Skip to contentSkip to conversation

Add support for write life time hints

PATCH Superseded Jens Axboe 2 participants 12 messages Jun 19, 2017 – Jun 20, 2017

Subsystems (7)

v8 of 10 Superseded Latest is v10 →

lore
Share & tools
lore

Jens Axboe

author maintainer
#0 lore raw

PATCH v8 1/1 Add support for write life time hints

A new iteration of this patchset, previously known as write streams. As before, this patchset aims at enabling applications split up writes into separate streams, based on the perceived life time of the data written. This is useful for a variety of reasons:

- For NVMe, this feature is ratified and released with the NVMe 1.3 spec. Devices implementing Directives can expose multiple streams. Separating data written into streams based on life time can drastically reduce the write amplification. This helps device endurance, and increases performance. Testing just performed internally at Facebook with these patches showed up to a 25% reduction in NAND writes in a RocksDB setup.

- Software caching solutions can make more intelligent decisions on how and where to place data.

Contrary to previous patches, we're not exposing numeric stream values anymore. I've previously advocated for just doing a set of hints that makes sense instead. See the coverage from the LSFMM summit this year:

https://lwn.net/Articles/717755/

This patchset attempts to do that. We add an fcntl(2) interface to get/set these types of hints. We define 4 hints that pertain to data write life times:

RWH_WRITE_LIFE_SHORT Data written with this flag is expected to have
a high overwrite rate, or life time.

RWH_WRITE_LIFE_MEDIUM Longer life time than SHORT

RWH_WRITE_LIFE_LONG Longer life time than MEDIUM

RWH_WRITE_LIFE_EXTREME Longer life time than LONG

The idea is that these are relative values, so an application can use them as they see fit. The underlying device can then place data appropriately, or be free to ignore the hint. It's just a hint.

A branch based on current master can be pulled
from here:

git://git.kernel.dk/linux-block write-stream.8

Changes since v7:

- NVMe: change 'streams' parameter to be a bool enable/disable. We hardwire the number of streams anyway and use the appropriate amount, so no point in exposing this value.
- NVMe: collapse stream values appropriately, instead of just doing a basic MOD.
- Get rid of pwritev2(2) flags. Just use the fcntl(2) interface.
- Collapse some patches
- Change fcntl(2) interface to get/set values from a user supplied 64-bit pointer.
- Move inode-to-iocb mask setting to iocb_flags().

Changes since v6:

- Rewrite NVMe write stream assignment
- Change NVMe stream assignment to be per-controller, not per-ns. Then we can use the same IDs across name spaces, and we don't have to do lazy setup of streams.
- If streams are enabled on nvme, set io min/opt and discard granularity based on the stream params reported.
- Fixup F_SET_RW_HINT definition, it was 20, should have been 12.

Changes since v5:

- Change enum write_hint to enum rw_hint.
- Change fcntl() interface to be read/write generic
- Bring enum rw_hint all the way to bio/request
- Change references to streams in changelogs and debugfs interface
- Rebase to master to resolve blkdev.h conflict
- Reshuffle patches so the WRITE_LIFE_* hints and type come first. Allowed me to merge two block patches as well.

Changes since v4:

- Add enum write_hint and the WRITE_HINT_* values. This is what we use internally (until transformed to req/bio flags), and what is exposed to user space with the fcntl() interface. Maps directly to the RWF_WRITE_LIFE_* values.
- Add fcntl() interface for getting/setting hint values.
- Get rid of inode ->i_write_hint, encode the 3 bits of hint info in the inode flags intead.
- Allow a write with no hint to clear the old hint. Previously we only changed the hint if a new valid hint was given, not if no hint was passed in.
- Shrink flag space grabbed from 4 to 3 bits for RWF_* and the inode flags.

Changes since v3:

- Change any naming of stream ID to write hint.
- Various little API changes, suggested by Christoph
- Cleanup the NVMe bits, dump the debug info.
- Change NVMe to lazily allocate the streams.
- Various NVMe error handling improvements and command checking.

Changes since v2:

- Get rid of bio->bi_stream and replace with four request/bio flags. These map directly to the RWF_WRITE_* flags that the user passes in.
- Cleanup the NVMe stream setting.
- Drivers now responsible for updating the queue stream write counter, as they determine what stream to map a given flag to.

Changes since v1:

- Guard queue stream stats to ensure we don't mess up memory, if bio_stream() ever were to return a larger value than we support.
- NVMe: ensure we set the stream modulo the name space defined count.
- Cleanup the RWF_ and IOCB_ flags. Set aside 4 bits, and just store the stream value in there. This makes the passing of stream ID from RWF_ space to IOCB_ (and IOCB_ to bio) more efficient, and cleans it up in general.
- Kill the block internal definitions of the stream type, we don't need them anymore. See above.

Signature · 1 line
Jens Axboe
author maintainer
#1 lore raw

PATCH 1/9 fs: add fcntl() interface for setting/getting write life time hints

Define a set of write life time hints:

and add an fcntl interface for querying these flags, and also for setting them as well:

F_GET_RW_HINT Returns the read/write hint set.

F_SET_RW_HINT Pass one of the above write hints.

The user passes in a 64-bit pointer to get/set these values, and the interface returns 0/-1 on success/error.

Sample program testing/implementing basic setting/getting of write hints is below.

Add support for storing the write life time hint in the inode flags, and pass them to the kiocb flags as well. This is in preparation for utilizing these hints in the block layer, to guide on-media data placement.

