Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Don't destroy user event fields when removal fails

   User event fields are destroyed before the event is removed from
   visibility. But that can fail leaving the still visible event with no
   fields. Move the destroying of the fields to after the event is
   successfully removed from visibility.

 - Initialize function graph state is fork before calling
   copy_exec_state()

   For non-CLONE_VM forks, copy_exec_state() allocates a new
   task_exec_state. If that allocation fails, ftrace_graph_exit_task()
   will free the tasks ret_stack pointer. Since that pointer is still
   using the parent's ret_stack, it mistakenly frees the parent's
   pointer too.

   Call ftrace_graph_init() on the task first which will NULL out the
   new tasks's ret_stack and if the copy fails, it will not free
   anything.

 - Remove FGRAPH_MAX_INDEX

   The macro FGRAPH_MAX_INDEX was added but never used. Remove it.

 - Save ent_size in function graph printing of nested functions

   The function graph tracer needs to look at the next event to see if
   the next event is the return of the current function entry. If it is,
   it prints a single line:

	ktime_get();

   Otherwise it prints it like a nested function:

	tick_nohz_irq_exit() {
	    ktime_get();
	    kcpustat_irq_exit();
	}

   In order to look at the next event, it must save the current event so
   that it has the information to print from it. It saves the event in
   the iterator descriptor called "ent". What it doesn't save is the
   ent_size of the event which is now used to know if the function graph
   arguments are to be printed. The peek doesn't save the size so the
   size used happens to be that of the size of the last event that was
   seen.

   Save the entry event size in the iterator descriptor so that the
   correct size is used.

 - Fix several errors with freeing data in the histogram code

   The histogram code had a lot of leaked or or incorrect accounting
   when failures happen. Correct them.

 - Fix histogram regression of .percent and .graph modifiers

   Up until 6.3 histogram values could have "percent" or "graph"
   modifiers that changed how they were printed. But a change that added
   restricting histograms values from being strings, stack traces and
   other modifiers inadvertently prevented them from using the percent
   and graph modifiers, which were legal use cases for values.

   Put back the percent and graph modifiers.

 - Fix various typos in the comments

 - Set the trace_clock before initializing a histogram with clock
   argument

   The histogram API allows the user to specific which trace clock to
   use via a "clock=" string. The histogram is set up first before the
   clock is checked. If the passed in clock is not valid, it exits
   without fully fixing up the histogram leaving it on the list and a
   use-after-free can trigger.

   Update the clock argument first and if it fails then exit gracefully
   before the histogram trigger is placed on any lists.

 - Restore :mod: trailer after parsing in ftrace_set_clr_event

   The function ftrace_set_clr_event() modifies the parse string and
   needs to put it back to what was passed in. It searches for ":mod:"
   via a strsep() but fails to put back the first ':' in the string.

   Add back the ':' in the passed in string.

 - Take trace_array reference when opening a tracer options file

   The options files are dynamically created and some tracers add their
   own options. When a tracer adds their own list of options, the
   trace_array holding them has an array to hold the list of options for
   each tracer. This array increases in size via a krealloc(), and the
   new entry gets a newly allocated array to hold the options of the new
   tracer being added.

   The element in each entry of the tracer's option array holds a
   pointer back to the trace_array, a pointer to the tracer it is
   associated to, a pointer to the flags of the option.

   The issue is that these arrays are freed when the trace_array is
   freed when its instance it represents is removed from the instances
   directory. There's a race that an open of one of these options files
   can happen when the instance is being removed.

   Add a new helper function to be called by the open function of the
   options file to iterate all existing trace_arrays under a lock and
   find the one that has the given option element in one of it's tracer
   arrays. If found, then update the associated trace_array's reference
   counter to keep it from being freed. If not found, have the open call
   return -ENODEV.

 - Disable interrupts when acquiring the lock in rb_wake_up_waiters()

   The function rb_wake_up_waiters() assumes it will be called in
   interrupt context and does not disable irqs when taking
   cpu_buffer->reader_lock, which can be called in hard interrupt
   context. The issue is in PREEMPT_RT, this function is called in
   thread context leaving this lock open to a deadlock.

