[dm-devel] Updated cluster log patch (take 3)

Jonathan Brassow jbrassow at redhat.com
Mon May 11 21:03:03 UTC 2009


Evgeniy, I've added the comments you requested and removed some of the
pseudo-kernel-doc style comments that you mentioned.

Agk, I've also gone through dm-log-cluster.c to remove some of the
kernel-doc style comments that didn't add any value.

 brassow

This patch contains the cluster-aware device-mapper log module.

The log types contained herein ("clustered-disk" and "clustered-core")
operate in the same way as their single machine counterparts (dm-log.c),
but they are cluster coherent.  This is done by forwarding most logging
requests to userspace, where a daemon processes the request in an ordered
fashion with the rest of the nodes in the cluster.  This is necessary to
prevent log state corruption.  (Imagine if two machines are writing to the
same region of a mirror.  They would both mark the region dirty, but you
need a cluster-aware entity that can handle properly marking the region
clean when they are done.  Otherwise, you might clear the region when the
first machine is done, not the second.)

The structure used for communication between kernel and userspace are
located in include/linux/dm-log-cluster.h.  Due to the frequency,
diversity, and 2-way communication nature of the exchanges between
kernel and userspace, 'connector' is used as the interface for
communication.  One existing userspace implementation of the daemon
uses openAIS/corosync in order to communicate with guaranteed ordering
and delivery.  This daemon, which is capable of coherently managing log
state from multiple cluster machines, can be found in the LVM2 code
repository.  Other implementations of the daemon, with no association to
LVM or openAIS/corosync, are certainly possible.

Providing a cluster-aware logging type gives device-mapper RAID1 (and
potentially other RAIDs) the ability to operate in a cluster-coherent
fashion.

Signed-off-by: Jonathan Brassow <jbrassow at redhat.com>

Index: linux-2.6/drivers/md/Kconfig
===================================================================
--- linux-2.6.orig/drivers/md/Kconfig
+++ linux-2.6/drivers/md/Kconfig
@@ -231,6 +231,16 @@ config DM_MIRROR
          Allow volume managers to mirror logical volumes, also
          needed for live data migration tools such as 'pvmove'.
 
+config DM_LOG_CLUSTERED
+	tristate "Mirror cluster logging (EXPERIMENTAL)"
+	depends on DM_MIRROR && EXPERIMENTAL
+	select CONNECTOR
+	---help---
+	  The cluster logging module provides a mechanism for
+	  keeping log state coherent amoung a cluster of machines.
+	  Device-mapper mirroring (RAID1) can leverage this log type
+	  to make mirrors that are cluster-aware.
+
 config DM_ZERO
 	tristate "Zero target"
 	depends on BLK_DEV_DM
Index: linux-2.6/drivers/md/Makefile
===================================================================
--- linux-2.6.orig/drivers/md/Makefile
+++ linux-2.6/drivers/md/Makefile
@@ -8,6 +8,7 @@ dm-multipath-y	+= dm-path-selector.o dm-
 dm-snapshot-y	+= dm-snap.o dm-exception.o dm-exception-store.o \
 		   dm-snap-transient.o dm-snap-persistent.o
 dm-mirror-y	+= dm-raid1.o
+dm-log-clustered-y += dm-log-cluster.o dm-log-cluster-transfer.o
 md-mod-y	+= md.o bitmap.o
 raid456-y	+= raid5.o
 raid6_pq-y	+= raid6algos.o raid6recov.o raid6tables.o \
@@ -38,6 +39,7 @@ obj-$(CONFIG_DM_DELAY)		+= dm-delay.o
 obj-$(CONFIG_DM_MULTIPATH)	+= dm-multipath.o dm-round-robin.o
 obj-$(CONFIG_DM_SNAPSHOT)	+= dm-snapshot.o
 obj-$(CONFIG_DM_MIRROR)		+= dm-mirror.o dm-log.o dm-region-hash.o
+obj-$(CONFIG_DM_LOG_CLUSTERED)	+= dm-log-clustered.o
 obj-$(CONFIG_DM_ZERO)		+= dm-zero.o
 