/*
* writehint.c: check or set a file/inode write hint */

static char *str[] = { "WRITE_LIFE_NONE", "WRITE_LIFE_SHORT",

			"WRITE_LIFE_MEDIUM", "WRITE_LIFE_LONG",
			"WRITE_LIFE_EXTREME" };

int main(int argc, char *argv[])

{
	uint64_t hint = -1ULL;
	int fd, ret;

	if (argc < 2) {
		fprintf(stderr, "%s: dev <hint>\n", argv[0]);
		return 1;
	}

	fd = open(argv[1], O_RDONLY);
	if (fd < 0) {
		perror("open");
		return 2;
	}

	if (argc > 2)
		hint = atoi(argv[2]);

	if (hint == -1ULL) {
		ret = fcntl(fd, F_RW_GET_HINT, &hint);
		if (ret < 0) {
			perror("fcntl: F_RW_GET_HINT");
			return 3;
		}
	} else {
		ret = fcntl(fd, F_RW_SET_HINT, &hint);
		if (ret < 0) {
			perror("fcntl: F_RW_SET_HINT");
			return 4;
		}
	}

	printf("%s: %shint %s\n", argv[1], hint != -1ULL ? "set " : "", str[hint]);
	close(fd);
	return 0;
}
---
 fs/fcntl.c                 | 43 ++++++++++++++++++++++++++++++++
 fs/inode.c                 | 11 +++++++++
 include/linux/fs.h         | 61 ++++++++++++++++++++++++++++++++++++++++++++++
 include/uapi/linux/fcntl.h | 15 ++++++++++++
 4 files changed, 130 insertions(+)
fs/fcntl.c +43 -0
index f4e7267d117f..113b78c11631 100644
@@ -243,6 +243,45 @@ static int f_getowner_uids(struct file *filp, unsigned long arg)
 }
 #endif
 
+static long fcntl_rw_hint(struct file *file, unsigned int cmd,
+			  u64 __user *ptr)
+{
+	struct inode *inode = file_inode(file);
+	long ret = 0;
+	u64 hint;
+
+	switch (cmd) {
+	case F_GET_RW_HINT:
+		hint = mask_to_write_hint(inode->i_flags, S_WRITE_LIFE_SHIFT);
+		if (put_user(hint, ptr))
+			ret = -EFAULT;
+		break;
+	case F_SET_RW_HINT:
+		if (get_user(hint, ptr)) {
+			ret = -EFAULT;
+			break;
+		}
+		switch (hint) {
+		case WRITE_LIFE_NONE:
+		case WRITE_LIFE_SHORT:
+		case WRITE_LIFE_MEDIUM:
+		case WRITE_LIFE_LONG:
+		case WRITE_LIFE_EXTREME:
+			inode_set_write_hint(inode, hint);
+			ret = 0;
+			break;
+		default:
+			ret = -EINVAL;
+		}
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	return ret;
+}
+
 static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
 		struct file *filp)
 {
@@ -337,6 +376,10 @@ static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
 	case F_GET_SEALS:
 		err = shmem_fcntl(filp, cmd, arg);
 		break;
+	case F_GET_RW_HINT:
+	case F_SET_RW_HINT:
+		err = fcntl_rw_hint(filp, cmd, (u64 __user *) arg);
+		break;
 	default:
 		break;
 	}
fs/inode.c +11 -0
index db5914783a71..defb015a2c6d 100644
@@ -2120,3 +2120,14 @@ struct timespec current_time(struct inode *inode)
 	return timespec_trunc(now, inode->i_sb->s_time_gran);
 }
 EXPORT_SYMBOL(current_time);
+
+void inode_set_write_hint(struct inode *inode, enum rw_hint hint)
+{
+	unsigned int flags = write_hint_to_mask(hint, S_WRITE_LIFE_SHIFT);
+
+	if (flags != mask_to_write_hint(inode->i_flags, S_WRITE_LIFE_SHIFT)) {
+		inode_lock(inode);
+		inode_set_flags(inode, flags, S_WRITE_LIFE_MASK);
+		inode_unlock(inode);
+	}
+}
include/linux/fs.h +61 -0

104 lines not on this page · raw message

include/uapi/linux/fcntl.h +15 -0
index 813afd6eee71..def8f70e8bae 100644
@@ -43,6 +43,21 @@
 /* (1U << 31) is reserved for signed error codes */
 
 /*
+ * Set/Get write life time hints.
+ */
+#define F_GET_RW_HINT		(F_LINUX_SPECIFIC_BASE + 11)
+#define F_SET_RW_HINT		(F_LINUX_SPECIFIC_BASE + 12)
+
+/*
+ * Valid hint values for F_{GET,SET}_RW_HINT
+ */
+#define RWH_WRITE_LIFE_NONE	0
+#define RWH_WRITE_LIFE_SHORT	1
+#define RWH_WRITE_LIFE_MEDIUM	2
+#define RWH_WRITE_LIFE_LONG	3
+#define RWH_WRITE_LIFE_EXTREME	4
+
+/*
  * Types of directory notifications that may be requested.
  */
 #define DN_ACCESS	0x00000001	/* File accessed */
Signature · 1 line
2.7.4
#2 lore raw

Re: [PATCH 1/9] fs: add fcntl() interface for setting/getting write life time hints

Quoted 28 lines · Jens Axboe
fs/fcntl.c +28 -0
+static long fcntl_rw_hint(struct file *file, unsigned int cmd,
+			  u64 __user *ptr)
+{
+	struct inode *inode =3D file_inode(file);
+	long ret =3D 0;
+	u64 hint;
+
+	switch (cmd) {
+	case F_GET_RW_HINT:
+		hint =3D mask_to_write_hint(inode->i_flags, S_WRITE_LIFE_SHIFT);
+		if (put_user(hint, ptr))
+			ret =3D -EFAULT;
+		break;
+	case F_SET_RW_HINT:
+		if (get_user(hint, ptr)) {
+			ret =3D -EFAULT;
+			break;
+		}
+		switch (hint) {
+		case WRITE_LIFE_NONE:
+		case WRITE_LIFE_SHORT:
+		case WRITE_LIFE_MEDIUM:
+		case WRITE_LIFE_LONG:
+		case WRITE_LIFE_EXTREME:
+			inode_set_write_hint(inode, hint);
+			ret =3D 0;
+			break;
+		default:

fs/fcntl.c static int f_getowner_uids(struct file *filp, unsigned long arg) · line 283

Quoted 10 lines
fs/fcntl.c +10 -0
+			ret =3D -EINVAL;
+		}
+		break;
+	default:
+		ret =3D -EINVAL;
+		break;
+	}
+
+	return ret;
+}

