[PATCH v6] Improve the performance of --num-threads -d 31

Zhou Wenjian zhouwj-fnst at cn.fujitsu.com
Wed Apr 13 22:42:35 PDT 2016


v6:
        1. address Atsushi Kumagai's comments about calculating and reserving memory
v5:
        1. remove unused variable check_out
        2. reset num_threads if memory is not enough
v4:
        1. fix a bug caused by the logic
v3:
        1. remove some unused variables
        2. fix a bug caused by the wrong logic
        3. fix a bug caused by optimising
        4. improve more performance by using Minoru Usui's code

multi-threads implementation will introduce extra cost when handling
each page. The origin implementation will also do the extra work for
filtered pages. So there is a big performance degradation in
--num-threads -d 31.
The new implementation won't do the extra work for filtered pages any
more. So the performance of -d 31 is close to that of serial processing.

The new implementation is just like the following:
        * The basic idea is producer producing page and consumer writing page.
        * Each producer have a page_flag_buf list which is used for storing
          page's description.
        * The size of page_flag_buf is little so it won't take too much memory.
        * And all producers will share a page_data_buf array which is
          used for storing page's compressed data.
        * The main thread is the consumer. It will find the next pfn and write
          it into file.
        * The next pfn is smallest pfn in all page_flag_buf.

Signed-off-by: Minoru Usui <min-usui at ti.jp.nec.com>
Signed-off-by: Zhou Wenjian <zhouwj-fnst at cn.fujitsu.com>
---
 makedumpfile.c | 320 +++++++++++++++++++++++++++++++++------------------------
 makedumpfile.h |  38 ++++---
 2 files changed, 210 insertions(+), 148 deletions(-)

diff --git a/makedumpfile.c b/makedumpfile.c
index fa0b779..a826321 100644
--- a/makedumpfile.c
+++ b/makedumpfile.c
@@ -3481,9 +3481,8 @@ initial_for_parallel()
 {
 	unsigned long len_buf_out;
 	unsigned long page_data_buf_size;
-	unsigned long limit_size;
-	int page_data_num;
-	int i;
+	struct page_flag *current;
+	int i, j;
 
 	len_buf_out = calculate_len_buf_out(info->page_size);
 
@@ -3553,20 +3552,7 @@ initial_for_parallel()
 #endif
 	}
 
-	/*
-	 * get a safe number of page_data
-	 */
-	page_data_buf_size = MAX(len_buf_out, info->page_size);
-
-	limit_size = (get_free_memory_size()
-		      - MAP_REGION * info->num_threads) * 0.6;
-
-	page_data_num = limit_size / page_data_buf_size;
-
-	info->num_buffers = MIN(NUM_BUFFERS, page_data_num);
-
-	DEBUG_MSG("Number of struct page_data for produce/consume: %d\n",
-			info->num_buffers);
+	info->num_buffers = PAGE_DATA_NUM * info->num_threads;
 
 	/*
 	 * allocate memory for page_data
@@ -3579,6 +3565,7 @@ initial_for_parallel()
 	}
 	memset(info->page_data_buf, 0, sizeof(struct page_data) * info->num_buffers);
 
+	page_data_buf_size = MAX(len_buf_out, info->page_size);
 	for (i = 0; i < info->num_buffers; i++) {
 		if ((info->page_data_buf[i].buf = malloc(page_data_buf_size)) == NULL) {
 			MSG("Can't allocate memory for buf of page_data_buf. %s\n",
@@ -3588,6 +3575,36 @@ initial_for_parallel()
 	}
 