 quiet_cmd_unroll = UNROLL  $@
Index: linux-2.6/include/linux/dm-log-cluster.h
===================================================================
--- /dev/null
+++ linux-2.6/include/linux/dm-log-cluster.h
@@ -0,0 +1,56 @@
+/*
+ * Copyright (C) 2006-2009 Red Hat, Inc.
+ *
+ * This file is released under the LGPL.
+ */
+
+#ifndef __DM_LOG_CLUSTER_H__
+#define __DM_LOG_CLUSTER_H__
+
+#include <linux/dm-ioctl.h> /* For DM_UUID_LEN */
+
+/*
+ * The following are the possible request types.
+ * They represent the various functions that make up the log API
+ */
+#define DM_CLOG_CTR                    1
+#define DM_CLOG_DTR                    2
+#define DM_CLOG_PRESUSPEND             3
+#define DM_CLOG_POSTSUSPEND            4
+#define DM_CLOG_RESUME                 5
+#define DM_CLOG_GET_REGION_SIZE        6
+#define DM_CLOG_IS_CLEAN               7
+#define DM_CLOG_IN_SYNC                8
+#define DM_CLOG_FLUSH                  9
+#define DM_CLOG_MARK_REGION           10
+#define DM_CLOG_CLEAR_REGION          11
+#define DM_CLOG_GET_RESYNC_WORK       12
+#define DM_CLOG_SET_REGION_SYNC       13
+#define DM_CLOG_GET_SYNC_COUNT        14
+#define DM_CLOG_STATUS_INFO           15
+#define DM_CLOG_STATUS_TABLE          16
+#define DM_CLOG_IS_REMOTE_RECOVERING  17
+
+/*
+ * (DM_CLOG_REQUEST_MASK & request_type) to get the request type
+ *
+ * We are reserving 8 bits of the 32-bit 'request_type' field for the
+ * various request types above.  The remaining 24-bits can be
+ * reserved for future use and compatibility concerns.
+ */
+#define DM_CLOG_REQUEST_MASK 0xFF
+
+struct dm_clog_request {
+	char uuid[DM_UUID_LEN]; /* Ties a request to a specific mirror log */
+	char padding[7];        /* Padding because DM_UUID_LEN = 129 */
+
+	int32_t error;          /* Used to report back processing errors */
+
+	uint32_t seq;           /* Sequence number for request */
+	uint32_t request_type;  /* DM_CLOG_* defined above */
+	uint32_t data_size;     /* How much data (not including this struct) */
+
+	char data[0];
+};
+
+#endif /* __DM_LOG_CLUSTER_H__ */
Index: linux-2.6/drivers/md/dm-log-cluster-transfer.c
===================================================================
--- /dev/null
+++ linux-2.6/drivers/md/dm-log-cluster-transfer.c
@@ -0,0 +1,276 @@
+/*
+ * Copyright (C) 2006-2009 Red Hat, Inc.
+ *
+ * This file is released under the LGPL.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <net/sock.h>
+#include <linux/workqueue.h>
+#include <linux/connector.h>
+#include <linux/device-mapper.h>
+
+#include <linux/dm-log-cluster.h>
+#include "dm-log-cluster-transfer.h"
+
+#include <asm/div64.h> /* Unnecessary */
+
+#define SHORT_UUID(x) (strlen(x) > 8) ? ((x) + (strlen(x) - 8)) : (x)
+
+static uint32_t seq;
+
+/*
+ * Pre-allocated space for speed
+ */
+#define DM_CLOG_PREALLOCED_SIZE 512
+static struct cn_msg *prealloced_cn_msg;
+static struct dm_clog_request *prealloced_clog_tfr;
+
+static struct cb_id cn_clog_id = { CN_IDX_DM, CN_VAL_DM_CLUSTER_LOG };
+static DEFINE_MUTEX(_lock);
+
+struct receiving_pkg {
+	struct list_head list;
+	struct completion complete;
+
+	uint32_t seq;
+
+	int error;
+	size_t *data_size;
+	char *data;
+};
+
+static DEFINE_SPINLOCK(receiving_list_lock);
+static struct list_head receiving_list;
+
+static int dm_clog_sendto_server(struct dm_clog_request *tfr)
+{
+	int r;
+	int size;
+	struct cn_msg *msg = prealloced_cn_msg;
+
+	if (tfr != prealloced_clog_tfr) {
+		size = sizeof(struct cn_msg) +
+			sizeof(struct dm_clog_request) + tfr->data_size;
+		msg = kmalloc(size, GFP_NOIO);
+		if (!msg)
+			return -ENOMEM;
+		memcpy((msg + 1), tfr,
+		       sizeof(struct dm_clog_request) + tfr->data_size);
+	}
+
+	memset(msg, 0, sizeof(struct cn_msg));
+
+	msg->id.idx = cn_clog_id.idx;
+	msg->id.val = cn_clog_id.val;
+	msg->ack = 0;
+	msg->seq = tfr->seq;
+	msg->len = sizeof(struct dm_clog_request) + tfr->data_size;
+
+	r = cn_netlink_send(msg, 0, gfp_any());
+
+	if (msg != prealloced_cn_msg)
+		kfree(msg);
+
+	return r;
+}
+
+/*
+ * Parameters for this function can be either msg or tfr, but not
+ * both.  This function fills in the reply for a waiting request.
+ * If just msg is given, then the reply is simply an ACK from userspace
+ * that the request was received.
+ *
+ * Returns: 0 on success, -ENOENT on failure
+ */
+static int fill_pkg(struct cn_msg *msg, struct dm_clog_request *tfr)
+{
+	uint32_t rtn_seq = (msg) ? msg->seq : (tfr) ? tfr->seq : 0;
+	struct receiving_pkg *pkg;
+
+	/*
+	 * The 'receiving_pkg' entries in this list are statically
+	 * allocated on the stack in 'dm_clog_consult_server'.
+	 * Each process that is waiting for a reply from the user
+	 * space server will have an entry in this list.
+	 *
+	 * We are safe to do it this way because the stack space