Hello Jens,

Do we need an (inline) helper function for checking the validity of a numerical WRITE_LIFE value next to the definition of the WRITE_LIFE_* constants, e.g. WRITE_LIFE_NONE <=3D hint && hint <=3D WRITE_LIFE_EXTREME?

include/linux/fs.h struct writeback_control; · line 277

Quoted 5 lines
include/linux/fs.h +5 -0
+/*
+ * Steal 3 bits for stream information, this allows 8 valid streams
+ */
+#define IOCB_WRITE_LIFE_SHIFT	7
+#define IOCB_WRITE_LIFE_MASK	(BIT(7) | BIT(8) | BIT(9))

A minor comment: how about making this easier to read by defining IOCB_WRITE_LIFE_MASK as (7 << IOCB_WRITE_LIFE_SHIFT)?

include/linux/fs.h struct super_operations { · line 1849

Quoted 7 lines

/*
+ * Expected life time hint of a write for this inode. This uses the
+ * WRITE_LIFE_* encoding, we just need to define the shift. We need
+ * 3 bits for this. Next S_* value is 131072, bit 17.
+ */

include/linux/fs.h +2 -0
+#define S_WRITE_LIFE_MASK	0x1c000	/* bits 14..16 */
+#define S_WRITE_LIFE_SHIFT	14	/* 16384, next bit */

Another minor comment: how about making this easier to read by defining S_WRITE_LIFE_MASK as (7 << S_WRITE_LIFE_SHIFT)?

Quoted 9 lines

/*
+ * Write life time hint values.
+ */

include/uapi/linux/fcntl.h +6 -0
+enum rw_hint {
+	WRITE_LIFE_NONE =3D RWH_WRITE_LIFE_NONE,
+	WRITE_LIFE_SHORT =3D RWH_WRITE_LIFE_SHORT,
+	WRITE_LIFE_MEDIUM =3D RWH_WRITE_LIFE_MEDIUM,
+	WRITE_LIFE_LONG =3D RWH_WRITE_LIFE_LONG,
+	WRITE_LIFE_EXTREME =3D RWH_WRITE_LIFE_EXTREME

include/uapi/linux/fcntl.h line 58

Quoted 10 lines
include/uapi/linux/fcntl.h +9 -0
+};
[ ... ]
+/*
+ * Valid hint values for F_{GET,SET}_RW_HINT
+ */
+#define RWH_WRITE_LIFE_NONE	0
+#define RWH_WRITE_LIFE_SHORT	1
+#define RWH_WRITE_LIFE_MEDIUM	2
+#define RWH_WRITE_LIFE_LONG	3
+#define RWH_WRITE_LIFE_EXTREME	4

Maybe I missed something, but it's not clear to me why we have both an enum=
and
defines with the same numerical values? BTW, I prefer an enum above #define= s.

Thanks,

Bart.=

author maintainer
#3 lore raw

Re: [PATCH 1/9] fs: add fcntl() interface for setting/getting write life time hints

Quoted 42 lines · Bart Van Assche
Quoted 28 lines · Jens Axboe
fs/fcntl.c +28 -0
+static long fcntl_rw_hint(struct file *file, unsigned int cmd,
+			  u64 __user *ptr)
+{
+	struct inode *inode = file_inode(file);
+	long ret = 0;
+	u64 hint;
+
+	switch (cmd) {
+	case F_GET_RW_HINT:
+		hint = mask_to_write_hint(inode->i_flags, S_WRITE_LIFE_SHIFT);
+		if (put_user(hint, ptr))
+			ret = -EFAULT;
+		break;
+	case F_SET_RW_HINT:
+		if (get_user(hint, ptr)) {
+			ret = -EFAULT;
+			break;
+		}
+		switch (hint) {
+		case WRITE_LIFE_NONE:
+		case WRITE_LIFE_SHORT:
+		case WRITE_LIFE_MEDIUM:
+		case WRITE_LIFE_LONG:
+		case WRITE_LIFE_EXTREME:
+			inode_set_write_hint(inode, hint);
+			ret = 0;
+			break;
+		default:

fs/fcntl.c static int f_getowner_uids(struct file *filp, unsigned long arg) · line 283

Quoted 10 lines
fs/fcntl.c +10 -0
+			ret = -EINVAL;
+		}
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	return ret;
+}

Hello Jens,

Quoted 3 lines

Do we need an (inline) helper function for checking the validity of a numerical WRITE_LIFE value next to the definition of the WRITE_LIFE_* constants, e.g. WRITE_LIFE_NONE <= hint && hint <= WRITE_LIFE_EXTREME?

Might not hurt in general, I can fold something like that in.

Quoted 8 lines

include/linux/fs.h struct writeback_control; · line 277

Quoted 5 lines
include/linux/fs.h +5 -0
+/*
+ * Steal 3 bits for stream information, this allows 8 valid streams
+ */
+#define IOCB_WRITE_LIFE_SHIFT	7
+#define IOCB_WRITE_LIFE_MASK	(BIT(7) | BIT(8) | BIT(9))

A minor comment: how about making this easier to read by defining IOCB_WRITE_LIFE_MASK as (7 << IOCB_WRITE_LIFE_SHIFT)?

Agree, that would be prettier.

Quoted 10 lines

include/linux/fs.h struct super_operations { · line 1849

Quoted 7 lines

/*
+ * Expected life time hint of a write for this inode. This uses the
+ * WRITE_LIFE_* encoding, we just need to define the shift. We need
+ * 3 bits for this. Next S_* value is 131072, bit 17.
+ */

include/linux/fs.h +2 -0
+#define S_WRITE_LIFE_MASK	0x1c000	/* bits 14..16 */
+#define S_WRITE_LIFE_SHIFT	14	/* 16384, next bit */

Another minor comment: how about making this easier to read by defining S_WRITE_LIFE_MASK as (7 << S_WRITE_LIFE_SHIFT)?

Agree, I'll make that change too.

