to_cpu(tval->count);
+	} else if (fic & FIC_FINALIZED) {
+		merg->fin_seq = seq;
+		merg->fin_total += le64_to_cpu(tval->total);
+		merg->fin_count += le64_to_cpu(tval->count);
+	} else {
+		merg->log_seq = seq;
+		merg->log_total += le64_to_cpu(tval->total);
+		merg->log_count += le64_to_cpu(tval->count);
+	}
+}
+
+/*
+ * .totl. item merging has to be careful because the log btree merging
+ * code can write partial results to the fs_root.  This means that a
+ * reader can see both cases where new finalized logs should be applied
+ * to the old fs items and where old finalized logs have already been
+ * applied to the partially merged fs items.  Currently active logged
+ * items are always applied on top of all cases.
+ *
+ * These cases are differentiated with a combination of sequence numbers
+ * in items, the count of contributing xattrs, and a flag
+ * differentiating finalized and active logged items.  This lets us
+ * recognize all cases, including when finalized logs were merged and
+ * deleted the fs item.
+ */
+void scoutfs_totl_merge_resolve(struct scoutfs_totl_merging *merg, __u64 *total, __u64 *count)
+{
+	*total = 0;
+	*count = 0;
+
+	/* start with the fs item if we have it */
+	if (merg->fs_seq != 0) {
+		*total = merg->fs_total;
+		*count = merg->fs_count;
+	}
+
+	/* apply finalized logs if they're newer or creating */
+	if (((merg->fs_seq != 0) && (merg->fin_seq > merg->fs_seq)) ||
+	    ((merg->fs_seq == 0) && (merg->fin_count > 0))) {
+		*total += merg->fin_total;
+		*count += merg->fin_count;
+	}
+
+	/* always apply active logs which must be newer than fs and finalized */
+	if (merg->log_seq > 0) {
+		*total += merg->log_total;
+		*count += merg->log_count;
+	}
+}

kmod/src/totl.h

@@ -0,0 +1,24 @@
+#ifndef _SCOUTFS_TOTL_H_
+#define _SCOUTFS_TOTL_H_
+
+#include "key.h"
+
+struct scoutfs_totl_merging {
+	u64 fs_seq;
+	u64 fs_total;
+	u64 fs_count;
+	u64 fin_seq;
+	u64 fin_total;
+	s64 fin_count;
+	u64 log_seq;
+	u64 log_total;
+	s64 log_count;
+};
+
+void scoutfs_totl_set_range(struct scoutfs_key *start, struct scoutfs_key *end);
+void scoutfs_totl_merge_init(struct scoutfs_totl_merging *merg);
+void scoutfs_totl_merge_contribute(struct scoutfs_totl_merging *merg,
+				   u64 seq, u8 flags, void *val, int val_len, int fic);
+void scoutfs_totl_merge_resolve(struct scoutfs_totl_merging *merg, __u64 *total, __u64 *count);
+
+#endif

kmod/src/trace/quota.h

@@ -0,0 +1,143 @@
+
+/*
+ * Tracing squota_input
+ */
+#define SQI_FMT "[%u %llu %llu %llu]"
+
+#define SQI_ARGS(i)						\
+	(i)->op, (i)->attrs[0], (i)->attrs[1], (i)->attrs[2]
+
+#define SQI_FIELDS(pref)					\
+	__array(__u64, pref##_attrs, SQ_NS__NR_SELECT)		\
+	__field(__u8, pref##_op)
+
+#define SQI_ASSIGN(pref, i)					\
+	__entry->pref##_attrs[0] = (i)->attrs[0];		\
+	__entry->pref##_attrs[1] = (i)->attrs[1];		\
+	__entry->pref##_attrs[2] = (i)->attrs[2];		\
+	__entry->pref##_op = (i)->op;
+
+#define SQI_ENTRY_ARGS(pref)					\
+	__entry->pref##_op, __entry->pref##_attrs[0],		\
+	__entry->pref##_attrs[1], __entry->pref##_attrs[2]
+
+/*
+ * Tracing squota_rule
+ */
+#define SQR_FMT "[%u %llu,%u,%x %llu,%u,%x %llu,%u,%x %u %llu]"
+
+#define SQR_ARGS(r)							\
+	(r)->prio,							\
+	(r)->name_val[0], (r)->name_source[0], (r)->name_flags[0],	\
+	(r)->name_val[1], (r)->name_source[1], (r)->name_flags[1],	\
+	(r)->name_val[2], (r)->name_source[2], (r)->name_flags[2],	\
+	(r)->op, (r)->limit						\
+
+#define SQR_FIELDS(pref)			\
+	__array(__u64, pref##_name_val, 3)	\
+	__field(__u64, pref##_limit)		\
+	__array(__u8, pref##_name_source, 3)	\
+	__array(__u8, pref##_name_flags, 3)	\
+	__field(__u8, pref##_prio)		\
+	__field(__u8, pref##_op)
+
+#define SQR_ASSIGN(pref, r)					\
+	__entry->pref##_name_val[0] = (r)->names[0].val;	\
+	__entry->pref##_name_val[1] = (r)->names[1].val;	\
+	__entry->pref##_name_val[2] = (r)->names[2].val;	\
+	__entry->pref##_limit = (r)->limit;			\
+	__entry->pref##_name_source[0] = (r)->names[0].source;	\
+	__entry->pref##_name_source[1] = (r)->names[1].source;	\
+	__entry->pref##_name_source[2] = (r)->names[2].source;	\
+	__entry->pref##_name_flags[0] = (r)->names[0].flags;	\
+	__entry->pref##_name_flags[1] = (r)->names[1].flags;	\
+	__entry->pref##_name_flags[2] = (r)->names[2].flags;	\
+	__entry->pref##_prio = (r)->prio;			\
+	__entry->pref##_op = (r)->op;
+
+#define SQR_ENTRY_ARGS(pref)						\
+	__entry->pref##_prio, __entry->pref##_name_val[0],		\
+	__entry->pref##_name_source[0], __entry->pref##_name_flags[0],	\
+	__entry->pref##_name_val[1], __entry->pref##_name_source[1],	\
+	__entry->pref##_name_flags[1], __entry->pref##_name_val[2],	\
+	__entry->pref##_name_source[2], __entry->pref##_name_flags[2],	\
+	__entry->pref##_op, __entry->pref##_limit
+
+TRACE_EVENT(scoutfs_quota_check,
+	TP_PROTO(struct super_block *sb, long rs_ptr, struct squota_input *inp, int ret),
+
+	TP_ARGS(sb, rs_ptr, inp, ret),
+
+	TP_STRUCT__entry(
+		SCSB_TRACE_FIELDS
+		__field(long, rs_ptr)
+		SQI_FIELDS(i)
+		__field(int, ret)
+	),
+
+	TP_fast_assign(
+		SCSB_TRACE_ASSIGN(sb);
+		__entry->rs_ptr = rs_ptr;
+		SQI_ASSIGN(i, inp);
+		__entry->ret = ret;
+	),
+
+	TP_printk(SCSBF" rs_ptr %ld ret %d inp "SQI_FMT,
+		  SCSB_TRACE_ARGS, __entry->rs_ptr, __entry->ret, SQI_ENTRY_ARGS(i))
+);
+
+DECLARE_EVENT_CLASS(scoutfs_quota_rule_op_class,
+	TP_PROTO(struct super_block *sb, struct squota_rule *rule, int ret),
+
+	TP_ARGS(sb, rule, ret),
+
+	TP_STRUCT__entry(
+		SCSB_TRACE_FIELDS
+		SQR_FIELDS(r)
+		__field(int, ret)
+	),
+
+	TP_fast_assign(
+		SCSB_TRACE_ASSIGN(sb);
+		SQR_ASSIGN(r, rule);
+		__entry->ret = ret;
+	),
+
+	TP_printk(SCSBF" "SQR_FMT" ret %d",
+		  SCSB_TRACE_ARGS, SQR_ENTRY_ARGS(r), __entry->ret)
+);
+DEFINE_EVENT(scoutfs_quota_rule_op_class, scoutfs_quota_add_rule,
+	TP_PROTO(struct super_block *sb, struct squota_rule *rule, int ret),
+	TP_ARGS(sb, rule, ret)
+);
+DEFINE_EVENT(scoutfs_quota_rule_op_class, scoutfs_quota_del_rule,
+	TP_PROTO(struct super_block *sb, struct squota_rule *rule, int ret),
+	TP_ARGS(sb, rule, ret)
+);
+
+TRACE_EVENT(scoutfs_quota_totl_check,
+	TP_PROTO(struct super_block *sb, struct squota_input *inp, struct scoutfs_key *key,
+		 u64 limit, int ret),
+
+	TP_ARGS(sb, inp, key, limit, ret),
+
+	TP_STRUCT__entry(
+		SCSB_TRACE_FIELDS
+		SQI_FIELDS(i)
+		sk_trace_define(k)
+		__field(__u64, limit)
+		__field(int, ret)
+	),
+
+	TP_fast_assign(
+		SCSB_TRACE_ASSIGN(sb);
+		SQI_ASSIGN(i, inp);
+		sk_trace_assign(k, key);
+		__entry->limit = limit;
+		__entry->ret = ret;
+	),
+
+	TP_printk(SCSBF" inp "SQI_FMT" key "SK_FMT" limit %llu ret %d",
+		  SCSB_TRACE_ARGS, SQI_ENTRY_ARGS(i), sk_trace_args(k), __entry->limit,
+		  __entry->ret)
+);

kmod/src/trace/wkic.h

@@ -0,0 +1,112 @@
+
+DECLARE_EVENT_CLASS(scoutfs_wkic_wpage_class,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end),
+
+	TP_STRUCT__entry(
+		SCSB_TRACE_FIELDS
+		__field(void *, ptr)
+		__field(int, which)
+		__field(bool, n0l)
+		__field(bool, n1l)
+		sk_trace_define(start)
+		sk_trace_define(end)
+	),
+
+	TP_fast_assign(
+		SCSB_TRACE_ASSIGN(sb);
+		__entry->ptr = ptr;
+		__entry->which = which;
+		__entry->n0l = n0l;
+		__entry->n1l = n1l;
+		sk_trace_assign(start, start);
+		sk_trace_assign(end, end);
+		__entry->which = which;
+	),
+
+	TP_printk(SCSBF" ptr %p wh %d nl %u,%u start "SK_FMT " end "SK_FMT, SCSB_TRACE_ARGS,
+			__entry->ptr, __entry->which, __entry->n0l, __entry->n1l,
+			sk_trace_args(start), sk_trace_args(end))
+);
+
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_alloced,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_freeing,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_found,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_trimmed,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_erased,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_inserting,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_inserted,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_shrinking,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_dropping,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_replaying,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+DEFINE_EVENT(scoutfs_wkic_wpage_class, scoutfs_wkic_wpage_filled,
+	TP_PROTO(struct super_block *sb, void *ptr, int which, bool n0l, bool n1l,
+		 struct scoutfs_key *start, struct scoutfs_key *end),
+	TP_ARGS(sb, ptr, which, n0l, n1l, start, end)
+);
+
+TRACE_EVENT(scoutfs_wkic_read_items,
+	TP_PROTO(struct super_block *sb, struct scoutfs_key *key, struct scoutfs_key *start,
+		 struct scoutfs_key *end),
+
+	TP_ARGS(sb, key, start, end),
+
+	TP_STRUCT__entry(
+		SCSB_TRACE_FIELDS
+		sk_trace_define(key)
+		sk_trace_define(start)
+		sk_trace_define(end)
+	),
+
+	TP_fast_assign(
+		SCSB_TRACE_ASSIGN(sb);
+		sk_trace_assign(key, start);
+		sk_trace_assign(start, start);
+		sk_trace_assign(end, end);
+	),
+
+	TP_printk(SCSBF" key "SK_FMT" start "SK_FMT " end "SK_FMT, SCSB_TRACE_ARGS,
+			sk_trace_args(key), sk_trace_args(start), sk_trace_args(end))
+);

kmod/src/wkic.h

@@ -0,0 +1,19 @@
+#ifndef _SCOUTFS_WKIC_H_
+#define _SCOUTFS_WKIC_H_
+
+#include "format.h"
+
+typedef int (*wkic_iter_cb_t)(struct scoutfs_key *key, void *val, unsigned int val_len,
+			      void *cb_arg);
+
+int scoutfs_wkic_iterate(struct super_block *sb, struct scoutfs_key *key, struct scoutfs_key *last,
+			 struct scoutfs_key *range_start, struct scoutfs_key *range_end,
+			 wkic_iter_cb_t cb, void *cb_arg);
+int scoutfs_wkic_iterate_stable(struct super_block *sb, struct scoutfs_key *key,
+				struct scoutfs_key *last, struct scoutfs_key *range_start,
+				struct scoutfs_key *range_end, wkic_iter_cb_t cb, void *cb_arg);
+
+int scoutfs_wkic_setup(struct super_block *sb);
+void scoutfs_wkic_destroy(struct super_block *sb);
+
+#endif

kmod/src/xattr.c

@@ -81,7 +81,20 @@ static void init_xattr_key(struct scoutfs_key *key, u64 ino, u32 name_hash,
 #define SCOUTFS_XATTR_PREFIX		"scoutfs."
 #define SCOUTFS_XATTR_PREFIX_LEN	(sizeof(SCOUTFS_XATTR_PREFIX) - 1)
 
+/*
+ * We could have hidden the logic that needs this in a user-prefix
+ * specific .set handler, but I wanted to make sure that we always
+ * applied that logic from any call chains to _xattr_set.  The
+ * additional strcmp isn't so expensive given all the rest of the work
+ * we're doing in here.
+ */
+static inline bool is_user(const char *name)
+{
+	return !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
+}
+
 #define HIDE_TAG	"hide."
+#define INDX_TAG	"indx."
 #define SRCH_TAG	"srch."
 #define TOTL_TAG	"totl."
 #define TAG_LEN		(sizeof(HIDE_TAG) - 1)
@@ -103,6 +116,9 @@ int scoutfs_xattr_parse_tags(const char *name, unsigned int name_len,
 		if (!strncmp(name, HIDE_TAG, TAG_LEN)) {
 			if (++tgs->hide == 0)
 				return -EINVAL;
+		} else if (!strncmp(name, INDX_TAG, TAG_LEN)) {
+			if (++tgs->indx == 0)
+				return -EINVAL;
 		} else if (!strncmp(name, SRCH_TAG, TAG_LEN)) {
 			if (++tgs->srch == 0)
 				return -EINVAL;
@@ -540,47 +556,57 @@ static int parse_totl_u64(const char *s, int len, u64 *res)
 }
 
 /*
- * non-destructive relatively quick parse of the last 3 dotted u64s that
- * make up the name of the xattr total.  -EINVAL is returned if there
- * are anything but 3 valid u64 encodings between single dots at the end
- * of the name.
+ * non-destructive relatively quick parse of final dotted u64s in an
+ * xattr name.  If the required number of values are found then we
+ * return the number of bytes in the name that are not the final dotted
+ * u64s with their dots.  -EINVAL is returned if we didn't find the
+ * required number of values.
  */
-static int parse_totl_key(struct scoutfs_key *key, const char *name, int name_len)
+static int parse_dotted_u64s(u64 *u64s, int nr, const char *name, int name_len)
 {
-	u64 tot_name[3];
 	int end = name_len;
-	int nr = 0;
 	int len;
 	int ret;
 	int i;
+	int u;
 
 	/* parse name elements in reserve order from end of xattr name string */
-	for (i = name_len - 1; i >= 0 && nr < ARRAY_SIZE(tot_name); i--) {
+	for (u = nr - 1, i = name_len - 1; u >= 0 && i >= 0; i--) {
 		if (name[i] != '.')
 			continue;
 
 		len = end - (i + 1);
-		ret = parse_totl_u64(&name[i + 1], len, &tot_name[nr]);
+		ret = parse_totl_u64(&name[i + 1], len, &u64s[u]);
 		if (ret < 0)
 			goto out;
 
 		end = i;
-		nr++;
+		u--;
 	}
 
-	if (nr == ARRAY_SIZE(tot_name)) {
-		/* swap to account for parsing in reverse */
-		swap(tot_name[0], tot_name[2]);
-		scoutfs_xattr_init_totl_key(key, tot_name);
-		ret = 0;
-	} else {
+	if (u == -1)
+		ret = end;
+	else
 		ret = -EINVAL;
-	}
 
 out:
 	return ret;
 }
 
+static int parse_totl_key(struct scoutfs_key *key, const char *name, int name_len)
+{
+	u64 u64s[3];
+	int ret;
+
+	ret = parse_dotted_u64s(u64s, ARRAY_SIZE(u64s), name, name_len);
+	if (ret >= 0) {
+		scoutfs_xattr_init_totl_key(key, u64s);
+		ret = 0;
+	}
+
+	return ret;
+}
+
 static int apply_totl_delta(struct super_block *sb, struct scoutfs_key *key,
 			    struct scoutfs_xattr_totl_val *tval, struct scoutfs_lock *lock)
 {
@@ -607,6 +633,72 @@ int scoutfs_xattr_combine_totl(void *dst, int dst_len, void *src, int src_len)
 	return SCOUTFS_DELTA_COMBINED;
 }
 
+void scoutfs_xattr_indx_get_range(struct scoutfs_key *start, struct scoutfs_key *end)
+{
+	scoutfs_key_set_zeros(start);
+	start->sk_zone = SCOUTFS_XATTR_INDX_ZONE;
+	scoutfs_key_set_ones(end);
+	end->sk_zone = SCOUTFS_XATTR_INDX_ZONE;
+}
+
+/*
+ * .indx. keys are a bit funny because we're iterating over index keys
+ * by major:minor:inode:xattr_id.  That doesn't map nicely to the
+ * comparison precedence of the key fields.  We have to mess around a
+ * little bit to get the major into the most significant key bits and
+ * the low bits of xattr id into the least significant key bits.
+ */
+void scoutfs_xattr_init_indx_key(struct scoutfs_key *key, u8 major, u64 minor, u64 ino, u64 xid)
+{
+	scoutfs_key_set_zeros(key);
+	key->sk_zone = SCOUTFS_XATTR_INDX_ZONE;
+
+	key->_sk_first = cpu_to_le64(((u64)major << 56) | (minor >> 8));
+	key->_sk_second = cpu_to_le64((minor << 56) | (ino >> 8));
+	key->_sk_third = cpu_to_le64((ino << 56) | (xid >> 8));
+	key->_sk_fourth = xid & 0xff;
+}
+
+void scoutfs_xattr_get_indx_key(struct scoutfs_key *key, u8 *major, u64 *minor, u64 *ino, u64 *xid)
+{
+	*major = le64_to_cpu(key->_sk_first) >> 56;
+	*minor = (le64_to_cpu(key->_sk_first) << 8) | (le64_to_cpu(key->_sk_second) >> 56);
+	*ino = (le64_to_cpu(key->_sk_second) << 8) | (le64_to_cpu(key->_sk_third) >> 56);
+	*xid = (le64_to_cpu(key->_sk_third) << 8) | key->_sk_fourth;
+}
+
+void scoutfs_xattr_set_indx_key_xid(struct scoutfs_key *key, u64 xid)
+{
+	u8 major;
+	u64 minor;
+	u64 ino;
+	u64 dummy;
+
+	scoutfs_xattr_get_indx_key(key, &major, &minor, &ino, &dummy);
+	scoutfs_xattr_init_indx_key(key, major, minor, ino, xid);
+}
+
+/*
+ * This initial parsing of the name doesn't yet have access to an xattr
+ * id to put in the key.  That's added later as the existing xattr is
+ * found or a new xattr's id is allocated.
+ */
+static int parse_indx_key(struct scoutfs_key *key, const char *name, int name_len, u64 ino)
+{
+	u64 u64s[2];
+	int ret;
+
+	ret = parse_dotted_u64s(u64s, ARRAY_SIZE(u64s), name, name_len);
+	if (ret < 0)
+		return ret;
+
+	if (u64s[0] > U8_MAX)
+		return -EINVAL;
+
+	scoutfs_xattr_init_indx_key(key, u64s[0], u64s[1], ino, 0);
+	return 0;
+}
+
 /*
  * The confusing swiss army knife of creating, modifying, and deleting
  * xattrs.
@@ -627,7 +719,7 @@ int scoutfs_xattr_combine_totl(void *dst, int dst_len, void *src, int src_len)
 int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_len,
 			     const void *value, size_t size, int flags,
 			     const struct scoutfs_xattr_prefix_tags *tgs,
-			     struct scoutfs_lock *lck, struct scoutfs_lock *totl_lock,
+			     struct scoutfs_lock *lck, struct scoutfs_lock *tag_lock,
 			     struct list_head *ind_locks)
 {
 	struct scoutfs_inode_info *si = SCOUTFS_I(inode);
@@ -635,10 +727,11 @@ int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_
 	const u64 ino = scoutfs_ino(inode);
 	struct scoutfs_xattr_totl_val tval = {0,};
 	struct scoutfs_xattr *xat = NULL;
-	struct scoutfs_key totl_key;
+	struct scoutfs_key tag_key;
 	struct scoutfs_key key;
 	bool undo_srch = false;
 	bool undo_totl = false;
+	bool undo_indx = false;
 	u8 found_parts;
 	unsigned int xat_bytes_totl;
 	unsigned int xat_bytes;
@@ -651,7 +744,8 @@ int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_
 
 	trace_scoutfs_xattr_set(sb, name_len, value, size, flags);
 
-	if (WARN_ON_ONCE(tgs->totl && !totl_lock))
+	if (WARN_ON_ONCE(tgs->totl && tgs->indx) ||
+	    WARN_ON_ONCE((tgs->totl | tgs->indx) && !tag_lock))
 		return -EINVAL;
 
 	/* mirror the syscall's errors for large names and values */
@@ -664,10 +758,22 @@ int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_
 	    (flags & ~(XATTR_CREATE | XATTR_REPLACE)))
 		return -EINVAL;
 
-	if ((tgs->hide | tgs->srch | tgs->totl) && !capable(CAP_SYS_ADMIN))
+	if ((tgs->hide | tgs->indx | tgs->srch | tgs->totl) && !capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	if (tgs->totl && ((ret = parse_totl_key(&totl_key, name, name_len)) != 0))
+	if (tgs->totl && ((ret = parse_totl_key(&tag_key, name, name_len)) != 0))
+		return ret;
+
+	if (tgs->indx &&
+	    (ret = scoutfs_fmt_vers_unsupported(sb, SCOUTFS_FORMAT_VERSION_FEAT_INDX_TAG)))
+		return ret;
+
+	if (tgs->indx && ((ret = parse_indx_key(&tag_key, name, name_len, ino)) != 0))
+		return ret;
+
+	/* retention blocks user. xattr modification, all else allowed */
+	ret = scoutfs_inode_check_retention(inode);
+	if (ret < 0 && is_user(name))
 		return ret;
 
 	/* allocate enough to always read an existing xattr's totl */
@@ -708,6 +814,12 @@ int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_
 	/* found fields in key will also be used */
 	found_parts = ret >= 0 ? xattr_nr_parts(xat) : 0;
 