+	 * is unique to each process, but still addressable by
+	 * other processes.
+	 */
+	list_for_each_entry(pkg, &receiving_list, list) {
+		if (rtn_seq != pkg->seq)
+			continue;
+
+		if (msg) {
+			pkg->error = -msg->ack;
+			/*
+			 * If we are trying again, we will need to know our
+			 * storage capacity.  Otherwise, along with the
+			 * error code, we make explicit that we have no data.
+			 */
+			if (pkg->error != -EAGAIN)
+				*(pkg->data_size) = 0;
+		} else if (tfr->data_size > *(pkg->data_size)) {
+			DMERR("Insufficient space to receive package [%u]::",
+			      tfr->request_type);
+			DMERR("  tfr->data_size    = %u", tfr->data_size);
+			DMERR("  *(pkg->data_size) = %lu", *(pkg->data_size));
+
+			*(pkg->data_size) = 0;
+			pkg->error = -ENOSPC;
+		} else {
+			pkg->error = tfr->error;
+			memcpy(pkg->data, tfr->data, tfr->data_size);
+			*(pkg->data_size) = tfr->data_size;
+		}
+		complete(&pkg->complete);
+		return 0;
+	}
+
+	return -ENOENT;
+}
+
+/*
+ * This is the connector callback that delivers data
+ * that was sent from userspace.
+ */
+static void cn_clog_callback(void *data)
+{
+	struct cn_msg *msg = (struct cn_msg *)data;
+	struct dm_clog_request *tfr = (struct dm_clog_request *)(msg + 1);
+
+	spin_lock(&receiving_list_lock);
+	if (msg->len == 0)
+		fill_pkg(msg, NULL);
+	else if (msg->len < sizeof(*tfr))
+		DMERR("Incomplete message received: [%u]", msg->seq);
+	else
+		fill_pkg(NULL, tfr);
+	spin_unlock(&receiving_list_lock);
+}
+
+/**
+ * dm_clog_consult_server
+ * @uuid: log's uuid (must be DM_UUID_LEN in size)
+ * @request_type:  found in include/linux/dm-log-cluster.h
+ * @data: data to tx to the server
+ * @data_size: size of data in bytes
+ * @rdata: place to put return data from server
+ * @rdata_size: value-result (amount of space given/amount of space used)
+ *
+ * Only one process at a time can communicate with the server.
+ * rdata_size is undefined on failure.
+ *
+ * Returns: 0 on success, -EXXX on failure
+ **/
+int dm_clog_consult_server(const char *uuid, int request_type,
+			   char *data, size_t data_size,
+			   char *rdata, size_t *rdata_size)
+{
+	int r = 0;
+	size_t dummy = 0;
+	int overhead_size =
+		sizeof(struct dm_clog_request *) + sizeof(struct cn_msg);
+	struct dm_clog_request *tfr = prealloced_clog_tfr;
+	struct receiving_pkg pkg;
+
+	if (data_size > (DM_CLOG_PREALLOCED_SIZE - overhead_size)) {
+		DMINFO("Size of tfr exceeds preallocated size");
+		/* FIXME: is kmalloc sufficient if we need this much space? */
+		tfr = kzalloc(data_size + sizeof(*tfr), GFP_NOIO);
+	}
+
+	if (!tfr)
+		return -ENOMEM;
+
+	if (!rdata_size)
+		rdata_size = &dummy;
+resend:
+	/*
+	 * We serialize the sending of requests so we can
+	 * use the preallocated space.
+	 */
+	mutex_lock(&_lock);
+
+	memset(tfr, 0, DM_CLOG_PREALLOCED_SIZE - overhead_size);
+	memcpy(tfr->uuid, uuid, DM_UUID_LEN);
+	tfr->seq = seq++;
+	tfr->request_type = request_type;
+	tfr->data_size = data_size;
+	if (data && data_size)
+		memcpy(tfr->data, data, data_size);
+
+	memset(&pkg, 0, sizeof(pkg));
+	init_completion(&pkg.complete);
+	pkg.seq = tfr->seq;
+	pkg.data_size = rdata_size;
+	pkg.data = rdata;
+	spin_lock(&receiving_list_lock);
+	list_add(&(pkg.list), &receiving_list);
+	spin_unlock(&receiving_list_lock);
+
+	r = dm_clog_sendto_server(tfr);
+
+	mutex_unlock(&_lock);
+
+	if (r) {
+		DMERR("Unable to send cluster log request [%u] to server: %d",
+		      request_type, r);
+		spin_lock(&receiving_list_lock);
+		list_del_init(&(pkg.list));
+		spin_unlock(&receiving_list_lock);
+
+		goto out;
+	}
+
+	r = wait_for_completion_timeout(&(pkg.complete), 15 * HZ);
+	spin_lock(&receiving_list_lock);
+	list_del_init(&(pkg.list));
+	spin_unlock(&receiving_list_lock);
+	if (!r) {
+		DMWARN("[%s] Request timed out: [%u/%u] - retrying",
+		       SHORT_UUID(uuid), request_type, pkg.seq);
+		goto resend;
+	}
+
+	r = pkg.error;
+	if (r == -EAGAIN)
+		goto resend;
+
+out:
+	if (tfr != (struct dm_clog_request *)prealloced_clog_tfr)
+		kfree(tfr);
+
+	return r;
+}
+
+int dm_clog_tfr_init(void)
+{
+	int r;
+	void *prealloced;
+
+	INIT_LIST_HEAD(&receiving_list);
+
+	prealloced = kmalloc(DM_CLOG_PREALLOCED_SIZE, GFP_KERNEL);
+	if (!prealloced)
+		return -ENOMEM;
+
+	prealloced_cn_msg = prealloced;
+	prealloced_clog_tfr = prealloced + sizeof(struct cn_msg);
+
+	r = cn_add_callback(&cn_clog_id, "clulog", cn_clog_callback);
+	if (r) {
+		cn_del_callback(&cn_clog_id);
+		return r;
+	}
+
+	return 0;
+}
+
+void dm_clog_tfr_exit(void)
+{
+	cn_del_callback(&cn_clog_id);
+	kfree(prealloced_cn_msg);
+}
Index: linux-2.6/drivers/md/dm-log-cluster-transfer.h
===================================================================
--- /dev/null