   Take the lock with interrupts disabled.

 - Use rcu_assign_pointer() for tmp_ops filter hash

   The tmp_ops used in update_ftrace_direct_mod() assigns its
   filter_hash field directly, but that field is annotated as __rcu and
   sparse complains. Assign it with rcu_assign_pointer()

 - Fix use-after-free in enable_trigger_private_data_free()

   The trace_event_call is accessed through the event_trigger_data's
   trace_event_file pointer to put the trace_event_call on freeing. The
   issue is that the trace_event_file data may have been freed already
   causing a use-after-free. Add a field to the event_trigger_data that
   points directly to the trace_event_call so that it can decrement its
   reference directly without needing to go through the
   trace_event_file.

 - Fix accounting of buffer data remote headers

   trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount
   the number of pages is needed for the asked for size as it doesn't
   take into account the meta data on each page. Add a helper function
   to do the calculation properly and use that in these functions.

 - Catch nr_page_va overflow in ring_buffer_desc sizing

   The number of pages per remote ring buffer is capped by
   ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to
   overflow that field would silently allocate a descriptor smaller than
   what was asked for.

 - Do not resize the subbuf order if any per_cpu buffer is disabled

   The mmapping of ring buffers disables resizing the subbuffers, but it
   is done per-cpu whereas the subbuf size change is done for all the
   per_cpu buffers under the buffer->mutex. It could change the size of
   some while the mapping is happening on others. Have the resize of the
   subbuf order check all the per_cpu buffers under the lock to see if
   any of them is disabled before starting and causing an inconsistency
   between buffers that are being mapped.

* tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits)
  ring-buffer: Check resize_disabled before publishing the new subbuf order
  tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing
  tracing/remotes: Account for ring buffer page header in size calculation
  tracing: Don't dereference trace_event_file in deferred trigger free
  ftrace: Use rcu_assign_pointer() for tmp_ops filter hash
  ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters()
  tracing: Take trace_array reference when opening a tracer options file
  tracing: Fix ring_buffer_read_page_size() kernel-doc
  tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event()
  tracing: Fix memory corruption from a "STACKTRACE" histogram key
  tracing: Fix memory corruption from the histogram stacktrace modifier
  tracing: Undo the registration when enabling the histogram trigger fails
  tracing: Take the reference before publishing the named histogram trigger
  tracing: Set the trace clock before registering the histogram trigger
  tracing: Fix typo "preceeded" in comment
  tracing: Fix typo "availabe" in comment
  tracing: Let histogram values keep the percent and graph modifiers
  tracing: Keep the entry count when the histogram stats allocation fails
  tracing: Free histogram the field rejected for a bad modifier
  tracing: Free histogram the var ref when its initialization fails
  ...