 	/*
+	 * initial page_flag for each thread
+	 */
+	if ((info->page_flag_buf = malloc(sizeof(void *) * info->num_threads))
+	    == NULL) {
+		MSG("Can't allocate memory for page_flag_buf. %s\n",
+				strerror(errno));
+		return FALSE;
+	}
+	memset(info->page_flag_buf, 0, sizeof(void *) * info->num_threads);
+
+	for (i = 0; i < info->num_threads; i++) {
+		if ((info->page_flag_buf[i] = calloc(1, sizeof(struct page_flag))) == NULL) {
+			MSG("Can't allocate memory for page_flag. %s\n",
+				strerror(errno));
+			return FALSE;
+		}
+		current = info->page_flag_buf[i];
+
+		for (j = 1; j < PAGE_FLAG_NUM; j++) {
+			if ((current->next = calloc(1, sizeof(struct page_flag))) == NULL) {
+				MSG("Can't allocate memory for page_flag. %s\n",
+					strerror(errno));
+				return FALSE;
+			}
+			current = current->next;
+		}
+		current->next = info->page_flag_buf[i];
+	}
+
+	/*
 	 * initial fd_memory for threads
 	 */
 	for (i = 0; i < info->num_threads; i++) {
@@ -3612,7 +3629,8 @@ initial_for_parallel()
 void
 free_for_parallel()
 {
-	int i;
+	int i, j;
+	struct page_flag *current;
 
 	if (info->threads != NULL) {
 		for (i = 0; i < info->num_threads; i++) {
@@ -3655,6 +3673,19 @@ free_for_parallel()
 		free(info->page_data_buf);
 	}
 
+	if (info->page_flag_buf != NULL) {
+		for (i = 0; i < info->num_threads; i++) {
+			for (j = 0; j < PAGE_FLAG_NUM; j++) {
+				if (info->page_flag_buf[i] != NULL) {
+					current = info->page_flag_buf[i];
+					info->page_flag_buf[i] = current->next;
+					free(current);
+				}
+			}
+		}
+		free(info->page_flag_buf);
+	}
+
 	if (info->parallel_info == NULL)
 		return;
 
@@ -7075,11 +7106,11 @@ void *
 kdump_thread_function_cyclic(void *arg) {
 	void *retval = PTHREAD_FAIL;
 	struct thread_args *kdump_thread_args = (struct thread_args *)arg;
-	struct page_data *page_data_buf = kdump_thread_args->page_data_buf;
+	volatile struct page_data *page_data_buf = kdump_thread_args->page_data_buf;
+	volatile struct page_flag *page_flag_buf = kdump_thread_args->page_flag_buf;
 	struct cycle *cycle = kdump_thread_args->cycle;
-	int page_data_num = kdump_thread_args->page_data_num;
-	mdf_pfn_t pfn;
-	int index;
+	mdf_pfn_t pfn = cycle->start_pfn;
+	int index = kdump_thread_args->thread_num;
 	int buf_ready;
 	int dumpable;
 	int fd_memory = 0;
@@ -7125,47 +7156,48 @@ kdump_thread_function_cyclic(void *arg) {
 						kdump_thread_args->thread_num);
 	}
 
-	while (1) {
-		/* get next pfn */
-		pthread_mutex_lock(&info->current_pfn_mutex);
-		pfn = info->current_pfn;
-		info->current_pfn++;
-		pthread_mutex_unlock(&info->current_pfn_mutex);
-
-		if (pfn >= kdump_thread_args->end_pfn)
-			break;
-
-		index = -1;
+	/*
+	 * filtered page won't take anything
+	 * unfiltered zero page will only take a page_flag_buf
+	 * unfiltered non-zero page will take a page_flag_buf and a page_data_buf
+	 */
+	while (pfn < cycle->end_pfn) {
 		buf_ready = FALSE;
 
+		pthread_mutex_lock(&info->page_data_mutex);
+		while (page_data_buf[index].used != FALSE) {
+			index = (index + 1) % info->num_buffers;
+		}
+		page_data_buf[index].used = TRUE;
+		pthread_mutex_unlock(&info->page_data_mutex);
+
 		while (buf_ready == FALSE) {
 			pthread_testcancel();
-
-			index = pfn % page_data_num;
-
-			if (pfn - info->consumed_pfn > info->num_buffers)
+			if (page_flag_buf->ready == FLAG_READY)
 				continue;
 