+++ linux-2.6/drivers/md/dm-log-cluster-transfer.h
@@ -0,0 +1,18 @@
+/*
+ * Copyright (C) 2006-2009 Red Hat, Inc.
+ *
+ * This file is released under the LGPL.
+ */
+
+#ifndef __DM_LOG_CLUSTER_TRANSFER_H__
+#define __DM_LOG_CLUSTER_TRANSFER_H__
+
+#define DM_MSG_PREFIX "dm-log-clustered"
+
+int dm_clog_tfr_init(void);
+void dm_clog_tfr_exit(void);
+int dm_clog_consult_server(const char *uuid, int request_type,
+			   char *data, size_t data_size,
+			   char *rdata, size_t *rdata_size);
+
+#endif /* __DM_LOG_CLUSTER_TRANSFER_H__ */
Index: linux-2.6/drivers/md/dm-log-cluster.c
===================================================================
--- /dev/null
+++ linux-2.6/drivers/md/dm-log-cluster.c
@@ -0,0 +1,745 @@
+/*
+ * Copyright (C) 2006-2009 Red Hat, Inc.
+ *
+ * This file is released under the LGPL.
+ */
+
+#include <linux/blkdev.h> /* for sector_div, which is used in dm-dirty-log.h */
+#include <linux/bio.h>
+#include <linux/dm-dirty-log.h>
+#include <linux/device-mapper.h>
+
+#include <linux/dm-log-cluster.h>
+#include "dm-log-cluster-transfer.h"
+
+struct flush_entry {
+	int type;
+	region_t region;
+	struct list_head list;
+};
+
+struct log_c {
+	struct dm_target *ti;
+	uint32_t region_size;
+	region_t region_count;
+	char uuid[DM_UUID_LEN];
+
+	char *ctr_str; /* Gives ability to restart if userspace dies */
+	uint32_t ctr_size;
+
+	/*
+	 * in_sync_hint gets set when doing is_remote_recovering.  It
+	 * represents the first region that needs recovery.  IOW, the
+	 * first zero bit of sync_bits.  This can be useful for to limit
+	 * traffic for calls like is_remote_recovering and get_resync_work,
+	 * but be take care in its use for anything else.
+	 */
+	uint64_t in_sync_hint;
+
+	spinlock_t flush_lock;
+	struct list_head flush_list;  /* only for clear and mark requests */
+
+	struct dm_dev *disk_log;
+};
+
+static mempool_t *flush_entry_pool;
+
+static void *flush_entry_alloc(gfp_t gfp_mask, void *pool_data)
+{
+	return kmalloc(sizeof(struct flush_entry), gfp_mask);
+}
+
+static void flush_entry_free(void *element, void *pool_data)
+{
+	kfree(element);
+}
+
+int cluster_do_request(struct log_c *lc, const char *uuid, int request_type,
+		       char *data, size_t data_size,
+		       char *rdata, size_t *rdata_size)
+{
+	int r;
+
+	/*
+	 * If the server isn't there, -ESRCH is returned,
+	 * and we must keep trying until the server is
+	 * restored.
+	 */
+retry:
+	r = dm_clog_consult_server(uuid, request_type, data,
+				   data_size, rdata, rdata_size);
+
+	if (r != -ESRCH)
+		return r;
+
+	DMERR(" Userspace cluster log server not found.");
+	while (1) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		schedule_timeout(2*HZ);
+		DMWARN("Attempting to contact cluster log server...");
+		r = dm_clog_consult_server(uuid, DM_CLOG_CTR, lc->ctr_str,
+					   lc->ctr_size, NULL, NULL);
+		if (!r)
+			break;
+	}
+	DMINFO("Reconnected to cluster log server... CTR complete");
+	r = dm_clog_consult_server(uuid, DM_CLOG_RESUME, NULL,
+				   0, NULL, NULL);
+	if (!r)
+		goto retry;
+
+	DMERR("Error trying to resume cluster log: %d", r);
+
+	return -ESRCH;
+}
+
+static int cluster_ctr(struct dm_dirty_log *log, struct dm_target *ti,
+		       unsigned int argc, char **argv,
+		       struct dm_dev *disk_log)
+{
+	int i;
+	int r = 0;
+	int str_size;
+	int offset = (disk_log) ? 1 : 0;
+	char *ctr_str = NULL;
+	struct log_c *lc = NULL;
+	uint32_t region_size;
+	region_t region_count;
+
+	/* Already checked argument count */
+
+	if (sscanf(argv[offset], "%u", &region_size) != 1) {
+		DMWARN("Invalid region size string");
+		return -EINVAL;
+	}
+
+	region_count = dm_sector_div_up(ti->len, region_size);
+
+	lc = kmalloc(sizeof(*lc), GFP_KERNEL);
+	if (!lc) {
+		DMWARN("Unable to allocate cluster log context.");
+		return -ENOMEM;
+	}
+
+	lc->ti = ti;
+	lc->region_size = region_size;
+	lc->region_count = region_count;
+	lc->disk_log = disk_log;
+
+	/* FIXME: Need to check size of uuid arg */
+	memcpy(lc->uuid, argv[1 + offset], DM_UUID_LEN);
+	spin_lock_init(&lc->flush_lock);
+	INIT_LIST_HEAD(&lc->flush_list);
+
+	for (i = 0, str_size = 0; i < argc; i++)
+		str_size += strlen(argv[i]) + 1; /* +1 for space between args */
+
+	str_size += 20; /* Max number of chars in a printed u64 number */
+
+	ctr_str = kzalloc(str_size, GFP_KERNEL);
+	if (!ctr_str) {
+		DMWARN("Unable to allocate memory for constructor string");
+		kfree(lc);
+		return -ENOMEM;
+	}
+
+	for (i = 0, str_size = 0; i < argc; i++)
+		str_size += sprintf(ctr_str + str_size, "%s ", argv[i]);
+	str_size += sprintf(ctr_str + str_size, "%llu",
+			    (unsigned long long)ti->len);
+
+	/* Send table string */
+	r = dm_clog_consult_server(lc->uuid, DM_CLOG_CTR,
+				   ctr_str, str_size, NULL, NULL);
+
