swap: change block allocation algorithm for SSD

PATCH Superseded Shaohua Li 6 participants 15 messages Feb 21, 2013 – Mar 21, 2013

Subsystems (2)

Memory management MM swap

lore

Share & tools

lore

Shaohua Li

author

2013-02-21 02:17 UTC #0 lore raw

PATCH v3 1/4 swap: change block allocation algorithm for SSD

I'm using a fast SSD to do swap. scan_swap_map() sometimes uses up to 20~30% CPU time (when cluster is hard to find, the CPU time can be up to 80%), which becomes a bottleneck. scan_swap_map() scans a byte array to search a 256 page cluster, which is very slow.

Here I introduced a simple algorithm to search cluster. Since we only care about 256 pages cluster, we can just use a counter to track if a cluster is free. Every 256 pages use one int to store the counter. If the counter of a cluster is 0, the cluster is free. All free clusters will be added to a list, so searching cluster is very efficient. With this, scap_swap_map() overhead disappears.

Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.

This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.

We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).

The patch slightly changes which cluster is choosen. It always adds free cluster to list tail. This can help wear leveling for low end SSD too. And if no cluster found, the scan_swap_map() will do search from the end of last cluster. So if no cluster found, the scan_swap_map() will do search from the end of last free cluster, which is random. For SSD, this isn't a problem at all.

Another downside is the cluster must be aligned to 256 pages, which will reduce the chance to find a cluster. I would expect this isn't a big problem for SSD because of the non-seek penality. (And this is the reason I only enable the algorithm for SSD).

V2 -> V3:
rebase to latest linux-next

V1 -> V2:
1. free cluster is added to a list, which makes searching cluster more efficient
2. only enable the algorithm for SSD.

Signed-off-by Shaohua Li <[email protected]>

---
 include/linux/swap.h |    3 
 mm/swapfile.c        |  181 +++++++++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 172 insertions(+), 12 deletions(-)

include/linux/swap.h +3 -0

Index: linux/include/linux/swap.h
===================================================================
@@ -185,6 +185,9 @@ struct swap_info_struct {
 	signed char	next;		/* next type on the swap list */
 	unsigned int	max;		/* extent of the swap_map */
 	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
+	unsigned int *cluster_info;	/* cluster info. Only for SSD */
+	unsigned int free_cluster_head;
+	unsigned int free_cluster_tail;
 	unsigned int lowest_bit;	/* index of first free in swap_map */
 	unsigned int highest_bit;	/* index of last free in swap_map */
 	unsigned int pages;		/* total of usable pages of swap */

mm/swapfile.c +169 ~~-12~~

325 lines not on this page · raw message

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Kyungmin Park

2013-02-21 08:13 UTC #1 lore raw

Hi,

It's not related topic with this patch, but now I'm integrating with zswap with this patch but zswap uses each own writeback codes so it can't use this cluster concept.

I'm still can't find proper approaches to integrate zswap (+writeback) with this concept.

Do you have any ideas or plan to work with zswap?

Thank you,
Kyungmin Park

Quoted 401 lines · Shaohua Li

Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.

This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.

We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).

V2 -> V3:
rebase to latest linux-next

V1 -> V2:
1. free cluster is added to a list, which makes searching cluster more efficient
2. only enable the algorithm for SSD.

Signed-off-by Shaohua Li <[email protected]>

---
 include/linux/swap.h |    3
 mm/swapfile.c        |  181 +++++++++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 172 insertions(+), 12 deletions(-)

Index: linux/include/linux/swap.h
===================================================================

include/linux/swap.h 2013-02-18 15:21:09.285317914 +0800 renamed from include/linux/swap.h 2013-02-18 15:06:06.000000000 +0800 +3 -0

@@ -185,6 +185,9 @@ struct swap_info_struct {
        signed char     next;           /* next type on the swap list */
        unsigned int    max;            /* extent of the swap_map */
        unsigned char *swap_map;        /* vmalloc'ed array of usage counts */
+       unsigned int *cluster_info;     /* cluster info. Only for SSD */
+       unsigned int free_cluster_head;
+       unsigned int free_cluster_tail;
        unsigned int lowest_bit;        /* index of first free in swap_map */
        unsigned int highest_bit;       /* index of last free in swap_map */
        unsigned int pages;             /* total of usable pages of swap */

Index: linux/mm/swapfile.c
===================================================================

mm/swapfile.c 2013-02-18 15:21:09.285317914 +0800 renamed from mm/swapfile.c 2013-02-18 15:06:06.000000000 +0800 +168 ~~-12~~

@@ -184,6 +184,85 @@ static int wait_for_discard(void *word)
 #define SWAPFILE_CLUSTER       256
 #define LATENCY_LIMIT          256
 
+/*
+ * cluster info is a unsigned int, the highest 8 bits stores flags, the low 24
+ * bits stores next cluster if the cluster is free or cluster counter otherwise
+ */
+#define CLUSTER_FLAG_FREE (1 << 0)
+#define CLUSTER_FLAG_NEXT_NULL (1 << 1)
+#define CLUSTER_NULL (CLUSTER_FLAG_NEXT_NULL << 24)
+#define cluster_flag(info) ((info) >> 24)
+#define cluster_set_flag(info, flag) \
+       do { info = ((info) & 0xffffff) | ((flag) << 24); } while (0)
+#define cluster_count(info) ((info) & 0xffffff)
+#define cluster_set_count(info, c) \
+       do { info = (cluster_flag(info) << 24) | (c); } while (0)
+#define cluster_next(info) ((info) & 0xffffff)
+#define cluster_set_next(info, n) \
+       do { info = (cluster_flag(info) << 24) | (n); } while (0)
+#define cluster_is_free(info) (cluster_flag(info) & CLUSTER_FLAG_FREE)
+
+static inline void inc_cluster_info_page(struct swap_info_struct *p,
+       unsigned int *cluster_info, unsigned long page_nr)
+{
+       unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+       if (!cluster_info)
+               return;
+       if (cluster_is_free(cluster_info[idx])) {
+               VM_BUG_ON(p->free_cluster_head != idx);
+               p->free_cluster_head = cluster_next(cluster_info[idx]);
+               if (p->free_cluster_tail == idx) {
+                       p->free_cluster_tail = CLUSTER_NULL;
+                       p->free_cluster_head = CLUSTER_NULL;
+               }
+               cluster_set_flag(cluster_info[idx], 0);
+               cluster_set_count(cluster_info[idx], 0);
+       }
+
+       VM_BUG_ON(cluster_count(cluster_info[idx]) >= SWAPFILE_CLUSTER);
+       cluster_set_count(cluster_info[idx],
+               cluster_count(cluster_info[idx]) + 1);
+}
+
+static inline void dec_cluster_info_page(struct swap_info_struct *p,
+       unsigned int *cluster_info, unsigned long page_nr)
+{
+       unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+       if (!cluster_info)
+               return;
+
+       VM_BUG_ON(cluster_count(cluster_info[idx]) == 0);
+       cluster_set_count(cluster_info[idx],
+               cluster_count(cluster_info[idx]) - 1);
+
+       if (cluster_count(cluster_info[idx]) == 0) {
+               cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+               if (p->free_cluster_head == CLUSTER_NULL) {
+                       p->free_cluster_head = idx;
+                       p->free_cluster_tail = idx;
+               } else {
+                       cluster_set_next(cluster_info[p->free_cluster_tail],
+                               idx);
+                       p->free_cluster_tail = idx;
+               }
+       }
+}
+
+/*
+ * It's possible scan_swap_map() uses a free cluster in the middle of free
+ * cluster list. Avoiding such abuse to avoid list corruption.
+ */
+static inline bool scan_swap_map_recheck_cluster(struct swap_info_struct *si,
+       unsigned long offset)
+{
+       offset /= SWAPFILE_CLUSTER;
+       return si->free_cluster_head != CLUSTER_NULL &&
+               offset != si->free_cluster_head &&
+               cluster_is_free(si->cluster_info[offset]);
+}
+
 static unsigned long scan_swap_map(struct swap_info_struct *si,
                                   unsigned char usage)
 {