+	/* use existing xattr's id or allocate new when creating */
+	if (found_parts)
+		id = le64_to_cpu(key.skx_id);
+	else if (value)
+		id = si->next_xattr_id++;
+
 	if (found_parts && tgs->totl) {
 		/* parse old totl value before we clobber xat buf */
 		val_len = ret - offsetof(struct scoutfs_xattr, name[xat->name_len]);
@@ -718,12 +830,25 @@ int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_
 		le64_add_cpu(&tval.total, -total);
 	}
 
+	/*
+	 * indx xattrs don't have a value.  After returning an error for
+	 * non-zero val length or short circuiting modifying with the
+	 * same 0 length, all we're left with is creating or deleting
+	 * the xattr.
+	 */
+	if (tgs->indx) {
+		if (size != 0) {
+			ret = -EINVAL;
+			goto out;
+		}
+		if (found_parts && value) {
+			ret = 0;
+			goto out;
+		}
+	}
+
 	/* prepare the xattr header, name, and start of value in first item */
 	if (value) {
-		if (found_parts)
-			id = le64_to_cpu(key.skx_id);
-		else
-			id = si->next_xattr_id++;
 		xat->name_len = name_len;
 		xat->val_len = cpu_to_le16(size);
 		memset(xat->__pad, 0, sizeof(xat->__pad));
@@ -741,9 +866,18 @@ int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_
 		le64_add_cpu(&tval.total, total);
 	}
 
+	if (tgs->indx) {
+		scoutfs_xattr_set_indx_key_xid(&tag_key, id);
+		if (value)
+			ret = scoutfs_item_create_force(sb, &tag_key, NULL, 0, tag_lock, NULL);
+		else
+			ret = scoutfs_item_delete_force(sb, &tag_key, tag_lock, NULL);
+		if (ret < 0)
+			goto out;
+		undo_indx = true;
+	}
+
 	if (tgs->srch && !(found_parts && value)) {
-		if (found_parts)
-			id = le64_to_cpu(key.skx_id);
 		hash = scoutfs_hash64(name, name_len);
 		ret = scoutfs_forest_srch_add(sb, hash, ino, id);
 		if (ret < 0)
@@ -752,7 +886,7 @@ int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_
 	}
 
 	if (tgs->totl) {
-		ret = apply_totl_delta(sb, &totl_key, &tval, totl_lock);
+		ret = apply_totl_delta(sb, &tag_key, &tval, tag_lock);
 		if (ret < 0)
 			goto out;
 		undo_totl = true;
@@ -777,6 +911,13 @@ int scoutfs_xattr_set_locked(struct inode *inode, const char *name, size_t name_
 	ret = 0;
 
 out:
+	if (ret < 0 && undo_indx) {
+		if (value)
+			err = scoutfs_item_delete_force(sb, &tag_key, tag_lock, NULL);
+		else
+			err = scoutfs_item_create_force(sb, &tag_key, NULL, 0, tag_lock, NULL);
+		BUG_ON(err); /* inconsistent */
+	}
 	if (ret < 0 && undo_srch) {
 		err = scoutfs_forest_srch_add(sb, hash, ino, id);
 		BUG_ON(err);
@@ -785,7 +926,7 @@ out:
 		/* _delta() on dirty items shouldn't fail */
 		tval.total = cpu_to_le64(-le64_to_cpu(tval.total));
 		tval.count = cpu_to_le64(-le64_to_cpu(tval.count));
-		err = apply_totl_delta(sb, &totl_key, &tval, totl_lock);
+		err = apply_totl_delta(sb, &tag_key, &tval, tag_lock);
 		BUG_ON(err);
 	}
 
@@ -801,7 +942,7 @@ static int scoutfs_xattr_set(struct dentry *dentry, const char *name, const void
 	struct inode *inode = dentry->d_inode;
 	struct super_block *sb = inode->i_sb;
 	struct scoutfs_xattr_prefix_tags tgs;
-	struct scoutfs_lock *totl_lock = NULL;
+	struct scoutfs_lock *tag_lock = NULL;
 	struct scoutfs_lock *lck = NULL;
 	size_t name_len = strlen(name);
 	LIST_HEAD(ind_locks);
@@ -816,8 +957,11 @@ static int scoutfs_xattr_set(struct dentry *dentry, const char *name, const void
 	if (ret)
 		goto unlock;
 
-	if (tgs.totl) {
-		ret = scoutfs_lock_xattr_totl(sb, SCOUTFS_LOCK_WRITE_ONLY, 0, &totl_lock);
+	if (tgs.totl || tgs.indx) {
+		if (tgs.totl)
+			ret = scoutfs_lock_xattr_totl(sb, SCOUTFS_LOCK_WRITE_ONLY, 0, &tag_lock);
+		else
+			ret = scoutfs_lock_xattr_indx(sb, SCOUTFS_LOCK_WRITE_ONLY, 0, &tag_lock);
 		if (ret)
 			goto unlock;
 	}
@@ -836,7 +980,7 @@ retry:
 		goto release;
 
 	ret = scoutfs_xattr_set_locked(dentry->d_inode, name, name_len, value, size, flags, &tgs,
-				       lck, totl_lock, &ind_locks);
+				       lck, tag_lock, &ind_locks);
 	if (ret == 0)
 		scoutfs_update_inode_item(inode, lck, &ind_locks);
 
@@ -845,7 +989,7 @@ release:
 	scoutfs_inode_index_unlock(sb, &ind_locks);
 unlock:
 	scoutfs_unlock(sb, lck, SCOUTFS_LOCK_WRITE);
-	scoutfs_unlock(sb, totl_lock, SCOUTFS_LOCK_WRITE_ONLY);
+	scoutfs_unlock(sb, tag_lock, SCOUTFS_LOCK_WRITE_ONLY);
 
 	return ret;
 }
@@ -1055,14 +1199,15 @@ int scoutfs_xattr_drop(struct super_block *sb, u64 ino,
 {
 	struct scoutfs_xattr_prefix_tags tgs;
 	struct scoutfs_xattr *xat = NULL;
-	struct scoutfs_lock *totl_lock = NULL;
+	struct scoutfs_lock *tag_lock = NULL;
 	struct scoutfs_xattr_totl_val tval;
-	struct scoutfs_key totl_key;
+	struct scoutfs_key tag_key;
 	struct scoutfs_key last;
 	struct scoutfs_key key;
 	bool release = false;
 	unsigned int bytes;
 	unsigned int val_len;
+	u8 locked_zone = 0;
 	void *value;
 	u64 total;
 	u64 hash;
@@ -1108,16 +1253,32 @@ int scoutfs_xattr_drop(struct super_block *sb, u64 ino,
 				goto out;
 			}
 
-			ret = parse_totl_key(&totl_key, xat->name, xat->name_len) ?:
+			ret = parse_totl_key(&tag_key, xat->name, xat->name_len) ?:
 			      parse_totl_u64(value, val_len, &total);
 			if (ret < 0)
 				break;
 		}
 
-		if (tgs.totl && totl_lock == NULL) {
-			ret = scoutfs_lock_xattr_totl(sb, SCOUTFS_LOCK_WRITE_ONLY, 0, &totl_lock);
+		if (tgs.indx) {
+			ret = parse_indx_key(&tag_key, xat->name, xat->name_len, ino);
+			if (ret < 0)
+				goto out;
+		}
+
+		if ((tgs.totl || tgs.indx) && locked_zone != tag_key.sk_zone) {
+			if (tag_lock) {
+				scoutfs_unlock(sb, tag_lock, SCOUTFS_LOCK_WRITE_ONLY);
+				tag_lock = NULL;
+			}
+			if (tgs.totl)
+				ret = scoutfs_lock_xattr_totl(sb, SCOUTFS_LOCK_WRITE_ONLY, 0,
+							      &tag_lock);
+			else
+				ret = scoutfs_lock_xattr_indx(sb, SCOUTFS_LOCK_WRITE_ONLY, 0,
+							      &tag_lock);
 			if (ret < 0)
 				break;
+			locked_zone = tag_key.sk_zone;
 		}
 
 		ret = scoutfs_hold_trans(sb, false);
@@ -1140,7 +1301,13 @@ int scoutfs_xattr_drop(struct super_block *sb, u64 ino,
 		if (tgs.totl) {
 			tval.total = cpu_to_le64(-total);
 			tval.count = cpu_to_le64(-1LL);
-			ret = apply_totl_delta(sb, &totl_key, &tval, totl_lock);
+			ret = apply_totl_delta(sb, &tag_key, &tval, tag_lock);
+			if (ret < 0)
+				break;
+		}
+
+		if (tgs.indx) {
+			ret = scoutfs_item_delete_force(sb, &tag_key, tag_lock, NULL);
 			if (ret < 0)
 				break;
 		}
@@ -1153,7 +1320,7 @@ int scoutfs_xattr_drop(struct super_block *sb, u64 ino,
 
 	if (release)
 		scoutfs_release_trans(sb);
-	scoutfs_unlock(sb, totl_lock, SCOUTFS_LOCK_WRITE_ONLY);
+	scoutfs_unlock(sb, tag_lock, SCOUTFS_LOCK_WRITE_ONLY);
 	kfree(xat);
 out:
 	return ret;

kmod/src/xattr.h

@@ -3,6 +3,7 @@
 
 struct scoutfs_xattr_prefix_tags {
 	unsigned long hide:1,
+		      indx:1,
 		      srch:1,
 		      totl:1;
 };
@@ -30,4 +31,9 @@ int scoutfs_xattr_parse_tags(const char *name, unsigned int name_len,
 void scoutfs_xattr_init_totl_key(struct scoutfs_key *key, u64 *name);
 int scoutfs_xattr_combine_totl(void *dst, int dst_len, void *src, int src_len);
 
+void scoutfs_xattr_indx_get_range(struct scoutfs_key *start, struct scoutfs_key *end);
+void scoutfs_xattr_init_indx_key(struct scoutfs_key *key, u8 major, u64 minor, u64 ino, u64 xid);
+void scoutfs_xattr_get_indx_key(struct scoutfs_key *key, u8 *major, u64 *minor, u64 *ino, u64 *xid);
+void scoutfs_xattr_set_indx_key_xid(struct scoutfs_key *key, u64 xid);
+
 #endif

tests/.gitignore

@@ -9,3 +9,4 @@ src/find_xattrs
 src/stage_tmpfile
 src/create_xattr_loop
 src/o_tmpfile_umask
+src/o_tmpfile_linkat

tests/Makefile

@@ -12,7 +12,8 @@ BIN := src/createmany			\
 	src/find_xattrs			\
 	src/create_xattr_loop		\
 	src/fragmented_data_extents	\
-	src/o_tmpfile_umask
+	src/o_tmpfile_umask		\
+	src/o_tmpfile_linkat
 
 DEPS := $(wildcard src/*.d)
 

tests/README.md

@@ -113,6 +113,7 @@ used during the test.
 | T\_EX\_META\_DEV | scratch meta bdev    | -f              | /dev/vdd          |
 | T\_EX\_DATA\_DEV | scratch meta bdev    | -e              | /dev/vdc          |
 | T\_M[0-9]        | mount paths          | mounted per run | /mnt/test.[0-9]/  |
+| T\_MODULE        | built kernel module  | created per run | ../kmod/src/..ko  |
 | T\_NR\_MOUNTS    | number of mounts     | -n              | 3                 |
 | T\_O[0-9]        | mount options        | created per run | -o server\_addr=  |
 | T\_QUORUM        | quorum count         | -q              | 2                 |

tests/funcs/filter.sh

@@ -147,6 +147,10 @@ t_filter_dmesg()
 	# ignore systemd-journal rotating
 	re="$re|systemd-journald.*"
 
+	# format vers back/compat tries bad mounts
+	re="$re|scoutfs .* error.*outside of supported version.*"
+	re="$re|scoutfs .* error.*could not get .*super.*"
+
 	egrep -v "($re)" | \
 		ignore_harmless_unwind_kasan_stack_oob
 }

tests/funcs/fs.sh

@@ -29,13 +29,12 @@ t_mount_rid()
 }
 
 #
-# Output the "f.$fsid.r.$rid" identifier string for the given mount
-# number, 0 is used by default if none is specified. 
+# Output the "f.$fsid.r.$rid" identifier string for the given path
+# in a mounted scoutfs volume.
 #
-t_ident()
+t_ident_from_mnt()
 {
-	local nr="${1:-0}"
-	local mnt="$(eval echo \$T_M$nr)"
+	local mnt="$1"
 	local fsid
 	local rid
 
@@ -45,6 +44,38 @@ t_ident()
 	echo "f.${fsid:0:6}.r.${rid:0:6}"
 }
 
+#
+# Output the "f.$fsid.r.$rid" identifier string for the given mount
+# number, 0 is used by default if none is specified.
+#
+t_ident()
+{
+	local nr="${1:-0}"
+	local mnt="$(eval echo \$T_M$nr)"
+
+	t_ident_from_mnt "$mnt"
+}
+
+#
+# Output the sysfs path for a path in a mounted fs.
+#
+t_sysfs_path_from_ident()
+{
+	local ident="$1"
+
+	echo "/sys/fs/scoutfs/$ident"
+}
+
+#
+# Output the sysfs path for a path in a mounted fs.
+#
+t_sysfs_path_from_mnt()
+{
+	local mnt="$1"
+
+	t_sysfs_path_from_ident $(t_ident_from_mnt $mnt)
+}
+
 #
 # Output the mount's sysfs path, defaulting to mount 0 if none is
 # specified.
@@ -53,7 +84,7 @@ t_sysfs_path()
 {
 	local nr="$1"
 
-	echo "/sys/fs/scoutfs/$(t_ident $nr)"
+	t_sysfs_path_from_ident $(t_ident $nr)
 }
 
 #

tests/golden/basic-posix-acl

@@ -0,0 +1,157 @@
+== setup test directory
+== getfacl
+directory drwxr-xr-x 0 0 0 '.'
+# file: .
+# owner: root
+# group: root
+user::rwx
+group::r-x
+other::r-x
+
+== basic non-acl access through permissions
+directory drwxr-xr-x 0 44444 0 'dir-testuid'
+touch: cannot touch 'dir-testuid/file-group-write': Permission denied
+touch: cannot touch 'symlinkdir-testuid/symlink-file-group-write': Permission denied
+regular empty file -rw-r--r-- 22222 44444 0 'dir-testuid/file-group-write'
+regular empty file -rw-r--r-- 22222 44444 0 'symlinkdir-testuid/symlink-file-group-write'
+== basic acl access
+directory drwxr-xr-x 0 0 0 'dir-root'
+touch: cannot touch 'dir-root/file-group-write': Permission denied
+touch: cannot touch 'symlinkdir-root/file-group-write': Permission denied
+# file: dir-root
+# owner: root
+# group: root
+user::rwx
+user:22222:rwx
+group::r-x
+mask::rwx
+other::r-x
+
+regular empty file -rw-r--r-- 22222 0 0 'dir-root/file-group-write'
+regular empty file -rw-r--r-- 22222 0 0 'symlinkdir-root/file-group-write'
+== directory exec
+Success
+Success
+# file: dir-root
+# owner: root
+# group: root
+user::rwx
+user:22222:rw-
+group::r-x
+mask::rwx
+other::r-x
+
+bash: line 0: cd: dir-root: Permission denied
+Failed
+bash: line 0: cd: symlinkdir-root: Permission denied
+Failed
+# file: dir-root
+# owner: root
+# group: root
+user::rwx
+user:22222:rw-
+group::r-x
+group:44444:rwx
+mask::rwx
+other::r-x
+
+Success
+Success
+== get/set attr
+regular empty file -rw-r--r-- 0 0 0 'file-root'
+setfattr: file-root: Permission denied
+# file: file-root
+# owner: root
+# group: root
+user::rw-
+user:22222:rw-
+group::r--
+mask::rw-
+other::r--
+
+# file: file-root
+user.test2="Success"
+
+# file: file-root
+# owner: root
+# group: root
+user::rw-
+group::r--
+mask::r--
+other::r--
+
+setfattr: file-root: Permission denied
+# file: file-root
+user.test2="Success"
+
+# file: file-root
+# owner: root
+# group: root
+user::rw-
+group::r--
+group:44444:rw-
+mask::rw-
+other::r--
+
+# file: file-root
+user.test2="Success"
+user.test4="Success"
+
+== inheritance / default acl
+directory drwxr-xr-x 0 0 0 'dir-root2'
+mkdir: cannot create directory 'dir-root2/dir': Permission denied
+touch: cannot touch 'dir-root2/dir/file': No such file or directory
+# file: dir-root2
+# owner: root
+# group: root
+user::rwx
+group::r-x
+other::r-x
+default:user::rwx
+default:user:22222:rwx
+default:group::r-x
+default:mask::rwx
+default:other::r-x
+
+mkdir: cannot create directory 'dir-root2/dir': Permission denied
+touch: cannot touch 'dir-root2/dir/file': No such file or directory
+# file: dir-root2
+# owner: root
+# group: root
+user::rwx
+user:22222:rwx
+group::r-x
+mask::rwx
+other::r-x
+default:user::rwx
+default:user:22222:rwx
+default:group::r-x
+default:mask::rwx
+default:other::r-x
+
+directory drwxrwxr-x 22222 0 4 'dir-root2/dir'
+# file: dir-root2/dir
+# owner: 22222
+# group: root
+user::rwx
+user:22222:rwx
+group::r-x
+mask::rwx
+other::r-x
+default:user::rwx
+default:user:22222:rwx
+default:group::r-x
+default:mask::rwx
+default:other::r-x
+
+regular empty file -rw-rw-r-- 22222 0 0 'dir-root2/dir/file'
+# file: dir-root2/dir/file
+# owner: 22222
+# group: root
+user::rw-
+user:22222:rwx	#effective:rw-
+group::r-x	#effective:r--
+mask::rw-
+other::r--
+
+== cleanup

tests/golden/basic-posix-consistency

@@ -56,3 +56,4 @@ mv: cannot move '/mnt/test/test/basic-posix-consistency/dir/c/clobber' to '/mnt/
 == inode indexes match after removing and syncing
 == concurrent creates make one file
 one-file
+== cleanup

tests/golden/change-devices

@@ -25,3 +25,4 @@ rc: 0
 equal_prepared
 large_prepared
 resized larger test rc: 0
+== cleanup

tests/golden/format-version-forward-back

@@ -0,0 +1,4 @@
+== ensuring utils and module for old versions
+== unmounting test fs and removing test module
+== testing combinations of old and new format versions
+== restoring test module and mount

tests/golden/projects

@@ -0,0 +1,24 @@
+== default new files don't have project
+0
+== set new project on files and dirs
+8675309
+8675309
+== non-root can see id
+8675309
+== can use IDs around long width limits
+2147483647
+2147483648
+4294967295
+9223372036854775807
+9223372036854775808
+18446744073709551615
+== created files and dirs inherit project id
+8675309
+8675309
+== inheritance continues
+8675309
+== clearing project id stops inheritance
+0
+0
+== o_tmpfile creations inherit dir
+8675309

tests/golden/quota

@@ -0,0 +1,41 @@
+== prepare dir with write perm for test ids
+== test assumes starting with no rules, empty list
+== add rule
+  7 13,L,- 15,L,- 17,L,- I 33 -
+== list is empty again after delete
+== can change limits without deleting
+  1 1,L,- 1,L,- 1,L,- I 100 -
+  1 1,L,- 1,L,- 1,L,- I 101 -
+  1 1,L,- 1,L,- 1,L,- I 99 -
+== wipe and restore rules in bulk
+  7 15,L,- 0,L,- 0,L,- I 33 -
+  7 14,L,- 0,L,- 0,L,- I 33 -
+  7 13,L,- 0,L,- 0,L,- I 33 -
+  7 12,L,- 0,L,- 0,L,- I 33 -
+  7 11,L,- 0,L,- 0,L,- I 33 -
+  7 10,L,- 0,L,- 0,L,- I 33 -
+  7 15,L,- 0,L,- 0,L,- I 33 -
+  7 14,L,- 0,L,- 0,L,- I 33 -
+  7 13,L,- 0,L,- 0,L,- I 33 -
+  7 12,L,- 0,L,- 0,L,- I 33 -
+  7 11,L,- 0,L,- 0,L,- I 33 -
+  7 10,L,- 0,L,- 0,L,- I 33 -
+== default rule prevents file creation
+touch: cannot touch '/mnt/test/test/quota/dir/file': Disk quota exceeded
+== decreasing totl allows file creation again
+== attr selecting rules prevent creation
+touch: cannot touch '/mnt/test/test/quota/dir/file': Disk quota exceeded
+touch: cannot touch '/mnt/test/test/quota/dir/file': Disk quota exceeded
+== multi attr selecting doesn't prevent partial
+touch: cannot touch '/mnt/test/test/quota/dir/file': Disk quota exceeded
+== op differentiates
+== higher priority rule applies
+touch: cannot touch '/mnt/test/test/quota/dir/file': Disk quota exceeded
+== data rules with total and count prevent write and fallocate
+dd: error writing '/mnt/test/test/quota/dir/file': Disk quota exceeded
+fallocate: fallocate failed: Disk quota exceeded
+dd: error writing '/mnt/test/test/quota/dir/file': Disk quota exceeded
+fallocate: fallocate failed: Disk quota exceeded
+== added rules work after bulk restore
+touch: cannot touch '/mnt/test/test/quota/dir/file': Disk quota exceeded
+== cleanup

tests/golden/retention-basic

@@ -0,0 +1,28 @@
+== setting retention on dir fails
+attr_x ioctl failed on '/mnt/test/test/retention-basic': Invalid argument (22)
+scoutfs: set-attr-x failed: Invalid argument (22)
+== set retention
+== get-attr-x shows retention
+1
+== unpriv can't clear retention
+attr_x ioctl failed on '/mnt/test/test/retention-basic/file-1': Operation not permitted (1)
+scoutfs: set-attr-x failed: Operation not permitted (1)
+== can set hidden scoutfs xattr in retention
+== setting user. xattr fails in retention
+setfattr: /mnt/test/test/retention-basic/file-1: Operation not permitted
+== file deletion fails in retention
+rm: cannot remove '/mnt/test/test/retention-basic/file-1': Operation not permitted
+== file rename fails in retention
+mv: cannot move '/mnt/test/test/retention-basic/file-1' to '/mnt/test/test/retention-basic/file-2': Operation not permitted
+== file write fails in retention
+date: write error: Operation not permitted
+== file truncate fails in retention
+truncate: failed to truncate '/mnt/test/test/retention-basic/file-1' at 0 bytes: Operation not permitted
+== setattr fails in retention
+touch: setting times of '/mnt/test/test/retention-basic/file-1': Operation not permitted
+== clear retention
+== file write
+== file rename
+== setattr
+== xattr deletion
+== cleanup

tests/golden/stage-release-race-alloc

@@ -1,2 +1,3 @@
 == create initial files
 == race stage and release
+== cleanup

tests/run-tests.sh

@@ -73,6 +73,7 @@ $(basename $0) options:
     -t        | Enabled trace events that match the given glob argument.
               | Multiple options enable multiple globbed events.
     -T <nr>   | Multiply the original trace buffer size by nr during the run.
+    -V <nr>   | Set mkfs device format version.
     -X        | xfstests git repo. Used by tests/xfstests.sh.
     -x        | xfstests git branch to checkout and track.
     -y        | xfstests ./check additional args
@@ -176,6 +177,11 @@ while true; do
 		T_TRACE_MULT="$2"
 		shift
 		;;
+	-V)
+		test -n "$2" || die "-V must have a format version argument"
+		T_MKFS_FORMAT_VERSION="-V $2"
+		shift
+		;;
 	-X)
 		test -n "$2" || die "-X requires xfstests git repo dir argument"
 		T_XFSTESTS_REPO="$2"
@@ -344,7 +350,7 @@ if [ -n "$T_MKFS" ]; then
 	done
 
 	msg "making new filesystem with $T_QUORUM quorum members"
-	cmd scoutfs mkfs -f $quo $T_DATA_ALLOC_ZONE_BLOCKS \
+	cmd scoutfs mkfs -f $quo $T_DATA_ALLOC_ZONE_BLOCKS $T_MKFS_FORMAT_VERSION \
 		"$T_META_DEVICE" "$T_DATA_DEVICE"
 fi
 