+	if (r == -ESRCH)
+		DMERR(" Userspace cluster log server not found");
+
+	if (r) {
+		kfree(lc);
+		kfree(ctr_str);
+	} else {
+		lc->ctr_str = ctr_str;
+		lc->ctr_size = str_size;
+		log->context = lc;
+	}
+
+	return r;
+}
+
+/*
+ * cluster_core_ctr
+ *
+ * argv contains:
+ *   <region_size> <uuid> [[no]sync]
+ *
+ * Returns: 0 on success, -XXX on failure
+ */
+static int cluster_core_ctr(struct dm_dirty_log *log, struct dm_target *ti,
+			    unsigned int argc, char **argv)
+{
+	int i, r;
+	if ((argc < 2) || (argc > 3)) {
+		DMERR("Too %s arguments to clustered-core mirror log type.",
+		      (argc < 2) ? "few" : "many");
+		DMERR("  %d arguments supplied:", argc);
+		for (i = 0; i < argc; i++)
+			DMERR("    %s", argv[i]);
+		return -EINVAL;
+	}
+
+	r = cluster_ctr(log, ti, argc, argv, NULL);
+
+	return r;
+}
+
+
+/*
+ * cluster_core_ctr
+ *
+ * argv contains:
+ *   <disk> <region_size> <uuid> [[no]sync]
+ *
+ * Returns: 0 on success, -XXX on failure
+ */
+static int cluster_disk_ctr(struct dm_dirty_log *log, struct dm_target *ti,
+			    unsigned int argc, char **argv)
+{
+	int r, i;
+	struct dm_dev *dev;
+
+	if ((argc < 3) || (argc > 4)) {
+		DMERR("Too %s arguments to clustered-disk mirror log type.",
+		      (argc < 3) ? "few" : "many");
+		DMERR("  %d arguments supplied:", argc);
+		for (i = 0; i < argc; i++)
+			DMERR("    %s", argv[i]);
+		return -EINVAL;
+	}
+
+	r = dm_get_device(ti, argv[0], 0, 0, FMODE_READ | FMODE_WRITE, &dev);
+	if (r)
+		return r;
+
+	r = cluster_ctr(log, ti, argc, argv, dev);
+	if (r)
+		dm_put_device(ti, dev);
+
+	return r;
+}
+
+static void cluster_dtr(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	r = dm_clog_consult_server(lc->uuid, DM_CLOG_DTR,
+				   NULL, 0,
+				   NULL, NULL);
+
+	if (lc->disk_log)
+		dm_put_device(lc->ti, lc->disk_log);
+	kfree(lc->ctr_str);
+	kfree(lc);
+
+	return;
+}
+
+static int cluster_presuspend(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	r = dm_clog_consult_server(lc->uuid, DM_CLOG_PRESUSPEND,
+				   NULL, 0,
+				   NULL, NULL);
+
+	return r;
+}
+
+static int cluster_postsuspend(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	r = dm_clog_consult_server(lc->uuid, DM_CLOG_POSTSUSPEND,
+				   NULL, 0,
+				   NULL, NULL);
+
+	return r;
+}
+
+static int cluster_resume(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	lc->in_sync_hint = 0;
+	r = dm_clog_consult_server(lc->uuid, DM_CLOG_RESUME,
+				   NULL, 0,
+				   NULL, NULL);
+
+	return r;
+}
+
+static uint32_t cluster_get_region_size(struct dm_dirty_log *log)
+{
+	struct log_c *lc = log->context;
+
+	return lc->region_size;
+}
+
+/*
+ * cluster_is_clean
+ *
+ * Check whether a region is clean.  If there is any sort of
+ * failure when consulting the server, we return not clean.
+ *
+ * Returns: 1 if clean, 0 otherwise
+ */
+static int cluster_is_clean(struct dm_dirty_log *log, region_t region)
+{
+	int r;
+	int is_clean;
+	size_t rdata_size;
+	struct log_c *lc = log->context;
+
+	rdata_size = sizeof(is_clean);
+	r = cluster_do_request(lc, lc->uuid, DM_CLOG_IS_CLEAN,
+			       (char *)&region, sizeof(region),
+			       (char *)&is_clean, &rdata_size);
+
+	return (r) ? 0 : is_clean;
+}
+
+/*
+ * cluster_in_sync
+ *
+ * Check if the region is in-sync.  If there is any sort
+ * of failure when consulting the server, we assume that
+ * the region is not in sync.
+ *
+ * If 'can_block' is set, return immediately
+ *
+ * Returns: 1 if in-sync, 0 if not-in-sync, -EWOULDBLOCK
+ */
+static int cluster_in_sync(struct dm_dirty_log *log, region_t region,
+			   int can_block)
+{
+	int r;
+	int in_sync;
+	size_t rdata_size;
+	struct log_c *lc = log->context;
+
+	/*
+	 * We can never respond directly - even if in_sync_hint is
+	 * set.  This is because another machine could see a device
+	 * failure and mark the region out-of-sync.  If we don't go
+	 * to userspace to ask, we might think the region is in-sync
+	 * and allow a read to pick up data that is stale.  (This is
+	 * very unlikely if a device actually fails; but it is very
+	 * likely if a connection to one device from one machine fails.)
+	 *
+	 * There still might be a problem if the mirror caches the region
+	 * state as in-sync... but then this call would not be made.  So,
+	 * that is a mirror problem.
+	 */
+	if (!can_block)
+		return -EWOULDBLOCK;
+
+	rdata_size = sizeof(in_sync);
+	r = cluster_do_request(lc, lc->uuid, DM_CLOG_IN_SYNC,
+			       (char *)&region, sizeof(region),
+			       (char *)&in_sync, &rdata_size);
+	return (r) ? 0 : in_sync;
+}
+
+/*
+ * cluster_flush
+ *
+ * This function is ok to block.
+ * The flush happens in two stages.  First, it sends all
+ * clear/mark requests that are on the list.  Then it
+ * tells the server to commit them.  This gives the
