[6.18] Track btrfs patches

.github/workflows/makefile.yml

@@ -0,0 +1,34 @@
+name: Makefile CI
+
+on:
+  push:
+    branches:
+      - 'rebase-*/btrfs-patches'
+  pull_request:
+    branches:
+      - 'rebase-*/btrfs-patches'
+  workflow_dispatch:
+
+jobs:
+  build:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: Configure minimal kernel
+      run: make tinyconfig
+
+    - name: Configure btrfs
+      run: |
+        echo "CONFIG_BLOCK=y" >>.config
+        echo "CONFIG_BTRFS_FS=y" >>.config
+        echo "CONFIG_BTRFS_FS_POSIX_ACL=y" >>.config
+        echo "CONFIG_BTRFS_ALLOCATOR_HINTS=y" >>.config
+        echo "CONFIG_BTRFS_PER_DEVICE_IO_STATS=y" >>.config
+        echo "CONFIG_BTRFS_READ_POLICIES=y" >>.config
+        make oldconfig
+
+    - name: Compile kernel
+      run: make -j$(nproc) all

.gitignore

@@ -183,3 +183,6 @@ sphinx_*/
 
 # Rust analyzer configuration
 /rust-project.json
+
+# Allow Github workflows
+!/.github

fs/btrfs/Kconfig

@@ -84,6 +84,78 @@ config BTRFS_ASSERT
 
 	  If unsure, say N.
 
+config BTRFS_ALLOCATOR_HINTS
+	bool "Btrfs allocator hints"
+	depends on BTRFS_FS
+	default n
+	help
+	  Enable support for allocator hints. This feature allows to select
+	  dedicated or preferred devices for meta data vs data, or prevent
+	  allocation from a device at all. This feature does not interact
+	  well with free space calculation because the formula expects to
+	  allocate space always from a device with most free space which is
+	  not true when hints are applied. It may also create issues if a
+	  device from the pool dies resulting in a situation where there are
+	  still enough RAID mirror members but the allocation hints don't
+	  allow to allocate from specific devices.
+
+	  You are advised to watch your free space closely with btrfs tools
+	  instead of relying on df only.
+
+	  Mounting a btrfs with this feature on or off is always possible,
+	  there are no incompatible changes to the file system. But running
+	  without this feature may place new chunks on unwanted devices and
+	  you may want to clean up later by balancing the affected chunks.
+
+	  Supported hint types in /sys/fs/btrfs/BTRFS-UUID/devinfo/ID/type:
+
+	  - type = 0 - allocate data chunks from this ID first (recommended
+	               for big disks with good sequential performance, e.g.
+	               HDDs), prefers data on this device
+	  - type = 1 - allocate meta data chunks from this ID first
+	               (recommended for fast and small disks with good
+	               latency, e.g. SSD/NVMe), prefers meta data on this
+	               device
+	  - type = 2 - allocate only meta data chunks from this ID, no data
+	               chunks will ever be allocated from this device
+	  - type = 3 - allocate only data chunks from this ID, no meta data
+	               chunks will ever be allocated from this device
+	  - type = 4 - allocate any chunks from this device last, will never
+	               allocate any space from this device unless there isn't
+	               enough space on other devices
+	  - type = 5 - never allocate any new chunks, useful when putting a
+	               device out of use and to avoid redundant chunk writes
+	               during balance/replace
+
+	  If unsure, say N.
+
+config BTRFS_PER_DEVICE_IO_STATS
+	bool "Btrfs per io devices stats"
+	depends on BTRFS_FS
+	default n
+	help
+	  Enable collecting io read stats per devices to evaluate the effects
+	  of different read policies better.
+
+	  This adds a new file /sys/fs/btrfs/BTRFS-UUID/devinfo/ID/read_stats.
+
+	  If unsure, say N.
+
+config BTRFS_READ_POLICIES
+	bool "Btrfs read policies"
+	depends on BTRFS_FS
+	default n
+	help
+	  This enables btrfs read policies to control how btrfs selects stripes
+	  from a mirror during read operations. This was originally part of
+	  the experimental feature set but it is safe to use and can provide
+	  huge performance benefits in certain scenarios without causing any
+	  performance regressions.
+
+	  This adds a new file /sys/fs/btrfs/BTRFS-UUID/read_policy.
+
+	  If unsure, say N.
+
 config BTRFS_EXPERIMENTAL
 	bool "Btrfs experimental features"
 	depends on BTRFS_FS

fs/btrfs/backref.c

@@ -1126,6 +1126,7 @@ static int add_inline_refs(struct btrfs_backref_walk_ctx *ctx,
 		if (ret)
 			return ret;
 		ptr += btrfs_extent_inline_ref_size(type);
+		cond_resched();
 	}
 
 	return 0;
@@ -1229,7 +1230,7 @@ static int add_keyed_refs(struct btrfs_backref_walk_ctx *ctx,
 		}
 		if (ret)
 			return ret;
-
+		cond_resched();
 	}
 
 	return ret;

fs/btrfs/super.c

@@ -2498,7 +2498,7 @@ static int __init btrfs_print_mod_info(void)
 #endif
 			;
 