@@ -352,7 +358,8 @@ if [ -n "$T_INSMOD" ]; then
 	msg "removing and reinserting scoutfs module"
 	test -e /sys/module/scoutfs && cmd rmmod scoutfs
 	cmd modprobe libcrc32c
-	cmd insmod "$T_KMOD/src/scoutfs.ko"
+	T_MODULE="$T_KMOD/src/scoutfs.ko"
+	cmd insmod "$T_MODULE"
 fi
 
 if [ -n "$T_TRACE_MULT" ]; then

tests/sequence

@@ -1,6 +1,7 @@
 export-get-name-parent.sh
 basic-block-counts.sh
 basic-bad-mounts.sh
+basic-posix-acl.sh
 inode-items-updated.sh
 simple-inode-index.sh
 simple-staging.sh
@@ -12,12 +13,16 @@ data-prealloc.sh
 setattr_more.sh
 offline-extent-waiting.sh
 move-blocks.sh
+projects.sh
 large-fragmented-free.sh
+format-version-forward-back.sh
 enospc.sh
 srch-safe-merge-pos.sh
 srch-basic-functionality.sh
 simple-xattr-unit.sh
+retention-basic.sh
 totl-xattr-tag.sh
+quota.sh
 lock-refleak.sh
 lock-shrink-consistency.sh
 lock-pr-cw-conflict.sh

tests/src/o_tmpfile_linkat.c

@@ -0,0 +1,71 @@
+/*
+ * Copyright (C) 2023 Versity Software, Inc.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+#include <unistd.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ioctl.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <assert.h>
+#include <limits.h>
+
+static void linkat_tmpfile(char *dir, char *lpath)
+{
+	char proc_self[PATH_MAX];
+	int ret;
+	int fd;
+
+	fd = open(dir, O_RDWR | O_TMPFILE, 0777);
+	if (fd < 0) {
+		perror("open(O_TMPFILE)");
+		exit(1);
+	}
+
+	snprintf(proc_self, sizeof(proc_self), "/proc/self/fd/%d", fd);
+
+	ret = linkat(AT_FDCWD, proc_self, AT_FDCWD, lpath, AT_SYMLINK_FOLLOW);
+	if (ret < 0) {
+		perror("linkat");
+		exit(1);
+	}
+
+	close(fd);
+}
+
+/*
+ * Use O_TMPFILE and linkat to create a new visible file, used to test
+ * the O_TMPFILE creation path by inspecting the created file.
+ */
+int main(int argc, char **argv)
+{
+	char *lpath;
+	char *dir;
+
+	if (argc < 3) {
+		printf("%s <open_dir> <linkat_path>\n", argv[0]);
+		return 1;
+	}
+
+	dir = argv[1];
+	lpath = argv[2];
+
+	linkat_tmpfile(dir, lpath);
+
+	return 0;
+}

tests/tests/basic-posix-acl.sh

@@ -0,0 +1,110 @@
+
+#
+# test basic POSIX acl functionality.
+#
+
+t_require_commands stat rm touch mkdir getfacl setfacl id sudo
+t_require_mounts 2
+
+# from quota.sh
+TEST_UID=22222
+TEST_GID=44444
+
+# sys_setreuid() set fs[uid] to e[ug]id
+SET_UID="--ruid=$TEST_UID --euid=$TEST_UID"
+SET_GID="--rgid=$TEST_GID --egid=$TEST_GID --clear-groups"
+
+# helper to avoid capturing dates from ls output
+L() {
+	stat -c "%F %A %u %g %s %N" $@
+}
+
+echo "== setup test directory"
+cd "$T_D0"
+
+echo "== getfacl"
+L .
+getfacl .
+
+echo "== basic non-acl access through permissions"
+rm -rf dir-testuid
+mkdir dir-testuid
+ln -sf dir-testuid symlinkdir-testuid
+chown root:44444 dir-testuid
+L dir-testuid
+setpriv $SET_UID $SET_GID touch dir-testuid/file-group-write
+setpriv $SET_UID $SET_GID touch symlinkdir-testuid/symlink-file-group-write
+chmod g+w dir-testuid
+setpriv $SET_UID $SET_GID touch dir-testuid/file-group-write
+setpriv $SET_UID $SET_GID touch symlinkdir-testuid/symlink-file-group-write
+L dir-testuid/file-group-write
+L symlinkdir-testuid/symlink-file-group-write
+
+echo "== basic acl access"
+rm -rf dir-root
+mkdir dir-root
+ln -sf dir-root symlinkdir-root
+L dir-root
+setpriv $SET_UID touch dir-root/file-group-write
+setpriv $SET_UID touch symlinkdir-root/file-group-write
+setfacl -m u:22222:rwx dir-root
+getfacl dir-root
+setpriv $SET_UID touch dir-root/file-group-write
+setpriv $SET_UID touch symlinkdir-root/file-group-write
+L dir-root/file-group-write
+L symlinkdir-root/file-group-write
+
+echo "== directory exec"
+setpriv $SET_UID bash -c "cd dir-root && echo Success"
+setpriv $SET_UID bash -c "cd symlinkdir-root && echo Success"
+setfacl -m u:22222:rw dir-root
+getfacl dir-root
+setpriv $SET_UID bash -c "cd dir-root || echo Failed"
+setpriv $SET_UID bash -c "cd symlinkdir-root || echo Failed"
+setfacl -m g:44444:rwx dir-root
+getfacl dir-root
+setpriv $SET_GID bash -c "cd dir-root && echo Success"
+setpriv $SET_GID bash -c "cd symlinkdir-root && echo Success"
+
+echo "== get/set attr"
+rm -rf file-root
+touch file-root
+L file-root
+setpriv $SET_UID getfattr -d file-root
+setpriv $SET_UID setfattr -n "user.test1" -v "Success" file-root
+setpriv $SET_UID getfattr -d file-root
+setfacl -m u:22222:rw file-root
+getfacl file-root
+setpriv $SET_UID setfattr -n "user.test2" -v "Success" file-root
+setpriv $SET_UID getfattr -d file-root
+setfacl -x u:22222 file-root
+getfacl file-root
+setpriv $SET_UID setfattr -n "user.test3" -v "Success" file-root
+setpriv $SET_UID getfattr -d file-root
+setfacl -m g:44444:rw file-root
+getfacl file-root
+setpriv $SET_GID setfattr -n "user.test4" -v "Success" file-root
+setpriv $SET_GID getfattr -d file-root
+
+echo "== inheritance / default acl"
+rm -rf dir-root2
+mkdir dir-root2
+L dir-root2
+setpriv $SET_UID mkdir dir-root2/dir
+setpriv $SET_UID touch dir-root2/dir/file
+setfacl -m d:u:22222:rwx dir-root2
+getfacl dir-root2
+setpriv $SET_UID mkdir dir-root2/dir
+setpriv $SET_UID touch dir-root2/dir/file
+setfacl -m u:22222:rwx dir-root2
+getfacl dir-root2
+setpriv $SET_UID mkdir dir-root2/dir
+setpriv $SET_UID touch dir-root2/dir/file
+L dir-root2/dir
+getfacl dir-root2/dir
+L dir-root2/dir/file
+getfacl dir-root2/dir/file
+
+echo "== cleanup"
+
+t_pass

tests/tests/basic-posix-consistency.sh

@@ -210,4 +210,7 @@ done
 wait
 ls "$T_D0/concurrent"
 
+echo "== cleanup"
+rm -f "$T_TMP.0" "$T_TMP.1"
+
 t_pass

tests/tests/change-devices.sh

@@ -73,4 +73,7 @@ test "$large_tot" -gt "$equal_tot" ; echo "resized larger test rc: $?"
 umount "$SCR"
 losetup -d "$scr_loop"
 
+echo "== cleanup"
+rm -f "$T_TMP.small" "$T_TMP.equal" "$T_TMP.large"
+
 t_pass

tests/tests/client-unmount-recovery.sh

@@ -28,7 +28,7 @@ while [ "$SECONDS" -lt "$END" ]; do
         for i in $(t_fs_nrs); do
                 if [ "$i" -ge "$quorum_nr" ]; then
                         t_umount $i &
-                        echo "umount $i pid $pid quo $quorum_nr" \
+                        echo "umount $i rid $rid quo $quorum_nr" \
                                 >> $T_TMP.log
                         mounted[$i]=0
                 fi
@@ -53,6 +53,9 @@ while [ "$SECONDS" -lt "$END" ]; do
 
         for i in "${lock_arr[@]}"; do
                 if [[ ! " ${rid_arr[*]} " =~ " $i " ]]; then
+                        echo -e "RID($i) exists" >> $T_TMP.log
+                        echo -e "rid_arr:\n${rid_arr[@]}" >> $T_TMP.log
+                        echo -e "lock_arr:\n${lock_arr[@]}" >> $T_TMP.log
                         t_fail "RID($i): exists when not mounted"
                 fi
         done

tests/tests/createmany-parallel-mounts.sh

@@ -2,7 +2,7 @@
 # Test clustered parallel createmany
 #
 
-t_require_commands mkdir createmany
+t_require_commands mkdir createmany bc
 t_require_mounts 2
 
 COUNT=50000
@@ -17,14 +17,14 @@ mkdir -p $T_D0/dir/0
 mkdir $T_D1/dir/1
 
 echo "== measure initial createmany"
-START=$SECONDS
+START=$(date +%s.%N)
 createmany -o "$T_D0/file_" $COUNT >> $T_TMP.full
 sync
-SINGLE=$((SECONDS - START))
-echo single $SINGLE >> $T_TMP.full
+END=$(date +%s.%N)
+SINGLE=$(echo "$END - $START" | bc)
 
 echo "== measure two concurrent createmany runs"
-START=$SECONDS
+START=$(date +%s.%N)
 (cd $T_D0/dir/0; createmany -o ./file_ $COUNT > /dev/null) &
 pids="$!"
 (cd $T_D1/dir/1; createmany -o ./file_ $COUNT > /dev/null) &
@@ -33,7 +33,9 @@ for p in $pids; do
         wait $p
 done
 sync
-BOTH=$((SECONDS - START))
+END=$(date +%s.%N)
+BOTH=$(echo "$END - $START" | bc)
+
 echo both $BOTH >> $T_TMP.full
 
 # Multi node still adds significant overhead, even with our CW locks
@@ -44,7 +46,7 @@ echo both $BOTH >> $T_TMP.full
 # exceed this factor should the CW locked items go back to fully
 # synchronized operation.
 FACTOR=200
-if [ "$BOTH" -gt $(($SINGLE*$FACTOR)) ]; then
+if [ $(echo "$BOTH > ( $SINGLE * $FACTOR )" | bc) == "1" ]; then
 	t_fail "both createmany took $BOTH sec, more than $FACTOR x single $SINGLE sec"
 fi
 

tests/tests/format-version-forward-back.sh

@@ -0,0 +1,179 @@
+#
+# Test our basic ability to work with different format versions.
+#
+# The current code being tested has a range of supported format
+# versions.   For each of the older supported format versions we have a
+# git hash of the commit before the next greater version was introduced.
+# We build versions of the scoutfs utility and kernel module for the
+# last commit in tree that had a lesser supported version as its max
+# supported version.   We use those binaries to test forward and back
+# compat as new and old code works with a persistent volume with a given
+# format version.
+#
+
+mount_has_format_version()
+{
+	local mnt="$1"
+	local vers="$2"
+	local sysfs_fmt_vers="$(t_sysfs_path_from_mnt $SCR)/format_version"
+
+	test "$(cat $sysfs_fmt_vers)" == "$vers"
+}
+
+SCR="/mnt/scoutfs.scratch"
+
+MIN=$(modinfo $T_MODULE | awk '($1 == "scoutfs_format_version_min:"){print $2}')
+MAX=$(modinfo $T_MODULE | awk '($1 == "scoutfs_format_version_max:"){print $2}')
+
+echo "min: $MIN max: $MAX" > "$T_TMP.log"
+
+test "$MIN" -gt 0 -a "$MAX" -gt 0 -a "$MIN" -le "$MAX" || \
+	t_fail "parsed bad versions, min: $MIN max: $MAX"
+
+test "$MIN" == "$MAX" && \
+	t_skip "only one supported format version: $MIN"
+
+# prepare dir and wipe any weird old partial state
+builds="$T_RESULTS/format_version_builds"
+mkdir -p "$builds"
+
+echo "== ensuring utils and module for old versions"
+declare -A commits
+commits[1]=c3c4b080
+for vers in $(seq $MIN $((MAX - 1))); do
+	dir="$builds/$vers"
+	platform=$(uname -rp)
+	buildmark="$dir/buildmark"
+	commit="${commits[$vers]}"
+
+	test -n "$commit" || \
+		t_fail "no commit for vers $vers"
+
+	# have our files for this version
+	test "$(cat $buildmark 2>&1)" == "$platform" && \
+		continue
+
+	# build as one big sequence of commands that can return failure
+	(
+		set -o pipefail
+
+		rm -rf $dir							&&
+		mkdir -p $dir/building						&&
+		cd "$T_TESTS/.."						&&
+		git archive --format=tar "$commit" | tar -C "$dir/building" -xf - &&
+		cd -								&&
+		find $dir							&&
+		make -C "$dir/building"						&&
+		mv $dir/building/utils/src/scoutfs $dir				&&
+		mv $dir/building/kmod/src/scoutfs.ko $dir			&&
+		rm -rf $dir/building						&&
+		echo "$platform" > $buildmark					&&
+		find $dir							&&
+		cat $buildmark
+	) >> "$T_TMP.log" 2>&1 || t_fail "version $vers build failed"
+done
+
+echo "== unmounting test fs and removing test module"
+t_quiet t_umount_all
+t_quiet rmmod scoutfs
+
+echo "== testing combinations of old and new format versions"
+mkdir -p "$SCR"
+for vers in $(seq $MIN $((MAX - 1))); do
+	old_scoutfs="$builds/$vers/scoutfs"
+	old_module="$builds/$vers/scoutfs.ko"
+
+	echo "mkfs $vers" >> "$T_TMP.log"
+	t_quiet $old_scoutfs mkfs -f -Q 0,127.0.0.1,53000 "$T_EX_META_DEV" "$T_EX_DATA_DEV" \
+		|| t_fail "mkfs $vers failed"
+
+	echo "mount $vers with $vers" >> "$T_TMP.log"
+	t_quiet insmod $old_module
+	t_quiet mount -t scoutfs -o metadev_path=$T_EX_META_DEV,quorum_slot_nr=0 \
+		"$T_EX_DATA_DEV" "$SCR"
+	t_quiet mount_has_format_version "$SCR" "$vers"
+
+	echo "creating files in $vers" >> "$T_TMP.log"
+	t_quiet touch "$SCR/file-"{1,2,3}
+	stat "$SCR"/file-* > "$T_TMP.stat" || \
+		t_fail "stat in $vers failed"
+
+	echo "remounting $vers fs with $MAX" >> "$T_TMP.log"
+	t_quiet umount "$SCR"
+	rmmod scoutfs
+	insmod "$T_MODULE"
+	t_quiet mount -t scoutfs -o metadev_path=$T_EX_META_DEV,quorum_slot_nr=0 \
+		"$T_EX_DATA_DEV" "$SCR"
+	t_quiet mount_has_format_version "$SCR" "$vers"
+
+	echo "verifying stat in $vers with $MAX" >> "$T_TMP.log"
+	diff -u "$T_TMP.stat" <(stat "$SCR"/file-*)
+
+	echo "keep/update/del existing, create new in $vers" >> "$T_TMP.log"
+	t_quiet touch "$SCR/file-2"
+	t_quiet rm -f "$SCR/file-3"
+	t_quiet touch "$SCR/file-4"
+	stat "$SCR"/file-* > "$T_TMP.stat" || \
+		t_fail "stat in $vers failed"
+
+	echo "remounting $vers fs with $vers" >> "$T_TMP.log"
+	t_quiet umount "$SCR"
+	rmmod scoutfs
+	insmod "$old_module"
+	t_quiet mount -t scoutfs -o metadev_path=$T_EX_META_DEV,quorum_slot_nr=0 \
+		"$T_EX_DATA_DEV" "$SCR"
+	t_quiet mount_has_format_version "$SCR" "$vers"
+
+	echo "verifying stat in $vers with $vers" >> "$T_TMP.log"
+	diff -u "$T_TMP.stat" <(stat "$SCR"/file-*)
+
+	echo "changing format vers to $MAX" >> "$T_TMP.log"
+	t_quiet umount "$SCR"
+	rmmod scoutfs
+	t_quiet scoutfs change-format-version -F -V $MAX $T_EX_META_DEV "$T_EX_DATA_DEV"
+
+	echo "mount fs $MAX with old $vers should fail" >> "$T_TMP.log"
+	insmod "$old_module"
+	mount -t scoutfs -o metadev_path=$T_EX_META_DEV,quorum_slot_nr=0 \
+		"$T_EX_DATA_DEV" "$SCR" >> "$T_TMP.log" 2>&1
+	if [ "$?" == "0" ]; then
+		umount "$SCR"
+		t_fail "old code ver $vers able to mount new ver $MAX"
+	fi
+
+	echo "remounting $MAX fs with $MAX" >> "$T_TMP.log"
+	rmmod scoutfs
+	insmod "$T_MODULE"
+	t_quiet mount -t scoutfs -o metadev_path=$T_EX_META_DEV,quorum_slot_nr=0 \
+		"$T_EX_DATA_DEV" "$SCR"
+	t_quiet mount_has_format_version "$SCR" "$MAX"
+
+	echo "verifying stat in $MAX with $MAX" >> "$T_TMP.log"
+	diff -u "$T_TMP.stat" <(stat "$SCR"/file-*)
+
+	echo "keep/update/del existing, create new in $MAX" >> "$T_TMP.log"
+	t_quiet touch "$SCR/file-2"
+	t_quiet rm -f "$SCR/file-4"
+	t_quiet touch "$SCR/file-5"
+	stat "$SCR"/file-* > "$T_TMP.stat" || \
+		t_fail "stat in $MAX failed"
+
+	echo "remounting $MAX fs with $MAX again" >> "$T_TMP.log"
+	t_quiet umount "$SCR"
+	t_quiet mount -t scoutfs -o metadev_path=$T_EX_META_DEV,quorum_slot_nr=0 \
+		"$T_EX_DATA_DEV" "$SCR"
+	t_quiet mount_has_format_version "$SCR" "$MAX"
+
+	echo "verifying stat in $MAX with $MAX again" >> "$T_TMP.log"
+	diff -u "$T_TMP.stat" <(stat "$SCR"/file-*)
+
+	echo "done with old vers $vers" >> "$T_TMP.log"
+	t_quiet umount "$SCR"
+	rmmod scoutfs
+done
+
+echo "== restoring test module and mount"
+insmod "$T_MODULE"
+t_mount_all
+
+t_pass

tests/tests/projects.sh

@@ -0,0 +1,52 @@
+
+# notable id to recognize in output
+ID=8675309
+
+echo "== default new files don't have project"
+touch "$T_D0/file"
+scoutfs get-attr-x -p "$T_D0/file"
+
+echo "== set new project on files and dirs"
+mkdir "$T_D0/dir"
+scoutfs set-attr-x -p $ID "$T_D0/file"
+scoutfs set-attr-x -p $ID "$T_D0/dir"
+scoutfs get-attr-x -p "$T_D0/file"
+scoutfs get-attr-x -p "$T_D0/dir"
+
+echo "== non-root can see id"
+chmod 644 "$T_D0/file"
+setpriv --ruid=12345 --euid=12345 scoutfs get-attr-x -p "$T_D0/file"
+
+echo "== can use IDs around long width limits"
+touch "$T_D0/ids"
+for id in         0x7FFFFFFF         0x80000000         0xFFFFFFFF \
+          0x7FFFFFFFFFFFFFFF 0x8000000000000000 0xFFFFFFFFFFFFFFFF; do
+	scoutfs set-attr-x -p $id "$T_D0/ids"
+	scoutfs get-attr-x -p "$T_D0/ids"
+done
+
+echo "== created files and dirs inherit project id"
+touch "$T_D0/dir/file"
+mkdir "$T_D0/dir/sub"
+scoutfs get-attr-x -p "$T_D0/dir/file"
+scoutfs get-attr-x -p "$T_D0/dir/sub"
+
+echo "== inheritance continues"
+mkdir "$T_D0/dir/sub/more"
+scoutfs get-attr-x -p "$T_D0/dir/sub/more"
+
+# .. just inherits 0 :) 
+echo "== clearing project id stops inheritance"
+scoutfs set-attr-x -p 0 "$T_D0/dir"
+touch "$T_D0/dir/another-file"
+mkdir "$T_D0/dir/another-sub"
+scoutfs get-attr-x -p "$T_D0/dir/another-file"
+scoutfs get-attr-x -p "$T_D0/dir/another-sub"
+
+echo "== o_tmpfile creations inherit dir"
+scoutfs set-attr-x -p $ID "$T_D0/dir"
+o_tmpfile_linkat "$T_D0/dir" "$T_D0/dir/tmpfile"
+scoutfs get-attr-x -p "$T_D0/dir/tmpfile"
+
+
+t_pass