1 files changed, 932 insertions, 0 deletions

diff --git a/drivers/md/dm-log-userspace-base.c b/drivers/md/dm-log-userspace-base.c
new file mode 100644
index 000000000..3ca35aa4e
--- /dev/null
+++ b/drivers/md/dm-log-userspace-base.c
@@ -0,0 +1,932 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2006-2009 Red Hat, Inc.
+ *
+ * This file is released under the LGPL.
+ */
+
+#include <linux/bio.h>
+#include <linux/slab.h>
+#include <linux/jiffies.h>
+#include <linux/dm-dirty-log.h>
+#include <linux/device-mapper.h>
+#include <linux/dm-log-userspace.h>
+#include <linux/module.h>
+#include <linux/workqueue.h>
+
+#include "dm-log-userspace-transfer.h"
+
+#define DM_LOG_USERSPACE_VSN "1.3.0"
+
+#define FLUSH_ENTRY_POOL_SIZE 16
+
+struct dm_dirty_log_flush_entry {
+	int type;
+	region_t region;
+	struct list_head list;
+};
+
+/*
+ * This limit on the number of mark and clear request is, to a degree,
+ * arbitrary.  However, there is some basis for the choice in the limits
+ * imposed on the size of data payload by dm-log-userspace-transfer.c:
+ * dm_consult_userspace().
+ */
+#define MAX_FLUSH_GROUP_COUNT 32
+
+struct log_c {
+	struct dm_target *ti;
+	struct dm_dev *log_dev;
+
+	char *usr_argv_str;
+	uint32_t usr_argc;
+
+	uint32_t region_size;
+	region_t region_count;
+	uint64_t luid;
+	char uuid[DM_UUID_LEN];
+
+	/*
+	 * Mark and clear requests are held until a flush is issued
+	 * so that we can group, and thereby limit, the amount of
+	 * network traffic between kernel and userspace.  The 'flush_lock'
+	 * is used to protect these lists.
+	 */
+	spinlock_t flush_lock;
+	struct list_head mark_list;
+	struct list_head clear_list;
+
+	/*
+	 * 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;
+
+	/*
+	 * Workqueue for flush of clear region requests.
+	 */
+	struct workqueue_struct *dmlog_wq;
+	struct delayed_work flush_log_work;
+	atomic_t sched_flush;
+
+	/*
+	 * Combine userspace flush and mark requests for efficiency.
+	 */
+	uint32_t integrated_flush;
+
+	mempool_t flush_entry_pool;
+};
+
+static struct kmem_cache *_flush_entry_cache;
+
+static int userspace_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_consult_userspace(uuid, lc->luid, request_type, data,
+				 data_size, rdata, rdata_size);
+
+	if (r != -ESRCH)
+		return r;
+
+	DMERR(" Userspace log server not found.");
+	while (1) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		schedule_timeout(2*HZ);
+		DMWARN("Attempting to contact userspace log server...");
+		r = dm_consult_userspace(uuid, lc->luid, DM_ULOG_CTR,
+					 lc->usr_argv_str,
+					 strlen(lc->usr_argv_str) + 1,
+					 NULL, NULL);
+		if (!r)
+			break;
+	}
+	DMINFO("Reconnected to userspace log server... DM_ULOG_CTR complete");
+	r = dm_consult_userspace(uuid, lc->luid, DM_ULOG_RESUME, NULL,
+				 0, NULL, NULL);
+	if (!r)
+		goto retry;
+
+	DMERR("Error trying to resume userspace log: %d", r);
+
+	return -ESRCH;
+}
+
+static int build_constructor_string(struct dm_target *ti,
+				    unsigned int argc, char **argv,
+				    char **ctr_str)
+{
+	int i, str_size;
+	char *str = NULL;
+
+	*ctr_str = NULL;
+
+	/*
+	 * Determine overall size of the string.
+	 */
+	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 */
+	str_size++; /* For NUL-terminator */
+
+	str = kzalloc(str_size, GFP_KERNEL);
+	if (!str) {
+		DMWARN("Unable to allocate memory for constructor string");