-			if (page_data_buf[index].ready != 0)
-				continue;
-
-			pthread_mutex_lock(&page_data_buf[index].mutex);
-
-			if (page_data_buf[index].ready != 0)
-				goto unlock;
-
-			buf_ready = TRUE;
+			/* get next dumpable pfn */
+			pthread_mutex_lock(&info->current_pfn_mutex);
+			for (pfn = info->current_pfn; pfn < cycle->end_pfn; pfn++) {
+				dumpable = is_dumpable(
+					info->fd_bitmap ? &bitmap_parallel : info->bitmap2,
+					pfn,
+					cycle);
+				if (dumpable)
+					break;
+			}
+			info->current_pfn = pfn + 1;
 
-			page_data_buf[index].pfn = pfn;
-			page_data_buf[index].ready = 1;
+			page_flag_buf->pfn = pfn;
+			page_flag_buf->ready = FLAG_FILLING;
+			pthread_mutex_unlock(&info->current_pfn_mutex);
+			sem_post(&info->page_flag_buf_sem);
 
-			dumpable = is_dumpable(
-				info->fd_bitmap ? &bitmap_parallel : info->bitmap2,
-				pfn,
-				cycle);
-			page_data_buf[index].dumpable = dumpable;
-			if (!dumpable)
-				goto unlock;
+			if (pfn >= cycle->end_pfn) {
+				info->current_pfn = cycle->end_pfn;
+				page_data_buf[index].used = FALSE;
+				break;
+			}
 
 			if (!read_pfn_parallel(fd_memory, pfn, buf,
 					       &bitmap_memory_parallel,
@@ -7178,11 +7210,11 @@ kdump_thread_function_cyclic(void *arg) {
 
 			if ((info->dump_level & DL_EXCLUDE_ZERO)
 			    && is_zero_page(buf, info->page_size)) {
-				page_data_buf[index].zero = TRUE;
-				goto unlock;
+				page_flag_buf->zero = TRUE;
+				goto next;
 			}
 
-			page_data_buf[index].zero = FALSE;
+			page_flag_buf->zero = FALSE;
 
 			/*
 			 * Compress the page data.
@@ -7232,12 +7264,14 @@ kdump_thread_function_cyclic(void *arg) {
 				page_data_buf[index].size  = info->page_size;
 				memcpy(page_data_buf[index].buf, buf, info->page_size);
 			}
-unlock:
-			pthread_mutex_unlock(&page_data_buf[index].mutex);
+			page_flag_buf->index = index;
+			buf_ready = TRUE;
+next:
+			page_flag_buf->ready = FLAG_READY;
+			page_flag_buf = page_flag_buf->next;
 
 		}
 	}
-
 	retval = NULL;
 
 fail:
@@ -7265,14 +7299,15 @@ write_kdump_pages_parallel_cyclic(struct cache_data *cd_header,
 	struct page_desc pd;
 	struct timeval tv_start;
 	struct timeval last, new;
-	unsigned long long consuming_pfn;
 	pthread_t **threads = NULL;
 	struct thread_args *kdump_thread_args = NULL;
 	void *thread_result;
-	int page_data_num;
+	int page_buf_num;
 	struct page_data *page_data_buf = NULL;
 	int i;
 	int index;
+	int end_count, consuming;
+	mdf_pfn_t current_pfn, temp_pfn;
 
 	if (info->flag_elf_dumpfile)
 		return FALSE;
@@ -7284,13 +7319,6 @@ write_kdump_pages_parallel_cyclic(struct cache_data *cd_header,
 		goto out;
 	}
 