tests/tests/quota.sh

@@ -0,0 +1,153 @@
+
+TEST_UID=22222
+TEST_GID=44444
+
+# sys_setreuid() set fs[uid] to e[ug]id
+SET_UID="--ruid=$TEST_UID --euid=$TEST_UID"
+SET_GID="--rgid=$TEST_GID --egid=$TEST_GID --clear-groups"
+
+FILE="$T_D0/dir/file"
+
+sync_and_drop()
+{
+	sync
+	echo 1 > $(t_debugfs_path)/drop_weak_item_cache
+	echo 1 > $(t_debugfs_path)/drop_quota_check_cache
+}
+
+reset_all()
+{
+	rm -f "$FILE"
+	scoutfs quota-wipe -p "$T_M0"
+	getfattr --absolute-names -d -m - "$T_D0" | \
+		grep "^scoutfs.totl." | \
+		cut -d '=' -f 1 | \
+		xargs -n 1 -I'{}'  setfattr -x '{}' "$T_D0"
+}
+
+echo "== prepare dir with write perm for test ids"
+mkdir "$T_D0/dir"
+chown --quiet $TEST_UID "$T_D0/dir"
+chgrp --quiet $TEST_GID "$T_D0/dir"
+
+echo "== test assumes starting with no rules, empty list"
+scoutfs quota-list -p "$T_M0"
+
+echo "== add rule"
+scoutfs quota-add -p "$T_M0" -r "7 13,L,- 15,L,- 17,L,- I 33 -"
+scoutfs quota-list -p "$T_M0"
+
+echo "== list is empty again after delete"
+scoutfs quota-del -p "$T_M0" -r "7 13,L,- 15,L,- 17,L,- I 33 -"
+scoutfs quota-list -p "$T_M0"
+
+echo "== can change limits without deleting"
+scoutfs quota-add -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- I 100 -"
+scoutfs quota-list -p "$T_M0"
+scoutfs quota-add -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- I 101 -"
+scoutfs quota-del -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- I 100 -"
+scoutfs quota-list -p "$T_M0"
+scoutfs quota-add -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- I 99 -"
+scoutfs quota-del -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- I 101 -"
+scoutfs quota-list -p "$T_M0"
+scoutfs quota-del -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- I 99 -"
+reset_all
+
+echo "== wipe and restore rules in bulk"
+for a in $(seq 10 15); do
+	scoutfs quota-add -p "$T_M0" -r "7 $a,L,- 0,L,- 0,L,- I 33 -"
+done
+scoutfs quota-list -p "$T_M0"
+scoutfs quota-list -p "$T_M0" > "$T_TMP.list"
+scoutfs quota-wipe -p "$T_M0"
+scoutfs quota-list -p "$T_M0"
+scoutfs quota-restore -p "$T_M0" < "$T_TMP.list"
+scoutfs quota-list -p "$T_M0"
+reset_all
+
+echo "== default rule prevents file creation"
+scoutfs quota-add -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- I 1 -"
+setfattr -n scoutfs.totl.test.1.1.1 -v 2 "$T_D0"
+sync_and_drop
+setpriv $SET_UID touch "$FILE" 2>&1 | t_filter_fs
+
+echo "== decreasing totl allows file creation again"
+setfattr -x scoutfs.totl.test.1.1.1 "$T_D0"
+sync_and_drop
+setpriv $SET_UID touch "$FILE"
+reset_all
+
+echo "== attr selecting rules prevent creation"
+scoutfs quota-add -p "$T_M0" -r "1 $TEST_UID,U,S 1,L,- 1,L,- I 1 -"
+scoutfs quota-add -p "$T_M0" -r "1 $TEST_GID,G,S 1,L,- 1,L,- I 1 -"
+setfattr -n scoutfs.totl.test.$TEST_UID.1.1 -v 2 "$T_D0"
+setfattr -n scoutfs.totl.test.$TEST_GID.1.1 -v 2 "$T_D0"
+sync_and_drop
+setpriv $SET_UID touch "$FILE" 2>&1 | t_filter_fs
+setpriv $SET_GID touch "$FILE" 2>&1 | t_filter_fs
+reset_all
+
+echo "== multi attr selecting doesn't prevent partial"
+scoutfs quota-add -p "$T_M0" -r "1 $TEST_UID,U,S $TEST_GID,G,S 1,L,- I 1 -"
+setfattr -n scoutfs.totl.test.$TEST_UID.$TEST_GID.1 -v 2 "$T_D0"
+sync_and_drop
+setpriv $SET_UID touch "$FILE"
+rm -f "$FILE"
+setpriv $SET_GID touch "$FILE"
+rm -f "$FILE"
+setpriv $SET_UID $SET_GID touch "$FILE" 2>&1 | t_filter_fs
+reset_all
+
+echo "== op differentiates"
+# inode ops succeed in presence of data rule
+scoutfs quota-add -p "$T_M0" -r "1 $TEST_UID,U,S 1,L,- 1,L,- D 1 -"
+setfattr -n scoutfs.totl.test.$TEST_UID.1.1 -v 2 "$T_D0"
+sync_and_drop
+setpriv $SET_UID touch "$FILE" 2>&1 | t_filter_fs
+reset_all
+# data ops succeed in presence of inode rule
+touch "$FILE"
+chown --quiet $TEST_UID "$FILE"
+scoutfs quota-add -p "$T_M0" -r "1 $TEST_UID,U,S 1,L,- 1,L,- I 1 -"
+setfattr -n scoutfs.totl.test.$TEST_UID.1.1 -v 2 "$T_D0"
+sync_and_drop
+setpriv $SET_UID fallocate -l 4096 "$FILE" 2>&1 | t_filter_fs
+reset_all
+
+echo "== higher priority rule applies"
+scoutfs quota-add -p "$T_M0" -r "1 $TEST_UID,U,S 1,L,- 1,L,- I 1000 -"
+scoutfs quota-add -p "$T_M0" -r "2 $TEST_UID,U,S 1,L,- 1,L,- I 1 -"
+setfattr -n scoutfs.totl.test.$TEST_UID.1.1 -v 2 "$T_D0"
+sync_and_drop
+setpriv $SET_UID touch "$FILE" 2>&1 | t_filter_fs
+reset_all
+
+echo "== data rules with total and count prevent write and fallocate" 
+touch "$FILE"
+scoutfs quota-add -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- D 1 -"
+setfattr -n scoutfs.totl.test.1.1.1 -v 2 "$T_D0"
+sync_and_drop
+dd if=/dev/zero of="$FILE" bs=4096 count=1 conv=notrunc status=none 2>&1 | t_filter_fs
+fallocate -l 4096 "$FILE" 2>&1 | t_filter_fs
+scoutfs quota-del -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- D 1 -"
+scoutfs quota-add -p "$T_M0" -r "1 1,L,- 1,L,- 1,L,- D 0 C"
+sync_and_drop
+dd if=/dev/zero of="$FILE" bs=4096 count=1 conv=notrunc status=none 2>&1 | t_filter_fs
+fallocate -l 4096 "$FILE" 2>&1 | t_filter_fs
+reset_all
+
+echo "== added rules work after bulk restore"
+seq -f "  1 %.0f,U,S 1,L,- 1,L,- I 1 -" 9000050000 -1 9000000000 > "$T_TMP.lots"
+scoutfs quota-restore -p "$T_M0" < "$T_TMP.lots"
+scoutfs quota-list -p "$T_M0" > "$T_TMP.list"
+diff -u "$T_TMP.lots" "$T_TMP.list"
+scoutfs quota-add -p "$T_M0" -r "1 $TEST_UID,U,S 1,L,- 1,L,- I 1 -"
+setfattr -n scoutfs.totl.test.$TEST_UID.1.1 -v 2 "$T_D0"
+sync_and_drop
+setpriv $SET_UID touch "$FILE" 2>&1 | t_filter_fs
+reset_all
+
+echo "== cleanup"
+rm -f "$T_TMP.lots" "$T_TMP.list"
+
+t_pass

tests/tests/retention-basic.sh

@@ -0,0 +1,57 @@
+t_require_commands scoutfs touch rm setfattr
+
+touch "$T_D0/file-1"
+
+echo "== setting retention on dir fails"
+scoutfs set-attr-x -t 1 "$T_D0" 2>&1 | t_filter_fs
+
+echo "== set retention"
+scoutfs set-attr-x -t 1 "$T_D0/file-1"
+
+echo "== get-attr-x shows retention"
+scoutfs get-attr-x -t "$T_D0/file-1"
+
+echo "== unpriv can't clear retention"
+setpriv --ruid=12345 --euid=12345 scoutfs set-attr-x -t 0 "$T_D0/file-1" 2>&1 | t_filter_fs
+
+echo "== can set hidden scoutfs xattr in retention"
+setfattr -n scoutfs.hide.srch.retention_test -v val "$T_D0/file-1"
+
+echo "== setting user. xattr fails in retention"
+setfattr -n user.retention_test -v val "$T_D0/file-1" 2>&1 | t_filter_fs
+
+echo "== file deletion fails in retention"
+rm -f "$T_D0/file-1" 2>&1 | t_filter_fs
+
+echo "== file rename fails in retention"
+mv $T_D0/file-1 $T_D0/file-2 2>&1 | t_filter_fs
+
+echo "== file write fails in retention"
+date >> $T_D0/file-1
+
+echo "== file truncate fails in retention"
+truncate -s 0 $T_D0/file-1 2>&1 | t_filter_fs
+
+echo "== setattr fails in retention"
+touch $T_D0/file-1 2>&1 | t_filter_fs
+
+echo "== clear retention"
+scoutfs set-attr-x -t 0 "$T_D0/file-1"
+
+echo "== file write"
+date >> $T_D0/file-1
+
+echo "== file rename"
+mv $T_D0/file-1 $T_D0/file-2
+mv $T_D0/file-2 $T_D0/file-1
+
+echo "== setattr"
+touch $T_D0/file-1 2>&1 | t_filter_fs
+
+echo "== xattr deletion"
+setfattr -x scoutfs.hide.srch.retention_test "$T_D0/file-1"
+
+echo "== cleanup"
+rm -f "$T_D0/file-1"
+
+t_pass

tests/tests/stage-release-race-alloc.sh

@@ -62,4 +62,7 @@ for r in $(seq 1 1000); do
 	done
 done
 
+echo "== cleanup"
+rm -f "$T_TMP.log"
+
 t_pass

tests/tests/totl-xattr-tag.sh

@@ -3,6 +3,7 @@ t_require_commands touch rm setfattr scoutfs find_xattrs
 read_xattr_totals()
 {
 	sync
+	echo 1 > $(t_debugfs_path)/drop_weak_item_cache
 	scoutfs read-xattr-totals -p "$T_M0"
 }
 
@@ -112,7 +113,6 @@ for phase in create update remove; do
 				echo "$k.0.0 = ${totals[$k]}, ${counts[$k]}"
 			  done ) | grep -v "= 0, 0$" | sort -n >> $T_TMP.check_arr
 
-			sync
 			read_xattr_totals | sort -n >> $T_TMP.check_read
 
 			diff -u $T_TMP.check_arr $T_TMP.check_read || \

utils/Makefile

@@ -7,7 +7,7 @@ FMTIOC_H := format.h ioctl.h
 FMTIOC_KMOD := $(addprefix ../kmod/src/,$(FMTIOC_H))
 
 CFLAGS := -Wall -O2 -Werror -D_FILE_OFFSET_BITS=64 -g -msse4.2 \
-	-I src/ -fno-strict-aliasing \
+	-fno-strict-aliasing \
 	-DSCOUTFS_FORMAT_HASH=0x$(SCOUTFS_FORMAT_HASH)LLU
 
 ifneq ($(wildcard $(firstword $(FMTIOC_KMOD))),)
@@ -15,9 +15,8 @@ CFLAGS += -I../kmod/src
 endif
 
 BIN := src/scoutfs
-OBJ_DIRS := src src/check
-OBJ := $(foreach dir,$(OBJ_DIRS),$(patsubst %.c,%.o,$(wildcard $(dir)/*.c)))
-DEPS := $(foreach dir,$(OBJ_DIRS),$(wildcard $(dir)/*.d))
+OBJ := $(patsubst %.c,%.o,$(wildcard src/*.c))
+DEPS := $(wildcard */*.d)
 
 all: $(BIN)
 

utils/man/scoutfs.5

@@ -270,6 +270,21 @@ metadata that is bound to a specific volume and should not be
 transferred with the file by tools that read extended attributes, like
 .BR tar(1) .
 .TP
+.B .indx.
+Attributes with the .indx. tag dd the inode containing the attribute to
+a filesystem-wide index.  The name of the extended attribute must end
+with strings representing two values separated by dots.  The first value
+is an unsigned 8bit value and the second is an unsigned 64bit value.
+These attributes can only be modified with root privileges and the
+attributes can not have a value.
+.sp
+The inodes in the index are stored in increasing sort order of the
+values, with the first u8 value being most significant.  Inodes can be
+at many positions as tracked by many extended attributes, and their
+position follows the creation, renaming, or deletion of the attributes.
+The index can be read with the read-xattr-index command which uses the
+underlying READ_XATTR_INDEX ioctl.
+.TP
 .B .srch.
 Attributes with the .srch. tag are indexed so that they can be
 found by the
@@ -295,6 +310,36 @@ with the
 ioctl.
 .RE
 
+.SH FILE RETENTION MODE
+A file can be set to retention mode by setting the
+.IB RETENTION
+attribute with the
+.IB SET_ATTR_X
+ioctl.  This flag can only be set on regular files and requires root
+permission (the
+.IB CAP_SYS_ADMIN
+capability).
+.sp
+Once in retention mode all modifications of the file will fail.  The
+only exceptions are that system extended attributes (all those without
+the "user." prefix) may be modified.  The retention bit may be cleared
+with sufficient priveledges to remove the retention restrictions on
+other modifications.
+.RE
+
+.SH PROJECT IDs
+All inodes have a project ID attribute that can be set via the
+SET_ATTR_X ioctl and displayed with the GET_ATTR_X ioctl.  Project IDs
+are an unsigned 64bit value and the value of 0 is reserved to indicate
+that no project ID is assigned.   If a project ID is set on a directory
+then all inodes created with it as the initial parent inheret that ID,
+for all file types.  This includes files initially unlinked from the
+namespace when created with O_TMPFILE.  Project IDs are only
+automatically inherited from the parent dir on initial creation.
+They're not changed as directory entry linkes to the inode are created
+or renamed.
+.RE
+
 .SH FORMAT VERSION
 The format version defines the layout and use of structures stored on
 devices and passed over the network.  The version is incremented for
@@ -373,6 +418,19 @@ The version that a mount is using is shown in the
 file in the mount's sysfs directory, typically
 .I /sys/fs/scoutfs/f.FSID.r.RID/
 .RE
+.sp
+The defined format versions are:
+.RS
+.TP
+.sp
+.B 1
+Initial format version.
+.TP
+.B 2
+Added retention mode by setting the retention attribute.  Added the
+project ID inode attribute.  Added quota rules and enforcement.  Added
+the .indx. extended attribute tag.
+.RE
 
 .SH CORRUPTION DETECTION
 A

utils/man/scoutfs.8

@@ -209,6 +209,16 @@ A path within a ScoutFS filesystem.
 .RE
 .PD
 
+.TP
+.BI "get-attr-x FILE"
+.sp
+Display ScoutFS-specific attributes from a file.  If no options are
+given than all the attributes that the command supports will be
+displayed.  If attributes are specified with options then only those
+attributes are displayed.  If only one attribute is specified then it
+will not have a label prefix in the display output.  The --help option
+will list the attributes that the command supports.  The file system may
+support a different set of attributes.
 .TP
 .BI "get-referring-entries [-p|--path PATH] INO"
 .sp
@@ -506,6 +516,15 @@ A path within a ScoutFS filesystem.
 .RE
 .PD
 
+.TP
+.BI "set-attr-x FILE"
+.sp
+Set ScoutFS-specific attributes on a file.  Only the attributes that are
+spcified by options will be set.  The --help option will list the
+attributes that the command understands.  The file system may support a
+different set of attributes.
+.PD
+
 .TP
 .BI "setattr FILE [-d, --data-version=VERSION [-s, --size=SIZE [-o, --offline]]] [-t, --ctime=TIMESPEC]"
 .sp

utils/src/attr_x.c

@@ -0,0 +1,304 @@
+#include <unistd.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/ioctl.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <string.h>
+#include <assert.h>
+#include <argp.h>
+#include <stdbool.h>
+
+#include "sparse.h"
+#include "util.h"
+#include "format.h"
+#include "ioctl.h"
+#include "parse.h"
+#include "cmd.h"
+
+struct attr_x_args {
+	bool set;
+	char *filename;
+	struct scoutfs_ioctl_inode_attr_x iax;
+};
+
+#define pr(iax, name, label, fmt, args...)			\
+do {								\
+	if ((iax->x_mask & SCOUTFS_IOC_IAX_##name)) {		\
+		if (__builtin_popcount(iax->x_mask) > 1)	\
+			printf(label ": " fmt "\n", ##args);	\
+		else						\
+			printf(fmt "\n", ##args);		\
+	}							\
+} while (0)
+
+#define prb(iax, name, label) \
+	pr(iax, name, label, "%u", !!((iax)->bits & SCOUTFS_IOC_IAX_B_##name))
+
+static int do_attr_x(struct attr_x_args *args)
+{
+	struct scoutfs_ioctl_inode_attr_x *iax = &args->iax;
+	int fd = -1;
+	int ret;
+	int op;
+
+	if (args->set) {
+		/* nothing to do if not setting */
+		if (iax->x_mask == 0)
+			return 0;
+		op = SCOUTFS_IOC_SET_ATTR_X;
+	} else {
+		/* get all known if none specified */
+		if (iax->x_mask == 0)
+			iax->x_mask = ~SCOUTFS_IOC_IAX__UNKNOWN;
+		op = SCOUTFS_IOC_GET_ATTR_X;
+	}
+
+	fd = open(args->filename, O_RDONLY);
+	if (fd < 0) {
+		ret = -errno;
+		fprintf(stderr, "failed to open '%s': %s (%d)\n",
+			args->filename, strerror(errno), errno);
+		goto out;
+	}
+
+	ret = ioctl(fd, op, iax);
+	if (ret < 0) {
+		ret = -errno;
+		fprintf(stderr, "attr_x ioctl failed on '%s': "
+			"%s (%d)\n", args->filename, strerror(errno), errno);
+		goto out;
+	}
+
+	if (!args->set) {
+		pr(iax, META_SEQ, "meta_seq", "%llu", iax->meta_seq);
+		pr(iax, DATA_SEQ, "data_seq", "%llu", iax->data_seq);
+		pr(iax, DATA_VERSION, "data_version", "%llu", iax->data_version);
+		pr(iax, ONLINE_BLOCKS, "online_blocks", "%llu", iax->online_blocks);
+		pr(iax, OFFLINE_BLOCKS, "offline_blocks", "%llu", iax->offline_blocks);
+		pr(iax, CTIME, "ctime", "%llu.%u", iax->ctime_sec, iax->ctime_nsec);
+		pr(iax, CRTIME, "crtime", "%llu.%u", iax->crtime_sec, iax->crtime_nsec);
+		pr(iax, SIZE, "size", "%llu", iax->size);
+		prb(iax, RETENTION, "retention");
+		pr(iax, PROJECT_ID, "project_id", "%llu", iax->project_id);
+	}
+
+	ret = 0;
+out:
+	if (fd >= 0)
+		close(fd);
+	return ret;
+}
+
+/*
+ * This is called for both get and set.  The get calls won't have
+ * arguments and are only setting the mask.  The set calls parse the
+ * value to set.  We could have defaults by making set option arguments
+ * optional, like setting the current time for timestamps, but that
+ * hasn't been needed.
+ *
+ * Option value parsing places no constraints on the attributes or
+ * values themselves once parsed.  This lets us use the set command to
+ * test the kernel's testing for invalid attribute combinations and
+ * values.
+ */
+static int parse_opt(int key, char *arg, struct argp_state *state)
+{
+	struct attr_x_args *args = state->input;
+	struct timespec ts;
+	int ret;
+	u64 x;
+
+	switch (key) {
+	case 'm':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_META_SEQ;
+		if (arg) {
+			ret = parse_u64(arg, &args->iax.meta_seq);
+			if (ret)
+				return ret;
+		}
+		break;
+	case 'd':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_DATA_SEQ;
+		if (arg) {
+			ret = parse_u64(arg, &args->iax.data_seq);
+			if (ret)
+				return ret;
+		}
+		break;
+	case 'v':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_DATA_VERSION;
+		if (arg) {
+			ret = parse_u64(arg, &args->iax.data_version);
+			if (ret)
+				return ret;
+			if (args->iax.data_version == 0)
+				argp_error(state, "data version must not be 0");
+		}
+		break;
+	case 'n':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_ONLINE_BLOCKS;
+		if (arg) {
+			ret = parse_u64(arg, &args->iax.online_blocks);
+			if (ret)
+				return ret;
+		}
+		break;
+	case 'f':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_OFFLINE_BLOCKS;
+		if (arg) {
+			ret = parse_u64(arg, &args->iax.offline_blocks);
+			if (ret)
+				return ret;
+		}
+		break;
+	case 'c':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_CTIME;
+		if (arg) {
+			ret = parse_timespec(arg, &ts);
+			if (ret)
+				return ret;
+			args->iax.ctime_sec = ts.tv_sec;
+			args->iax.ctime_nsec = ts.tv_nsec;
+		}
+		break;
+	case 'r':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_CRTIME;
+		if (arg) {
+			ret = parse_timespec(arg, &ts);
+			if (ret)
+				return ret;
+			args->iax.crtime_sec = ts.tv_sec;
+			args->iax.crtime_nsec = ts.tv_nsec;
+		}
+		break;
+	case 's':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_SIZE;
+		if (arg) {
+			ret = parse_u64(arg, &args->iax.size);
+			if (ret)
+				return ret;
+		}
+		break;
+	case 't':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_RETENTION;
+		if (arg) {
+			ret = parse_u64(arg, &x);
+			if (ret)
+				return ret;
+			if (x)
+				args->iax.bits |= SCOUTFS_IOC_IAX_B_RETENTION;
+		}
+		break;
+	case 'p':
+		args->iax.x_mask |= SCOUTFS_IOC_IAX_PROJECT_ID;
+		if (arg) {
+			ret = parse_u64(arg, &args->iax.project_id);
+			if (ret)
+				return ret;
+		}
+		break;
+	case ARGP_KEY_ARG:
+		if (!args->filename)
+			args->filename = strdup_or_error(state, arg);
+		else
+			argp_error(state, "more than one argument given");
+		break;
+	case ARGP_KEY_FINI:
+		if (!args->filename)
+			argp_error(state, "no filename given");
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+/*
+ * The get options are derived from these by copying the struct and
+ * modifying fields.
+ */
+static struct argp_option set_options[] = {
+	{ "meta_seq", 'm', "SEQ", 0, "Inode Metadata change index sequence number"},
+	{ "data_seq", 'd', "SEQ", 0, "File Data change index sequence number"},
+	{ "data_version", 'v', "VERSION", 0, "File Data contents version"},
+	{ "online_blocks", 'n', "COUNT", 0, "Online data block count"},
+	{ "offline_blocks", 'f', "COUNT", 0, "Offline data block count"},
+	{ "ctime", 'c', "SECS.NSECS", 0, "Inode change time (posix ctime)"},
+	{ "crtime", 'r', "SECS.NSECS", 0, "ScoutFS creation time"},
+	{ "size", 's', "SIZE", 0, "Inode i_size field"},
+	{ "retention", 't', "0|1", 0, "Retention flag"},
+	{ "project_id", 'p', "PROJECT_ID", 0, "Project ID"},
+	{ NULL }
+};
+
+static struct argp get_argp = {
+	NULL, /* dynamically built */
+	parse_opt,
+	"FILE",
+	"get extensible file attributes"
+};
+
+static int get_attr_x_cmd(int argc, char **argv)
+{
+	struct attr_x_args args = {0,};
+	int ret;
+
+	ret = argp_parse(&get_argp, argc, argv, 0, NULL, &args);
+	if (ret)
+		return ret;
+
+	return do_attr_x(&args);
+}
+
+/*
+ * The set options match the get arguments but don't take argument
+ * values to set.
+ */
+static void build_get_options(void)
+{
+	struct argp_option **opts = (struct argp_option **)&get_argp.options;
+	int i;
+
+	*opts = calloc(array_size(set_options), sizeof(set_options[0]));
+	assert(*opts);
+
+	memcpy(*opts, set_options, array_size(set_options) * sizeof(set_options[0]));
+
+	for (i = 0; i < array_size(set_options) - 1; i++)
+		(*opts)[i].arg = NULL;
+}
+
+static void __attribute__((constructor)) get_ctor(void)
+{
+	build_get_options();
+
+	cmd_register_argp("get-attr-x", &get_argp, GROUP_AGENT, get_attr_x_cmd);
+}
+
+static struct argp set_argp = {
+	set_options,
+	parse_opt,
+	"FILE",
+	"Set extensible file attributes"
+};
+
+static int set_attr_x_cmd(int argc, char **argv)
+{
+	struct attr_x_args args = {.set = true,};
+	int ret;
+
+	ret = argp_parse(&set_argp, argc, argv, 0, NULL, &args);
+	if (ret)
+		return ret;
+
+	return do_attr_x(&args);
+}
+
+static void __attribute__((constructor)) set_ctor(void)
+{
+	cmd_register_argp("set-attr-x", &set_argp, GROUP_AGENT, set_attr_x_cmd);
+}

utils/src/bitmap.c

@@ -10,11 +10,6 @@
  * Just a quick simple native bitmap.
  */
 
-int test_bit(unsigned long *bits, u64 nr)
-{
-	return !!(bits[nr / BITS_PER_LONG] & (1UL << (nr & (BITS_PER_LONG - 1))));
-}
-
 void set_bit(unsigned long *bits, u64 nr)
 {
 	bits[nr / BITS_PER_LONG] |= 1UL << (nr & (BITS_PER_LONG - 1));

utils/src/bitmap.h

@@ -1,7 +1,6 @@
 #ifndef _BITMAP_H_
 #define _BITMAP_H_
 
-int test_bit(unsigned long *bits, u64 nr);
 void set_bit(unsigned long *bits, u64 nr);
 void clear_bit(unsigned long *bits, u64 nr);
 u64 find_next_set_bit(unsigned long *start, u64 from, u64 total);

utils/src/change_format_version.c

@@ -94,6 +94,18 @@ static int do_change_fmt_vers(struct change_fmt_vers_args *args)
 		goto out;
 	}
 