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 	if (btrfs_get_mod_read_policy() == NULL)
 		pr_info("Btrfs loaded%s\n", options);
 	else
@@ -2565,7 +2565,7 @@ static const struct init_sequence mod_init_seq[] = {
 	}, {
 		.init_func = btrfs_extent_map_init,
 		.exit_func = btrfs_extent_map_exit,
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 	}, {
 		.init_func = btrfs_read_policy_init,
 		.exit_func = NULL,

fs/btrfs/sysfs.c

@@ -26,6 +26,10 @@
 #include "fs.h"
 #include "accessors.h"
 
+#ifdef CONFIG_BTRFS_PER_DEVICE_IO_STATS
+#include <linux/part_stat.h>
+#endif
+
 /*
  * Structure name                       Path
  * --------------------------------------------------------------------------
@@ -1319,13 +1323,14 @@ BTRFS_ATTR(, temp_fsid, btrfs_temp_fsid_show);
 
 static const char *btrfs_read_policy_name[] = {
 	"pid",
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 	"round-robin",
+	"queue",
 	"devid",
 #endif
 };
 
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 
 /* Global module configuration parameters. */
 static char *read_policy;
@@ -1337,7 +1342,7 @@ char *btrfs_get_mod_read_policy(void)
 /* Set perms to 0, disable /sys/module/btrfs/parameter/read_policy interface. */
 module_param(read_policy, charp, 0);
 MODULE_PARM_DESC(read_policy,
-"Global read policy: pid (default), round-robin[:<min_contig_read>], devid[:<devid>]");
+"Global read policy: pid (default), round-robin[:<min_contig_read>], queue, devid[:<devid>]");
 #endif
 
 int btrfs_read_policy_to_enum(const char *str, s64 *value_ret)
@@ -1350,7 +1355,7 @@ int btrfs_read_policy_to_enum(const char *str, s64 *value_ret)
 
 	strscpy(param, str);
 
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 	/* Separate value from input in policy:value format. */
 	value_str = strchr(param, ':');
 	if (value_str) {
@@ -1372,7 +1377,7 @@ int btrfs_read_policy_to_enum(const char *str, s64 *value_ret)
 	return sysfs_match_string(btrfs_read_policy_name, param);
 }
 
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 int __init btrfs_read_policy_init(void)
 {
 	s64 value;
@@ -1403,7 +1408,7 @@ static ssize_t btrfs_read_policy_show(struct kobject *kobj,
 
 		ret += sysfs_emit_at(buf, ret, "%s", btrfs_read_policy_name[i]);
 
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 		if (i == BTRFS_READ_POLICY_RR)
 			ret += sysfs_emit_at(buf, ret, ":%u",
 					     READ_ONCE(fs_devices->rr_min_contig_read));
@@ -1433,7 +1438,7 @@ static ssize_t btrfs_read_policy_store(struct kobject *kobj,
 	if (index < 0)
 		return -EINVAL;
 
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 	/* If moving from RR then disable collecting fs stats. */
 	if (fs_devices->read_policy == BTRFS_READ_POLICY_RR && index != BTRFS_READ_POLICY_RR)
 		fs_devices->collect_fs_stats = false;
@@ -2140,19 +2145,120 @@ static ssize_t btrfs_devinfo_error_stats_show(struct kobject *kobj,
 }
 BTRFS_ATTR(devid, error_stats, btrfs_devinfo_error_stats_show);
 