-	res = pthread_mutex_init(&info->consumed_pfn_mutex, NULL);
-	if (res != 0) {
-		ERRMSG("Can't initialize consumed_pfn_mutex. %s\n",
-				strerror(res));
-		goto out;
-	}
-
 	res = pthread_mutex_init(&info->filter_mutex, NULL);
 	if (res != 0) {
 		ERRMSG("Can't initialize filter_mutex. %s\n", strerror(res));
@@ -7314,36 +7342,23 @@ write_kdump_pages_parallel_cyclic(struct cache_data *cd_header,
 	end_pfn   = cycle->end_pfn;
 
 	info->current_pfn = start_pfn;
-	info->consumed_pfn = start_pfn - 1;
 
 	threads = info->threads;
 	kdump_thread_args = info->kdump_thread_args;
 
-	page_data_num = info->num_buffers;
+	page_buf_num = info->num_buffers;
 	page_data_buf = info->page_data_buf;
+	pthread_mutex_init(&info->page_data_mutex, NULL);
+	sem_init(&info->page_flag_buf_sem, 0, 0);
 
-	for (i = 0; i < page_data_num; i++) {
-		/*
-		 * producer will use pfn in page_data_buf to decide the
-		 * consumed pfn
-		 */
-		page_data_buf[i].pfn = start_pfn - 1;
-		page_data_buf[i].ready = 0;
-		res = pthread_mutex_init(&page_data_buf[i].mutex, NULL);
-		if (res != 0) {
-			ERRMSG("Can't initialize mutex of page_data_buf. %s\n",
-					strerror(res));
-			goto out;
-		}
-	}
+	for (i = 0; i < page_buf_num; i++)
+		page_data_buf[i].used = FALSE;
 
 	for (i = 0; i < info->num_threads; i++) {
 		kdump_thread_args[i].thread_num = i;
 		kdump_thread_args[i].len_buf_out = len_buf_out;
-		kdump_thread_args[i].start_pfn = start_pfn;
-		kdump_thread_args[i].end_pfn = end_pfn;
-		kdump_thread_args[i].page_data_num = page_data_num;
 		kdump_thread_args[i].page_data_buf = page_data_buf;
+		kdump_thread_args[i].page_flag_buf = info->page_flag_buf[i];
 		kdump_thread_args[i].cycle = cycle;
 
 		res = pthread_create(threads[i], NULL,
@@ -7356,55 +7371,86 @@ write_kdump_pages_parallel_cyclic(struct cache_data *cd_header,
 		}
 	}
 
-	consuming_pfn = start_pfn;
-	index = -1;
+	end_count = 0;
+	while (1) {
+		consuming = 0;
 
-	gettimeofday(&last, NULL);
+		/*
+		 * The basic idea is producer producing page and consumer writing page.
+		 * Each producer have a page_flag_buf list which is used for storing page's description.
+		 * The size of page_flag_buf is little so it won't take too much memory.
+		 * And all producers will share a page_data_buf array which is used for storing page's compressed data.
+		 * The main thread is the consumer. It will find the next pfn and write it into file.
+		 * The next pfn is smallest pfn in all page_flag_buf.
+		 */
+		sem_wait(&info->page_flag_buf_sem);
+		gettimeofday(&last, NULL);
+		while (1) {
+			current_pfn = end_pfn;
 
-	while (consuming_pfn < end_pfn) {
-		index = consuming_pfn % page_data_num;
+			/*
+			 * page_flag_buf is in circular linked list.
+			 * The array info->page_flag_buf[] records the current page_flag_buf in each thread's
+			 * page_flag_buf list.
+			 * consuming is used for recording in which thread the pfn is the smallest.
+			 * current_pfn is used for recording the value of pfn when checking the pfn.
+			 */
+			for (i = 0; i < info->num_threads; i++) {
+				if (info->page_flag_buf[i]->ready == FLAG_UNUSED)
+					continue;
+				temp_pfn = info->page_flag_buf[i]->pfn;
 
-		gettimeofday(&new, NULL);
-		if (new.tv_sec - last.tv_sec > WAIT_TIME) {
-			ERRMSG("Can't get data of pfn %llx.\n", consuming_pfn);
-			goto out;
-		}
+				/*
+				 * count how many threads have reached the end.
+				 */
+				if (temp_pfn >= end_pfn) {
+					info->page_flag_buf[i]->ready = FLAG_UNUSED;
+					end_count++;
+					continue;
+				}
 