Quoted 15 lines
Quoted 15 lines

/*
+ * Write life time hint values.
+ */

include/linux/fs.h +11 -0
+enum rw_hint {
+	WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE,
+	WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT,
+	WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM,
+	WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG,
+	WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME
+};
[ ... ]
+/*
+ * Valid hint values for F_{GET,SET}_RW_HINT
+ */
+#define RWH_WRITE_LIFE_NONE	0
Quoted 8 lines

include/uapi/linux/fcntl.h line 58

Quoted 4 lines
include/uapi/linux/fcntl.h +4 -0
+#define RWH_WRITE_LIFE_SHORT	1
+#define RWH_WRITE_LIFE_MEDIUM	2
+#define RWH_WRITE_LIFE_LONG	3
+#define RWH_WRITE_LIFE_EXTREME	4

Maybe I missed something, but it's not clear to me why we have both an enum and defines with the same numerical values? BTW, I prefer an enum above #defines.

We use the enum internally, that's the hint that the fs and block layer sees. The reason for the defines is for the user interface, where we don't want that to be an enum. So the mapping between the two is the definition of the enum rw_hint values.

Signature · 1 line
Jens Axboe
author maintainer
#4 lore raw

PATCH 2/9 block: add support for write hints in a bio

In reply to Jens Axboe

No functional changes in this patch, we just set aside 3 bits in the bio/request flags, which can be used to hold a WRITE_LIFE_* life time hint.

Ensure that we don't merge requests that have different life time hints assigned to them.

---
 block/blk-merge.c         | 16 ++++++++++++++++
 include/linux/blk_types.h | 20 ++++++++++++++++++++
 2 files changed, 36 insertions(+)
block/blk-merge.c +16 -0
index cea544ec5d96..17cd0c3f872d 100644
@@ -670,6 +670,14 @@ static struct request *attempt_merge(struct request_queue *q,
 		return NULL;
 
 	/*
+	 * Don't allow merge of different streams, or for a stream with
+	 * non-stream IO.
+	 */
+	if ((req->cmd_flags & REQ_WRITE_LIFE_MASK) !=
+	    (next->cmd_flags & REQ_WRITE_LIFE_MASK))
+		return NULL;
+
+	/*
 	 * If we are allowed to merge, then append bio list
 	 * from next to rq and release next. merge_requests_fn
 	 * will have updated segment counts, update sector
@@ -788,6 +796,14 @@ bool blk_rq_merge_ok(struct request *rq, struct bio *bio)
 	    !blk_write_same_mergeable(rq->bio, bio))
 		return false;
 
+	/*
+	 * Don't allow merge of different streams, or for a stream with
+	 * non-stream IO.
+	 */
+	if ((rq->cmd_flags & REQ_WRITE_LIFE_MASK) !=
+	    (bio->bi_opf & REQ_WRITE_LIFE_MASK))
+		return false;
+
 	return true;
 }
 
include/linux/blk_types.h +20 -0
index dcd45b15a3a5..fde216924aa3 100644
@@ -7,6 +7,7 @@
 
 #include <linux/types.h>
 #include <linux/bvec.h>
+#include <linux/fs.h>
 
 struct bio_set;
 struct bio;
@@ -220,6 +221,9 @@ enum req_flag_bits {
 	__REQ_PREFLUSH,		/* request for cache flush */
 	__REQ_RAHEAD,		/* read ahead, can fail anytime */
 	__REQ_BACKGROUND,	/* background IO */
+	__REQ_WRITE_HINT_SHIFT,	/* 3 bits for life time hint */
+	__REQ_WRITE_HINT_PAD1,
+	__REQ_WRITE_HINT_PAD2,
 
 	/* command specific flags for REQ_OP_WRITE_ZEROES: */
 	__REQ_NOUNMAP,		/* do not free blocks when zeroing */
@@ -240,6 +244,12 @@ enum req_flag_bits {
 #define REQ_PREFLUSH		(1ULL << __REQ_PREFLUSH)
 #define REQ_RAHEAD		(1ULL << __REQ_RAHEAD)
 #define REQ_BACKGROUND		(1ULL << __REQ_BACKGROUND)
+#define REQ_WRITE_SHORT		(WRITE_LIFE_SHORT << __REQ_WRITE_HINT_SHIFT)
+#define REQ_WRITE_MEDIUM	(WRITE_LIFE_MEDIUM << __REQ_WRITE_HINT_SHIFT)
+#define REQ_WRITE_LONG		(WRITE_LIFE_LONG << __REQ_WRITE_HINT_SHIFT)
+#define REQ_WRITE_EXTREME	(WRITE_LIFE_EXTREME << __REQ_WRITE_HINT_SHIFT)
+
+#define REQ_WRITE_LIFE_MASK	(0x7 << __REQ_WRITE_HINT_SHIFT)
 
 #define REQ_NOUNMAP		(1ULL << __REQ_NOUNMAP)
 
@@ -331,4 +341,14 @@ struct blk_rq_stat {
 	u64 batch;
 };
 
+static inline unsigned int write_hint_to_opf(enum rw_hint hint)
+{
+	return hint << __REQ_WRITE_HINT_SHIFT;
+}
+
+static inline enum rw_hint opf_to_write_hint(unsigned int opf)
+{
+	return (opf & REQ_WRITE_LIFE_MASK) >> __REQ_WRITE_HINT_SHIFT;
+}
+
 #endif /* __LINUX_BLK_TYPES_H */
Signature · 1 line
2.7.4
author maintainer
#5 lore raw

PATCH 3/9 blk-mq: expose stream write hints through debugfs

In reply to Jens Axboe

Useful to verify that things are working the way they should. Reading the file will return number of kb written with each write hint. Writing the file will reset the statistics. No care is taken to ensure that we don't race on updates.

Drivers will write to q->write_hints[] if they handle a given write hint.