@@ -225,6 +304,24 @@ static unsigned long scan_swap_map(struc
                        si->lowest_alloc = si->max;
                        si->highest_alloc = 0;
                }
+check_cluster:
+               if (si->free_cluster_head != CLUSTER_NULL) {
+                       offset = si->free_cluster_head * SWAPFILE_CLUSTER;
+                       last_in_cluster = offset + SWAPFILE_CLUSTER - 1;
+                       si->cluster_next = offset;
+                       si->cluster_nr = SWAPFILE_CLUSTER - 1;
+                       found_free_cluster = 1;
+                       goto checks;
+               } else if (si->cluster_info) {
+                       /*
+                        * Checking free cluster is fast enough, we can do the
+                        * check every time
+                        */
+                       si->cluster_nr = 0;
+                       si->lowest_alloc = 0;
+                       goto checks;
+               }
+
                spin_unlock(&si->lock);
 
                /*

@@ -285,6 +382,8 @@ static unsigned long scan_swap_map(struc
        }
 
 checks:
+       if (scan_swap_map_recheck_cluster(si, offset))
+               goto check_cluster;
        if (!(si->flags & SWP_WRITEOK))
                goto no_page;
        if (!si->highest_bit)

@@ -317,6 +416,7 @@ checks:
                si->highest_bit = 0;
        }
        si->swap_map[offset] = usage;
+       inc_cluster_info_page(si, si->cluster_info, offset);
        si->cluster_next = offset + 1;
        si->flags -= SWP_SCANNING;

@@ -600,6 +700,7 @@ static unsigned char swap_entry_free(str
 
        /* free if no reference */
        if (!usage) {
+               dec_cluster_info_page(p, p->cluster_info, offset);
                if (offset < p->lowest_bit)
                        p->lowest_bit = offset;
                if (offset > p->highest_bit)

@@ -1497,6 +1598,7 @@ static int setup_swap_extents(struct swa
 
 static void _enable_swap_info(struct swap_info_struct *p, int prio,
                                unsigned char *swap_map,
+                               unsigned int *cluster_info,
                                unsigned long *frontswap_map)
 {
        int i, prev;

@@ -1506,6 +1608,7 @@ static void _enable_swap_info(struct swa
        else
                p->prio = --least_priority;
        p->swap_map = swap_map;
+       p->cluster_info = cluster_info;
        frontswap_map_set(p, frontswap_map);
        p->flags |= SWP_WRITEOK;
        atomic_long_add(p->pages, &nr_swap_pages);

@@ -1527,11 +1630,12 @@ static void _enable_swap_info(struct swa
 
 static void enable_swap_info(struct swap_info_struct *p, int prio,
                                unsigned char *swap_map,
+                               unsigned int *cluster_info,
                                unsigned long *frontswap_map)
 {
        spin_lock(&swap_lock);
        spin_lock(&p->lock);
-       _enable_swap_info(p, prio, swap_map, frontswap_map);
+       _enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
        frontswap_init(p->type);
        spin_unlock(&p->lock);
        spin_unlock(&swap_lock);

@@ -1541,7 +1645,8 @@ static void reinsert_swap_info(struct sw
 {
        spin_lock(&swap_lock);
        spin_lock(&p->lock);
-       _enable_swap_info(p, p->prio, p->swap_map, frontswap_map_get(p));
+       _enable_swap_info(p, p->prio, p->swap_map, p->cluster_info,
+                                       frontswap_map_get(p));
        spin_unlock(&p->lock);
        spin_unlock(&swap_lock);
 }

@@ -1550,6 +1655,7 @@ SYSCALL_DEFINE1(swapoff, const char __us
 {
        struct swap_info_struct *p = NULL;
        unsigned char *swap_map;
+       unsigned int *cluster_info;
        struct file *swap_file, *victim;
        struct address_space *mapping;
        struct inode *inode;

@@ -1648,12 +1754,15 @@ SYSCALL_DEFINE1(swapoff, const char __us
        p->max = 0;
        swap_map = p->swap_map;
        p->swap_map = NULL;
+       cluster_info = p->cluster_info;
+       p->cluster_info = NULL;
        p->flags = 0;
        frontswap_invalidate_area(type);
        spin_unlock(&p->lock);
        spin_unlock(&swap_lock);
        mutex_unlock(&swapon_mutex);
        vfree(swap_map);
+       vfree(cluster_info);
        vfree(frontswap_map_get(p));
        /* Destroy swap account informatin */
        swap_cgroup_swapoff(type);

@@ -1966,15 +2075,21 @@ static unsigned long read_swap_header(st
 static int setup_swap_map_and_extents(struct swap_info_struct *p,
                                        union swap_header *swap_header,
                                        unsigned char *swap_map,
+                                       unsigned int *cluster_info,
                                        unsigned long maxpages,
                                        sector_t *span)
 {
        int i;
        unsigned int nr_good_pages;
        int nr_extents;
+       unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);
+       unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
 
        nr_good_pages = maxpages - 1;   /* omit header page */
 
+       p->free_cluster_head = CLUSTER_NULL;
+       p->free_cluster_tail = CLUSTER_NULL;
+
        for (i = 0; i < swap_header->info.nr_badpages; i++) {
                unsigned int page_nr = swap_header->info.badpages[i];
                if (page_nr == 0 || page_nr > swap_header->info.last_page)

@@ -1982,11 +2097,25 @@ static int setup_swap_map_and_extents(st
                if (page_nr < maxpages) {
                        swap_map[page_nr] = SWAP_MAP_BAD;
                        nr_good_pages--;
+                       /*
+                        * Not mark the cluster free yet, no list
+                        * operation involved
+                        */
+                       inc_cluster_info_page(p, cluster_info, page_nr);
                }
        }
 
+       /* Not mark the cluster free yet, no list operation involved */
+       for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
+               inc_cluster_info_page(p, cluster_info, i);
+
        if (nr_good_pages) {
                swap_map[0] = SWAP_MAP_BAD;
+               /*
+                * Not mark the cluster free yet, no list
+                * operation involved
+                */
+               inc_cluster_info_page(p, cluster_info, 0);
                p->max = maxpages;
                p->pages = nr_good_pages;
                nr_extents = setup_swap_extents(p, span);

@@ -1999,6 +2128,27 @@ static int setup_swap_map_and_extents(st
                return -EINVAL;
        }
 
+       if (!cluster_info)
+               return nr_extents;
+
+       for (i = 0; i < nr_clusters; i++) {
+               if (!cluster_count(cluster_info[idx])) {
+                       cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+                       if (p->free_cluster_head == CLUSTER_NULL) {
+                               p->free_cluster_head = idx;
+                               p->free_cluster_tail = idx;
+                       } else {
+                               cluster_set_next(
+                                       cluster_info[p->free_cluster_tail],
+                                       idx);
+                               p->free_cluster_tail = idx;
+                       }
+               }
+               idx++;
+               if (idx == nr_clusters)
+                       idx = 0;
+       }
+
        return nr_extents;
 }

@@ -2016,6 +2166,7 @@ SYSCALL_DEFINE2(swapon, const char __use
        sector_t span;
        unsigned long maxpages;
        unsigned char *swap_map = NULL;
+       unsigned int *cluster_info = NULL;
        unsigned long *frontswap_map = NULL;
        struct page *page = NULL;
        struct inode *inode = NULL;

@@ -2089,13 +2240,24 @@ SYSCALL_DEFINE2(swapon, const char __use
                error = -ENOMEM;
                goto bad_swap;
        }
+       if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
+               p->flags |= SWP_SOLIDSTATE;
+               p->cluster_next = 1 + (random32() % p->highest_bit);
+
+               cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
+                       SWAPFILE_CLUSTER) * sizeof(*cluster_info));
+               if (!cluster_info) {
+                       error = -ENOMEM;
+                       goto bad_swap;
+               }
+       }
 
        error = swap_cgroup_swapon(p->type, maxpages);
        if (error)
                goto bad_swap;
 
        nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
-               maxpages, &span);
+               cluster_info, maxpages, &span);
        if (unlikely(nr_extents < 0)) {
                error = nr_extents;
                goto bad_swap;