-		/*
-		 * check pfn first without mutex locked to reduce the time
-		 * trying to lock the mutex
-		 */
-		if (page_data_buf[index].pfn != consuming_pfn)
-			continue;
+				if (current_pfn < temp_pfn)
+					continue;
 
-		if (pthread_mutex_trylock(&page_data_buf[index].mutex) != 0)
-			continue;
+				consuming = i;
+				current_pfn = temp_pfn;
+			}
 
-		/* check whether the found one is ready to be consumed */
-		if (page_data_buf[index].pfn != consuming_pfn ||
-		    page_data_buf[index].ready != 1) {
-			goto unlock;
+			/*
+			 * If all the threads have reached the end, we will finish writing.
+			 */
+			if (end_count >= info->num_threads)
+				goto finish;
+
+			/*
+			 * If the page_flag_buf is not ready, the pfn recorded may be changed.
+			 * So we should recheck.
+			 */
+			if (info->page_flag_buf[consuming]->ready != FLAG_READY) {
+				gettimeofday(&new, NULL);
+				if (new.tv_sec - last.tv_sec > WAIT_TIME) {
+					ERRMSG("Can't get data of pfn.\n");
+					goto out;
+				}
+				continue;
+			}
+
+			if (current_pfn == info->page_flag_buf[consuming]->pfn)
+				break;
 		}
 
 		if ((num_dumped % per) == 0)
 			print_progress(PROGRESS_COPY, num_dumped, info->num_dumpable);
 
-		/* next pfn is found, refresh last here */
-		last = new;
-		consuming_pfn++;
-		info->consumed_pfn++;
-		page_data_buf[index].ready = 0;
-
-		if (page_data_buf[index].dumpable == FALSE)
-			goto unlock;
-
 		num_dumped++;
 
-		if (page_data_buf[index].zero == TRUE) {
+
+		if (info->page_flag_buf[consuming]->zero == TRUE) {
 			if (!write_cache(cd_header, pd_zero, sizeof(page_desc_t)))
 				goto out;
 			pfn_zero++;
 		} else {
+			index = info->page_flag_buf[consuming]->index;
 			pd.flags      = page_data_buf[index].flags;
 			pd.size       = page_data_buf[index].size;
 			pd.page_flags = 0;
@@ -7420,12 +7466,12 @@ write_kdump_pages_parallel_cyclic(struct cache_data *cd_header,
 			 */
 			if (!write_cache(cd_page, page_data_buf[index].buf, pd.size))
 				goto out;
-
+			page_data_buf[index].used = FALSE;
 		}
-unlock:
-		pthread_mutex_unlock(&page_data_buf[index].mutex);
+		info->page_flag_buf[consuming]->ready = FLAG_UNUSED;
+		info->page_flag_buf[consuming] = info->page_flag_buf[consuming]->next;
 	}
-
+finish:
 	ret = TRUE;
 	/*
 	 * print [100 %]
@@ -7463,15 +7509,9 @@ out:
 		}
 	}
 
-	if (page_data_buf != NULL) {
-		for (i = 0; i < page_data_num; i++) {
-			pthread_mutex_destroy(&page_data_buf[i].mutex);
-		}
-	}
-
+	sem_destroy(&info->page_flag_buf_sem);
 	pthread_rwlock_destroy(&info->usemmap_rwlock);
 	pthread_mutex_destroy(&info->filter_mutex);
-	pthread_mutex_destroy(&info->consumed_pfn_mutex);
 	pthread_mutex_destroy(&info->current_pfn_mutex);
 
 	return ret;
@@ -7564,6 +7604,7 @@ write_kdump_pages_cyclic(struct cache_data *cd_header, struct cache_data *cd_pag
 		num_dumped++;
 		if (!read_pfn(pfn, buf))
 			goto out;
+
 		filter_data_buffer(buf, pfn_to_paddr(pfn), info->page_size);
 