+	if ((le64_to_cpu(meta_super->flags) & SCOUTFS_FLAG_IS_META_BDEV) == 0) {
+		printf("device argument #1 is not a meta device (swap arguments?)\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if ((le64_to_cpu(data_super->flags) & SCOUTFS_FLAG_IS_META_BDEV) != 0) {
+		printf("device argument #2 is not a data device (swap arguments?)\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
 	if (le64_to_cpu(meta_super->fmt_vers) < SCOUTFS_FORMAT_VERSION_MIN ||
 	    le64_to_cpu(meta_super->fmt_vers) > SCOUTFS_FORMAT_VERSION_MAX) {
 		fprintf(stderr, "meta super block has format version %llu outside of supported version range %u-%u",
@@ -119,6 +131,16 @@ static int do_change_fmt_vers(struct change_fmt_vers_args *args)
 		goto out;
 	}
 
+	if (le64_to_cpu(meta_super->fmt_vers) > args->fmt_vers ||
+	    le64_to_cpu(data_super->fmt_vers) > args->fmt_vers) {
+		ret = -EPERM;
+		printf("Downgrade of Meta Format Version: %llu and Data Format Version: %llu to Format Version: %llu is not allowed\n",
+		       le64_to_cpu(meta_super->fmt_vers),
+		       le64_to_cpu(data_super->fmt_vers),
+		       args->fmt_vers);
+		goto out;
+	}
+
 	if (le64_to_cpu(meta_super->fmt_vers) != args->fmt_vers) {
 		meta_super->fmt_vers = cpu_to_le64(args->fmt_vers);
 

utils/src/check/alloc.c

@@ -1,159 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdbool.h>
-#include <sys/mman.h>
-#include <errno.h>
-
-#include "sparse.h"
-#include "util.h"
-#include "format.h"
-#include "bitmap.h"
-#include "key.h"
-
-#include "alloc.h"
-#include "block.h"
-#include "btree.h"
-#include "extent.h"
-#include "iter.h"
-#include "sns.h"
-
-/*
- * We check the list blocks serially.
- *
- * XXX:
- *  - compare ref seqs
- *  - detect cycles?
- */
-int alloc_list_meta_iter(struct scoutfs_alloc_list_head *lhead, extent_cb_t cb, void *cb_arg)
-{
-	struct scoutfs_alloc_list_block *lblk;
-	struct scoutfs_block_ref ref;
-	struct block *blk = NULL;
-	u64 blkno;
-	int ret;
-
-	ref = lhead->ref;
-
-	while (ref.blkno) {
-		blkno = le64_to_cpu(ref.blkno);
-
-		ret = cb(blkno, 1, cb_arg);
-		if (ret < 0) {
-			ret = xlate_iter_errno(ret);
-			goto out;
-		}
-
-		ret = block_get(&blk, blkno, 0);
-		if (ret < 0)
-			goto out;
-
-		lblk = block_buf(blk);
-		/* XXX verify block */
-		/* XXX sort?   maybe */
-
-		ref = lblk->next;
-
-		block_put(&blk);
-	}
-
-	ret = 0;
-out:
-	return ret;
-}
-
-int alloc_root_meta_iter(struct scoutfs_alloc_root *root, extent_cb_t cb, void *cb_arg)
-{
-	return btree_meta_iter(&root->root, cb, cb_arg);
-}
-
-int alloc_list_extent_iter(struct scoutfs_alloc_list_head *lhead, extent_cb_t cb, void *cb_arg)
-{
-	struct scoutfs_alloc_list_block *lblk;
-	struct scoutfs_block_ref ref;
-	struct block *blk = NULL;
-	u64 blkno;
-	int ret;
-	int i;
-
-	ref = lhead->ref;
-
-	while (ref.blkno) {
-		blkno = le64_to_cpu(ref.blkno);
-
-		ret = block_get(&blk, blkno, 0);
-		if (ret < 0)
-			goto out;
-
-		sns_push("alloc_list_block", blkno, 0);
-
-		lblk = block_buf(blk);
-		/* XXX verify block */
-		/* XXX sort?   maybe */
-
-		ret = 0;
-		for (i = 0; i < le32_to_cpu(lblk->nr); i++) {
-			blkno = le64_to_cpu(lblk->blknos[le32_to_cpu(lblk->start) + i]);
-
-			ret = cb(blkno, 1, cb_arg);
-			if (ret < 0)
-				break;
-		}
-
-		ref = lblk->next;
-
-		block_put(&blk);
-		sns_pop();
-		if (ret < 0) {
-			ret = xlate_iter_errno(ret);
-			goto out;
-		}
-	}
-
-	ret = 0;
-out:
-	return ret;
-}
-
-static bool valid_free_extent_key(struct scoutfs_key *key)
-{
-	return (key->sk_zone == SCOUTFS_FREE_EXTENT_BLKNO_ZONE ||
-	        key->sk_zone == SCOUTFS_FREE_EXTENT_ORDER_ZONE) &&
-	       (!key->_sk_fourth && !key->sk_type &&
-		(key->sk_zone == SCOUTFS_FREE_EXTENT_ORDER_ZONE || !key->_sk_third));
-}
-
-static int free_item_cb(struct scoutfs_key *key, void *val, u16 val_len, void *cb_arg)
-{
-	struct extent_cb_arg_t *ecba = cb_arg;
-	u64 start;
-	u64 len;
-
-	/* XXX not sure these eios are what we want */
-
-	if (val_len != 0)
-		return -EIO;
-
-	if (!valid_free_extent_key(key))
-		return -EIO;
-
-	if (key->sk_zone == SCOUTFS_FREE_EXTENT_ORDER_ZONE)
-		return -ECHECK_ITER_DONE;
-
-	start = le64_to_cpu(key->skfb_end) - le64_to_cpu(key->skfb_len) + 1;
-	len = le64_to_cpu(key->skfb_len);
-
-	return ecba->cb(start, len, ecba->cb_arg);
-}
-
-/*
- * Call the callback with each of the primary BLKNO free extents stored
- * in item in the given alloc root.  It doesn't visit the secondary
- * ORDER extents.
- */
-int alloc_root_extent_iter(struct scoutfs_alloc_root *root, extent_cb_t cb, void *cb_arg)
-{
-	struct extent_cb_arg_t ecba = { .cb = cb, .cb_arg = cb_arg };
-
-	return btree_item_iter(&root->root, free_item_cb, &ecba);
-}

utils/src/check/alloc.h

@@ -1,12 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_ALLOC_H
-#define _SCOUTFS_UTILS_CHECK_ALLOC_H
-
-#include "extent.h"
-
-int alloc_list_meta_iter(struct scoutfs_alloc_list_head *lhead, extent_cb_t cb, void *cb_arg);
-int alloc_root_meta_iter(struct scoutfs_alloc_root *root, extent_cb_t cb, void *cb_arg);
-
-int alloc_list_extent_iter(struct scoutfs_alloc_list_head *lhead, extent_cb_t cb, void *cb_arg);
-int alloc_root_extent_iter(struct scoutfs_alloc_root *root, extent_cb_t cb, void *cb_arg);
-
-#endif

utils/src/check/block.c

@@ -1,564 +0,0 @@
-#define _ISOC11_SOURCE /* aligned_alloc */
-#define _DEFAULT_SOURCE /* syscall() */
-#include <stdlib.h>
-#include <unistd.h>
-#include <stdbool.h>
-#include <stdio.h>
-#include <errno.h>
-#include <sys/syscall.h>
-#include <linux/aio_abi.h>
-
-#include "sparse.h"
-#include "util.h"
-#include "format.h"
-#include "list.h"
-#include "cmp.h"
-#include "hash.h"
-
-#include "block.h"
-#include "debug.h"
-#include "eno.h"
-
-static struct block_data {
-	struct list_head *hash_lists;
-	size_t hash_nr;
-
-	struct list_head active_head;
-	struct list_head inactive_head;
-	struct list_head dirty_list;
-	size_t nr_active;
-	size_t nr_inactive;
-	size_t nr_dirty;
-
-	int meta_fd;
-	size_t max_cached;
-	size_t nr_events;
-
-	aio_context_t ctx;
-	struct iocb *iocbs;
-	struct iocb **iocbps;
-	struct io_event *events;
-} global_bdat;
-
-struct block {
-	struct list_head hash_head;
-	struct list_head lru_head;
-	struct list_head dirty_head;
-	struct list_head submit_head;
-	unsigned long refcount;
-	unsigned long uptodate:1,
-		      active:1;
-	u64 blkno;
-	void *buf;
-	size_t size;
-};
-
-#define BLK_FMT \
-	"blkno %llu rc %ld d %u a %u"
-#define BLK_ARG(blk) \
-	(blk)->blkno, (blk)->refcount, !list_empty(&(blk)->dirty_head), blk->active
-#define debug_blk(blk, fmt, args...) \
-	debug(fmt " " BLK_FMT, ##args, BLK_ARG(blk))
-
-/*
- * This just allocates and initialzies the block.  The caller is
- * responsible for putting it on the appropriate initial lists and
- * managing refcounts.
- */
-static struct block *alloc_block(struct block_data *bdat, u64 blkno, size_t size)
-{
-	struct block *blk;
-
-	blk = calloc(1, sizeof(struct block));
-	if (blk) {
-		blk->buf = aligned_alloc(4096, size); /* XXX static alignment :/ */
-		if (!blk->buf) {
-			free(blk);
-			blk = NULL;
-		} else {
-			INIT_LIST_HEAD(&blk->hash_head);
-			INIT_LIST_HEAD(&blk->lru_head);
-			INIT_LIST_HEAD(&blk->dirty_head);
-			INIT_LIST_HEAD(&blk->submit_head);
-			blk->blkno = blkno;
-			blk->size = size;
-		}
-	}
-
-	return blk;
-}
-
-static void free_block(struct block_data *bdat, struct block *blk)
-{
-	debug_blk(blk, "free");
-
-	if (!list_empty(&blk->lru_head)) {
-		if (blk->active)
-			bdat->nr_active--;
-		else
-			bdat->nr_inactive--;
-		list_del(&blk->lru_head);
-	}
-
-	if (!list_empty(&blk->dirty_head)) {
-		bdat->nr_dirty--;
-		list_del(&blk->dirty_head);
-	}
-
-	if (!list_empty(&blk->hash_head))
-		list_del(&blk->hash_head);
-
-	if (!list_empty(&blk->submit_head))
-		list_del(&blk->submit_head);
-
-	free(blk->buf);
-	free(blk);
-}
-
-static bool blk_is_dirty(struct block *blk)
-{
-	return !list_empty(&blk->dirty_head);
-}
-
-/*
- * Rebalance the cache.
- *
- * First we shrink the cache to limit it to max_cached blocks.
- * Logically, we walk from oldest to newest in the inactive list and
- * then in the active list.  Since these lists are physically one
- * list_head list we achieve this with a reverse walk starting from the
- * active head.
- *
- * Then we rebalnace the size of the two lists.  The constraint is that
- * we don't let the active list grow larger than the inactive list.  We
- * move blocks from the oldest tail of the active list to the newest
- * head of the inactive list.
- *
- * <- [active head] <-> [ .. active list .. ] <-> [inactive head] <-> [ .. inactive list .. ] ->
- */
-static void rebalance_cache(struct block_data *bdat)
-{
-	struct block *blk;
-	struct block *blk_;
-
-	list_for_each_entry_safe_reverse(blk, blk_, &bdat->active_head, lru_head) {
-		if ((bdat->nr_active + bdat->nr_inactive) < bdat->max_cached)
-			break;
-
-		if (&blk->lru_head == &bdat->inactive_head || blk->refcount > 0 ||
-		    blk_is_dirty(blk))
-			continue;
-
-		free_block(bdat, blk);
-	}
-
-	list_for_each_entry_safe_reverse(blk, blk_, &bdat->inactive_head, lru_head) {
-		if (bdat->nr_active <= bdat->nr_inactive || &blk->lru_head == &bdat->active_head)
-			break;
-
-		list_move(&blk->lru_head, &bdat->inactive_head);
-		blk->active = 0;
-		bdat->nr_active--;
-		bdat->nr_inactive++;
-	}
-}
-
-static void make_active(struct block_data *bdat, struct block *blk)
-{
-	if (!blk->active) {
-		if (!list_empty(&blk->lru_head)) {
-			list_move(&blk->lru_head, &bdat->active_head);
-			bdat->nr_inactive--;
-		} else {
-			list_add(&blk->lru_head, &bdat->active_head);
-		}
-
-		blk->active = 1;
-		bdat->nr_active++;
-	}
-}
-
-static int compar_iocbp(const void *A, const void *B)
-{
-	struct iocb *a = *(struct iocb **)A;
-	struct iocb *b = *(struct iocb **)B;
-
-	return scoutfs_cmp(a->aio_offset, b->aio_offset);
-}
-
-static int submit_and_wait(struct block_data *bdat, struct list_head *list)
-{
-	struct io_event *event;
-	struct iocb *iocb;
-	struct block *blk;
-	int ret;
-	int err;
-	int nr;
-	int i;
-
-	err = 0;
-	nr = 0;
-	list_for_each_entry(blk, list, submit_head) {
-		iocb = &bdat->iocbs[nr];
-		bdat->iocbps[nr] = iocb;
-
-		memset(iocb, 0, sizeof(struct iocb));
-
-		iocb->aio_data = (intptr_t)blk;
-		iocb->aio_lio_opcode = blk_is_dirty(blk) ? IOCB_CMD_PWRITE : IOCB_CMD_PREAD;
-		iocb->aio_fildes = bdat->meta_fd;
-		iocb->aio_buf = (intptr_t)blk->buf;
-		iocb->aio_nbytes = blk->size;
-		iocb->aio_offset = blk->blkno * blk->size;
-
-		nr++;
-
-		debug_blk(blk, "submit");
-
-		if ((nr < bdat->nr_events) && blk->submit_head.next != list)
-			continue;
-
-		qsort(bdat->iocbps, nr, sizeof(bdat->iocbps[0]), compar_iocbp);
-
-		ret = syscall(__NR_io_submit, bdat->ctx, nr, bdat->iocbps);
-		if (ret != nr) {
-			if (ret >= 0)
-				errno = EIO;
-			ret = -errno;
-			printf("fatal system error submitting async IO: "ENO_FMT"\n",
-				ENO_ARG(-ret));
-			goto out;
-		}
-
-		ret = syscall(__NR_io_getevents, bdat->ctx, nr, nr, bdat->events, NULL);
-		if (ret != nr) {
-			if (ret >= 0)
-				errno = EIO;
-			ret = -errno;
-			printf("fatal system error getting IO events: "ENO_FMT"\n",
-				ENO_ARG(-ret));
-			goto out;
-		}
-
-		ret = 0;
-		for (i = 0; i < nr; i++) {
-			event = &bdat->events[i];
-			iocb = (struct iocb *)(intptr_t)event->obj;
-			blk = (struct block *)(intptr_t)event->data;
-
-			debug_blk(blk, "complete res %lld", (long long)event->res);
-
-			if (event->res >= 0 && event->res != blk->size)
-				event->res = -EIO;
-
-			/* io errors are fatal */
-			if (event->res < 0) {
-				ret = event->res;
-				goto out;
-			}
-
-			if (iocb->aio_lio_opcode == IOCB_CMD_PREAD) {
-				blk->uptodate = 1;
-			} else {
-				list_del_init(&blk->dirty_head);
-				bdat->nr_dirty--;
-			}
-		}
-		nr = 0;
-	}
-
-	ret = 0;
-out:
-	return ret ?: err;
-}
-
-static void inc_refcount(struct block *blk)
-{
-	blk->refcount++;
-}
-
-void block_put(struct block **blkp)
-{
-	struct block_data *bdat = &global_bdat;
-	struct block *blk = *blkp;
-
-	if (blk) {
-		blk->refcount--;
-		*blkp = NULL;
-
-		rebalance_cache(bdat);
-	}
-}
-
-static struct list_head *hash_bucket(struct block_data *bdat, u64 blkno)
-{
-	u32 hash = scoutfs_hash32(&blkno, sizeof(blkno));
-
-	return &bdat->hash_lists[hash % bdat->hash_nr];
-}
-
-static struct block *get_or_alloc(struct block_data *bdat, u64 blkno, int bf)
-{
-	struct list_head *bucket = hash_bucket(bdat, blkno);
-	struct block *search;
-	struct block *blk;
-	size_t size;
-
-	size = (bf & BF_SM) ? SCOUTFS_BLOCK_SM_SIZE : SCOUTFS_BLOCK_LG_SIZE;
-
-	blk = NULL;
-	list_for_each_entry(search, bucket, hash_head) {
-		if (search->blkno && blkno && search->size == size) {
-			blk = search;
-			break;
-		}
-	}
-
-	if (!blk) {
-		blk = alloc_block(bdat, blkno, size);
-		if (blk) {
-			list_add(&blk->hash_head, bucket);
-			list_add(&blk->lru_head, &bdat->inactive_head);
-			bdat->nr_inactive++;
-		}
-	}
-	if (blk)
-		inc_refcount(blk);
-
-	return blk;
-}
-
-/*
- * Get a block.
- *
- * The caller holds a refcount to the block while it's in use that
- * prevents it from being removed from the cache.  It must be dropped
- * with block_put();
- */
-int block_get(struct block **blk_ret, u64 blkno, int bf)
-{
-	struct block_data *bdat = &global_bdat;
-	struct block *blk;
-	LIST_HEAD(list);
-	int ret;
-
-	blk = get_or_alloc(bdat, blkno, bf);
-	if (!blk) {
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	if ((bf & BF_ZERO)) {
-		memset(blk->buf, 0, blk->size);
-		blk->uptodate = 1;
-	}
-
-	if (bf & BF_OVERWRITE)
-		blk->uptodate = 1;
-
-	if (!blk->uptodate) {
-		list_add(&blk->submit_head, &list);
-		ret = submit_and_wait(bdat, &list);
-		list_del_init(&blk->submit_head);
-		if (ret < 0)
-			goto out;
-	}
-
-	if ((bf & BF_DIRTY) && !blk_is_dirty(blk)) {
-		list_add_tail(&bdat->dirty_list, &blk->dirty_head);
-		bdat->nr_dirty++;
-	}
-
-	make_active(bdat, blk);
-
-	rebalance_cache(bdat);
-	ret = 0;
-out:
-	if (ret < 0)
-		block_put(&blk);
-	*blk_ret = blk;
-	return ret;
-}
-
-void *block_buf(struct block *blk)
-{
-	return blk->buf;
-}
-
-size_t block_size(struct block *blk)
-{
-	return blk->size;
-}
-
-/*
- * Drop the block from the cache, regardless of if it was free or not.
- * This is used to avoid writing blocks which were dirtied but then
- * later freed.
- *
- * The block is immediately freed and can't be referenced after this
- * returns.
- */
-void block_drop(struct block **blkp)
-{
-	struct block_data *bdat = &global_bdat;
-
-	free_block(bdat, *blkp);
-	*blkp = NULL;
-	rebalance_cache(bdat);
-}
-
-/*
- * This doesn't quite work for mixing large and small blocks, but that's
- * fine, we never do that.
- */
-static int compar_u64(const void *A, const void *B)
-{
-	u64 a = *((u64 *)A);
-	u64 b = *((u64 *)B);
-
-	return scoutfs_cmp(a, b);
-}
-
-/*
- * This read-ahead is synchronous and errors are ignored.  If any of the
- * blknos aren't present in the cache then we issue concurrent reads for
- * them and wait.  Any existing cached blocks will be left as is.
- *
- * We might be trying to read a lot more than the number of events so we
- * sort the caller's blknos before iterating over them rather than
- * relying on submission sorting the blocks in each submitted set.
- */
-void block_readahead(u64 *blknos, size_t nr)
-{
-	struct block_data *bdat = &global_bdat;
-	struct block *blk;
-	struct block *blk_;
-	LIST_HEAD(list);
-	size_t i;
-
-	if (nr == 0)
-		return;
-
-	qsort(blknos, nr, sizeof(blknos[0]), compar_u64);
-
-	for (i = 0; i < nr; i++) {
-		blk = get_or_alloc(bdat, blknos[i], 0);
-		if (blk) {
-			if (!blk->uptodate)
-				list_add_tail(&blk->submit_head, &list);
-			else
-				block_put(&blk);
-		}
-	}
-
-	(void)submit_and_wait(bdat, &list);
-
-	list_for_each_entry_safe(blk, blk_, &list, submit_head) {
-		list_del_init(&blk->submit_head);
-		block_put(&blk);
-	}
-
-	rebalance_cache(bdat);
-}
-
-/*
- * The caller's block changes form a consistent transaction.  If the amount of dirty
- * blocks is large enough we issue a write.
- */
-int block_try_commit(bool force)
-{
-	struct block_data *bdat = &global_bdat;
-	struct block *blk;
-	struct block *blk_;
-	LIST_HEAD(list);
-	int ret;
-
-	if (!force && bdat->nr_dirty < bdat->nr_events)
-		return 0;
-
-	list_for_each_entry(blk, &bdat->dirty_list, dirty_head) {
-		list_add_tail(&blk->submit_head, &list);
-		inc_refcount(blk);
-	}
-
-	ret = submit_and_wait(bdat, &list);
-
-	list_for_each_entry_safe(blk, blk_, &list, submit_head) {
-		list_del_init(&blk->submit_head);
-		block_put(&blk);
-	}
-
-	if (ret < 0) {
-		printf("error writing dirty transaction blocks\n");
-		goto out;
-	}
-
-	ret = block_get(&blk, SCOUTFS_SUPER_BLKNO, BF_SM | BF_OVERWRITE | BF_DIRTY);
-	if (ret == 0) {
-		list_add(&blk->submit_head, &list);
-		ret = submit_and_wait(bdat, &list);
-		list_del_init(&blk->submit_head);
-		block_put(&blk);
-	} else {
-		ret = -ENOMEM;
-	}
-	if (ret < 0)
-		printf("error writing super block to commit transaction\n");
-
-out:
-	rebalance_cache(bdat);
-	return ret;
-}
-
-int block_setup(int meta_fd, size_t max_cached_bytes, size_t max_dirty_bytes)
-{
-	struct block_data *bdat = &global_bdat;
-	size_t i;
-	int ret;
-
-	bdat->max_cached = DIV_ROUND_UP(max_cached_bytes, SCOUTFS_BLOCK_LG_SIZE);
-	bdat->hash_nr = bdat->max_cached / 4;
-	bdat->nr_events = DIV_ROUND_UP(max_dirty_bytes, SCOUTFS_BLOCK_LG_SIZE);
-
-	bdat->iocbs = calloc(bdat->nr_events, sizeof(bdat->iocbs[0]));
-	bdat->iocbps = calloc(bdat->nr_events, sizeof(bdat->iocbps[0]));
-	bdat->events = calloc(bdat->nr_events, sizeof(bdat->events[0]));
-	bdat->hash_lists = calloc(bdat->hash_nr, sizeof(bdat->hash_lists[0]));
-	if (!bdat->iocbs || !bdat->iocbps || !bdat->events || !bdat->hash_lists) {
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	INIT_LIST_HEAD(&bdat->active_head);
-	INIT_LIST_HEAD(&bdat->inactive_head);
-	INIT_LIST_HEAD(&bdat->dirty_list);
-	bdat->meta_fd = meta_fd;
-	list_add(&bdat->inactive_head, &bdat->active_head);
-
-	for (i = 0; i < bdat->hash_nr; i++)
-		INIT_LIST_HEAD(&bdat->hash_lists[i]);
-
-	ret = syscall(__NR_io_setup, bdat->nr_events, &bdat->ctx);
-
-out:
-	if (ret < 0) {
-		free(bdat->iocbs);
-		free(bdat->iocbps);
-		free(bdat->events);
-		free(bdat->hash_lists);
-	}
-
-	return ret;
-}
-
-void block_shutdown(void)
-{
-	struct block_data *bdat = &global_bdat;
-
-	syscall(SYS_io_destroy, bdat->ctx);
-
-	free(bdat->iocbs);
-	free(bdat->iocbps);
-	free(bdat->events);
-	free(bdat->hash_lists);
-}