+		return -ENOMEM;
+	}
+
+	str_size = sprintf(str, "%llu", (unsigned long long)ti->len);
+	for (i = 0; i < argc; i++)
+		str_size += sprintf(str + str_size, " %s", argv[i]);
+
+	*ctr_str = str;
+	return str_size;
+}
+
+static void do_flush(struct work_struct *work)
+{
+	int r;
+	struct log_c *lc = container_of(work, struct log_c, flush_log_work.work);
+
+	atomic_set(&lc->sched_flush, 0);
+
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_FLUSH, NULL, 0, NULL, NULL);
+
+	if (r)
+		dm_table_event(lc->ti->table);
+}
+
+/*
+ * userspace_ctr
+ *
+ * argv contains:
+ *	<UUID> [integrated_flush] <other args>
+ * Where 'other args' are the userspace implementation-specific log
+ * arguments.
+ *
+ * Example:
+ *	<UUID> [integrated_flush] clustered-disk <arg count> <log dev>
+ *	<region_size> [[no]sync]
+ *
+ * This module strips off the <UUID> and uses it for identification
+ * purposes when communicating with userspace about a log.
+ *
+ * If integrated_flush is defined, the kernel combines flush
+ * and mark requests.
+ *
+ * The rest of the line, beginning with 'clustered-disk', is passed
+ * to the userspace ctr function.
+ */
+static int userspace_ctr(struct dm_dirty_log *log, struct dm_target *ti,
+			 unsigned int argc, char **argv)
+{
+	int r = 0;
+	int str_size;
+	char *ctr_str = NULL;
+	struct log_c *lc = NULL;
+	uint64_t rdata;
+	size_t rdata_size = sizeof(rdata);
+	char *devices_rdata = NULL;
+	size_t devices_rdata_size = DM_NAME_LEN;
+
+	if (argc < 3) {
+		DMWARN("Too few arguments to userspace dirty log");
+		return -EINVAL;
+	}
+
+	lc = kzalloc_obj(*lc);
+	if (!lc) {
+		DMWARN("Unable to allocate userspace log context.");
+		return -ENOMEM;
+	}
+
+	/* The ptr value is sufficient for local unique id */
+	lc->luid = (unsigned long)lc;
+
+	lc->ti = ti;
+
+	if (strlen(argv[0]) > (DM_UUID_LEN - 1)) {
+		DMWARN("UUID argument too long.");
+		kfree(lc);
+		return -EINVAL;
+	}
+
+	lc->usr_argc = argc;
+
+	strscpy(lc->uuid, argv[0], sizeof(lc->uuid));
+	argc--;
+	argv++;
+	spin_lock_init(&lc->flush_lock);
+	INIT_LIST_HEAD(&lc->mark_list);
+	INIT_LIST_HEAD(&lc->clear_list);
+
+	if (!strcasecmp(argv[0], "integrated_flush")) {
+		lc->integrated_flush = 1;
+		argc--;
+		argv++;
+	}
+
+	str_size = build_constructor_string(ti, argc, argv, &ctr_str);
+	if (str_size < 0) {
+		kfree(lc);
+		return str_size;
+	}
+
+	devices_rdata = kzalloc(devices_rdata_size, GFP_KERNEL);
+	if (!devices_rdata) {
+		DMERR("Failed to allocate memory for device information");
+		r = -ENOMEM;
+		goto out;
+	}
+
+	r = mempool_init_slab_pool(&lc->flush_entry_pool, FLUSH_ENTRY_POOL_SIZE,
+				   _flush_entry_cache);
+	if (r) {
+		DMERR("Failed to create flush_entry_pool");
+		goto out;
+	}
+
+	/*
+	 * Send table string and get back any opened device.
+	 */
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_CTR,
+				 ctr_str, str_size,
+				 devices_rdata, &devices_rdata_size);
+
+	if (r < 0) {
+		if (r == -ESRCH)
+			DMERR("Userspace log server not found");
+		else
+			DMERR("Userspace log server failed to create log");
+		goto out;
+	}
+
+	/* Since the region size does not change, get it now */
+	rdata_size = sizeof(rdata);
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_GET_REGION_SIZE,
+				 NULL, 0, (char *)&rdata, &rdata_size);
+
+	if (r) {