+ * server a chance to optimise the commit to the cluster
+ * and/or disk, instead of doing it for every request.
+ *
+ * Additionally, we could implement another thread that
+ * sends the requests up to the server - reducing the
+ * load on flush.  Then the flush would have less in
+ * the list and be responsible for the finishing commit.
+ *
+ * Returns: 0 on success, < 0 on failure
+ */
+static int cluster_flush(struct dm_dirty_log *log)
+{
+	int r = 0;
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	LIST_HEAD(flush_list);
+	struct flush_entry *fe, *tmp_fe;
+
+	spin_lock_irqsave(&lc->flush_lock, flags);
+	list_splice_init(&lc->flush_list, &flush_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+
+	if (list_empty(&flush_list))
+		return 0;
+
+	/*
+	 * FIXME: Count up requests, group request types,
+	 * allocate memory to stick all requests in and
+	 * send to server in one go.  Failing the allocation,
+	 * do it one by one.
+	 */
+
+	list_for_each_entry(fe, &flush_list, list) {
+		r = cluster_do_request(lc, lc->uuid, fe->type,
+				       (char *)&fe->region,
+				       sizeof(fe->region),
+				       NULL, NULL);
+		if (r)
+			goto fail;
+	}
+
+	r = cluster_do_request(lc, lc->uuid, DM_CLOG_FLUSH,
+			       NULL, 0, NULL, NULL);
+
+fail:
+	/*
+	 * We can safely remove these entries, even if failure.
+	 * Calling code will receive an error and will know that
+	 * the log facility has failed.
+	 */
+	list_for_each_entry_safe(fe, tmp_fe, &flush_list, list) {
+		list_del(&fe->list);
+		mempool_free(fe, flush_entry_pool);
+	}
+
+	if (r)
+		dm_table_event(lc->ti->table);
+
+	return r;
+}
+
+/*
+ * cluster_mark_region
+ *
+ * This function should avoid blocking unless absolutely required.
+ * (Memory allocation is valid for blocking.)
+ */
+static void cluster_mark_region(struct dm_dirty_log *log, region_t region)
+{
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	struct flush_entry *fe;
+
+	/* Wait for an allocation, but _never_ fail */
+	fe = mempool_alloc(flush_entry_pool, GFP_NOIO);
+	BUG_ON(!fe);
+
+	spin_lock_irqsave(&lc->flush_lock, flags);
+	fe->type = DM_CLOG_MARK_REGION;
+	fe->region = region;
+	list_add(&fe->list, &lc->flush_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+
+	return;
+}
+
+/*
+ * cluster_clear_region
+ *
+ * This function must not block.
+ * So, the alloc can't block.  In the worst case, it is ok to
+ * fail.  It would simply mean we can't clear the region.
+ * Does nothing to current sync context, but does mean
+ * the region will be re-sync'ed on a reload of the mirror
+ * even though it is in-sync.
+ */
+static void cluster_clear_region(struct dm_dirty_log *log, region_t region)
+{
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	struct flush_entry *fe;
+
+	/*
+	 * If we fail to allocate, we skip the clearing of
+	 * the region.  This doesn't hurt us in any way, except
+	 * to cause the region to be resync'ed when the
+	 * device is activated next time.
+	 */
+	fe = mempool_alloc(flush_entry_pool, GFP_ATOMIC);
+	if (!fe) {
+		DMERR("Failed to allocate memory to clear region.");
+		return;
+	}
+
+	spin_lock_irqsave(&lc->flush_lock, flags);
+	fe->type = DM_CLOG_CLEAR_REGION;
+	fe->region = region;
+	list_add(&fe->list, &lc->flush_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+
+	return;
+}
+
+/*
+ * cluster_get_resync_work
+ *
+ * Get a region that needs recovery.  It is valid to return
+ * an error for this function.
+ *
+ * Returns: 1 if region filled, 0 if no work, <0 on error
+ */
+static int cluster_get_resync_work(struct dm_dirty_log *log, region_t *region)
+{
+	int r;
+	size_t rdata_size;
+	struct log_c *lc = log->context;
+	struct {
+		int i;
+		region_t r;
+	} pkg;
+
+	if (lc->in_sync_hint >= lc->region_count)
+		return 0;
+
+	rdata_size = sizeof(pkg);
+	r = cluster_do_request(lc, lc->uuid, DM_CLOG_GET_RESYNC_WORK,
+			       NULL, 0,
+			       (char *)&pkg, &rdata_size);
+
+	*region = pkg.r;
+	return (r) ? r : pkg.i;
+}
+
+/*
+ * cluster_set_region_sync
+ *
+ * Set the sync status of a given region.  This function
+ * must not fail.
+ */
+static void cluster_set_region_sync(struct dm_dirty_log *log,
+				    region_t region, int in_sync)
+{
+	int r;
+	struct log_c *lc = log->context;
+	struct {
+		region_t r;
+		int i;
+	} pkg;
+
+	pkg.r = region;
+	pkg.i = in_sync;
+
+	r = cluster_do_request(lc, lc->uuid, DM_CLOG_SET_REGION_SYNC,
+			       (char *)&pkg, sizeof(pkg),
+			       NULL, NULL);
+
+	/*
+	 * It would be nice to be able to report failures.
+	 * However, it is easy emough to detect and resolve.
+	 */
+	return;
+}
+
+/*
+ * cluster_get_sync_count
+ *
+ * If there is any sort of failure when consulting the server,
+ * we assume that the sync count is zero.