---
 block/blk-mq-debugfs.c | 24 ++++++++++++++++++++++++
 include/linux/blkdev.h |  3 +++
 2 files changed, 27 insertions(+)
block/blk-mq-debugfs.c +24 -0
index 9edebbdce0bd..9ebc2945f991 100644
@@ -135,6 +135,29 @@ static void print_stat(struct seq_file *m, struct blk_rq_stat *stat)
 	}
 }
 
+static int queue_write_hint_show(void *data, struct seq_file *m)
+{
+	struct request_queue *q = data;
+	int i;
+
+	for (i = 0; i < BLK_MAX_WRITE_HINTS; i++)
+		seq_printf(m, "hint%d: %llu\n", i, q->write_hints[i]);
+
+	return 0;
+}
+
+static ssize_t queue_write_hint_store(void *data, const char __user *buf,
+				      size_t count, loff_t *ppos)
+{
+	struct request_queue *q = data;
+	int i;
+
+	for (i = 0; i < BLK_MAX_WRITE_HINTS; i++)
+		q->write_hints[i] = 0;
+
+	return count;
+}
+
 static int queue_poll_stat_show(void *data, struct seq_file *m)
 {
 	struct request_queue *q = data;
@@ -730,6 +753,7 @@ static const struct blk_mq_debugfs_attr blk_mq_debugfs_queue_attrs[] = {
 	{"poll_stat", 0400, queue_poll_stat_show},
 	{"requeue_list", 0400, .seq_ops = &queue_requeue_list_seq_ops},
 	{"state", 0600, queue_state_show, queue_state_write},
+	{"write_hints", 0600, queue_write_hint_show, queue_write_hint_store},
 	{},
 };
 
include/linux/blkdev.h +3 -0
index 22cfba64ce81..687394b70924 100644
@@ -586,6 +586,9 @@ struct request_queue {
 
 	size_t			cmd_size;
 	void			*rq_alloc_data;
+
+#define BLK_MAX_WRITE_HINTS	5
+	u64			write_hints[BLK_MAX_WRITE_HINTS];
 };
 
 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
Signature · 1 line
2.7.4
author maintainer
#6 lore raw

PATCH 4/9 fs: add O_DIRECT support for sending down write life time hints

In reply to Jens Axboe
---
 fs/block_dev.c | 2 ++
 fs/direct-io.c | 2 ++
 fs/iomap.c     | 1 +
 3 files changed, 5 insertions(+)
fs/block_dev.c +2 -0
index dd91c99e9ba0..30e1fb65c2fa 100644
@@ -183,6 +183,8 @@ static unsigned int dio_bio_write_op(struct kiocb *iocb)
 	/* avoid the need for a I/O completion work item */
 	if (iocb->ki_flags & IOCB_DSYNC)
 		op |= REQ_FUA;
+
+	op |= write_hint_to_opf(iocb_write_hint(iocb));
 	return op;
 }
 
fs/direct-io.c +2 -0
index e8baaabebf13..9e9adca0c592 100644
@@ -385,6 +385,8 @@ dio_bio_alloc(struct dio *dio, struct dio_submit *sdio,
 	else
 		bio->bi_end_io = dio_bio_end_io;
 
+	bio->bi_opf |= write_hint_to_opf(iocb_write_hint(dio->iocb));
+
 	sdio->bio = bio;
 	sdio->logical_offset_in_bio = sdio->cur_page_fs_offset;
 }
fs/iomap.c +1 -0
index 18f2f2b8ba2c..63b9f87d9461 100644
@@ -804,6 +804,7 @@ iomap_dio_actor(struct inode *inode, loff_t pos, loff_t length,
 
 		if (dio->flags & IOMAP_DIO_WRITE) {
 			bio_set_op_attrs(bio, REQ_OP_WRITE, REQ_SYNC | REQ_IDLE);
+			bio->bi_opf |= write_hint_to_opf(inode_write_hint(inode));
 			task_io_account_write(bio->bi_iter.bi_size);
 		} else {
 			bio_set_op_attrs(bio, REQ_OP_READ, 0);
Signature · 1 line
2.7.4
author maintainer
#7 lore raw

PATCH 5/9 fs: add support for buffered writeback to pass down write hints

In reply to Jens Axboe
---
 fs/buffer.c | 14 +++++++++-----
 fs/mpage.c  |  1 +
 2 files changed, 10 insertions(+), 5 deletions(-)
fs/buffer.c +9 -5
index 306b720f7383..1259524715c8 100644
@@ -49,7 +49,7 @@
 
 static int fsync_buffers_list(spinlock_t *lock, struct list_head *list);
 static int submit_bh_wbc(int op, int op_flags, struct buffer_head *bh,
-			 struct writeback_control *wbc);
+			 unsigned int stream, struct writeback_control *wbc);
 
 #define BH_ENTRY(list) list_entry((list), struct buffer_head, b_assoc_buffers)
 
@@ -1829,7 +1829,8 @@ int __block_write_full_page(struct inode *inode, struct page *page,
 	do {
 		struct buffer_head *next = bh->b_this_page;
 		if (buffer_async_write(bh)) {
-			submit_bh_wbc(REQ_OP_WRITE, write_flags, bh, wbc);
+			submit_bh_wbc(REQ_OP_WRITE, write_flags, bh,
+					inode_write_hint(inode), wbc);
 			nr_underway++;
 		}
 		bh = next;
@@ -1883,7 +1884,8 @@ int __block_write_full_page(struct inode *inode, struct page *page,
 		struct buffer_head *next = bh->b_this_page;
 		if (buffer_async_write(bh)) {
 			clear_buffer_dirty(bh);
-			submit_bh_wbc(REQ_OP_WRITE, write_flags, bh, wbc);
+			submit_bh_wbc(REQ_OP_WRITE, write_flags, bh,
+					inode_write_hint(inode), wbc);
 			nr_underway++;
 		}
 		bh = next;
@@ -3091,7 +3093,7 @@ void guard_bio_eod(int op, struct bio *bio)
 }
 
 static int submit_bh_wbc(int op, int op_flags, struct buffer_head *bh,
-			 struct writeback_control *wbc)
+			 unsigned int write_hint, struct writeback_control *wbc)
 {
 	struct bio *bio;
 