utils/src/check/block.h

@@ -1,32 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_BLOCK_H_
-#define _SCOUTFS_UTILS_CHECK_BLOCK_H_
-
-#include <unistd.h>
-#include <stdbool.h>
-
-struct block;
-
-#include "sparse.h"
-
-/* block flags passed to block_get() */
-enum {
-	BF_ZERO      = (1 << 0), /* zero contents buf as block is returned */
-	BF_DIRTY     = (1 << 1), /* block will be written with transaction */
-	BF_SM        = (1 << 2), /* small 4k block instead of large 64k block */
-	BF_OVERWRITE = (1 << 3), /* caller will overwrite contents, don't read */
-};
-
-int block_get(struct block **blk_ret, u64 blkno, int bf);
-void block_put(struct block **blkp);
-
-void *block_buf(struct block *blk);
-size_t block_size(struct block *blk);
-void block_drop(struct block **blkp);
-
-void block_readahead(u64 *blknos, size_t nr);
-int block_try_commit(bool force);
-
-int block_setup(int meta_fd, size_t max_cached_bytes, size_t max_dirty_bytes);
-void block_shutdown(void);
-
-#endif

utils/src/check/btree.c

@@ -1,209 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdbool.h>
-#include <errno.h>
-
-#include "sparse.h"
-#include "util.h"
-#include "format.h"
-#include "key.h"
-#include "avl.h"
-
-#include "block.h"
-#include "btree.h"
-#include "extent.h"
-#include "iter.h"
-#include "sns.h"
-#include "meta.h"
-#include "problem.h"
-
-static inline void *item_val(struct scoutfs_btree_block *bt, struct scoutfs_btree_item *item)
-{
-	return (void *)bt + le16_to_cpu(item->val_off);
-}
-
-static void readahead_refs(struct scoutfs_btree_block *bt)
-{
-	struct scoutfs_btree_item *item;
-	struct scoutfs_avl_node *node;
-	struct scoutfs_block_ref *ref;
-	u64 *blknos;
-	u64 blkno;
-	u16 valid = 0;
-	u16 nr = le16_to_cpu(bt->nr_items);
-	int i;
-
-	blknos = calloc(nr, sizeof(blknos[0]));
-	if (!blknos)
-		return;
-
-	node = avl_first(&bt->item_root);
-
-	for (i = 0; i < nr; i++) {
-		item = container_of(node, struct scoutfs_btree_item, node);
-		ref = item_val(bt, item);
-		blkno = le64_to_cpu(ref->blkno);
-
-		if (valid_meta_blkno(blkno))
-			blknos[valid++] = blkno;
-
-		node = avl_next(&bt->item_root, &item->node);
-	}
-
-	if (valid > 0)
-		block_readahead(blknos, valid);
-	free(blknos);
-}
-
-/*
- * Call the callback on the referenced block.  Then if the block
- * contains referneces read it and recurse into all its references.
- */
-static int btree_ref_meta_iter(struct scoutfs_block_ref *ref, unsigned level, extent_cb_t cb,
-			       void *cb_arg)
-{
-	struct scoutfs_btree_item *item;
-	struct scoutfs_btree_block *bt;
-	struct scoutfs_avl_node *node;
-	struct block *blk = NULL;
-	u64 blkno;
-	int ret;
-	int i;
-
-	blkno = le64_to_cpu(ref->blkno);
-	if (!blkno)
-		return 0;
-
-	ret = cb(blkno, 1, cb_arg);
-	if (ret < 0) {
-		ret = xlate_iter_errno(ret);
-		return 0;
-	}
-
-	if (level == 0)
-		return 0;
-
-	ret = block_get(&blk, blkno, 0);
-	if (ret < 0)
-		return ret;
-
-	sns_push("btree_parent", blkno, 0);
-
-	bt = block_buf(blk);
-
-	/* XXX integrate verification with block cache */
-	if (bt->level != level) {
-		problem(PB_BTREE_BLOCK_BAD_LEVEL, "expected %u level %u", level, bt->level);
-		ret = -EINVAL;
-		goto out;
-	}
-
-	/* read-ahead last level of parents */
-	if (level == 2)
-		readahead_refs(bt);
-
-	node = avl_first(&bt->item_root);
-
-	for (i = 0; i < le16_to_cpu(bt->nr_items); i++) {
-		item = container_of(node, struct scoutfs_btree_item, node);
-		ref = item_val(bt, item);
-
-		ret = btree_ref_meta_iter(ref, level - 1, cb, cb_arg);
-		if (ret < 0)
-			goto out;
-
-		node = avl_next(&bt->item_root, &item->node);
-	}
-
-	ret = 0;
-out:
-	block_put(&blk);
-	sns_pop();
-
-	return ret;
-}
-
-int btree_meta_iter(struct scoutfs_btree_root *root, extent_cb_t cb, void *cb_arg)
-{
-	/* XXX check root */
-	if (root->height == 0)
-		return 0;
-
-	return btree_ref_meta_iter(&root->ref, root->height - 1, cb, cb_arg);
-}
-
-static int btree_ref_item_iter(struct scoutfs_block_ref *ref, unsigned level,
-			       btree_item_cb_t cb, void *cb_arg)
-{
-	struct scoutfs_btree_item *item;
-	struct scoutfs_btree_block *bt;
-	struct scoutfs_avl_node *node;
-	struct block *blk = NULL;
-	u64 blkno;
-	int ret;
-	int i;
-
-	blkno = le64_to_cpu(ref->blkno);
-	if (!blkno)
-		return 0;
-
-	ret = block_get(&blk, blkno, 0);
-	if (ret < 0)
-		return ret;
-
-	if (level)
-		sns_push("btree_parent", blkno, 0);
-	else
-		sns_push("btree_leaf", blkno, 0);
-
-	bt = block_buf(blk);
-
-	/* XXX integrate verification with block cache */
-	if (bt->level != level) {
-		problem(PB_BTREE_BLOCK_BAD_LEVEL, "expected %u level %u", level, bt->level);
-		ret = -EINVAL;
-		goto out;
-	}
-
-	/* read-ahead leaves that contain items */
-	if (level == 1)
-		readahead_refs(bt);
-
-	node = avl_first(&bt->item_root);
-
-	for (i = 0; i < le16_to_cpu(bt->nr_items); i++) {
-		item = container_of(node, struct scoutfs_btree_item, node);
-
-		if (level) {
-			ref = item_val(bt, item);
-			ret = btree_ref_item_iter(ref, level - 1, cb, cb_arg);
-		} else {
-			ret = cb(&item->key, item_val(bt, item),
-				 le16_to_cpu(item->val_len), cb_arg);
-			debug("free item key "SK_FMT" ret %d", SK_ARG(&item->key), ret);
-		}
-		if (ret < 0) {
-			ret = xlate_iter_errno(ret);
-			goto out;
-		}
-
-		node = avl_next(&bt->item_root, &item->node);
-	}
-
-	ret = 0;
-out:
-	block_put(&blk);
-	sns_pop();
-
-	return ret;
-}
-
-int btree_item_iter(struct scoutfs_btree_root *root, btree_item_cb_t cb, void *cb_arg)
-{
-	/* XXX check root */
-	if (root->height == 0)
-		return 0;
-
-	return btree_ref_item_iter(&root->ref, root->height - 1, cb, cb_arg);
-}

utils/src/check/btree.h

@@ -1,14 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_BTREE_H_
-#define _SCOUTFS_UTILS_CHECK_BTREE_H_
-
-#include "util.h"
-#include "format.h"
-
-#include "extent.h"
-
-typedef int (*btree_item_cb_t)(struct scoutfs_key *key, void *val, u16 val_len, void *cb_arg);
-
-int btree_meta_iter(struct scoutfs_btree_root *root, extent_cb_t cb, void *cb_arg);
-int btree_item_iter(struct scoutfs_btree_root *root, btree_item_cb_t cb, void *cb_arg);
-
-#endif

utils/src/check/check.c

@@ -1,149 +0,0 @@
-#define _GNU_SOURCE /* O_DIRECT */
-#include <unistd.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <sys/ioctl.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <string.h>
-#include <assert.h>
-#include <stdbool.h>
-#include <argp.h>
-
-#include "sparse.h"
-#include "parse.h"
-#include "util.h"
-#include "format.h"
-#include "ioctl.h"
-#include "cmd.h"
-#include "dev.h"
-
-#include "alloc.h"
-#include "block.h"
-#include "debug.h"
-#include "meta.h"
-#include "super.h"
-
-struct check_args {
-	char *meta_device;
-	char *data_device;
-	char *debug_path;
-};
-
-static int do_check(struct check_args *args)
-{
-	int debug_fd = -1;
-	int meta_fd = -1;
-	int data_fd = -1;
-	int ret;
-
-	if (args->debug_path) {
-		debug_fd = open(args->debug_path, O_WRONLY | O_CREAT | O_TRUNC, 0644);
-		if (debug_fd < 0) {
-			ret = -errno;
-			fprintf(stderr, "error opening debug output file '%s': %s (%d)\n",
-				args->debug_path, strerror(errno), errno);
-			goto out;
-		}
-
-		debug_enable(debug_fd);
-	}
-
-	meta_fd = open(args->meta_device, O_DIRECT | O_RDWR | O_EXCL);
-	if (meta_fd < 0) {
-		ret = -errno;
-		fprintf(stderr, "failed to open meta device '%s': %s (%d)\n",
-			args->meta_device, strerror(errno), errno);
-		goto out;
-	}
-
-	data_fd = open(args->data_device, O_DIRECT | O_RDWR | O_EXCL);
-	if (data_fd < 0) {
-		ret = -errno;
-		fprintf(stderr, "failed to open data device '%s': %s (%d)\n",
-			args->data_device, strerror(errno), errno);
-		goto out;
-	}
-
-	ret = block_setup(meta_fd, 128 * 1024 * 1024, 32 * 1024 * 1024);
-	if (ret < 0)
-		goto out;
-
-	ret = check_supers() ?:
-	      check_meta_alloc();
-out:
-	/* and tear it all down */
-	block_shutdown();
-	super_shutdown();
-	debug_disable();
-
-	if (meta_fd >= 0)
-		close(meta_fd);
-	if (data_fd >= 0)
-		close(data_fd);
-	if (debug_fd >= 0)
-		close(debug_fd);
-
-	return ret;
-}
-
-static int parse_opt(int key, char *arg, struct argp_state *state)
-{
-	struct check_args *args = state->input;
-
-	switch (key) {
-	case 'd':
-		args->debug_path = strdup_or_error(state, arg);
-		break;
-	case 'e':
-	case ARGP_KEY_ARG:
-		if (!args->meta_device)
-			args->meta_device = strdup_or_error(state, arg);
-		else if (!args->data_device)
-			args->data_device = strdup_or_error(state, arg);
-		else
-			argp_error(state, "more than two device arguments given");
-		break;
-	case ARGP_KEY_FINI:
-		if (!args->meta_device)
-			argp_error(state, "no metadata device argument given");
-		if (!args->data_device)
-			argp_error(state, "no data device argument given");
-		break;
-	default:
-		break;
-	}
-
-	return 0;
-}
-
-static struct argp_option options[] = {
-	{ "debug", 'd', "FILE_PATH", 0, "Path to debug output file, will be created or truncated"},
-	{ NULL }
-};
-
-static struct argp argp = {
-	options,
-	parse_opt,
-	"META-DEVICE DATA-DEVICE",
-	"Check filesystem consistency"
-};
-
-static int check_cmd(int argc, char **argv)
-{
-	struct check_args check_args = {NULL};
-	int ret;
-
-	ret = argp_parse(&argp, argc, argv, 0, NULL, &check_args);
-	if (ret)
-		return ret;
-
-	return do_check(&check_args);
-}
-
-static void __attribute__((constructor)) check_ctor(void)
-{
-	cmd_register_argp("check", &argp, GROUP_CORE, check_cmd);
-}

utils/src/check/debug.c

@@ -1,16 +0,0 @@
-#include <stdlib.h>
-
-#include "debug.h"
-
-int debug_fd = -1;
-
-void debug_enable(int fd)
-{
-	debug_fd = fd;
-}
-
-void debug_disable(void)
-{
-	if (debug_fd >= 0)
-		debug_fd = -1;
-}

utils/src/check/debug.h

@@ -1,17 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_DEBUG_H_
-#define _SCOUTFS_UTILS_CHECK_DEBUG_H_
-
-#include <stdio.h>
-
-#define debug(fmt, args...)				\
-do {							\
-	if (debug_fd >= 0)				\
-		dprintf(debug_fd, fmt"\n", ##args);	\
-} while (0)
-
-extern int debug_fd;
-
-void debug_enable(int fd);
-void debug_disable(void);
-
-#endif

utils/src/check/eno.h

@@ -1,9 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_ENO_H_
-#define _SCOUTFS_UTILS_CHECK_ENO_H_
-
-#include <errno.h>
-
-#define ENO_FMT		"%d (%s)"
-#define ENO_ARG(eno)	eno, strerror(eno)
-
-#endif

utils/src/check/extent.c

@@ -1,312 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdbool.h>
-#include <errno.h>
-
-#include "util.h"
-#include "lk_rbtree_wrapper.h"
-
-#include "debug.h"
-#include "extent.h"
-
-/*
- * In-memory extent management in rbtree nodes.
- */
-
-bool extents_overlap(u64 a_start, u64 a_len, u64 b_start, u64 b_len)
-{
-	u64 a_end = a_start + a_len;
-	u64 b_end = b_start + b_len;
-
-	return !((a_end <= b_start) || (b_end <= a_start));
-}
-
-static int ext_contains(struct extent_node *ext, u64 start, u64 len)
-{
-	return ext->start <= start && ext->start + ext->len >= start + len;
-}
-
-/*
- * True if the given extent is bisected by the given range; there's
- * leftover containing extents on both the left and right sides of the
- * range in the extent.
- */
-static int ext_bisected(struct extent_node *ext, u64 start, u64 len)
-{
-	return ext->start < start && ext->start + ext->len > start + len;
-}
-
-static struct extent_node *ext_from_rbnode(struct rb_node *rbnode)
-{
-	return rbnode ? container_of(rbnode, struct extent_node, rbnode) : NULL;
-}
-
-static struct extent_node *next_ext(struct extent_node *ext)
-{
-	return ext ? ext_from_rbnode(rb_next(&ext->rbnode)) : NULL;
-}
-
-static struct extent_node *prev_ext(struct extent_node *ext)
-{
-	return ext ? ext_from_rbnode(rb_prev(&ext->rbnode)) : NULL;
-}
-
-struct walk_results {
-	unsigned bisect_to_leaf:1;
-	struct extent_node *found;
-	struct extent_node *next;
-	struct rb_node *parent;
-	struct rb_node **node;
-};
-
-static void walk_extents(struct extent_root *root, u64 start, u64 len, struct walk_results *wlk)
-{
-	struct rb_node **node = &root->rbroot.rb_node;
-	struct extent_node *ext;
-	u64 end = start + len;
-	int cmp;
-
-	wlk->found = NULL;
-	wlk->next = NULL;
-	wlk->parent = NULL;
-
-	while (*node) {
-		wlk->parent = *node;
-		ext = ext_from_rbnode(*node);
-		cmp = end <= ext->start ? -1 :
-		      start >= ext->start + ext->len ? 1 : 0;
-
-		if (cmp < 0) {
-			node = &ext->rbnode.rb_left;
-			wlk->next = ext;
-		} else if (cmp > 0) {
-			node = &ext->rbnode.rb_right;
-		} else {
-			wlk->found = ext;
-			if (!(wlk->bisect_to_leaf && ext_bisected(ext, start, len)))
-				break;
-			/* walk right so we can insert greater right from bisection */
-			node = &ext->rbnode.rb_right;
-		}
-	}
-
-	wlk->node = node;
-}
-
-/*
- * Return an extent that overlaps with the given range.
- */
-int extent_lookup(struct extent_root *root, u64 start, u64 len, struct extent_node *found)
-{
-	struct walk_results wlk = { 0, };
-	int ret;
-
-	walk_extents(root, start, len, &wlk);
-	if (wlk.found) {
-		memset(found, 0, sizeof(struct extent_node));
-		found->start = wlk.found->start;
-		found->len = wlk.found->len;
-		ret = 0;
-	} else {
-		ret = -ENOENT;
-	}
-
-	return ret;
-}
-
-/*
- * Callers can iterate through direct node references and are entirely
- * responsible for consistency when doing so.
- */
-struct extent_node *extent_first(struct extent_root *root)
-{
-	struct walk_results wlk = { 0, };
-
-	walk_extents(root, 0, 1, &wlk);
-
-	return wlk.found ?: wlk.next;
-}
-
-struct extent_node *extent_next(struct extent_node *ext)
-{
-	return next_ext(ext);
-}
-
-struct extent_node *extent_prev(struct extent_node *ext)
-{
-	return prev_ext(ext);
-}
-
-/*
- * Insert a new extent into the tree.  We can extend existing nodes,
- * merge with neighbours, or remove existing extents entirely if we
- * insert a range that fully spans existing nodes.
- */
-static int walk_insert(struct extent_root *root, u64 start, u64 len, int found_err)
-{
-	struct walk_results wlk = { 0, };
-	struct extent_node *ext;
-	struct extent_node *nei;
-	int ret;
-
-	walk_extents(root, start, len, &wlk);
-
-	ext = wlk.found;
-	if (ext && found_err) {
-		ret = found_err;
-		goto out;
-	}
-
-	if (!ext) {
-		ext = malloc(sizeof(struct extent_node));
-		if (!ext) {
-			ret = -ENOMEM;
-			goto out;
-		}
-
-		ext->start = start;
-		ext->len = len;
-
-		rb_link_node(&ext->rbnode, wlk.parent, wlk.node);
-		rb_insert_color(&ext->rbnode, &root->rbroot);
-	}
-
-	/* start by expanding an existing extent if our range is larger */
-	if (start < ext->start) {
-		ext->len += ext->start - start;
-		ext->start = start;
-	}
-	if (ext->start + ext->len < start + len)
-		ext->len += (start + len) - (ext->start + ext->len);
-
-	/* drop any fully spanned neighbors, possibly merging with a final adjacent one */
-
-	while ((nei = prev_ext(ext))) {
-		if (nei->start + nei->len < ext->start)
-			break;
-
-		if (nei->start < ext->start) {
-			ext->len += ext->start - nei->start;
-			ext->start = nei->start;
-		}
-
-		rb_erase(&nei->rbnode, &root->rbroot);
-		free(nei);
-	}
-
-	while ((nei = next_ext(ext))) {
-		if (ext->start + ext->len < nei->start)
-			break;
-
-		if (ext->start + ext->len < nei->start + nei->len)
-			ext->len += (nei->start + nei->len) - (ext->start + ext->len);
-
-		rb_erase(&nei->rbnode, &root->rbroot);
-		free(nei);
-	}
-
-	ret = 0;
-out:
-	debug("start %llu len %llu ret %d", start, len, ret);
-	return ret;
-}
-
-/*
- * Insert a new extent.  The specified extent must not overlap with any
- * existing extents or -EEXIST is returned.
- */
-int extent_insert_new(struct extent_root *root, u64 start, u64 len)
-{
-	return walk_insert(root, start, len, true);
-}
-
-/*
- * Insert an extent, extending any existing extents that may overlap.
- */
-int extent_insert_extend(struct extent_root *root, u64 start, u64 len)
-{
-	return walk_insert(root, start, len, false);
-}
-
-/*
- * Remove the specified extent from an existing node.  The given extent must be fully
- * contained in a single node or -ENOENT is returned.
- */
-int extent_remove(struct extent_root *root, u64 start, u64 len)
-{
-	struct extent_node *ext;
-	struct extent_node *ins;
-	struct walk_results wlk = {
-		.bisect_to_leaf = 1,
-	};
-	int ret;
-
-	walk_extents(root, start, len, &wlk);
-
-	if (!(ext = wlk.found) || !ext_contains(ext, start, len)) {
-		ret = -ENOENT;
-		goto out;
-	}
-
-	if (ext_bisected(ext, start, len)) {
-		debug("found bisected start %llu len %llu", ext->start, ext->len);
-		ins = malloc(sizeof(struct extent_node));
-		if (!ins) {
-			ret = -ENOMEM;
-			goto out;
-		}
-
-		ins->start = start + len;
-		ins->len = (ext->start + ext->len) - ins->start;
-
-		rb_link_node(&ins->rbnode, wlk.parent, wlk.node);
-		rb_insert_color(&ins->rbnode, &root->rbroot);
-	}
-
-	if (start > ext->start) {
-		ext->len = start - ext->start;
-	} else if (len < ext->len) {
-		ext->start += len;
-		ext->len -= len;
-	} else {
-		rb_erase(&ext->rbnode, &root->rbroot);
-	}
-
-	ret = 0;
-out:
-	debug("start %llu len %llu ret %d", start, len, ret);
-
-	return ret;
-}
-
-void extent_root_init(struct extent_root *root)
-{
-	root->rbroot = RB_ROOT;
-	root->total = 0;
-}
-
-void extent_root_free(struct extent_root *root)
-{
-	struct extent_node *ext;
-	struct rb_node *node;
-	struct rb_node *tmp;
-
-	for (node = rb_first(&root->rbroot); node && ((tmp = rb_next(node)), 1); node = tmp) {
-		ext = rb_entry(node, struct extent_node, rbnode);
-		rb_erase(&ext->rbnode, &root->rbroot);
-		free(ext);
-	}
-}
-
-void extent_root_print(struct extent_root *root)
-{
-	struct extent_node *ext;
-	struct rb_node *node;
-	struct rb_node *tmp;
-
-	for (node = rb_first(&root->rbroot); node && ((tmp = rb_next(node)), 1); node = tmp) {
-		ext = rb_entry(node, struct extent_node, rbnode);
-		debug("  start %llu len %llu", ext->start, ext->len);
-	}
-}