 		/*
@@ -10201,7 +10242,6 @@ out:
 	return free_size;
 }
 
-
 /*
  * Choose the less value of the three below as the size of cyclic buffer.
  *  - the size enough for storing the 1st or 2nd bitmap for the whole of vmcore
@@ -10223,6 +10263,20 @@ calculate_cyclic_buffer_size(void) {
 	 *  free memory for safety.
 	 */
 	limit_size = get_free_memory_size() * 0.6;
+
+	/*
+	 * Recalculate the limit_size according to num_threads.
+	 * And reset num_threads if there is not enough memory.
+	 */
+	if (limit_size - maximum_size < THREAD_REGION * info->num_threads) {
+		MSG("There isn't enough memory for %d threads.\n", info->num_threads);
+
+		info->num_threads = MAX(0, limit_size - maximum_size) / THREAD_REGION;
+		MSG("--num_threads is set to %d.\n", info->num_threads);
+	}
+
+	limit_size = limit_size - THREAD_REGION * info->num_threads;
+
 	/* Try to keep both 1st and 2nd bitmap at the same time. */
 	bitmap_size = info->max_mapnr * 2 / BITPERBYTE;
 
diff --git a/makedumpfile.h b/makedumpfile.h
index e0b5bbf..322d31b 100644
--- a/makedumpfile.h
+++ b/makedumpfile.h
@@ -44,6 +44,7 @@
 #include "print_info.h"
 #include "sadump_mod.h"
 #include <pthread.h>
+#include <semaphore.h>
 
 /*
  * Result of command
@@ -974,10 +975,11 @@ typedef unsigned long long int ulonglong;
  * for parallel process
  */
 
-#define PAGE_DATA_NUM	(50)
+#define PAGE_FLAG_NUM	(20)
+#define PAGE_DATA_NUM	(5)
 #define WAIT_TIME	(60 * 10)
 #define PTHREAD_FAIL	((void *)-2)
-#define NUM_BUFFERS	(50)
+#define THREAD_REGION	(200 * 1024)
 
 struct mmap_cache {
 	char	*mmap_buf;
@@ -985,28 +987,33 @@ struct mmap_cache {
 	off_t   mmap_end_offset;
 };
 
+enum {
+	FLAG_UNUSED,
+	FLAG_READY,
+	FLAG_FILLING
+};
+struct page_flag {
+	mdf_pfn_t pfn;
+	char zero;
+	char ready;
+	short index;
+	struct page_flag *next;
+};
+
 struct page_data
 {
-	mdf_pfn_t pfn;
-	int dumpable;
-	int zero;
-	unsigned int flags;
 	long size;
 	unsigned char *buf;
-	pthread_mutex_t mutex;
-	/*
-	 * whether the page_data is ready to be consumed
-	 */
-	int ready;
+	int flags;
+	int used;
 };
 
 struct thread_args {
 	int thread_num;
 	unsigned long len_buf_out;
-	mdf_pfn_t start_pfn, end_pfn;
-	int page_data_num;
 	struct cycle *cycle;
 	struct page_data *page_data_buf;
+	struct page_flag *page_flag_buf;
 };
 
 /*
@@ -1295,11 +1302,12 @@ struct DumpInfo {
 	pthread_t **threads;
 	struct thread_args *kdump_thread_args;
 	struct page_data *page_data_buf;
+	struct page_flag **page_flag_buf;
+	sem_t page_flag_buf_sem;
 	pthread_rwlock_t usemmap_rwlock;
 	mdf_pfn_t current_pfn;
 	pthread_mutex_t current_pfn_mutex;
-	mdf_pfn_t consumed_pfn;
-	pthread_mutex_t consumed_pfn_mutex;
+	pthread_mutex_t page_data_mutex;
 	pthread_mutex_t filter_mutex;
 };
 extern struct DumpInfo		*info;
-- 
1.8.3.1






More information about the kexec mailing list