+ *
+ * Returns: sync count on success, 0 on failure
+ */
+static region_t cluster_get_sync_count(struct dm_dirty_log *log)
+{
+	int r;
+	size_t rdata_size;
+	region_t sync_count;
+	struct log_c *lc = (struct log_c *)log->context;
+
+	rdata_size = sizeof(sync_count);
+	r = cluster_do_request(lc, lc->uuid, DM_CLOG_GET_SYNC_COUNT,
+			       NULL, 0,
+			       (char *)&sync_count, &rdata_size);
+
+	if (r)
+		return 0;
+
+	if (sync_count >= lc->region_count)
+		lc->in_sync_hint = lc->region_count;
+
+	return sync_count;
+}
+
+/*
+ * cluster_status
+ *
+ * Returns: amount of space consumed
+ */
+static int cluster_status(struct dm_dirty_log *log, status_type_t status_type,
+			  char *result, unsigned int maxlen)
+{
+	int r = 0;
+	size_t sz = (size_t)maxlen;
+	struct log_c *lc = log->context;
+
+	switch (status_type) {
+	case STATUSTYPE_INFO:
+		r = cluster_do_request(lc, lc->uuid, DM_CLOG_STATUS_INFO,
+				       NULL, 0,
+				       result, &sz);
+		/*
+		 * FIXME: If we fail to contact server, we should still
+		 * populate this with parsible results
+		 */
+		break;
+	case STATUSTYPE_TABLE:
+		r = cluster_do_request(lc, lc->uuid, DM_CLOG_STATUS_TABLE,
+				       NULL, 0,
+				       result, &sz);
+		break;
+	}
+	return (r) ? 0 : (int)sz;
+}
+
+/*
+ * cluster_is_remote_recovering
+ *
+ * Returns: 1 if region recovering, 0 otherwise
+ */
+static int cluster_is_remote_recovering(struct dm_dirty_log *log,
+					region_t region)
+{
+	int r;
+	struct log_c *lc = log->context;
+	static unsigned long long limit;
+	struct {
+		int is_recovering;
+		uint64_t in_sync_hint;
+	} pkg;
+	size_t rdata_size = sizeof(pkg);
+
+	/*
+	 * Once the mirror has been reported to be in-sync,
+	 * it will never again ask for recovery work.  So,
+	 * we can safely say there is not a remote machine
+	 * recovering if the device is in-sync.  (in_sync_hint
+	 * must be reset at resume time.)
+	 */
+	if (region < lc->in_sync_hint)
+		return 0;
+	else if (jiffies < limit)
+		return 1;
+
+	limit = jiffies + (HZ / 4);
+	r = cluster_do_request(lc, lc->uuid, DM_CLOG_IS_REMOTE_RECOVERING,
+			       (char *)&region, sizeof(region),
+			       (char *)&pkg, &rdata_size);
+	if (r)
+		return 1;
+
+	lc->in_sync_hint = pkg.in_sync_hint;
+
+	return pkg.is_recovering;
+}
+
+static struct dm_dirty_log_type _clustered_core_type = {
+	.name = "clustered-core",
+	.module = THIS_MODULE,
+	.ctr = cluster_core_ctr,
+	.dtr = cluster_dtr,
+	.presuspend = cluster_presuspend,
+	.postsuspend = cluster_postsuspend,
+	.resume = cluster_resume,
+	.get_region_size = cluster_get_region_size,
+	.is_clean = cluster_is_clean,
+	.in_sync = cluster_in_sync,
+	.flush = cluster_flush,
+	.mark_region = cluster_mark_region,
+	.clear_region = cluster_clear_region,
+	.get_resync_work = cluster_get_resync_work,
+	.set_region_sync = cluster_set_region_sync,
+	.get_sync_count = cluster_get_sync_count,
+	.status = cluster_status,
+	.is_remote_recovering = cluster_is_remote_recovering,
+};
+
+static struct dm_dirty_log_type _clustered_disk_type = {
+	.name = "clustered-disk",
+	.module = THIS_MODULE,
+	.ctr = cluster_disk_ctr,
+	.dtr = cluster_dtr,
+	.presuspend = cluster_presuspend,
+	.postsuspend = cluster_postsuspend,
+	.resume = cluster_resume,
+	.get_region_size = cluster_get_region_size,
+	.is_clean = cluster_is_clean,
+	.in_sync = cluster_in_sync,
+	.flush = cluster_flush,
+	.mark_region = cluster_mark_region,
+	.clear_region = cluster_clear_region,
+	.get_resync_work = cluster_get_resync_work,
+	.set_region_sync = cluster_set_region_sync,
+	.get_sync_count = cluster_get_sync_count,
+	.status = cluster_status,
+	.is_remote_recovering = cluster_is_remote_recovering,
+};
+
+static int __init cluster_dirty_log_init(void)
+{
+	int r = 0;
+
+	flush_entry_pool = mempool_create(100, flush_entry_alloc,
+					  flush_entry_free, NULL);
+
+	if (!flush_entry_pool) {
+		DMWARN("Unable to create flush_entry_pool:  No memory.");
+		return -ENOMEM;
+	}
+
+	r = dm_clog_tfr_init();
+	if (r) {
+		DMWARN("Unable to initialize cluster log communications");
+		mempool_destroy(flush_entry_pool);
+		return r;
+	}
+
+	r = dm_dirty_log_type_register(&_clustered_core_type);
+	if (r) {
+		DMWARN("Couldn't register clustered-core dirty log type");
+		dm_clog_tfr_exit();
+		mempool_destroy(flush_entry_pool);
+		return r;
+	}
+
+	r = dm_dirty_log_type_register(&_clustered_disk_type);
+	if (r) {
+		DMWARN("Couldn't register clustered-disk dirty log type");
+		dm_dirty_log_type_unregister(&_clustered_core_type);
+		dm_clog_tfr_exit();
+		mempool_destroy(flush_entry_pool);
+		return r;
+	}
+
+	DMINFO("(built %s %s) installed", __DATE__, __TIME__);
+	return 0;
+}
+
+static void __exit cluster_dirty_log_exit(void)
+{
+	dm_dirty_log_type_unregister(&_clustered_disk_type);