@@ -3134,6 +3136,8 @@ static int submit_bh_wbc(int op, int op_flags, struct buffer_head *bh,
 		op_flags |= REQ_META;
 	if (buffer_prio(bh))
 		op_flags |= REQ_PRIO;
+
+	op_flags |= write_hint_to_opf(write_hint);
 	bio_set_op_attrs(bio, op, op_flags);
 
 	submit_bio(bio);
@@ -3142,7 +3146,7 @@ static int submit_bh_wbc(int op, int op_flags, struct buffer_head *bh,
 
 int submit_bh(int op, int op_flags, struct buffer_head *bh)
 {
-	return submit_bh_wbc(op, op_flags, bh, NULL);
+	return submit_bh_wbc(op, op_flags, bh, 0, NULL);
 }
 EXPORT_SYMBOL(submit_bh);
 
fs/mpage.c +1 -0
index 9524fdde00c2..d8a750873bf4 100644
@@ -615,6 +615,7 @@ static int __mpage_writepage(struct page *page, struct writeback_control *wbc,
 			goto confused;
 
 		wbc_init_bio(wbc, bio);
+		bio->bi_opf |= write_hint_to_opf(inode_write_hint(inode));
 	}
 
 	/*
Signature · 1 line
2.7.4
author maintainer
#8 lore raw

PATCH 6/9 ext4: add support for passing in write hints for buffered writes

In reply to Jens Axboe
---
 fs/ext4/page-io.c | 2 ++
 1 file changed, 2 insertions(+)
fs/ext4/page-io.c +2 -0
index 930ca0fc9a0f..92834b702728 100644
@@ -350,6 +350,7 @@ void ext4_io_submit(struct ext4_io_submit *io)
 	if (bio) {
 		int io_op_flags = io->io_wbc->sync_mode == WB_SYNC_ALL ?
 				  REQ_SYNC : 0;
+		io_op_flags |= write_hint_to_opf(inode_write_hint(io->io_end->inode));
 		bio_set_op_attrs(io->io_bio, REQ_OP_WRITE, io_op_flags);
 		submit_bio(io->io_bio);
 	}
@@ -397,6 +398,7 @@ static int io_submit_add_bh(struct ext4_io_submit *io,
 		ret = io_submit_init_bio(io, bh);
 		if (ret)
 			return ret;
+		io->io_bio->bi_opf |= write_hint_to_opf(inode_write_hint(inode));
 	}
 	ret = bio_add_page(io->io_bio, page, bh->b_size, bh_offset(bh));
 	if (ret != bh->b_size)
Signature · 1 line
2.7.4
author maintainer
#9 lore raw

PATCH 7/9 xfs: add support for passing in write hints for buffered writes

In reply to Jens Axboe
---
 fs/xfs/xfs_aops.c | 2 ++
 1 file changed, 2 insertions(+)
fs/xfs/xfs_aops.c +2 -0
index 76b6f988e2fa..e4d9d470402c 100644
@@ -506,6 +506,7 @@ xfs_submit_ioend(
 		return status;
 	}
 
+	ioend->io_bio->bi_opf |= write_hint_to_opf(inode_write_hint(ioend->io_inode));
 	submit_bio(ioend->io_bio);
 	return 0;
 }
@@ -565,6 +566,7 @@ xfs_chain_bio(
 	bio_chain(ioend->io_bio, new);
 	bio_get(ioend->io_bio);		/* for xfs_destroy_ioend */
 	ioend->io_bio->bi_opf = REQ_OP_WRITE | wbc_to_write_flags(wbc);
+	ioend->io_bio->bi_opf |= write_hint_to_opf(inode_write_hint(ioend->io_inode));
 	submit_bio(ioend->io_bio);
 	ioend->io_bio = new;
 }
Signature · 1 line
2.7.4
author maintainer
#10 lore raw

PATCH 8/9 btrfs: add support for passing in write hints for buffered writes

In reply to Jens Axboe
---
 fs/btrfs/extent_io.c | 1 +
 1 file changed, 1 insertion(+)
fs/btrfs/extent_io.c +1 -0
index 19eedf2e630b..3e57cfaa6dd6 100644
@@ -2830,6 +2830,7 @@ static int submit_extent_page(int op, int op_flags, struct extent_io_tree *tree,
 	bio_add_page(bio, page, page_size, offset);
 	bio->bi_end_io = end_io_func;
 	bio->bi_private = tree;
+	op_flags |= write_hint_to_opf(inode_write_hint(page->mapping->host));
 	bio_set_op_attrs(bio, op, op_flags);
 	if (wbc) {
 		wbc_init_bio(wbc, bio);
Signature · 1 line
2.7.4
author maintainer
#11 lore raw

PATCH 9/9 nvme: add support for streams and directives

In reply to Jens Axboe

This adds support for Directives in NVMe, particular for the Streams directive. Support for Directives is a new feature in NVMe 1.3. It allows a user to pass in information about where to store the data, so that it the device can do so most effiently. If an application is managing and writing data with different life times, mixing differently retentioned data onto the same locations on flash can cause write amplification to grow. This, in turn, will reduce performance and life time of the device.

We default to allocating 4 streams, controller wide, so we can use them on all name spaces. This is configurable with the 'streams' module parameter. If a write stream is set in a write, flag is as such before sending it to the device.

---
 drivers/nvme/host/core.c | 189 +++++++++++++++++++++++++++++++++++++++++++++--
 drivers/nvme/host/nvme.h |   4 +
 include/linux/nvme.h     |  48 ++++++++++++
 3 files changed, 236 insertions(+), 5 deletions(-)
drivers/nvme/host/core.c +184 -5
index aee37b73231d..fa1acfa27fa4 100644
@@ -68,6 +68,10 @@ MODULE_PARM_DESC(force_apst, "allow APST for newly enumerated devices even if qu
 struct workqueue_struct *nvme_wq;
 EXPORT_SYMBOL_GPL(nvme_wq);
 
+static bool streams = true;
+module_param(streams, bool, 0644);
+MODULE_PARM_DESC(streams, "use streams, if available");
+
 static LIST_HEAD(nvme_ctrl_list);
 static DEFINE_SPINLOCK(dev_list_lock);
 