@@ -2104,21 +2266,15 @@ SYSCALL_DEFINE2(swapon, const char __use
        if (frontswap_enabled)
                frontswap_map = vzalloc(maxpages / sizeof(long));
 
-       if (p->bdev) {
-               if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
-                       p->flags |= SWP_SOLIDSTATE;
-                       p->cluster_next = 1 + (random32() % p->highest_bit);
-               }
-               if ((swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
-                       p->flags |= SWP_DISCARDABLE;
-       }
+       if (p->bdev && (swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
+               p->flags |= SWP_DISCARDABLE;
 
        mutex_lock(&swapon_mutex);
        prio = -1;
        if (swap_flags & SWAP_FLAG_PREFER)
                prio =
                  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
-       enable_swap_info(p, prio, swap_map, frontswap_map);
+       enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 
        printk(KERN_INFO "Adding %uk swap on %s.  "
                        "Priority:%d extents:%d across:%lluk %s%s%s\n",

mm/swapfile.c 2013-02-18 15:21:09.285317914 +0800 bad_swap: · line 2310

Quoted 7 lines

mm/swapfile.c 2013-02-18 15:21:09.285317914 +0800 renamed from mm/swapfile.c 2013-02-18 15:06:06.000000000 +0800 +1 -0

@@ -2148,6 +2304,7 @@ bad_swap:
        p->flags = 0;
        spin_unlock(&swap_lock);
        vfree(swap_map);
+       vfree(cluster_info);
        if (swap_file) {
                if (inode && S_ISREG(inode->i_mode)) {
                        mutex_unlock(&inode->i_mutex);

Shaohua Li

author

2013-02-21 09:35 UTC #2 lore raw

Quoted 10 lines · Kyungmin Park

Hi,
It's not related topic with this patch, but now I'm integrating with zswap with this patch but zswap uses each own writeback codes so it can't use this cluster concept.
I'm still can't find proper approaches to integrate zswap (+writeback) with this concept.
Do you have any ideas or plan to work with zswap?

I didn't look at zswap. At first glance, when zswap fallbacks to writeback, it will make swap very sparse (so cause bad IO pattern), since some pages are compressed, some not. Is this the problem you are trying to solve? This should exist without my patch too. Sorry, I have no idea. I'm afraid zswap need manage the swap partion by itself.

Thanks,
Shaohua

Rafael Aquini

2013-03-12 15:13 UTC #3 lore raw

In reply to Shaohua Li

Quoted 5 lines · Shaohua Li

I'm using a fast SSD to do swap. scan_swap_map() sometimes uses up to 20~30% CPU time (when cluster is hard to find, the CPU time can be up to 80%), which becomes a bottleneck. scan_swap_map() scans a byte array to search a 256 page cluster, which is very slow.

Quoted 7 lines

Here I introduced a simple algorithm to search cluster. Since we only care about 256 pages cluster, we can just use a counter to track if a cluster is free. Every 256 pages use one int to store the counter. If the counter of a cluster is 0, the cluster is free. All free clusters will be added to a list, so searching cluster is very efficient. With this, scap_swap_map() overhead disappears.

Nice work! I've been testing your series with the attached test-prog, as it was triggering a swapin_readahead() softlockup storm due to a race window scan_swap_map() was opening when discarding clusters and swap device was saturated by the swap thrashing going on. Your patch 3 closes that window, and your series is still playing very well as I'm writing this message.

/me just thinks you could use some numbers of yours to actually show the benefits instead of just claim them on the commit message.

Other than that, I have little to say about your work, and as long as Hugh is happy with it, you can surely stick my Acked-by on your series.

Regards!

Rafael

Quoted 398 lines

Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.

This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.

We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).

V2 -> V3:
rebase to latest linux-next

V1 -> V2:
1. free cluster is added to a list, which makes searching cluster more efficient
2. only enable the algorithm for SSD.

Signed-off-by Shaohua Li <[email protected]>

---
 include/linux/swap.h |    3 
 mm/swapfile.c        |  181 +++++++++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 172 insertions(+), 12 deletions(-)

Index: linux/include/linux/swap.h
===================================================================

include/linux/swap.h +3 -0

@@ -185,6 +185,9 @@ struct swap_info_struct {
 	signed char	next;		/* next type on the swap list */
 	unsigned int	max;		/* extent of the swap_map */
 	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
+	unsigned int *cluster_info;	/* cluster info. Only for SSD */
+	unsigned int free_cluster_head;
+	unsigned int free_cluster_tail;
 	unsigned int lowest_bit;	/* index of first free in swap_map */
 	unsigned int highest_bit;	/* index of last free in swap_map */
 	unsigned int pages;		/* total of usable pages of swap */

Index: linux/mm/swapfile.c
===================================================================

mm/swapfile.c +169 ~~-13~~

@@ -184,6 +184,85 @@ static int wait_for_discard(void *word)
 #define SWAPFILE_CLUSTER	256
 #define LATENCY_LIMIT		256
 
+/*
+ * cluster info is a unsigned int, the highest 8 bits stores flags, the low 24
+ * bits stores next cluster if the cluster is free or cluster counter otherwise
+ */
+#define CLUSTER_FLAG_FREE (1 << 0)
+#define CLUSTER_FLAG_NEXT_NULL (1 << 1)
+#define CLUSTER_NULL (CLUSTER_FLAG_NEXT_NULL << 24)
+#define cluster_flag(info) ((info) >> 24)
+#define cluster_set_flag(info, flag) \
+	do { info = ((info) & 0xffffff) | ((flag) << 24); } while (0)
+#define cluster_count(info) ((info) & 0xffffff)
+#define cluster_set_count(info, c) \
+	do { info = (cluster_flag(info) << 24) | (c); } while (0)
+#define cluster_next(info) ((info) & 0xffffff)
+#define cluster_set_next(info, n) \
+	do { info = (cluster_flag(info) << 24) | (n); } while (0)
+#define cluster_is_free(info) (cluster_flag(info) & CLUSTER_FLAG_FREE)
+
+static inline void inc_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+	if (cluster_is_free(cluster_info[idx])) {
+		VM_BUG_ON(p->free_cluster_head != idx);
+		p->free_cluster_head = cluster_next(cluster_info[idx]);
+		if (p->free_cluster_tail == idx) {
+			p->free_cluster_tail = CLUSTER_NULL;
+			p->free_cluster_head = CLUSTER_NULL;
+		}
+		cluster_set_flag(cluster_info[idx], 0);
+		cluster_set_count(cluster_info[idx], 0);
+	}
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) >= SWAPFILE_CLUSTER);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) + 1);
+}
+
+static inline void dec_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) == 0);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) - 1);
+
+	if (cluster_count(cluster_info[idx]) == 0) {
+		cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+		if (p->free_cluster_head == CLUSTER_NULL) {
+			p->free_cluster_head = idx;
+			p->free_cluster_tail = idx;
+		} else {
+			cluster_set_next(cluster_info[p->free_cluster_tail],
+				idx);
+			p->free_cluster_tail = idx;
+		}
+	}
+}
+
+/*
+ * It's possible scan_swap_map() uses a free cluster in the middle of free
+ * cluster list. Avoiding such abuse to avoid list corruption.
+ */
+static inline bool scan_swap_map_recheck_cluster(struct swap_info_struct *si,
+	unsigned long offset)
+{
+	offset /= SWAPFILE_CLUSTER;
+	return si->free_cluster_head != CLUSTER_NULL &&
+		offset != si->free_cluster_head &&
+		cluster_is_free(si->cluster_info[offset]);
+}
+
 static unsigned long scan_swap_map(struct swap_info_struct *si,
 				   unsigned char usage)
 {