utils/src/check/extent.h

@@ -1,38 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_EXTENT_H_
-#define _SCOUTFS_UTILS_CHECK_EXTENT_H_
-
-#include "lk_rbtree_wrapper.h"
-
-struct extent_root {
-	struct rb_root rbroot;
-	u64 total;
-};
-
-struct extent_node {
-	struct rb_node rbnode;
-	u64 start;
-	u64 len;
-};
-
-typedef int (*extent_cb_t)(u64 start, u64 len, void *arg);
-
-struct extent_cb_arg_t {
-	extent_cb_t cb;
-	void *cb_arg;
-};
-
-bool extents_overlap(u64 a_start, u64 a_len, u64 b_start, u64 b_len);
-
-int extent_lookup(struct extent_root *root, u64 start, u64 len, struct extent_node *found);
-struct extent_node *extent_first(struct extent_root *root);
-struct extent_node *extent_next(struct extent_node *ext);
-struct extent_node *extent_prev(struct extent_node *ext);
-int extent_insert_new(struct extent_root *root, u64 start, u64 len);
-int extent_insert_extend(struct extent_root *root, u64 start, u64 len);
-int extent_remove(struct extent_root *root, u64 start, u64 len);
-
-void extent_root_init(struct extent_root *root);
-void extent_root_free(struct extent_root *root);
-void extent_root_print(struct extent_root *root);
-
-#endif

utils/src/check/image.c

@@ -1,540 +0,0 @@
-#define _GNU_SOURCE /* O_DIRECT */
-#include <unistd.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <fcntl.h>
-#include <errno.h>
-#include <string.h>
-#include <stdbool.h>
-#include <argp.h>
-
-#include "sparse.h"
-#include "bitmap.h"
-#include "parse.h"
-#include "util.h"
-#include "format.h"
-#include "crc.h"
-#include "cmd.h"
-#include "dev.h"
-
-#include "alloc.h"
-#include "block.h"
-#include "btree.h"
-#include "log_trees.h"
-#include "super.h"
-
-/* huh. */
-#define OFF_MAX (off_t)((u64)((off_t)~0ULL) >> 1)
-
-#define SCOUTFS_META_IMAGE_HEADER_MAGIC		0x8aee00d098fa60c5ULL
-#define SCOUTFS_META_IMAGE_BLOCK_HEADER_MAGIC	0x70bd5e9269effd86ULL
-
-struct scoutfs_meta_image_header {
-	__le64 magic;
-	__le64 total_bytes;
-	__le32 version;
-} __packed;
-
-struct scoutfs_meta_image_block_header {
-	__le64 magic;
-	__le64 offset;
-	__le32 size;
-	__le32 crc;
-} __packed;
-
-struct image_args {
-	char *meta_device;
-	bool is_read;
-	bool show_header;
-	u64 ra_window;
-};
-
-struct block_bitmaps {
-	unsigned long *bits;
-	u64 size;
-	u64 count;
-};
-
-#define errf(fmt, args...) \
-	dprintf(STDERR_FILENO, fmt, ##args)
-
-static int set_meta_bit(u64 start, u64 len, void *arg)
-{
-	struct block_bitmaps *bm = arg;
-	int ret;
-
-	if (len != 1) {
-		ret = -EINVAL;
-	} else {
-		if (!test_bit(bm->bits, start)) {
-			set_bit(bm->bits, start);
-			bm->count++;
-		}
-		ret = 0;
-	}
-
-	return ret;
-}
-
-static int get_ref_bits(struct block_bitmaps *bm)
-{
-	struct scoutfs_super_block *super = global_super;
-	int ret;
-	u64 i;
-
-	/*
-	 * There are almost no small blocks we need to read, so we read
-	 * them as the large blocks that contain them to simplify the
-	 * block reading process.
-	 */
-	set_meta_bit(SCOUTFS_SUPER_BLKNO >> SCOUTFS_BLOCK_SM_LG_SHIFT, 1, bm);
-
-	for (i = 0; i < SCOUTFS_QUORUM_BLOCKS; i++)
-		set_meta_bit((SCOUTFS_QUORUM_BLKNO + i) >> SCOUTFS_BLOCK_SM_LG_SHIFT, 1, bm);
-
-	ret = alloc_root_meta_iter(&super->meta_alloc[0], set_meta_bit, bm) ?:
-	      alloc_root_meta_iter(&super->meta_alloc[1], set_meta_bit, bm) ?:
-	      alloc_root_meta_iter(&super->data_alloc, set_meta_bit, bm) ?:
-	      alloc_list_meta_iter(&super->server_meta_avail[0], set_meta_bit, bm) ?:
-	      alloc_list_meta_iter(&super->server_meta_avail[1], set_meta_bit, bm) ?:
-	      alloc_list_meta_iter(&super->server_meta_freed[0], set_meta_bit, bm) ?:
-	      alloc_list_meta_iter(&super->server_meta_freed[1], set_meta_bit, bm) ?:
-	      btree_meta_iter(&super->fs_root, set_meta_bit, bm) ?:
-	      btree_meta_iter(&super->logs_root, set_meta_bit, bm) ?:
-	      btree_meta_iter(&super->log_merge, set_meta_bit, bm) ?:
-	      btree_meta_iter(&super->mounted_clients, set_meta_bit, bm) ?:
-	      btree_meta_iter(&super->srch_root, set_meta_bit, bm) ?:
-	      log_trees_meta_iter(set_meta_bit, bm);
-
-	return ret;
-}
-
-/*
- * Note that this temporarily modifies the header that it's given.
- */
-static __le32 calc_crc(struct scoutfs_meta_image_block_header *bh, void *buf, size_t size)
-{
-	__le32 saved = bh->crc;
-	u32 crc = ~0;
-
-	bh->crc = 0;
-	crc = crc32c(crc, bh, sizeof(*bh));
-	crc = crc32c(crc, buf, size);
-	bh->crc = saved;
-
-	return cpu_to_le32(crc);
-}
-
-static void printf_header(struct scoutfs_meta_image_header *hdr)
-{
-	errf("magic: 0x%016llx\n"
-	     "total_bytes: %llu\n"
-	     "version: %u\n",
-	       le64_to_cpu(hdr->magic),
-	       le64_to_cpu(hdr->total_bytes),
-	       le32_to_cpu(hdr->version));
-}
-
-typedef ssize_t (*rw_func_t)(int fd, void *buf, size_t count, off_t offset);
-
-static inline ssize_t rw_read(int fd, void *buf, size_t count, off_t offset)
-{
-	return read(fd, buf, count);
-}
-
-static inline ssize_t rw_pread(int fd, void *buf, size_t count, off_t offset)
-{
-	return pread(fd, buf, count, offset);
-}
-
-static inline ssize_t rw_write(int fd, void *buf, size_t count, off_t offset)
-{
-	return write(fd, buf, count);
-}
-
-static inline ssize_t rw_pwrite(int fd, void *buf, size_t count, off_t offset)
-{
-	return pwrite(fd, buf, count, offset);
-}
-
-static int rw_full_count(rw_func_t func, u64 *tot, int fd, void *buf, size_t count, off_t offset)
-{
-	ssize_t sret;
-
-	while (count > 0) {
-		sret = func(fd, buf, count, offset);
-		if (sret <= 0 || sret > count) {
-			if (sret < 0)
-				return -errno;
-			else
-				return -EIO;
-		}
-
-		if (tot)
-			*tot += sret;
-		buf += sret;
-		count -= sret;
-	}
-
-	return 0;
-}
-
-static int read_image(struct image_args *args, int fd, struct block_bitmaps *bm)
-{
-	struct scoutfs_meta_image_block_header bh;
-	struct scoutfs_meta_image_header hdr;
-	u64 opening;
-	void *buf;
-	off_t off;
-	u64 bit;
-	u64 ra;
-	int ret;
-
-	buf = malloc(SCOUTFS_BLOCK_LG_SIZE);
-	if (!buf) {
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	hdr.magic = cpu_to_le64(SCOUTFS_META_IMAGE_HEADER_MAGIC);
-	hdr.total_bytes = cpu_to_le64(sizeof(hdr) +
-				      (bm->count * (SCOUTFS_BLOCK_LG_SIZE + sizeof(bh))));
-	hdr.version = cpu_to_le32(1);
-
-	if (args->show_header) {
-		printf_header(&hdr);
-		ret = 0;
-		goto out;
-	}
-
-	ret = rw_full_count(rw_write, NULL, STDOUT_FILENO, &hdr, sizeof(hdr), 0);
-	if (ret < 0)
-		goto out;
-
-	opening = args->ra_window;
-	ra = 0;
-	bit = 0;
-
-	for (bit = 0; (bit = find_next_set_bit(bm->bits, bit, bm->size)) < bm->size; bit++) {
-
-		/* readahead to open the full window, then a block at a time */
-		do {
-			ra = find_next_set_bit(bm->bits, ra, bm->size);
-			if (ra < bm->size) {
-				off = ra << SCOUTFS_BLOCK_LG_SHIFT;
-				posix_fadvise(fd, off, SCOUTFS_BLOCK_LG_SIZE, POSIX_FADV_WILLNEED);
-				ra++;
-				if (opening)
-					opening -= min(opening, SCOUTFS_BLOCK_LG_SIZE);
-			}
-		} while (opening > 0);
-
-		off = bit << SCOUTFS_BLOCK_LG_SHIFT;
-		ret = rw_full_count(rw_pread, NULL, fd, buf, SCOUTFS_BLOCK_LG_SIZE, off);
-		if (ret < 0)
-			goto out;
-
-		/*
-		 * Might as well try to drop the pages we've used to
-		 * reduce memory pressure on our read-ahead pages that
-		 * are waiting.
-		 */
-		posix_fadvise(fd, off, SCOUTFS_BLOCK_LG_SIZE, POSIX_FADV_DONTNEED);
-
-		bh.magic = SCOUTFS_META_IMAGE_BLOCK_HEADER_MAGIC;
-		bh.offset = cpu_to_le64(off);
-		bh.size = cpu_to_le32(SCOUTFS_BLOCK_LG_SIZE);
-		bh.crc = calc_crc(&bh, buf, SCOUTFS_BLOCK_LG_SIZE);
-
-		ret = rw_full_count(rw_write, NULL, STDOUT_FILENO, &bh, sizeof(bh), 0) ?:
-		      rw_full_count(rw_write, NULL, STDOUT_FILENO, buf, SCOUTFS_BLOCK_LG_SIZE, 0);
-		if (ret < 0)
-			goto out;
-	}
-
-out:
-	free(buf);
-
-	return ret;
-}
-
-static int invalid_header(struct scoutfs_meta_image_header *hdr)
-{
-	if (le64_to_cpu(hdr->magic) != SCOUTFS_META_IMAGE_HEADER_MAGIC) {
-		errf("bad image header magic 0x%016llx (!= expected %016llx)\n",
-		       le64_to_cpu(hdr->magic), SCOUTFS_META_IMAGE_HEADER_MAGIC);
-
-	} else if (le32_to_cpu(hdr->version) != 1) {
-		errf("unknown image header version %u\n", le32_to_cpu(hdr->version));
-
-	} else {
-		return 0;
-	}
-
-	return -EIO;
-}
-
-/*
- * Doesn't catch offset+size overflowing, presumes pwrite() will return
- * an error.
- */
-static int invalid_block_header(struct scoutfs_meta_image_block_header *bh)
-{
-	if (le64_to_cpu(bh->magic) != SCOUTFS_META_IMAGE_BLOCK_HEADER_MAGIC) {
-		errf("bad block header magic 0x%016llx (!= expected %016llx)\n",
-		       le64_to_cpu(bh->magic), SCOUTFS_META_IMAGE_BLOCK_HEADER_MAGIC);
-
-	} else if (le32_to_cpu(bh->size) == 0) {
-		errf("invalid block header size %u\n", le32_to_cpu(bh->size));
-
-	} else if (le32_to_cpu(bh->size) > SIZE_MAX) {
-		errf("block header size %u too large for size_t (> %zu)\n",
-		       le32_to_cpu(bh->size), (size_t)SIZE_MAX);
-
-	} else if (le64_to_cpu(bh->offset) > OFF_MAX) {
-		errf("block header offset %llu too large for off_t (> %llu)\n",
-		       le64_to_cpu(bh->offset), (u64)OFF_MAX);
-
-	} else {
-		return 0;
-	}
-
-	return -EIO;
-}
-
-static int write_image(struct image_args *args, int fd, struct block_bitmaps *bm)
-{
-	struct scoutfs_meta_image_block_header bh;
-	struct scoutfs_meta_image_header hdr;
-	size_t writeback_batch = (2 * 1024 * 1024);
-	size_t buf_size;
-	size_t dirty;
-	size_t size;
-	off_t first;
-	off_t last;
-	off_t off;
-	__le32 calc;
-	void *buf;
-	u64 tot;
-	int ret;
-
-	tot = 0;
-
-	ret = rw_full_count(rw_read, &tot, STDIN_FILENO, &hdr, sizeof(hdr), 0);
-	if (ret < 0)
-		goto out;
-
-	if (args->show_header) {
-		printf_header(&hdr);
-		ret = 0;
-		goto out;
-	}
-
-	ret = invalid_header(&hdr);
-	if (ret < 0)
-		goto out;
-
-	dirty = 0;
-	first = OFF_MAX;
-	last = 0;
-	buf = NULL;
-	buf_size = 0;
-
-	while (tot < le64_to_cpu(hdr.total_bytes)) {
-
-		ret = rw_full_count(rw_read, &tot, STDIN_FILENO, &bh, sizeof(bh), 0);
-		if (ret < 0)
-			goto out;
-
-		ret = invalid_block_header(&bh);
-		if (ret < 0)
-			goto out;
-
-		size = le32_to_cpu(bh.size);
-		if (buf_size < size) {
-			buf = realloc(buf, size);
-			if (!buf) {
-				ret = -ENOMEM;
-				goto out;
-			}
-
-			buf_size = size;
-		}
-
-		ret = rw_full_count(rw_read, &tot, STDIN_FILENO, buf, size, 0);
-		if (ret < 0)
-			goto out;
-
-		calc = calc_crc(&bh, buf, size);
-		if (calc != bh.crc) {
-			errf("crc err");
-			ret = -EIO;
-			goto out;
-		}
-
-		off = le64_to_cpu(bh.offset);
-
-		ret = rw_full_count(rw_pwrite, NULL, fd, buf, size, off);
-		if (ret < 0)
-			goto out;
-
-		dirty += size;
-		first = min(first, off);
-		last = max(last, off);
-		if (dirty >= writeback_batch) {
-			posix_fadvise(fd, first, last, POSIX_FADV_DONTNEED);
-			dirty = 0;
-			first = OFF_MAX;
-			last = 0;
-		}
-	}
-
-	ret = fsync(fd);
-	if (ret < 0) {
-		ret = -errno;
-		goto out;
-	}
-
-out:
-	return ret;
-}
-
-static int do_image(struct image_args *args)
-{
-	struct block_bitmaps bm = { .bits = NULL };
-	int meta_fd = -1;
-	u64 dev_size;
-	mode_t mode;
-	int ret;
-
-	mode = args->is_read ? O_RDONLY : O_RDWR;
-
-	meta_fd = open(args->meta_device, mode);
-	if (meta_fd < 0) {
-		ret = -errno;
-		errf("failed to open meta device '%s': %s (%d)\n",
-		     args->meta_device, strerror(errno), errno);
-		goto out;
-	}
-
-	if (args->is_read) {
-		ret = flush_device(meta_fd);
-		if (ret < 0)
-			goto out;
-
-		ret = get_device_size(args->meta_device, meta_fd, &dev_size);
-		if (ret < 0)
-			goto out;
-
-		bm.size = DIV_ROUND_UP(dev_size, SCOUTFS_BLOCK_LG_SIZE);
-		bm.bits = calloc(1, round_up(bm.size, BITS_PER_LONG) / 8);
-		if (!bm.bits) {
-			ret = -ENOMEM;
-			goto out;
-		}
-
-		ret = block_setup(meta_fd, 128 * 1024 * 1024, 32 * 1024 * 1024) ?:
-		      check_supers() ?:
-		      get_ref_bits(&bm) ?:
-		      read_image(args, meta_fd, &bm);
-		block_shutdown();
-	} else {
-		ret = write_image(args, meta_fd, &bm);
-	}
-out:
-	free(bm.bits);
-
-	if (meta_fd >= 0)
-		close(meta_fd);
-
-	return ret;
-}
-
-static int parse_opt(int key, char *arg, struct argp_state *state)
-{
-	struct image_args *args = state->input;
-	int ret;
-
-	switch (key) {
-	case 'h':
-		args->show_header = true;
-		break;
-	case 'r':
-		ret = parse_u64(arg, &args->ra_window);
-		if (ret)
-			argp_error(state, "readahead winddoe parse error");
-		break;
-	case ARGP_KEY_ARG:
-		if (!args->meta_device)
-			args->meta_device = strdup_or_error(state, arg);
-		else
-			argp_error(state, "more than two device arguments given");
-		break;
-	case ARGP_KEY_FINI:
-		if (!args->meta_device)
-			argp_error(state, "no metadata device argument given");
-		break;
-	default:
-		break;
-	}
-
-	return 0;
-}
-
-static struct argp_option options[] = {
-	{ "show-header", 'h', NULL, 0, "Print image header and exit without processing stream" },
-	{ "readahead", 'r', "NR", 0, "Maintain read-ahead window of NR blocks" },
-	{ NULL }
-};
-
-static struct argp read_image_argp = {
-	options,
-	parse_opt,
-	"META-DEVICE",
-	"Read metadata image stream from metadata device file"
-};
-
-#define DEFAULT_RA_WINDOW (512 * 1024)
-
-static int read_image_cmd(int argc, char **argv)
-{
-	struct image_args image_args = {
-		.is_read = true,
-		.ra_window = DEFAULT_RA_WINDOW,
-	};
-	int ret;
-
-	ret = argp_parse(&read_image_argp, argc, argv, 0, NULL, &image_args);
-	if (ret)
-		return ret;
-
-	return do_image(&image_args);
-}
-
-static struct argp write_image_argp = {
-	options,
-	parse_opt,
-	"META-DEVICE",
-	"Write metadata image stream to metadata device file"
-};
-
-static int write_image_cmd(int argc, char **argv)
-{
-	struct image_args image_args = {
-		.is_read = false,
-		.ra_window = DEFAULT_RA_WINDOW,
-	};
-	int ret;
-
-	ret = argp_parse(&write_image_argp, argc, argv, 0, NULL, &image_args);
-	if (ret)
-		return ret;
-
-	return do_image(&image_args);
-}
-
-static void __attribute__((constructor)) image_ctor(void)
-{
-	cmd_register_argp("read-metadata-image", &read_image_argp, GROUP_CORE, read_image_cmd);
-	cmd_register_argp("write-metadata-image", &write_image_argp, GROUP_CORE, write_image_cmd);
-}

utils/src/check/iter.h

@@ -1,15 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_ITER_H_
-#define _SCOUTFS_UTILS_CHECK_ITER_H_
-
-/*
- * Callbacks can return a weird -errno that we'll never use to indicate
- * that iteration can stop and return 0 for success.
- */
-#define ECHECK_ITER_DONE EL2HLT
-
-static inline int xlate_iter_errno(int ret)
-{
-	return ret == -ECHECK_ITER_DONE ? 0 : ret;
-}
-
-#endif