+#ifdef CONFIG_BTRFS_ALLOCATOR_HINTS
+static ssize_t btrfs_devinfo_type_show(struct kobject *kobj,
+					    struct kobj_attribute *a, char *buf)
+{
+	struct btrfs_device *device = container_of(kobj, struct btrfs_device,
+						   devid_kobj);
+
+	return scnprintf(buf, PAGE_SIZE, "0x%08llx\n", device->type);
+}
+
+static ssize_t btrfs_devinfo_type_store(struct kobject *kobj,
+				 struct kobj_attribute *a,
+				 const char *buf, size_t len)
+{
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_root *root;
+	struct btrfs_device *device;
+	int ret;
+	struct btrfs_trans_handle *trans;
+
+	u64 type, prev_type;
+
+	device = container_of(kobj, struct btrfs_device, devid_kobj);
+	fs_info = device->fs_info;
+	if (!fs_info)
+		return -EPERM;
+
+	/*
+	 * Changing the type field requires starting a transaction which will cause a NULL derefernce in
+	 * __reserve_bytes if the file system is not fully open. Thus, return EBUSY if the file system is not fully
+	 * initialized.
+	 */
+	if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))
+		return -EBUSY;
+
+	root = fs_info->chunk_root;
+	if (sb_rdonly(fs_info->sb))
+		return -EROFS;
+
+	ret = kstrtou64(buf, 0, &type);
+	if (ret < 0)
+		return -EINVAL;
+
+	/* for now, only allow touching the 'allocation hint' bits */
+	if (type & ~((1 << BTRFS_DEV_ALLOCATION_MASK_BIT_COUNT) - 1))
+		return -EINVAL;
+
+	trans = btrfs_start_transaction(root, 1);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
+	prev_type = device->type;
+	device->type = type;
+
+	ret = btrfs_update_device(trans, device);
+
+	if (ret < 0) {
+		btrfs_abort_transaction(trans, ret);
+		btrfs_end_transaction(trans);
+		goto abort;
+	}
+
+	ret = btrfs_commit_transaction(trans);
+	if (ret < 0)
+		goto abort;
+
+	return len;
+abort:
+	device->type = prev_type;
+	return  ret;
+}
+BTRFS_ATTR_RW(devid, type, btrfs_devinfo_type_show, btrfs_devinfo_type_store);
+#endif
+
+#ifdef CONFIG_BTRFS_PER_DEVICE_IO_STATS
+static ssize_t btrfs_devinfo_read_stats_show(struct kobject *kobj,
+                                                       struct kobj_attribute *a, char *buf)
+{
+	struct btrfs_device *device = container_of(kobj, struct btrfs_device,
+						   devid_kobj);
+	u64 read_wait = part_stat_read(device->bdev, nsecs[READ]);
+	unsigned long read_ios = part_stat_read(device->bdev, ios[READ]);
+
+	u64 avg_wait = 0;
+	if (read_wait && read_ios && read_wait >= read_ios)
+		avg_wait = div_u64(read_wait, read_ios);
+
+	return scnprintf(buf, PAGE_SIZE, "ios %lu wait %llu avg %llu age %llu ignored %llu\n",
+	                 read_ios, read_wait, avg_wait,
+	                 (u64)atomic64_read(&device->last_io_age),
+	                 (u64)atomic64_read(&device->stripe_ignored));
+}
+BTRFS_ATTR(devid, read_stats, btrfs_devinfo_read_stats_show);
+#endif
+
 /*
  * Information about one device.
  *
  * Path: /sys/fs/btrfs/<uuid>/devinfo/<devid>/
  */
 static struct attribute *devid_attrs[] = {
+#ifdef CONFIG_BTRFS_PER_DEVICE_IO_STATS
+	BTRFS_ATTR_PTR(devid, read_stats),
+#endif
 	BTRFS_ATTR_PTR(devid, error_stats),
 	BTRFS_ATTR_PTR(devid, fsid),
 	BTRFS_ATTR_PTR(devid, in_fs_metadata),
 	BTRFS_ATTR_PTR(devid, missing),
 	BTRFS_ATTR_PTR(devid, replace_target),
 	BTRFS_ATTR_PTR(devid, scrub_speed_max),
 	BTRFS_ATTR_PTR(devid, writeable),
+#ifdef CONFIG_BTRFS_ALLOCATOR_HINTS
+	BTRFS_ATTR_PTR(devid, type),
+#endif
 	NULL
 };
 ATTRIBUTE_GROUPS(devid);

fs/btrfs/sysfs.h

@@ -50,7 +50,7 @@ void btrfs_sysfs_del_one_qgroup(struct btrfs_fs_info *fs_info,
 				struct btrfs_qgroup *qgroup);
 int btrfs_read_policy_to_enum(const char *str, s64 *value);
 
-#ifdef CONFIG_BTRFS_EXPERIMENTAL
+#ifdef CONFIG_BTRFS_READ_POLICIES
 int __init btrfs_read_policy_init(void);
 char *btrfs_get_mod_read_policy(void);
 #endif