+		DMERR("Failed to get region size of dirty log");
+		goto out;
+	}
+
+	lc->region_size = (uint32_t)rdata;
+	lc->region_count = dm_sector_div_up(ti->len, lc->region_size);
+
+	if (devices_rdata_size) {
+		if (devices_rdata[devices_rdata_size - 1] != '\0') {
+			DMERR("DM_ULOG_CTR device return string not properly terminated");
+			r = -EINVAL;
+			goto out;
+		}
+		r = dm_get_device(ti, devices_rdata,
+				  dm_table_get_mode(ti->table), &lc->log_dev);
+		if (r)
+			DMERR("Failed to register %s with device-mapper",
+			      devices_rdata);
+	}
+
+	if (lc->integrated_flush) {
+		lc->dmlog_wq = alloc_workqueue("dmlogd",
+					       WQ_MEM_RECLAIM | WQ_PERCPU, 0);
+		if (!lc->dmlog_wq) {
+			DMERR("couldn't start dmlogd");
+			r = -ENOMEM;
+			goto out;
+		}
+
+		INIT_DELAYED_WORK(&lc->flush_log_work, do_flush);
+		atomic_set(&lc->sched_flush, 0);
+	}
+
+out:
+	kfree(devices_rdata);
+	if (r) {
+		mempool_exit(&lc->flush_entry_pool);
+		kfree(lc);
+		kfree(ctr_str);
+	} else {
+		lc->usr_argv_str = ctr_str;
+		log->context = lc;
+	}
+
+	return r;
+}
+
+static void userspace_dtr(struct dm_dirty_log *log)
+{
+	struct log_c *lc = log->context;
+
+	if (lc->integrated_flush) {
+		/* flush workqueue */
+		if (atomic_read(&lc->sched_flush))
+			flush_delayed_work(&lc->flush_log_work);
+
+		destroy_workqueue(lc->dmlog_wq);
+	}
+
+	(void) dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_DTR,
+				    NULL, 0, NULL, NULL);
+
+	if (lc->log_dev)
+		dm_put_device(lc->ti, lc->log_dev);
+
+	mempool_exit(&lc->flush_entry_pool);
+
+	kfree(lc->usr_argv_str);
+	kfree(lc);
+}
+
+static int userspace_presuspend(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_PRESUSPEND,
+				 NULL, 0, NULL, NULL);
+
+	return r;
+}
+
+static int userspace_postsuspend(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	/*
+	 * Run planned flush earlier.
+	 */
+	if (lc->integrated_flush && atomic_read(&lc->sched_flush))
+		flush_delayed_work(&lc->flush_log_work);
+
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_POSTSUSPEND,
+				 NULL, 0, NULL, NULL);
+
+	return r;
+}
+
+static int userspace_resume(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	lc->in_sync_hint = 0;
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_RESUME,
+				 NULL, 0, NULL, NULL);
+
+	return r;
+}
+
+static uint32_t userspace_get_region_size(struct dm_dirty_log *log)
+{
+	struct log_c *lc = log->context;
+
+	return lc->region_size;
+}
+
+/*
+ * userspace_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 userspace_is_clean(struct dm_dirty_log *log, region_t region)
+{
+	int r;
+	uint64_t region64 = (uint64_t)region;
+	int64_t is_clean;
+	size_t rdata_size;
+	struct log_c *lc = log->context;
+
+	rdata_size = sizeof(is_clean);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_IS_CLEAN,
+				 (char *)&region64, sizeof(region64),
+				 (char *)&is_clean, &rdata_size);
+
+	return (r) ? 0 : (int)is_clean;
+}
+
+/*
+ * userspace_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 userspace_in_sync(struct dm_dirty_log *log, region_t region,
+			     int can_block)
+{
+	int r;
+	uint64_t region64 = region;
+	int64_t 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 = userspace_do_request(lc, lc->uuid, DM_ULOG_IN_SYNC,
+				 (char *)&region64, sizeof(region64),
+				 (char *)&in_sync, &rdata_size);
+	return (r) ? 0 : (int)in_sync;
+}
+