@@ -225,6 +304,24 @@ static unsigned long scan_swap_map(struc
 			si->lowest_alloc = si->max;
 			si->highest_alloc = 0;
 		}
+check_cluster:
+		if (si->free_cluster_head != CLUSTER_NULL) {
+			offset = si->free_cluster_head * SWAPFILE_CLUSTER;
+			last_in_cluster = offset + SWAPFILE_CLUSTER - 1;
+			si->cluster_next = offset;
+			si->cluster_nr = SWAPFILE_CLUSTER - 1;
+			found_free_cluster = 1;
+			goto checks;
+		} else if (si->cluster_info) {
+			/*
+			 * Checking free cluster is fast enough, we can do the
+			 * check every time
+			 */
+			si->cluster_nr = 0;
+			si->lowest_alloc = 0;
+			goto checks;
+		}
+
 		spin_unlock(&si->lock);
 
 		/*

@@ -285,6 +382,8 @@ static unsigned long scan_swap_map(struc
 	}
 
 checks:
+	if (scan_swap_map_recheck_cluster(si, offset))
+		goto check_cluster;
 	if (!(si->flags & SWP_WRITEOK))
 		goto no_page;
 	if (!si->highest_bit)

@@ -317,6 +416,7 @@ checks:
 		si->highest_bit = 0;
 	}
 	si->swap_map[offset] = usage;
+	inc_cluster_info_page(si, si->cluster_info, offset);
 	si->cluster_next = offset + 1;
 	si->flags -= SWP_SCANNING;

@@ -600,6 +700,7 @@ static unsigned char swap_entry_free(str
 
 	/* free if no reference */
 	if (!usage) {
+		dec_cluster_info_page(p, p->cluster_info, offset);
 		if (offset < p->lowest_bit)
 			p->lowest_bit = offset;
 		if (offset > p->highest_bit)

@@ -1497,6 +1598,7 @@ static int setup_swap_extents(struct swa
 
 static void _enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	int i, prev;

@@ -1506,6 +1608,7 @@ static void _enable_swap_info(struct swa
 	else
 		p->prio = --least_priority;
 	p->swap_map = swap_map;
+	p->cluster_info = cluster_info;
 	frontswap_map_set(p, frontswap_map);
 	p->flags |= SWP_WRITEOK;
 	atomic_long_add(p->pages, &nr_swap_pages);

@@ -1527,11 +1630,12 @@ static void _enable_swap_info(struct swa
 
 static void enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, prio, swap_map, frontswap_map);
+	_enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 	frontswap_init(p->type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);

@@ -1541,7 +1645,8 @@ static void reinsert_swap_info(struct sw
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, p->prio, p->swap_map, frontswap_map_get(p));
+	_enable_swap_info(p, p->prio, p->swap_map, p->cluster_info,
+					frontswap_map_get(p));
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 }

@@ -1550,6 +1655,7 @@ SYSCALL_DEFINE1(swapoff, const char __us
 {
 	struct swap_info_struct *p = NULL;
 	unsigned char *swap_map;
+	unsigned int *cluster_info;
 	struct file *swap_file, *victim;
 	struct address_space *mapping;
 	struct inode *inode;

@@ -1648,12 +1754,15 @@ SYSCALL_DEFINE1(swapoff, const char __us
 	p->max = 0;
 	swap_map = p->swap_map;
 	p->swap_map = NULL;
+	cluster_info = p->cluster_info;
+	p->cluster_info = NULL;
 	p->flags = 0;
 	frontswap_invalidate_area(type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 	mutex_unlock(&swapon_mutex);
 	vfree(swap_map);
+	vfree(cluster_info);
 	vfree(frontswap_map_get(p));
 	/* Destroy swap account informatin */
 	swap_cgroup_swapoff(type);

@@ -1966,15 +2075,21 @@ static unsigned long read_swap_header(st
 static int setup_swap_map_and_extents(struct swap_info_struct *p,
 					union swap_header *swap_header,
 					unsigned char *swap_map,
+					unsigned int *cluster_info,
 					unsigned long maxpages,
 					sector_t *span)
 {
 	int i;
 	unsigned int nr_good_pages;
 	int nr_extents;
+	unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);
+	unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
 
 	nr_good_pages = maxpages - 1;	/* omit header page */
 
+	p->free_cluster_head = CLUSTER_NULL;
+	p->free_cluster_tail = CLUSTER_NULL;
+
 	for (i = 0; i < swap_header->info.nr_badpages; i++) {
 		unsigned int page_nr = swap_header->info.badpages[i];
 		if (page_nr == 0 || page_nr > swap_header->info.last_page)

@@ -1982,11 +2097,25 @@ static int setup_swap_map_and_extents(st
 		if (page_nr < maxpages) {
 			swap_map[page_nr] = SWAP_MAP_BAD;
 			nr_good_pages--;
+			/*
+			 * Not mark the cluster free yet, no list
+			 * operation involved
+			 */
+			inc_cluster_info_page(p, cluster_info, page_nr);
 		}
 	}
 
+	/* Not mark the cluster free yet, no list operation involved */
+	for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
+		inc_cluster_info_page(p, cluster_info, i);
+
 	if (nr_good_pages) {
 		swap_map[0] = SWAP_MAP_BAD;
+		/*
+		 * Not mark the cluster free yet, no list
+		 * operation involved
+		 */
+		inc_cluster_info_page(p, cluster_info, 0);
 		p->max = maxpages;
 		p->pages = nr_good_pages;
 		nr_extents = setup_swap_extents(p, span);

@@ -1999,6 +2128,27 @@ static int setup_swap_map_and_extents(st
 		return -EINVAL;
 	}
 
+	if (!cluster_info)
+		return nr_extents;
+
+	for (i = 0; i < nr_clusters; i++) {
+		if (!cluster_count(cluster_info[idx])) {
+			cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+			if (p->free_cluster_head == CLUSTER_NULL) {
+				p->free_cluster_head = idx;
+				p->free_cluster_tail = idx;
+			} else {
+				cluster_set_next(
+					cluster_info[p->free_cluster_tail],
+					idx);
+				p->free_cluster_tail = idx;
+			}
+		}
+		idx++;
+		if (idx == nr_clusters)
+			idx = 0;
+	}
+
 	return nr_extents;
 }

@@ -2016,6 +2166,7 @@ SYSCALL_DEFINE2(swapon, const char __use
 	sector_t span;
 	unsigned long maxpages;
 	unsigned char *swap_map = NULL;
+	unsigned int *cluster_info = NULL;
 	unsigned long *frontswap_map = NULL;
 	struct page *page = NULL;
 	struct inode *inode = NULL;

@@ -2089,13 +2240,24 @@ SYSCALL_DEFINE2(swapon, const char __use
 		error = -ENOMEM;
 		goto bad_swap;
 	}
+	if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
+		p->flags |= SWP_SOLIDSTATE;
+		p->cluster_next = 1 + (random32() % p->highest_bit);
+
+		cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
+			SWAPFILE_CLUSTER) * sizeof(*cluster_info));
+		if (!cluster_info) {
+			error = -ENOMEM;
+			goto bad_swap;
+		}
+	}
 
 	error = swap_cgroup_swapon(p->type, maxpages);
 	if (error)
 		goto bad_swap;
 
 	nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
-		maxpages, &span);
+		cluster_info, maxpages, &span);
 	if (unlikely(nr_extents < 0)) {
 		error = nr_extents;
 		goto bad_swap;

@@ -2104,21 +2266,15 @@ SYSCALL_DEFINE2(swapon, const char __use
 	if (frontswap_enabled)
 		frontswap_map = vzalloc(maxpages / sizeof(long));
 