@@ -297,6 +301,141 @@ struct request *nvme_alloc_request(struct request_queue *q,
 }
 EXPORT_SYMBOL_GPL(nvme_alloc_request);
 
+/*
+
+ * Returns number of streams allocated for use by, or -1 on error.
+ */
+static int nvme_streams_allocate(struct nvme_ctrl *ctrl, unsigned int nstreams)
+{
+	struct nvme_command c;
+	union nvme_result res;
+	int ret;
+
+	memset(&c, 0, sizeof(c));
+
+	c.directive.opcode = nvme_admin_directive_recv;
+	c.directive.nsid = cpu_to_le32(0xffffffff);
+	c.directive.doper = NVME_DIR_RCV_ST_OP_RESOURCE;
+	c.directive.dtype = NVME_DIR_STREAMS;
+	c.directive.endir = nstreams;
+
+	ret = __nvme_submit_sync_cmd(ctrl->admin_q, &c, &res, NULL, 0, 0,
+					NVME_QID_ANY, 0, 0);
+	if (ret)
+		return -1;
+
+	return le32_to_cpu(res.u32) & 0xffff;
+}
+
+static int nvme_enable_streams(struct nvme_ctrl *ctrl)
+{
+	struct nvme_command c;
+
+	memset(&c, 0, sizeof(c));
+
+	c.directive.opcode = nvme_admin_directive_send;
+	c.directive.nsid = cpu_to_le32(0xffffffff);
+	c.directive.doper = NVME_DIR_SND_ID_OP_ENABLE;
+	c.directive.dtype = NVME_DIR_IDENTIFY;
+	c.directive.tdtype = NVME_DIR_STREAMS;
+	c.directive.endir = NVME_DIR_ENDIR;
+
+	return nvme_submit_sync_cmd(ctrl->admin_q, &c, NULL, 0);
+}
+
+static int nvme_get_stream_params(struct nvme_ctrl *ctrl,
+				  struct streams_directive_params *s, u32 nsid)
+{
+	struct nvme_command c;
+
+	memset(&c, 0, sizeof(c));
+	memset(s, 0, sizeof(*s));
+
+	c.directive.opcode = nvme_admin_directive_recv;
+	c.directive.nsid = cpu_to_le32(nsid);
+	c.directive.numd = sizeof(*s);
+	c.directive.doper = NVME_DIR_RCV_ST_OP_PARAM;
+	c.directive.dtype = NVME_DIR_STREAMS;
+
+	return nvme_submit_sync_cmd(ctrl->admin_q, &c, s, sizeof(*s));
+}
+
+static int nvme_setup_directives(struct nvme_ctrl *ctrl)
+{
+	struct streams_directive_params s;
+	unsigned int nstreams;
+	int ret;
+
+	if (!(ctrl->oacs & NVME_CTRL_OACS_DIRECTIVES))
+		return 0;
+	if (!streams)
+		return 0;
+
+	ret = nvme_enable_streams(ctrl);
+	if (ret)
+		return ret;
+
+	ret = nvme_get_stream_params(ctrl, &s, 0xffffffff);
+	if (ret)
+		return ret;
+
+	ctrl->nssa = le16_to_cpu(s.nssa);
+
+	nstreams = min_t(unsigned int, ctrl->nssa, BLK_MAX_WRITE_HINTS - 1);
+	ret = nvme_streams_allocate(ctrl, nstreams);
+	if (ret < 0)
+		return ret;
+
+	/* require at least 2 streams to use them effectively */
+	if (ret > 1) {
+		ret = min(ret, BLK_MAX_WRITE_HINTS - 1);
+		ctrl->nr_streams = ret;
+		dev_info(ctrl->device, "successfully enabled %d streams\n", ret);
+	}
+
+	return 0;
+}
+
+/*
+ * Write hint number to stream mappings
+ */
+static const unsigned int stream_mappings[BLK_MAX_WRITE_HINTS][BLK_MAX_WRITE_HINTS] = {
+	/* 0 or 1 stream, we don't use streams */
+	{ 0, },
+	{ 0, },
+	/* collapse short+medium to short, and long+extreme to medium */
+	{ WRITE_LIFE_NONE, WRITE_LIFE_SHORT, WRITE_LIFE_SHORT,
+		WRITE_LIFE_MEDIUM, WRITE_LIFE_MEDIUM },
+	/* collapse long+extreme to long */
+	{ WRITE_LIFE_NONE, WRITE_LIFE_SHORT, WRITE_LIFE_MEDIUM,
+		WRITE_LIFE_LONG, WRITE_LIFE_LONG },
+	/* 4 streams, no collapsing needed */
+	{ WRITE_LIFE_NONE, WRITE_LIFE_SHORT, WRITE_LIFE_MEDIUM,
+		WRITE_LIFE_LONG, WRITE_LIFE_EXTREME },
+};
+
+/*
+ * Check if 'req' has a write hint associated with it. If it does, assign
+ * a valid namespace stream to the write. If we haven't setup streams yet,
+ * kick off configuration and ignore the hints until that has completed.
+ */
+static void nvme_assign_write_stream(struct nvme_ctrl *ctrl,
+				     struct request *req, u16 *control,
+				     u32 *dsmgmt)
+{
+	enum rw_hint streamid;
+
+	streamid = opf_to_write_hint(req->cmd_flags);
+	if (streamid != WRITE_LIFE_NONE) {
+		streamid = stream_mappings[ctrl->nr_streams][streamid];
+		*control |= NVME_RW_DTYPE_STREAMS;
+		*dsmgmt |= streamid << 16;
+	}
+
+	if (streamid < ARRAY_SIZE(req->q->write_hints))
+		req->q->write_hints[streamid] += blk_rq_bytes(req) >> 9;
+}
+
 static inline void nvme_setup_flush(struct nvme_ns *ns,
 		struct nvme_command *cmnd)
 {
@@ -348,6 +487,7 @@ static blk_status_t nvme_setup_discard(struct nvme_ns *ns, struct request *req,
 static inline blk_status_t nvme_setup_rw(struct nvme_ns *ns,
 		struct request *req, struct nvme_command *cmnd)
 {
+	struct nvme_ctrl *ctrl = ns->ctrl;
 	u16 control = 0;
 	u32 dsmgmt = 0;
 