+	dm_dirty_log_type_unregister(&_clustered_core_type);
+	dm_clog_tfr_exit();
+	mempool_destroy(flush_entry_pool);
+	DMINFO("(built %s %s) removed", __DATE__, __TIME__);
+	return;
+}
+
+module_init(cluster_dirty_log_init);
+module_exit(cluster_dirty_log_exit);
+
+MODULE_DESCRIPTION(DM_NAME " cluster-aware dirty log");
+MODULE_AUTHOR("Jonathan Brassow <dm-devel at redhat.com>");
+MODULE_LICENSE("GPL");
Index: linux-2.6/Documentation/dm-log.txt
===================================================================
--- /dev/null
+++ linux-2.6/Documentation/dm-log.txt
@@ -0,0 +1,63 @@
+Device-Mapper Logging
+=====================
+The device-mapper logging code is used by some of the device-mapper
+RAID targets to track regions of the disk that are not consistent.
+A region (or portion of the address space) of the disk may be
+inconsistent because a RAID stripe is currently being operated on or
+a machine died while the region was being altered.  In the case of
+mirrors, a region would be considered dirty/inconsistent while you
+are writing to it because the writes need to be replicated for all
+the legs of the mirror and may not reach the legs at the same time.
+Once all writes are complete, the region is considered clean again.
+
+There is a generic logging interface that the device-mapper RAID
+implementations use to perform logging operations (see
+dm_dirty_log_type in include/linux/dm-dirty-log.h).  Various different
+logging implementations are available and provide different
+capabilities.  The list includes:
+
+Type	       Files
+====	       =====
+disk	       drivers/md/dm-log.c
+core	       drivers/md/dm-log.c
+cluster-disk   drivers/md/dm-log-cluster* include/linux/dm-log-cluster.h
+cluster-core   drivers/md/dm-log-cluster* include/linux/dm-log-cluster.h
+
+The "disk" log type
+-------------------
+This log implementation commits the log state to disk.  This way, the
+logging state survives reboots/crashes.
+
+The "core" log type
+-------------------
+This log implementation keeps the log state in memory.  The log state
+will not survive a reboot or crash, but there may be a small boost in
+performance.  This method can also be used if no storage device is
+available for storing log state.
+
+The "cluster-*" log types
+-------------------------
+These types operate in the same way as their single machine counterparts,
+but they are cluster-aware.  This is done by forwarding most logging
+requests to userspace, where a daemon processes the request in an ordered
+fashion with the rest of the nodes in the cluster.  This is necessary to
+prevent log state corruption.  (Imagine if two machines are writing to the
+same region of a mirror.  They would both mark the region dirty, but you
+need a cluster-aware entity that can handle properly marking the region
+clean when they are done.  Otherwise, you might clear the region when the
+first machine is done, not the second.)
+
+The structure used for communication between kernel and userspace are
+located in include/linux/dm-log-cluster.h.  Due to the frequency,
+diversity, and 2-way communication nature of the exchanges between
+kernel and userspace, 'connector' is used as the interface for
+communication.  One existing userspace implementation of the daemon
+uses openAIS/corosync in order to communicate with guaranteed ordering
+and delivery.  This daemon, which is capable of coherently managing log
+state from multiple cluster machines, can be found in the LVM2 code
+repository.  Other implementations with no association to LVM or
+openAIS/corosync are certainly possible.
+
+Providing a cluster-aware logging type gives device-mapper RAID1 (and
+potentially other RAIDs) the ability to operate in a cluster-aware
+fashion.
Index: linux-2.6/include/linux/connector.h
===================================================================
--- linux-2.6.orig/include/linux/connector.h
+++ linux-2.6/include/linux/connector.h
@@ -41,8 +41,10 @@
 #define CN_IDX_BB			0x5	/* BlackBoard, from the TSP GPL sampling framework */
 #define CN_DST_IDX			0x6
 #define CN_DST_VAL			0x1
+#define CN_IDX_DM			0x7	/* Device Mapper */
+#define CN_VAL_DM_CLUSTER_LOG		0x1
 
-#define CN_NETLINK_USERS		7
+#define CN_NETLINK_USERS		8
 
 /*
  * Maximum connector's message size.
Index: linux-2.6/include/linux/Kbuild
===================================================================
--- linux-2.6.orig/include/linux/Kbuild
+++ linux-2.6/include/linux/Kbuild
@@ -57,6 +57,7 @@ header-y += dlmconstants.h
 header-y += dlm_device.h
 header-y += dlm_netlink.h
 header-y += dm-ioctl.h
+header-y += dm-log-cluster.h
 header-y += dn.h
 header-y += dqblk_xfs.h
 header-y += efs_fs_sb.h





More information about the dm-devel mailing list