-	if (p->bdev) {
-		if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
-			p->flags |= SWP_SOLIDSTATE;
-			p->cluster_next = 1 + (random32() % p->highest_bit);
-		}
-		if ((swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
-			p->flags |= SWP_DISCARDABLE;
-	}
+	if (p->bdev && (swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
+		p->flags |= SWP_DISCARDABLE;
 
 	mutex_lock(&swapon_mutex);
 	prio = -1;
 	if (swap_flags & SWAP_FLAG_PREFER)
 		prio =
 		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
-	enable_swap_info(p, prio, swap_map, frontswap_map);
+	enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 
 	printk(KERN_INFO "Adding %uk swap on %s.  "
 			"Priority:%d extents:%d across:%lluk %s%s%s\n",

@@ -2148,6 +2304,7 @@ bad_swap:
 	p->flags = 0;
 	spin_unlock(&swap_lock);
 	vfree(swap_map);
+	vfree(cluster_info);
 	if (swap_file) {
 		if (inode && S_ISREG(inode->i_mode)) {
 			mutex_unlock(&inode->i_mutex);
 
--

Quoted 4 lines

To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Rafael Aquini

2013-03-12 15:19 UTC #4 lore raw

Quoted 1 line · Rafael Aquini

Nice work! I've been testing your series with the attached test-prog, as it was

Sorry, I forgot to attach the test-prog on my last message. here it is.

It just takes running it for a couple of minutes to get to that softlockup storm I told you.

If your box has, lets say 4GB of ram, trigger it as follows:
./threaded_memtest -qpv -m4096m

Rafael

/* $Id: threaded_memtest.c,v 1.7 2008/02/12 01:17:07 gnichols Exp $
*
* A scalable, threaded memory exerciser/tester.
*
* Author: Will Woods <[email protected]>
* Copyright (C) 2006 Red Hat, Inc. All Rights Reserved.
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. *
* Notes:
* This program uses sched_setaffinity(), which is Linux-specific. This could
* probably be ported to other systems with a fairly simple #ifdef / #define
* of setaffinity(), below. You might also have to find a replacement for
* sysconf(), which (while a POSIX function) is not available on some other
* systems (e.g. OSX).
*/

#include <unistd.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <errno.h>
#include <sys/sysinfo.h>
#include <sys/mman.h>
#include <sys/time.h>
#include <signal.h>
#define __USE_GNU 1
#include <pthread.h>
#include <sched.h>
#ifdef OLD_SCHED_SETAFFINITY
#define setaffinity(mask) sched_setaffinity(0,&mask) #else
#define setaffinity(mask) sched_setaffinity(0,sizeof(mask),&mask) #endif

#define VERSION "$Revision: 1.7 $" /* CVS version info */ #define DEFAULT_THREADS 2
#define DEFAULT_RUNTIME 60*15
#define DEFAULT_MEMPCT 0.95
#define BARLEN 40

/* configurable values used by the threads */
int verbose = 0;
int quiet = 0;
int parallel = 0;
unsigned num_threads, default_threads = DEFAULT_THREADS; unsigned runtime, default_runtime = DEFAULT_RUNTIME; unsigned long memsize, default_memsize;
/* system info */
unsigned num_cpus;
unsigned long total_ram;
/* statistic gathering */
struct timeval start={0,0}, finish={0,0}, duration={0,0}; unsigned long *loop_counters = NULL;
/* pointers for threads and their memory regions */ pthread_t *threads;
char **mmap_regions = NULL;
/* Thread mutexes and conditions */
unsigned created_threads = 0;
pthread_mutex_t ct_mutex = PTHREAD_MUTEX_INITIALIZER; unsigned live_threads = 0;
pthread_mutex_t lt_mutex = PTHREAD_MUTEX_INITIALIZER; unsigned mmap_done = 0;
pthread_mutex_t init_mutex = PTHREAD_MUTEX_INITIALIZER; pthread_cond_t init_cond = PTHREAD_COND_INITIALIZER; pthread_mutex_t mmap_mutex = PTHREAD_MUTEX_INITIALIZER; pthread_cond_t mmap_cond = PTHREAD_COND_INITIALIZER; pthread_mutex_t test_mutex = PTHREAD_MUTEX_INITIALIZER; pthread_cond_t test_start = PTHREAD_COND_INITIALIZER; pthread_mutex_t finish_mutex = PTHREAD_MUTEX_INITIALIZER; pthread_cond_t finish_cond = PTHREAD_COND_INITIALIZER; pthread_mutex_t running_threads_mutex = PTHREAD_MUTEX_INITIALIZER; pthread_cond_t running_threads_cond = PTHREAD_COND_INITIALIZER; unsigned done = 0;
long unsigned running_threads = 0;
/* short name of the program */
char *basename = NULL;

#ifdef CPU_ALLOC
/* RHEL6+ set the affinity for the current task to the given CPU */ /* This now uses dynamic cpu_sets as the convention cpu_set_t
was limited to 1024p */

int on_cpu(unsigned cpu){
    cpu_set_t* mask;
    size_t masksize;
    mask = CPU_ALLOC(num_cpus);
    masksize = CPU_ALLOC_SIZE(num_cpus);

    CPU_ZERO_S(masksize, mask);
    CPU_SET_S(cpu, masksize, mask);
    if (sched_setaffinity(0, masksize, mask) < 0) {
        perror("sched_setaffinity");
        return -1;
    }
    return 0;
}
#else
/* use old setup - RHEL5*/
int on_cpu(unsigned cpu){
    cpu_set_t mask;
    CPU_ZERO(&mask);
    CPU_SET(cpu,&mask);
    if (setaffinity(mask) <  0){
        perror("sched_setaffinity");
        return -1;
    }
    return 0;
}

#endif

/* Parse a memsize string like '34m' or '128k' into a long int */

long unsigned parse_memsize(const char *str) {
    long unsigned size;
    char okchars[] = "GgMmKk%";
    char unit;
    size=atoi(str); /* ignores trailing non-digit chars */
    unit=str[strlen(str)-1];
    if (index(okchars,unit)) {
       switch (unit) {
           case 'G':
           case 'g':size *= 1024;
           case 'M':
           case 'm':size *= 1024;
           case 'K':
           case 'k':size *= 1024; break;
           case '%':size = (size/100.0)*total_ram; break;
        }
    }
    return size;
}    
char memsize_str[22]; /* a 64-bit int is 20 digits long */
/* print a nice human-readable string for a large number of bytes */
char *human_memsize(long unsigned size) {
    char unit=' ';
    if (size > 10240) { unit='K'; size /= 1024; }
    if (size > 10240) { unit='M'; size /= 1024; }
    if (size > 10240) { unit='G'; size /= 1024; }
    snprintf(memsize_str,22,"%ld%c",size,unit);
    return memsize_str;
}

/* A cute little progress bar */
void progressbar(char *label, unsigned cur, unsigned total) {

    unsigned pos;
    char bar[BARLEN+1],spinner[]="-\\|/";
    pos=(BARLEN*cur)/total;
    memset(bar,'.',BARLEN);
    memset(bar,'#',pos);
    bar[BARLEN]='\0';
    if ((pos < BARLEN) && (total >= BARLEN*2))
        bar[pos]=spinner[cur%4];
    printf("\r%18s [%s] %u/%u",label,bar,cur,total);
    fflush(stdout);
}

/* This is the function that the threads run */

void *mem_twiddler(void *arg) {
    unsigned long thread_id, pages, pagesize, i, p;
    volatile long garbage;
    long *lp;
    int t,offset;
    char *my_region;
    unsigned long mapsize = *(unsigned long *)arg; 
    
    /* Make sure each thread gets a unique ID */
    pthread_mutex_lock(&ct_mutex);
    thread_id=created_threads++;
    pthread_mutex_unlock(&ct_mutex);
    if (parallel) {
        /* let main() go as soon as the thread is created */
        pthread_mutex_lock(&mmap_mutex);
        mmap_done=1;
        pthread_cond_signal(&mmap_cond); 
        pthread_mutex_unlock(&mmap_mutex);
    }
    
    on_cpu(thread_id % num_cpus);
    pagesize=getpagesize();
    pages=mapsize/pagesize;
    