@@ -375,6 +515,9 @@ static inline blk_status_t nvme_setup_rw(struct nvme_ns *ns,
 	cmnd->rw.slba = cpu_to_le64(nvme_block_nr(ns, blk_rq_pos(req)));
 	cmnd->rw.length = cpu_to_le16((blk_rq_bytes(req) >> ns->lba_shift) - 1);
 
+	if (req_op(req) == REQ_OP_WRITE && ctrl->nr_streams)
+		nvme_assign_write_stream(ctrl, req, &control, &dsmgmt);
+
 	if (ns->ms) {
 		switch (ns->pi_type) {
 		case NVME_NS_DPS_PI_TYPE3:
@@ -1088,14 +1231,21 @@ static void nvme_set_chunk_size(struct nvme_ns *ns)
 
 static void nvme_config_discard(struct nvme_ns *ns)
 {
-	struct nvme_ctrl *ctrl = ns->ctrl;
 	u32 logical_block_size = queue_logical_block_size(ns->queue);
+	struct nvme_ctrl *ctrl = ns->ctrl;
 
 	BUILD_BUG_ON(PAGE_SIZE / sizeof(struct nvme_dsm_range) <
 			NVME_DSM_MAX_RANGES);
 
-	ns->queue->limits.discard_alignment = logical_block_size;
-	ns->queue->limits.discard_granularity = logical_block_size;
+	if (ctrl->nr_streams && ns->sws && ns->sgs) {
+		unsigned int sz = logical_block_size * ns->sws * ns->sgs;
+
+		ns->queue->limits.discard_alignment = sz;
+		ns->queue->limits.discard_granularity = sz;
+	} else {
+		ns->queue->limits.discard_alignment = logical_block_size;
+		ns->queue->limits.discard_granularity = logical_block_size;
+	}
 	blk_queue_max_discard_sectors(ns->queue, UINT_MAX);
 	blk_queue_max_discard_segments(ns->queue, NVME_DSM_MAX_RANGES);
 	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, ns->queue);
@@ -1135,6 +1285,7 @@ static int nvme_revalidate_ns(struct nvme_ns *ns, struct nvme_id_ns **id)
 static void __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id)
 {
 	struct nvme_ns *ns = disk->private_data;
+	struct nvme_ctrl *ctrl = ns->ctrl;
 	u16 bs;
 
 	/*
@@ -1149,7 +1300,7 @@ static void __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id)
 
 	blk_mq_freeze_queue(disk->queue);
 
-	if (ns->ctrl->ops->flags & NVME_F_METADATA_SUPPORTED)
+	if (ctrl->ops->flags & NVME_F_METADATA_SUPPORTED)
 		nvme_prep_integrity(disk, id, bs);
 	blk_queue_logical_block_size(ns->queue, bs);
 	if (ns->noiob)
@@ -1161,7 +1312,7 @@ static void __nvme_revalidate_disk(struct gendisk *disk, struct nvme_id_ns *id)
 	else
 		set_capacity(disk, le64_to_cpup(&id->nsze) << (ns->lba_shift - 9));
 
-	if (ns->ctrl->oncs & NVME_CTRL_ONCS_DSM)
+	if (ctrl->oncs & NVME_CTRL_ONCS_DSM)
 		nvme_config_discard(ns);
 	blk_mq_unfreeze_queue(disk->queue);
 }
@@ -1766,6 +1917,7 @@ int nvme_init_identify(struct nvme_ctrl *ctrl)
 		dev_pm_qos_hide_latency_tolerance(ctrl->device);
 
 	nvme_configure_apst(ctrl);
+	nvme_setup_directives(ctrl);
 
 	ctrl->identified = true;
 
@@ -2158,6 +2310,32 @@ static struct nvme_ns *nvme_find_get_ns(struct nvme_ctrl *ctrl, unsigned nsid)
 	return ret;
 }
 
+static int nvme_setup_streams_ns(struct nvme_ctrl *ctrl, struct nvme_ns *ns)
+{
+	struct streams_directive_params s;
+	int ret;
+
+	if (!ctrl->nr_streams)
+		return 0;
+
+	ret = nvme_get_stream_params(ctrl, &s, ns->ns_id);
+	if (ret)
+		return ret;
+
+	ns->sws = le32_to_cpu(s.sws);
+	ns->sgs = le16_to_cpu(s.sgs);
+
+	if (ns->sws) {
+		unsigned int bs = 1 << ns->lba_shift;
+
+		blk_queue_io_min(ns->queue, bs * ns->sws);
+		if (ns->sgs)
+			blk_queue_io_opt(ns->queue, bs * ns->sws * ns->sgs);
+	}
+
+	return 0;
+}
+
 static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid)
 {
 	struct nvme_ns *ns;
@@ -2187,6 +2365,7 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid)
 
 	blk_queue_logical_block_size(ns->queue, 1 << ns->lba_shift);
 	nvme_set_queue_limits(ctrl, ns->queue);
+	nvme_setup_streams_ns(ctrl, ns);
 
 	sprintf(disk_name, "nvme%dn%d", ctrl->instance, ns->instance);
 
drivers/nvme/host/nvme.h +4 -0
index ec8c7363934d..f616835afc4c 100644
@@ -147,6 +147,8 @@ struct nvme_ctrl {
 	u16 oncs;
 	u16 vid;
 	u16 oacs;
+	u16 nssa;
+	u16 nr_streams;
 	atomic_t abort_limit;
 	u8 event_limit;
 	u8 vwc;
@@ -199,6 +201,8 @@ struct nvme_ns {
 	unsigned ns_id;
 	int lba_shift;
 	u16 ms;
+	u16 sgs;
+	u32 sws;
 	bool ext;
 	u8 pi_type;
 	unsigned long flags;
include/linux/nvme.h +48 -0

90 lines not on this page · raw message

Signature · 1 line
2.7.4