+static int flush_one_by_one(struct log_c *lc, struct list_head *flush_list)
+{
+	int r = 0;
+	struct dm_dirty_log_flush_entry *fe;
+
+	list_for_each_entry(fe, flush_list, list) {
+		r = userspace_do_request(lc, lc->uuid, fe->type,
+					 (char *)&fe->region,
+					 sizeof(fe->region),
+					 NULL, NULL);
+		if (r)
+			break;
+	}
+
+	return r;
+}
+
+static int flush_by_group(struct log_c *lc, struct list_head *flush_list,
+			  int flush_with_payload)
+{
+	int r = 0;
+	int count;
+	uint32_t type = 0;
+	struct dm_dirty_log_flush_entry *fe, *tmp_fe;
+	LIST_HEAD(tmp_list);
+	uint64_t group[MAX_FLUSH_GROUP_COUNT];
+
+	/*
+	 * Group process the requests
+	 */
+	while (!list_empty(flush_list)) {
+		count = 0;
+
+		list_for_each_entry_safe(fe, tmp_fe, flush_list, list) {
+			group[count] = fe->region;
+			count++;
+
+			list_move(&fe->list, &tmp_list);
+
+			type = fe->type;
+			if (count >= MAX_FLUSH_GROUP_COUNT)
+				break;
+		}
+
+		if (flush_with_payload) {
+			r = userspace_do_request(lc, lc->uuid, DM_ULOG_FLUSH,
+						 (char *)(group),
+						 count * sizeof(uint64_t),
+						 NULL, NULL);
+			/*
+			 * Integrated flush failed.
+			 */
+			if (r)
+				break;
+		} else {
+			r = userspace_do_request(lc, lc->uuid, type,
+						 (char *)(group),
+						 count * sizeof(uint64_t),
+						 NULL, NULL);
+			if (r) {
+				/*
+				 * Group send failed.  Attempt one-by-one.
+				 */
+				list_splice_init(&tmp_list, flush_list);
+				r = flush_one_by_one(lc, flush_list);
+				break;
+			}
+		}
+	}
+
+	/*
+	 * Must collect flush_entrys that were successfully processed
+	 * as a group so that they will be free'd by the caller.
+	 */
+	list_splice_init(&tmp_list, flush_list);
+
+	return r;
+}
+
+/*
+ * userspace_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, 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 userspace_flush(struct dm_dirty_log *log)
+{
+	int r = 0;
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	LIST_HEAD(mark_list);
+	LIST_HEAD(clear_list);
+	int mark_list_is_empty;
+	int clear_list_is_empty;
+	struct dm_dirty_log_flush_entry *fe, *tmp_fe;
+	mempool_t *flush_entry_pool = &lc->flush_entry_pool;
+
+	spin_lock_irqsave(&lc->flush_lock, flags);
+	list_splice_init(&lc->mark_list, &mark_list);
+	list_splice_init(&lc->clear_list, &clear_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+
+	mark_list_is_empty = list_empty(&mark_list);
+	clear_list_is_empty = list_empty(&clear_list);
+
+	if (mark_list_is_empty && clear_list_is_empty)
+		return 0;
+
+	r = flush_by_group(lc, &clear_list, 0);
+	if (r)
+		goto out;
+
+	if (!lc->integrated_flush) {
+		r = flush_by_group(lc, &mark_list, 0);
+		if (r)
+			goto out;
+		r = userspace_do_request(lc, lc->uuid, DM_ULOG_FLUSH,
+					 NULL, 0, NULL, NULL);
+		goto out;
+	}
+
+	/*
+	 * Send integrated flush request with mark_list as payload.
+	 */
+	r = flush_by_group(lc, &mark_list, 1);
+	if (r)
+		goto out;
+
+	if (mark_list_is_empty && !atomic_read(&lc->sched_flush)) {
+		/*
+		 * When there are only clear region requests,
+		 * we schedule a flush in the future.
+		 */
+		queue_delayed_work(lc->dmlog_wq, &lc->flush_log_work, 3 * HZ);
+		atomic_set(&lc->sched_flush, 1);
+	} else {
+		/*
+		 * Cancel pending flush because we
+		 * have already flushed in mark_region.
+		 */
+		cancel_delayed_work(&lc->flush_log_work);