    /* Map a chunk of memory */
    if (verbose) printf("thread %ld: mapping %s RAM\n",
                        thread_id,human_memsize(mapsize));
    my_region=mmap(NULL,mapsize,PROT_READ|PROT_WRITE,
                   MAP_ANONYMOUS|MAP_PRIVATE,-1,0);
    if (my_region == MAP_FAILED) { perror("mmap"); exit(1); }
    mmap_regions[thread_id] = my_region;
    /* Dirty each page of the mem region to fault them into existence */
    for (i=0;i<pages;i++) {
        lp=(long *)&(my_region[i*pagesize]);
        lp[0]=0xDEADBEEF; /* magic number */
        lp[1]=thread_id;
        lp[2]=i;
    }
    /* Okay, we have grabbed our memory */
    if (verbose) printf("thread %ld: mapping complete\n",thread_id);
    
    /* let main() go now that the thread is finished initializing. */
    if (!parallel) {
        mmap_done=1;
        pthread_cond_signal(&mmap_cond); 
	pthread_mutex_unlock(&mmap_mutex);
    }

    /*
     * Incrementing live_threads inside test_mutex avoids a timing
     * sensitive race -- otherwise some threads could miss the
     * pthread_cond_broadcast of test_start.
     */
    pthread_mutex_lock(&test_mutex);
    live_threads++;
    if (live_threads == num_threads) {
        /* if this is the last thread to init, let main() know we're done */
        /* NOTE: only the last thread sends this signal */
        pthread_cond_signal(&init_cond);
    }

    /* Wait for the signal to begin testing */
    while (start.tv_sec == 0) { 
        pthread_cond_wait(&test_start,&test_mutex);
    }
    pthread_mutex_unlock(&test_mutex);

    pthread_mutex_lock(&running_threads_mutex);
    running_threads++;
    if (verbose)
        printf("thread %lu (%lu): test start\n",thread_id,running_threads);
    pthread_mutex_unlock(&running_threads_mutex);

    loop_counters[thread_id]=0;
    while (!done) {
        /* Choose a random thread and a random page */
        t = rand() % num_threads;
        p = rand() % pages; 
        lp = (long *)&(mmap_regions[t][p*pagesize]);
        /* Check the info we wrote there earlier */
        if (lp[0] != 0xDEADBEEF || lp[1] != t || lp[2] != p) {
            fprintf(stderr,"MEMORY CORRUPTION DETECTED\n");
            fprintf(stderr,"thread %lu (CPU %lu) reading map %u, page %lu\n",
                    thread_id,thread_id % num_cpus,t,p);
            fprintf(stderr,"read: %#lx %lu %lu  should be: %#x %i %lu\n",
                    lp[0],lp[1],lp[2],0xDEADBEEF,t,p);
        }
        /* choose a random word (other than the first 3 */
        offset = (rand() % ((pagesize/sizeof(long))-3))+3;
        if (rand() % 2) {
            lp[offset] = rand();
        } else {
            garbage = lp[offset];
        }
        loop_counters[thread_id]++;
    }
    


    /* make sure everyone's finished before we unmap */
    pthread_mutex_lock(&finish_mutex);

    running_threads--;
    if (verbose)
        printf("thread %lu (%lu): test start\n",thread_id,running_threads);
    if (running_threads == 0)
        pthread_cond_broadcast(&finish_cond);   /* This is the cleanup thread */
    else {
        while (running_threads > 0) {
	    pthread_cond_wait(&finish_cond,&finish_mutex);
	}
    }
    pthread_mutex_unlock(&finish_mutex);

    /* Clean up and exit. */
    if (verbose) printf("thread %lu unmapping and exiting\n",thread_id);
    if (munmap(my_region,mapsize) != 0) {
        perror("munmap"); exit(2);
    }
    return NULL;
}

/* Function to be called on interrupt */
void int_handler(int signum) { done=1; }

/* print usage info (with name of binary) */

void usage(void) {
    printf("usage: %s [-h] [-v] [-q] [-p] [-t sec] [-n threads] [-m size]\n",
            basename);
    printf("  -h: show this help\n");
    printf("  -v: verbose\n");
    printf("  -q: quiet (do not show progress meters)\n");
    printf("  -p: parallel thread startup\n");
    printf("  -t: test time, in seconds. default: %u\n",default_runtime);
    printf("  -n: number of threads. default: %u (2*num_cpus)\n",default_threads);
    printf("  -m: memory usage. default: %s (%.0f%% of free RAM)\n",
            human_memsize(default_memsize),DEFAULT_MEMPCT*100.0);
    printf("memory size may use k/m/g suffixes, or may be a percentage of total RAM.\n");
}

int main(int argc, char **argv) {
    struct sysinfo info;
    struct sigaction mysig;
    int i,rv=0;
    float duration_f, loops_per_sec;
    unsigned long free_mem, mapsize;

    basename=strrchr(argv[0],'/');
    if (basename) basename++; else basename=argv[0];

    /* Calculate default values */
    /* Get processor count. */
    num_cpus = sysconf(_SC_NPROCESSORS_ONLN);
    /* Ensure we have at least two threads per CPU */
    if (num_cpus*2 > default_threads)
        default_threads = num_cpus*2;
    /* Get memory info */
    if (sysinfo(&info) != 0) { perror("sysinfo"); return -1; }
    free_mem=(info.freeram+info.bufferram)*info.mem_unit;
    total_ram=info.totalram*info.mem_unit;
    /* default to using most of free_mem */
    default_memsize = free_mem * DEFAULT_MEMPCT; 
    
    /* Set configurable values to reasonable defaults */
    runtime = default_runtime;
    num_threads = default_threads;
    memsize = default_memsize;
    
    /* parse options */
    while ((i = getopt(argc,argv,"hvqpt:n:m:")) != -1) {
        switch (i) {
            case 'h':
                usage();
                return 0;
            case 'v':
                verbose=1;
                break;
            case 'q':
                quiet=1;
                break;
            case 'p':
                parallel=1;
                break;
            case 't':
                runtime=atoi(optarg);
                if (!runtime) {
                    printf("%s: error: bad runtime \"%s\"\n",basename,optarg);
                    return 1;
                }
                break;
            case 'n':
                num_threads=atoi(optarg);
                if (!num_threads) {
                    printf("%s: error: bad thread count \"%s\"\n",basename,optarg);
                    return 1;
                }
                break;
            case 'm':
                memsize=parse_memsize(optarg);
                if (!memsize) {
                    printf("%s: error: bad memory size \"%s\"\n",basename,optarg);
                    return 1;
                }
                break;
        }
    }

    /* calculate mapsize now that memsize/num_threads is set */
    mapsize = memsize/num_threads;
    /* sanity checks */
    if (num_threads < num_cpus)
        printf("Warning: num_threads < num_cpus. This isn't usually a good idea.\n");
    if (memsize > free_mem)
        printf("Warning: memsize > free_mem. You will probably hit swap.\n");
    /* A little information */
    if (verbose) {
        printf("Detected %u processors.\n",num_cpus);
        printf("RAM: %.1f%% free (%s/",
                100.0*(double)free_mem/(double)total_ram,
                human_memsize(free_mem));
        printf("%s)\n",human_memsize(total_ram));
    }

    printf("Testing %s RAM for %u seconds using %u threads:\n",
            human_memsize(memsize),runtime,num_threads);
    