utils/src/check/log_trees.c

@@ -1,98 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <errno.h>
-
-#include "sparse.h"
-#include "util.h"
-#include "format.h"
-#include "key.h"
-
-#include "alloc.h"
-#include "btree.h"
-#include "debug.h"
-#include "extent.h"
-#include "iter.h"
-#include "sns.h"
-#include "log_trees.h"
-#include "super.h"
-
-struct iter_args {
-	extent_cb_t cb;
-	void *cb_arg;
-};
-
-static int lt_meta_iter(struct scoutfs_key *key, void *val, u16 val_len, void *cb_arg)
-{
-	struct iter_args *ia = cb_arg;
-	struct scoutfs_log_trees *lt;
-	int ret;
-
-	if (val_len != sizeof(struct scoutfs_log_trees))
-		; /* XXX */
-
-	lt = val;
-
-	sns_push("log_trees", le64_to_cpu(lt->rid), le64_to_cpu(lt->nr));
-
-	debug("lt rid 0x%16llx nr %llu", le64_to_cpu(lt->rid), le64_to_cpu(lt->nr));
-
-	sns_push("meta_avail", 0, 0);
-	ret = alloc_list_meta_iter(&lt->meta_avail, ia->cb, ia->cb_arg);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("meta_freed", 0, 0);
-	ret = alloc_list_meta_iter(&lt->meta_freed, ia->cb, ia->cb_arg);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("item_root", 0, 0);
-	ret = btree_meta_iter(&lt->item_root, ia->cb, ia->cb_arg);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	if (lt->bloom_ref.blkno) {
-		sns_push("bloom_ref", 0, 0);
-		ret = ia->cb(le64_to_cpu(lt->bloom_ref.blkno), 1, ia->cb_arg);
-		sns_pop();
-		if (ret < 0) {
-			ret = xlate_iter_errno(ret);
-			goto out;
-		}
-	}
-
-	sns_push("data_avail", 0, 0);
-	ret = alloc_root_meta_iter(&lt->data_avail, ia->cb, ia->cb_arg);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("data_freed", 0, 0);
-	ret = alloc_root_meta_iter(&lt->data_freed, ia->cb, ia->cb_arg);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	ret = 0;
-out:
-	sns_pop();
-
-	return ret;
-}
-
-/*
- * Call the callers callback with the extent of all the metadata block references contained
- * in log btrees.  We walk the logs_root btree items and walk all the metadata structures
- * they reference.
- */
-int log_trees_meta_iter(extent_cb_t cb, void *cb_arg)
-{
-	struct scoutfs_super_block *super = global_super;
-	struct iter_args ia = { .cb = cb, .cb_arg = cb_arg };
-
-	return btree_item_iter(&super->logs_root, lt_meta_iter, &ia);
-}

utils/src/check/log_trees.h

@@ -1,8 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_LOG_TREES_H_
-#define _SCOUTFS_UTILS_CHECK_LOG_TREES_H_
-
-#include "extent.h"
-
-int log_trees_meta_iter(extent_cb_t cb, void *cb_arg);
-
-#endif

utils/src/check/meta.c

@@ -1,367 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdbool.h>
-#include <sys/mman.h>
-#include <errno.h>
-
-#include "sparse.h"
-#include "util.h"
-#include "format.h"
-#include "bitmap.h"
-#include "key.h"
-
-#include "alloc.h"
-#include "btree.h"
-#include "debug.h"
-#include "extent.h"
-#include "sns.h"
-#include "log_trees.h"
-#include "meta.h"
-#include "problem.h"
-#include "super.h"
-
-static struct meta_data {
-	struct extent_root meta_refed;
-	struct extent_root meta_free;
-	struct {
-		u64 ref_blocks;
-		u64 free_extents;
-		u64 free_blocks;
-	} stats;
-} global_mdat;
-
-bool valid_meta_blkno(u64 blkno)
-{
-	u64 tot = le64_to_cpu(global_super->total_meta_blocks);
-
-	return blkno >= SCOUTFS_META_DEV_START_BLKNO && blkno < tot;
-}
-
-static bool valid_meta_extent(u64 start, u64 len)
-{
-	u64 tot = le64_to_cpu(global_super->total_meta_blocks);
-	bool valid;
-
-	valid = len > 0 &&
-		start >= SCOUTFS_META_DEV_START_BLKNO &&
-		start < tot &&
-		len <= tot &&
-		((start + len) <= tot) &&
-		((start + len) > start);
-
-	debug("start %llu len %llu valid %u", start, len, !!valid);
-
-	if (!valid)
-		problem(PB_META_EXTENT_INVALID, "start %llu len %llu", start, len);
-
-	return valid;
-}
-
-/*
- * Track references to individual metadata blocks.  This uses the extent
- * callback type but is only ever called for single block references.
- * Any reference to a block that has already been referenced is
- * considered invalid and is ignored.  Later repair will resolve
- * duplicate references.
- */
-static int insert_meta_ref(u64 start, u64 len, void *arg)
-{
-	struct meta_data *mdat = &global_mdat;
-	struct extent_root *root = arg;
-	int ret = 0;
-
-	/* this is tracking single metadata block references */
-	if (len != 1) {
-		ret = -EINVAL;
-		goto out;
-	}
-
-	if (valid_meta_blkno(start)) {
-		ret = extent_insert_new(root, start, len);
-		if (ret == 0)
-			mdat->stats.ref_blocks++;
-		else if (ret == -EEXIST)
-			problem(PB_META_REF_OVERLAPS_EXISTING, "blkno %llu", start);
-	}
-
-out:
-	return ret;
-}
-
-static int insert_meta_free(u64 start, u64 len, void *arg)
-{
-	struct meta_data *mdat = &global_mdat;
-	struct extent_root *root = arg;
-	int ret = 0;
-
-	if (valid_meta_extent(start, len)) {
-		ret = extent_insert_new(root, start, len);
-		if (ret == 0) {
-			mdat->stats.free_extents++;
-			mdat->stats.free_blocks++;
-
-		} else if (ret == -EEXIST) {
-			problem(PB_META_FREE_OVERLAPS_EXISTING,
-				"start %llu llen %llu", start, len);
-		}
-
-	}
-
-	return ret;
-}
-
-/*
- * Walk all metadata references in the system.  This walk doesn't need
- * to read metadata that doesn't contain any metadata references so it
- * can skip the bulk of metadata blocks.  This gives us the set of
- * referenced metadata blocks which we can then use to repair metadata
- * allocator structures.
- */
-static int get_meta_refs(void)
-{
-	struct meta_data *mdat = &global_mdat;
-	struct scoutfs_super_block *super = global_super;
-	int ret;
-
-	extent_root_init(&mdat->meta_refed);
-
-	/* XXX record reserved blocks around super as referenced */
-
-	sns_push("meta_alloc", 0, 0);
-	ret = alloc_root_meta_iter(&super->meta_alloc[0], insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("meta_alloc", 1, 0);
-	ret = alloc_root_meta_iter(&super->meta_alloc[1], insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("data_alloc", 1, 0);
-	ret = alloc_root_meta_iter(&super->data_alloc, insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("server_meta_avail", 0, 0);
-	ret = alloc_list_meta_iter(&super->server_meta_avail[0],
-				   insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("server_meta_avail", 1, 0);
-	ret = alloc_list_meta_iter(&super->server_meta_avail[1],
-				   insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("server_meta_freed", 0, 0);
-	ret = alloc_list_meta_iter(&super->server_meta_freed[0],
-				   insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("server_meta_freed", 1, 0);
-	ret = alloc_list_meta_iter(&super->server_meta_freed[1],
-				   insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("fs_root", 0, 0);
-	ret = btree_meta_iter(&super->fs_root, insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("logs_root", 0, 0);
-	ret = btree_meta_iter(&super->logs_root, insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("log_merge", 0, 0);
-	ret = btree_meta_iter(&super->log_merge, insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("mounted_clients", 0, 0);
-	ret = btree_meta_iter(&super->mounted_clients, insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("srch_root", 0, 0);
-	ret = btree_meta_iter(&super->srch_root, insert_meta_ref, &mdat->meta_refed);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	ret = log_trees_meta_iter(insert_meta_ref, &mdat->meta_refed);
-	if (ret < 0)
-		goto out;
-
-	printf("found %llu referenced metadata blocks\n", mdat->stats.ref_blocks);
-	ret = 0;
-out:
-	return ret;
-}
-
-static int get_meta_free(void)
-{
-	struct meta_data *mdat = &global_mdat;
-	struct scoutfs_super_block *super = global_super;
-	int ret;
-
-	extent_root_init(&mdat->meta_free);
-
-	sns_push("meta_alloc", 0, 0);
-	ret = alloc_root_extent_iter(&super->meta_alloc[0], insert_meta_free, &mdat->meta_free);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("meta_alloc", 1, 0);
-	ret = alloc_root_extent_iter(&super->meta_alloc[1], insert_meta_free, &mdat->meta_free);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("server_meta_avail", 0, 0);
-	ret = alloc_list_extent_iter(&super->server_meta_avail[0],
-				     insert_meta_free, &mdat->meta_free);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("server_meta_avail", 1, 0);
-	ret = alloc_list_extent_iter(&super->server_meta_avail[1],
-				     insert_meta_free, &mdat->meta_free);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("server_meta_freed", 0, 0);
-	ret = alloc_list_extent_iter(&super->server_meta_freed[0],
-				     insert_meta_free, &mdat->meta_free);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	sns_push("server_meta_freed", 1, 0);
-	ret = alloc_list_extent_iter(&super->server_meta_freed[1],
-				     insert_meta_free, &mdat->meta_free);
-	sns_pop();
-	if (ret < 0)
-		goto out;
-
-	printf("found %llu free metadata blocks in %llu extents\n",
-	       mdat->stats.free_blocks, mdat->stats.free_extents);
-	ret = 0;
-out:
-	return ret;
-}
-
-/*
- * All the space between referenced blocks must be recorded in the free
- * extents.  The free extent walk didn't check that the extents
- * overlapped with references, we do that here.  Remember that metadata
- * block references were merged into extents here, the refed extents
- * aren't necessarily all a single block.
- */
-static int compare_refs_and_free(void)
-{
-	struct meta_data *mdat = &global_mdat;
-	struct extent_node *ref;
-	struct extent_node *free;
-	struct extent_node *next;
-	struct extent_node *prev;
-	u64 expect;
-	u64 start;
-	u64 end;
-
-	expect = 0;
-	ref = extent_first(&mdat->meta_refed);
-	free = extent_first(&mdat->meta_free);
-	while (ref || free) {
-
-		printf("exp %llu ref %llu.%llu free %llu.%llu\n",
-			expect, ref ? ref->start : 0, ref ? ref->len : 0,
-			free ? free->start : 0, free ? free->len : 0);
-
-		/* referenced marked free, remove ref from free and continue from same point */
-		if (ref && free && extents_overlap(ref->start, ref->len, free->start, free->len)) {
-			printf("ref extent %llu.%llu overlaps free %llu %llu\n",
-				ref->start, ref->len, free->start, free->len);
-
-			start = max(ref->start, free->start);
-			end = min(ref->start + ref->len, free->start + free->len);
-
-			prev = extent_prev(free);
-
-			extent_remove(&mdat->meta_free, start, end - start);
-
-			if (prev)
-				free = extent_next(prev);
-			else
-				free = extent_first(&mdat->meta_free);
-			continue;
-		}
-
-		/* see which extent starts earlier */
-		if (!free || (ref && ref->start <= free->start))
-			next = ref;
-		else
-			next = free;
-
-		/* untracked region before next extent */
-		if (expect < next->start) {
-			printf("missing free extent %llu.%llu\n", expect, next->start - expect);
-			expect = next->start;
-			continue;
-		}
-
-
-		/* didn't overlap, advance past next extent */
-		expect = next->start + next->len;
-		if (next == ref)
-			ref = extent_next(ref);
-		else
-			free = extent_next(free);
-	}
-
-	return 0;
-}
-
-/*
- * Check the metadata allocators by comparing the set of referenced
- * blocks with the set of free blocks that are stored in free btree
- * items and alloc list blocks.
- */
-int check_meta_alloc(void)
-{
-	int ret;
-
-	ret = get_meta_refs();
-	if (ret < 0)
-		goto out;
-
-	ret = get_meta_free();
-	if (ret < 0)
-		goto out;
-
-	ret = compare_refs_and_free();
-	if (ret < 0)
-		goto out;
-
-	ret = 0;
-out:
-	return ret;
-}

utils/src/check/meta.h

@@ -1,9 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_META_H_
-#define _SCOUTFS_UTILS_CHECK_META_H_
-
-bool valid_meta_blkno(u64 blkno);
-
-int check_meta_alloc(void);
-
-#endif
-

utils/src/check/padding.c

@@ -1,23 +0,0 @@
-#include <string.h>
-#include <stdbool.h>
-
-#include "util.h"
-#include "padding.h"
-
-bool padding_is_zeros(const void *data, size_t sz)
-{
-	static char zeros[32] = {0,};
-	const size_t batch = array_size(zeros);
-
-	while (sz >= batch) {
-		if (memcmp(data, zeros, batch))
-			return false;
-		data += batch;
-		sz -= batch;
-	}
-
-	if (sz > 0 && memcmp(data, zeros, sz))
-		return false;
-
-	return true;
-}

utils/src/check/padding.h

@@ -1,6 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_PADDING_H_
-#define _SCOUTFS_UTILS_CHECK_PADDING_H_
-
-bool padding_is_zeros(const void *data, size_t sz);
-
-#endif

utils/src/check/problem.c

@@ -1,23 +0,0 @@
-#include <stdio.h>
-#include <stdint.h>
-
-#include "problem.h"
-
-#if 0
-#define PROB_STR(pb) [pb] = #pb
-static char *prob_strs[] = {
-	PROB_STR(PB_META_EXTENT_INVALID),
-	PROB_STR(PB_META_EXTENT_OVERLAPS_EXISTING),
-};
-#endif
-
-static struct problem_data {
-	uint64_t counts[PB__NR];
-} global_pdat;
-
-void problem_record(prob_t pb)
-{
-	struct problem_data *pdat = &global_pdat;
-
-	pdat->counts[pb]++;
-}

utils/src/check/problem.h

@@ -1,23 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_PROBLEM_H_
-#define _SCOUTFS_UTILS_CHECK_PROBLEM_H_
-
-#include "debug.h"
-#include "sns.h"
-
-typedef enum {
-	PB_META_EXTENT_INVALID,
-	PB_META_REF_OVERLAPS_EXISTING,
-	PB_META_FREE_OVERLAPS_EXISTING,
-	PB_BTREE_BLOCK_BAD_LEVEL,
-	PB__NR,
-} prob_t;
-
-#define problem(pb, fmt, ...)							\
-do {										\
-	debug("problem found: "#pb": %s: "fmt, sns_str(), __VA_ARGS__);	\
-	problem_record(pb);							\
-} while (0)
-
-void problem_record(prob_t pb);
-
-#endif

utils/src/check/sns.c

@@ -1,118 +0,0 @@
-#include <stdlib.h>
-#include <string.h>
-
-#include "sns.h"
-
-/*
- * This "str num stack" is used to describe our location in metadata at
- * any given time.
- *
- * As we descend into structures we pop a string on decribing them,
- * perhaps with associated numbers.  Pushing and popping is very cheap
- * and only rarely do we format the stack into a string, as an arbitrary
- * example:
- *   super.fs_root.btree_parent:1231.btree_leaf:3231"
- */
-
-#define SNS_MAX_DEPTH	1000
-#define SNS_STR_SIZE	(SNS_MAX_DEPTH * (SNS_MAX_STR_LEN + 1 + 16 + 1))
-
-static struct sns_data {
-	unsigned int depth;
-
-	struct sns_entry {
-		char *str;
-		size_t len;
-		u64 a;
-		u64 b;
-	} ents[SNS_MAX_DEPTH];
-
-	char str[SNS_STR_SIZE];
-
-} global_lsdat;
-
-void _sns_push(char *str, size_t len, u64 a, u64 b)
-{
-	struct sns_data *lsdat = &global_lsdat;
-
-	if (lsdat->depth < SNS_MAX_DEPTH) {
-		lsdat->ents[lsdat->depth++] = (struct sns_entry) {
-			.str = str,
-			.len = len,
-			.a = a,
-			.b = b,
-		};
-	}
-}
-
-void sns_pop(void)
-{
-	struct sns_data *lsdat = &global_lsdat;
-
-	if (lsdat->depth > 0)
-		lsdat->depth--;
-}
-
-static char *append_str(char *pos, char *str, size_t len)
-{
-	memcpy(pos, str, len);
-	return pos + len;
-}
-
-/*
- * This is not called for x = 0 so we don't need to emit an initial 0.
- * We could by using do {} while instead of while {}.
- */
-static char *append_u64x(char *pos, u64 x)
-{
-	static char hex[] = "0123456789abcdef";
-
-	while (x) {
-		*pos++ = hex[x & 0xf];
-		x >>= 4;
-	}
-
-	return pos;
-}
-
-static char *append_char(char *pos, char c)
-{
-	*(pos++) = c;
-	return pos;
-}
-
-/*
- * Return a pointer to a null terminated string that describes the
- * current location stack.  The string buffer is global.
- */
-char *sns_str(void)
-{
-	struct sns_data *lsdat = &global_lsdat;
-	struct sns_entry *ent;
-	char *pos;
-	int i;
-
-	pos = lsdat->str;
-	for (i = 0; i < lsdat->depth; i++) {
-		ent = &lsdat->ents[i];
-
-		if (i)
-			pos = append_char(pos, '.');
-
-		pos = append_str(pos, ent->str, ent->len);
-
-		if (ent->a) {
-			pos = append_char(pos, ':');
-			pos = append_u64x(pos, ent->a);
-		}
-
-		if (ent->b) {
-			pos = append_char(pos, ':');
-			pos = append_u64x(pos, ent->b);
-		}
-	}
-
-	*pos = '\0';
-
-	return lsdat->str;
-}

utils/src/check/sns.h

@@ -1,20 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_SNS_H_
-#define _SCOUTFS_UTILS_CHECK_SNS_H_
-
-#include <assert.h>
-
-#include "sparse.h"
-
-#define SNS_MAX_STR_LEN 20
-
-#define sns_push(str, a, b)					\
-do {								\
-	build_assert(sizeof(str) - 1 <= SNS_MAX_STR_LEN);	\
-	_sns_push((str), sizeof(str) - 1, a, b);		\
-} while (0)
-
-void _sns_push(char *str, size_t len, u64 a, u64 b);
-void sns_pop(void);
-char *sns_str(void);
-
-#endif

utils/src/check/super.c

@@ -1,57 +0,0 @@
-#include <unistd.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <errno.h>
-
-#include "sparse.h"
-#include "util.h"
-#include "format.h"
-
-#include "block.h"
-#include "super.h"
-
-/*
- * After we check the super blocks we provide a global buffer to track
- * the current super block.  It is referenced to get static information
- * about the system and is also modified and written as part of
- * transactions.
- */
-struct scoutfs_super_block *global_super;
-
-/*
- * After checking the supers we save a copy of it in a global buffer that's used by
- * other modules to track the current super.  It can be modified and written during commits.
- */
-int check_supers(void)
-{
-	struct scoutfs_super_block *super = NULL;
-	struct block *blk = NULL;
-	int ret;
-
-	global_super = malloc(sizeof(struct scoutfs_super_block));
-	if (!global_super) {
-		printf("error allocating super block buffer\n");
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	ret = block_get(&blk, SCOUTFS_SUPER_BLKNO, BF_SM);
-	if (ret < 0) {
-		printf("error reading super block\n");
-		goto out;
-	}
-
-	super = block_buf(blk);
-
-	memcpy(global_super, super, sizeof(struct scoutfs_super_block));
-	ret = 0;
-out:
-	block_put(&blk);
-
-	return ret;
-}
-
-void super_shutdown(void)
-{
-	free(global_super);
-}

utils/src/check/super.h

@@ -1,9 +0,0 @@
-#ifndef _SCOUTFS_UTILS_CHECK_SUPER_H_
-#define _SCOUTFS_UTILS_CHECK_SUPER_H_
-
-extern struct scoutfs_super_block *global_super;
-
-int check_supers(void);
-void super_shutdown(void);
-
-#endif

utils/src/key.h

@@ -141,4 +141,13 @@ static inline void scoutfs_key_dec(struct scoutfs_key *key)
 	key->sk_zone--;
 }
 
+static inline void scoutfs_xattr_get_indx_key(struct scoutfs_key *key, u8 *major, u64 *minor,
+					      u64 *ino, u64 *xid)
+{
+	*major = le64_to_cpu(key->_sk_first) >> 56;
+	*minor = (le64_to_cpu(key->_sk_first) << 8) | (le64_to_cpu(key->_sk_second) >> 56);
+	*ino = (le64_to_cpu(key->_sk_second) << 8) | (le64_to_cpu(key->_sk_third) >> 56);
+	*xid = (le64_to_cpu(key->_sk_third) << 8) | key->_sk_fourth;
+}
+
 #endif

utils/src/list.h

@@ -156,16 +156,6 @@ static inline void list_move_tail(struct list_head *list,
         list_add_tail(list, head);
 }
 
-/**
- * list_is_head - tests whether @list is the list @head
- * @list: the entry to test
- * @head: the head of the list
- */
-static inline int list_is_head(const struct list_head *list, const struct list_head *head)
-{
-	return list == head;
-}
-
 /**
  * list_empty - tests whether a list is empty
  * @head: the list to test.
@@ -252,15 +242,6 @@ static inline void list_splice_init(struct list_head *list,
         for (pos = (head)->next, n = pos->next; pos != (head); \
                 pos = n, n = pos->next)
 
-/**
- * list_entry_is_head - test if the entry points to the head of the list
- * @pos:	the type * to cursor
- * @head:	the head for your list.
- * @member:	the name of the list_head within the struct.
- */
-#define list_entry_is_head(pos, head, member)				\
-	(&pos->member == (head))
-
 /**
  * list_for_each_entry        -        iterate over list of given type
  * @pos:        the type * to use as a loop counter.
@@ -326,28 +307,4 @@ static inline void list_splice_init(struct list_head *list,
 #define list_next_entry(pos, member) \
         list_entry((pos)->member.next, typeof(*(pos)), member)
 
-/**
- * list_prev_entry - get the prev element in list
- * @pos:	the type * to cursor
- * @member:	the name of the list_head within the struct.
- */
-#define list_prev_entry(pos, member) \
-	list_entry((pos)->member.prev, typeof(*(pos)), member)
-
-/**
- * list_for_each_entry_safe_reverse - iterate backwards over list safe against removal
- * @pos:	the type * to use as a loop cursor.
- * @n:		another type * to use as temporary storage
- * @head:	the head for your list.
- * @member:	the name of the list_head within the struct.
- *
- * Iterate backwards over list of given type, safe against removal
- * of list entry.
- */
-#define list_for_each_entry_safe_reverse(pos, n, head, member)		\
-	for (pos = list_last_entry(head, typeof(*pos), member),		\
-		n = list_prev_entry(pos, member);			\
-	     !list_entry_is_head(pos, head, member); 			\
-	     pos = n, n = list_prev_entry(n, member))
-
 #endif

utils/src/lk_rbtree_wrapper.h

@@ -1,24 +0,0 @@
-#ifndef _LK_RBTREE_WRAPPER_H_
-#define _LK_RBTREE_WRAPPER_H_
-
-/*
- * We're using this lame hack to build and use the kernel's rbtree in
- * userspace.  We drop the kernel's rbtree*[ch] implementation in and
- * use them with this wrapper.  We only have to remove the kernel
- * includes from the imported files.
- */
-
-#include <stdbool.h>
-#include "util.h"
-
-#define rcu_assign_pointer(a, b)	do { a = b; } while (0)
-#define READ_ONCE(a)			({ a; })
-#define WRITE_ONCE(a, b)		do { a = b; } while (0)
-#define unlikely(a)			({ a; })
-#define EXPORT_SYMBOL(a)		/* nop */
-
-#include "rbtree_types.h"
-#include "rbtree.h"
-#include "rbtree_augmented.h"
-
-#endif