+		atomic_set(&lc->sched_flush, 0);
+	}
+
+out:
+	/*
+	 * We can safely remove these entries, even after failure.
+	 * Calling code will receive an error and will know that
+	 * the log facility has failed.
+	 */
+	list_for_each_entry_safe(fe, tmp_fe, &mark_list, list) {
+		list_del(&fe->list);
+		mempool_free(fe, flush_entry_pool);
+	}
+	list_for_each_entry_safe(fe, tmp_fe, &clear_list, list) {
+		list_del(&fe->list);
+		mempool_free(fe, flush_entry_pool);
+	}
+
+	if (r)
+		dm_table_event(lc->ti->table);
+
+	return r;
+}
+
+/*
+ * userspace_mark_region
+ *
+ * This function should avoid blocking unless absolutely required.
+ * (Memory allocation is valid for blocking.)
+ */
+static void userspace_mark_region(struct dm_dirty_log *log, region_t region)
+{
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	struct dm_dirty_log_flush_entry *fe;
+
+	/* Wait for an allocation, but _never_ fail */
+	fe = mempool_alloc(&lc->flush_entry_pool, GFP_NOIO);
+	BUG_ON(!fe);
+
+	spin_lock_irqsave(&lc->flush_lock, flags);
+	fe->type = DM_ULOG_MARK_REGION;
+	fe->region = region;
+	list_add(&fe->list, &lc->mark_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+}
+
+/*
+ * userspace_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 userspace_clear_region(struct dm_dirty_log *log, region_t region)
+{
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	struct dm_dirty_log_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(&lc->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_ULOG_CLEAR_REGION;
+	fe->region = region;
+	list_add(&fe->list, &lc->clear_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+}
+
+/*
+ * userspace_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 userspace_get_resync_work(struct dm_dirty_log *log, region_t *region)
+{
+	int r;
+	size_t rdata_size;
+	struct log_c *lc = log->context;
+	struct {
+		int64_t i; /* 64-bit for mix arch compatibility */
+		region_t r;
+	} pkg;
+
+	if (lc->in_sync_hint >= lc->region_count)
+		return 0;
+
+	rdata_size = sizeof(pkg);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_GET_RESYNC_WORK,
+				 NULL, 0, (char *)&pkg, &rdata_size);
+
+	*region = pkg.r;
+	return (r) ? r : (int)pkg.i;
+}
+
+/*
+ * userspace_set_region_sync
+ *
+ * Set the sync status of a given region.  This function
+ * must not fail.
+ */
+static void userspace_set_region_sync(struct dm_dirty_log *log,
+				      region_t region, int in_sync)
+{
+	struct log_c *lc = log->context;
+	struct {
+		region_t r;
+		int64_t i;
+	} pkg;
+
+	pkg.r = region;
+	pkg.i = (int64_t)in_sync;
+
+	(void) userspace_do_request(lc, lc->uuid, DM_ULOG_SET_REGION_SYNC,
+				    (char *)&pkg, sizeof(pkg), NULL, NULL);
+
+	/*
+	 * It would be nice to be able to report failures.
+	 * However, it is easy enough to detect and resolve.
+	 */
+}
+
+/*
+ * userspace_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 userspace_get_sync_count(struct dm_dirty_log *log)
+{
+	int r;
+	size_t rdata_size;
+	uint64_t sync_count;
+	struct log_c *lc = log->context;
+
+	rdata_size = sizeof(sync_count);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_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 (region_t)sync_count;
+}
+
+/*
+ * userspace_status
+ *
+ * Returns: amount of space consumed
+ */
+static int userspace_status(struct dm_dirty_log *log, status_type_t status_type,