    /* Allocate room for thread info */
    threads=(pthread_t *)malloc(num_threads*sizeof(pthread_t));
    mmap_regions=(char **)malloc(num_threads*sizeof(char *));
    loop_counters=(unsigned long *)malloc(num_threads*sizeof(unsigned long *));

    for (i = 0; i < num_threads; i++) {
	mmap_regions[i] = NULL;
	loop_counters[i] = 0;
    }
    /* Create all our threads! */
    while (created_threads < num_threads) {
        pthread_mutex_lock(&mmap_mutex);
        mmap_done=0;
        if (pthread_create(&threads[created_threads],NULL,
                    mem_twiddler,(void*)&mapsize) != 0) {
            perror("pthread_create"); exit(1);
        }
        /* Wait for it to finish initializing */
        while (!mmap_done) { pthread_cond_wait(&mmap_cond,&mmap_mutex); }
        pthread_mutex_unlock(&mmap_mutex);
        if (!verbose && !quiet) 
            progressbar("Starting threads",created_threads,num_threads);
    }

    if (parallel) {
        /* Wait for the signal that everyone is finished initializing */        
        pthread_mutex_lock(&init_mutex);
        while (live_threads < num_threads) {
	    pthread_cond_wait(&init_cond,&init_mutex);
	}
        pthread_mutex_unlock(&init_mutex);
    }

    /* Let the testing begin! */
    if (!verbose && !quiet) printf("\n");
    gettimeofday(&start,NULL);
    pthread_cond_broadcast(&test_start);

    /* catch ^C signal */
    mysig.sa_handler=int_handler;
    sigemptyset(&mysig.sa_mask);
    mysig.sa_flags=0;
    sigaction(SIGINT,&mysig,NULL);

    /* Wait until all threads are actually running otherwise some threads
       never get started before done is set on large UV systems */

    while (running_threads < num_threads) {
	if (!quiet) progressbar("Running Threads", running_threads, num_threads);
	sleep((num_threads - running_threads) / 60 + 1);
    }
    /* Wait for the allotted time */
    i=0;
    while (!done && (i<runtime)) {
        if (sleep(1) == 0) i++;
        if (!quiet) progressbar("Testing RAM",i,runtime);
    }
    if (i != runtime)
        rv=1;
    
    /* Signal completion and join all threads */
    done=1;
    while (live_threads) {
        pthread_join(threads[live_threads-1],NULL);
        live_threads--;
	if (!quiet) progressbar("Joined Threads", (num_threads - live_threads), num_threads);
    }
    gettimeofday(&finish,NULL);
    if (!quiet) printf("\n");
    /* Test is officially complete. Calculate run speed. */
    timersub(&finish,&start,&duration);
    duration_f=(float)duration.tv_sec + (float)duration.tv_usec / 1000000.0;
    loops_per_sec=0;
    if (verbose) printf("Runtime was %.2fs\n",duration_f);
    for (i=0;i<num_threads;i++) {
        if (verbose) printf("thread %i: %lu loops\n",i,loop_counters[i]);
        loops_per_sec += (float)loop_counters[i]/duration_f;
    }
    printf("Total loops per second: %.2f\n",loops_per_sec);
    
    /* All done. Return success. */
    printf("Testing complete.\n");
    return rv;
}

Shaohua Li

author

2013-03-18 05:09 UTC #5 lore raw

In reply to Shaohua Li

Ping! are there any comments for this series?

Quoted 401 lines · Shaohua Li

Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.

This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.

We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).

V2 -> V3:
rebase to latest linux-next

V1 -> V2:
1. free cluster is added to a list, which makes searching cluster more efficient
2. only enable the algorithm for SSD.

Signed-off-by Shaohua Li <[email protected]>

---
 include/linux/swap.h |    3 
 mm/swapfile.c        |  181 +++++++++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 172 insertions(+), 12 deletions(-)

Index: linux/include/linux/swap.h
===================================================================

include/linux/swap.h +3 -0

@@ -185,6 +185,9 @@ struct swap_info_struct {
 	signed char	next;		/* next type on the swap list */
 	unsigned int	max;		/* extent of the swap_map */
 	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
+	unsigned int *cluster_info;	/* cluster info. Only for SSD */
+	unsigned int free_cluster_head;
+	unsigned int free_cluster_tail;
 	unsigned int lowest_bit;	/* index of first free in swap_map */
 	unsigned int highest_bit;	/* index of last free in swap_map */
 	unsigned int pages;		/* total of usable pages of swap */

Index: linux/mm/swapfile.c
===================================================================

mm/swapfile.c +168 ~~-12~~

@@ -184,6 +184,85 @@ static int wait_for_discard(void *word)
 #define SWAPFILE_CLUSTER	256
 #define LATENCY_LIMIT		256
 
+/*
+ * cluster info is a unsigned int, the highest 8 bits stores flags, the low 24
+ * bits stores next cluster if the cluster is free or cluster counter otherwise
+ */
+#define CLUSTER_FLAG_FREE (1 << 0)
+#define CLUSTER_FLAG_NEXT_NULL (1 << 1)
+#define CLUSTER_NULL (CLUSTER_FLAG_NEXT_NULL << 24)
+#define cluster_flag(info) ((info) >> 24)
+#define cluster_set_flag(info, flag) \
+	do { info = ((info) & 0xffffff) | ((flag) << 24); } while (0)
+#define cluster_count(info) ((info) & 0xffffff)
+#define cluster_set_count(info, c) \
+	do { info = (cluster_flag(info) << 24) | (c); } while (0)
+#define cluster_next(info) ((info) & 0xffffff)
+#define cluster_set_next(info, n) \
+	do { info = (cluster_flag(info) << 24) | (n); } while (0)
+#define cluster_is_free(info) (cluster_flag(info) & CLUSTER_FLAG_FREE)
+
+static inline void inc_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+	if (cluster_is_free(cluster_info[idx])) {
+		VM_BUG_ON(p->free_cluster_head != idx);
+		p->free_cluster_head = cluster_next(cluster_info[idx]);
+		if (p->free_cluster_tail == idx) {
+			p->free_cluster_tail = CLUSTER_NULL;
+			p->free_cluster_head = CLUSTER_NULL;
+		}
+		cluster_set_flag(cluster_info[idx], 0);
+		cluster_set_count(cluster_info[idx], 0);
+	}
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) >= SWAPFILE_CLUSTER);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) + 1);
+}
+
+static inline void dec_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) == 0);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) - 1);
+
+	if (cluster_count(cluster_info[idx]) == 0) {
+		cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+		if (p->free_cluster_head == CLUSTER_NULL) {
+			p->free_cluster_head = idx;
+			p->free_cluster_tail = idx;
+		} else {
+			cluster_set_next(cluster_info[p->free_cluster_tail],
+				idx);
+			p->free_cluster_tail = idx;
+		}
+	}
+}
+
+/*
+ * It's possible scan_swap_map() uses a free cluster in the middle of free
+ * cluster list. Avoiding such abuse to avoid list corruption.
+ */
+static inline bool scan_swap_map_recheck_cluster(struct swap_info_struct *si,
+	unsigned long offset)
+{
+	offset /= SWAPFILE_CLUSTER;
+	return si->free_cluster_head != CLUSTER_NULL &&
+		offset != si->free_cluster_head &&
+		cluster_is_free(si->cluster_info[offset]);
+}
+
 static unsigned long scan_swap_map(struct swap_info_struct *si,
 				   unsigned char usage)
 {

@@ -225,6 +304,24 @@ static unsigned long scan_swap_map(struc
 			si->lowest_alloc = si->max;
 			si->highest_alloc = 0;
 		}
+check_cluster:
+		if (si->free_cluster_head != CLUSTER_NULL) {
+			offset = si->free_cluster_head * SWAPFILE_CLUSTER;
+			last_in_cluster = offset + SWAPFILE_CLUSTER - 1;
+			si->cluster_next = offset;
+			si->cluster_nr = SWAPFILE_CLUSTER - 1;
+			found_free_cluster = 1;
+			goto checks;
+		} else if (si->cluster_info) {
+			/*
+			 * Checking free cluster is fast enough, we can do the
+			 * check every time
+			 */
+			si->cluster_nr = 0;
+			si->lowest_alloc = 0;
+			goto checks;
+		}
+
 		spin_unlock(&si->lock);
 