+			    char *result, unsigned int maxlen)
+{
+	int r = 0;
+	char *table_args;
+	size_t sz = (size_t)maxlen;
+	struct log_c *lc = log->context;
+
+	switch (status_type) {
+	case STATUSTYPE_INFO:
+		r = userspace_do_request(lc, lc->uuid, DM_ULOG_STATUS_INFO,
+					 NULL, 0, result, &sz);
+
+		if (r) {
+			sz = 0;
+			DMEMIT("%s 1 COM_FAILURE", log->type->name);
+		}
+		break;
+	case STATUSTYPE_TABLE:
+		sz = 0;
+		table_args = strchr(lc->usr_argv_str, ' ');
+		BUG_ON(!table_args); /* There will always be a ' ' */
+		table_args++;
+
+		DMEMIT("%s %u %s ", log->type->name, lc->usr_argc, lc->uuid);
+		if (lc->integrated_flush)
+			DMEMIT("integrated_flush ");
+		DMEMIT("%s ", table_args);
+		break;
+	case STATUSTYPE_IMA:
+		*result = '\0';
+		break;
+	}
+	return (r) ? 0 : (int)sz;
+}
+
+/*
+ * userspace_is_remote_recovering
+ *
+ * Returns: 1 if region recovering, 0 otherwise
+ */
+static int userspace_is_remote_recovering(struct dm_dirty_log *log,
+					  region_t region)
+{
+	int r;
+	uint64_t region64 = region;
+	struct log_c *lc = log->context;
+	static unsigned long limit;
+	struct {
+		int64_t 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 (time_after(limit, jiffies))
+		return 1;
+
+	limit = jiffies + (HZ / 4);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_IS_REMOTE_RECOVERING,
+				 (char *)&region64, sizeof(region64),
+				 (char *)&pkg, &rdata_size);
+	if (r)
+		return 1;
+
+	lc->in_sync_hint = pkg.in_sync_hint;
+
+	return (int)pkg.is_recovering;
+}
+
+static struct dm_dirty_log_type _userspace_type = {
+	.name = "userspace",
+	.module = THIS_MODULE,
+	.ctr = userspace_ctr,
+	.dtr = userspace_dtr,
+	.presuspend = userspace_presuspend,
+	.postsuspend = userspace_postsuspend,
+	.resume = userspace_resume,
+	.get_region_size = userspace_get_region_size,
+	.is_clean = userspace_is_clean,
+	.in_sync = userspace_in_sync,
+	.flush = userspace_flush,
+	.mark_region = userspace_mark_region,
+	.clear_region = userspace_clear_region,
+	.get_resync_work = userspace_get_resync_work,
+	.set_region_sync = userspace_set_region_sync,
+	.get_sync_count = userspace_get_sync_count,
+	.status = userspace_status,
+	.is_remote_recovering = userspace_is_remote_recovering,
+};
+
+static int __init userspace_dirty_log_init(void)
+{
+	int r = 0;
+
+	_flush_entry_cache = KMEM_CACHE(dm_dirty_log_flush_entry, 0);
+	if (!_flush_entry_cache) {
+		DMWARN("Unable to create flush_entry_cache: No memory.");
+		return -ENOMEM;
+	}
+
+	r = dm_ulog_tfr_init();
+	if (r) {
+		DMWARN("Unable to initialize userspace log communications");
+		kmem_cache_destroy(_flush_entry_cache);
+		return r;
+	}
+
+	r = dm_dirty_log_type_register(&_userspace_type);
+	if (r) {
+		DMWARN("Couldn't register userspace dirty log type");
+		dm_ulog_tfr_exit();
+		kmem_cache_destroy(_flush_entry_cache);
+		return r;
+	}
+
+	DMINFO("version " DM_LOG_USERSPACE_VSN " loaded");
+	return 0;
+}
+
+static void __exit userspace_dirty_log_exit(void)
+{
+	dm_dirty_log_type_unregister(&_userspace_type);
+	dm_ulog_tfr_exit();
+	kmem_cache_destroy(_flush_entry_cache);
+
+	DMINFO("version " DM_LOG_USERSPACE_VSN " unloaded");
+}
+
+module_init(userspace_dirty_log_init);
+module_exit(userspace_dirty_log_exit);
+
+MODULE_DESCRIPTION(DM_NAME " userspace dirty log link");
+MODULE_AUTHOR("Jonathan Brassow <dm-devel@lists.linux.dev>");
+MODULE_LICENSE("GPL");