 		/*

@@ -285,6 +382,8 @@ static unsigned long scan_swap_map(struc
 	}
 
 checks:
+	if (scan_swap_map_recheck_cluster(si, offset))
+		goto check_cluster;
 	if (!(si->flags & SWP_WRITEOK))
 		goto no_page;
 	if (!si->highest_bit)

@@ -317,6 +416,7 @@ checks:
 		si->highest_bit = 0;
 	}
 	si->swap_map[offset] = usage;
+	inc_cluster_info_page(si, si->cluster_info, offset);
 	si->cluster_next = offset + 1;
 	si->flags -= SWP_SCANNING;

@@ -600,6 +700,7 @@ static unsigned char swap_entry_free(str
 
 	/* free if no reference */
 	if (!usage) {
+		dec_cluster_info_page(p, p->cluster_info, offset);
 		if (offset < p->lowest_bit)
 			p->lowest_bit = offset;
 		if (offset > p->highest_bit)

@@ -1497,6 +1598,7 @@ static int setup_swap_extents(struct swa
 
 static void _enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	int i, prev;

@@ -1506,6 +1608,7 @@ static void _enable_swap_info(struct swa
 	else
 		p->prio = --least_priority;
 	p->swap_map = swap_map;
+	p->cluster_info = cluster_info;
 	frontswap_map_set(p, frontswap_map);
 	p->flags |= SWP_WRITEOK;
 	atomic_long_add(p->pages, &nr_swap_pages);

@@ -1527,11 +1630,12 @@ static void _enable_swap_info(struct swa
 
 static void enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, prio, swap_map, frontswap_map);
+	_enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 	frontswap_init(p->type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);

@@ -1541,7 +1645,8 @@ static void reinsert_swap_info(struct sw
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, p->prio, p->swap_map, frontswap_map_get(p));
+	_enable_swap_info(p, p->prio, p->swap_map, p->cluster_info,
+					frontswap_map_get(p));
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 }

@@ -1550,6 +1655,7 @@ SYSCALL_DEFINE1(swapoff, const char __us
 {
 	struct swap_info_struct *p = NULL;
 	unsigned char *swap_map;
+	unsigned int *cluster_info;
 	struct file *swap_file, *victim;
 	struct address_space *mapping;
 	struct inode *inode;

@@ -1648,12 +1754,15 @@ SYSCALL_DEFINE1(swapoff, const char __us
 	p->max = 0;
 	swap_map = p->swap_map;
 	p->swap_map = NULL;
+	cluster_info = p->cluster_info;
+	p->cluster_info = NULL;
 	p->flags = 0;
 	frontswap_invalidate_area(type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 	mutex_unlock(&swapon_mutex);
 	vfree(swap_map);
+	vfree(cluster_info);
 	vfree(frontswap_map_get(p));
 	/* Destroy swap account informatin */
 	swap_cgroup_swapoff(type);

@@ -1966,15 +2075,21 @@ static unsigned long read_swap_header(st
 static int setup_swap_map_and_extents(struct swap_info_struct *p,
 					union swap_header *swap_header,
 					unsigned char *swap_map,
+					unsigned int *cluster_info,
 					unsigned long maxpages,
 					sector_t *span)
 {
 	int i;
 	unsigned int nr_good_pages;
 	int nr_extents;
+	unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);
+	unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
 
 	nr_good_pages = maxpages - 1;	/* omit header page */
 
+	p->free_cluster_head = CLUSTER_NULL;
+	p->free_cluster_tail = CLUSTER_NULL;
+
 	for (i = 0; i < swap_header->info.nr_badpages; i++) {
 		unsigned int page_nr = swap_header->info.badpages[i];
 		if (page_nr == 0 || page_nr > swap_header->info.last_page)

@@ -1982,11 +2097,25 @@ static int setup_swap_map_and_extents(st
 		if (page_nr < maxpages) {
 			swap_map[page_nr] = SWAP_MAP_BAD;
 			nr_good_pages--;
+			/*
+			 * Not mark the cluster free yet, no list
+			 * operation involved
+			 */
+			inc_cluster_info_page(p, cluster_info, page_nr);
 		}
 	}
 
+	/* Not mark the cluster free yet, no list operation involved */
+	for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
+		inc_cluster_info_page(p, cluster_info, i);
+
 	if (nr_good_pages) {
 		swap_map[0] = SWAP_MAP_BAD;
+		/*
+		 * Not mark the cluster free yet, no list
+		 * operation involved
+		 */
+		inc_cluster_info_page(p, cluster_info, 0);
 		p->max = maxpages;
 		p->pages = nr_good_pages;
 		nr_extents = setup_swap_extents(p, span);

@@ -1999,6 +2128,27 @@ static int setup_swap_map_and_extents(st
 		return -EINVAL;
 	}
 
+	if (!cluster_info)
+		return nr_extents;
+
+	for (i = 0; i < nr_clusters; i++) {
+		if (!cluster_count(cluster_info[idx])) {
+			cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+			if (p->free_cluster_head == CLUSTER_NULL) {
+				p->free_cluster_head = idx;
+				p->free_cluster_tail = idx;
+			} else {
+				cluster_set_next(
+					cluster_info[p->free_cluster_tail],
+					idx);
+				p->free_cluster_tail = idx;
+			}
+		}
+		idx++;
+		if (idx == nr_clusters)
+			idx = 0;
+	}
+
 	return nr_extents;
 }

@@ -2016,6 +2166,7 @@ SYSCALL_DEFINE2(swapon, const char __use
 	sector_t span;
 	unsigned long maxpages;
 	unsigned char *swap_map = NULL;
+	unsigned int *cluster_info = NULL;
 	unsigned long *frontswap_map = NULL;
 	struct page *page = NULL;
 	struct inode *inode = NULL;

@@ -2089,13 +2240,24 @@ SYSCALL_DEFINE2(swapon, const char __use
 		error = -ENOMEM;
 		goto bad_swap;
 	}
+	if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
+		p->flags |= SWP_SOLIDSTATE;
+		p->cluster_next = 1 + (random32() % p->highest_bit);
+
+		cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
+			SWAPFILE_CLUSTER) * sizeof(*cluster_info));
+		if (!cluster_info) {
+			error = -ENOMEM;
+			goto bad_swap;
+		}
+	}
 
 	error = swap_cgroup_swapon(p->type, maxpages);
 	if (error)
 		goto bad_swap;
 
 	nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
-		maxpages, &span);
+		cluster_info, maxpages, &span);
 	if (unlikely(nr_extents < 0)) {
 		error = nr_extents;
 		goto bad_swap;

@@ -2104,21 +2266,15 @@ SYSCALL_DEFINE2(swapon, const char __use
 	if (frontswap_enabled)
 		frontswap_map = vzalloc(maxpages / sizeof(long));
 
-	if (p->bdev) {
-		if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
-			p->flags |= SWP_SOLIDSTATE;
-			p->cluster_next = 1 + (random32() % p->highest_bit);
-		}
-		if ((swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
-			p->flags |= SWP_DISCARDABLE;
-	}
+	if (p->bdev && (swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
+		p->flags |= SWP_DISCARDABLE;
 
 	mutex_lock(&swapon_mutex);
 	prio = -1;
 	if (swap_flags & SWAP_FLAG_PREFER)
 		prio =
 		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
-	enable_swap_info(p, prio, swap_map, frontswap_map);
+	enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 
 	printk(KERN_INFO "Adding %uk swap on %s.  "
 			"Priority:%d extents:%d across:%lluk %s%s%s\n",

mm/swapfile.c bad_swap: · line 2310

Quoted 7 lines

mm/swapfile.c +1 -0

@@ -2148,6 +2304,7 @@ bad_swap:
 	p->flags = 0;
 	spin_unlock(&swap_lock);
 	vfree(swap_map);
+	vfree(cluster_info);
 	if (swap_file) {
 		if (inode && S_ISREG(inode->i_mode)) {
 			mutex_unlock(&inode->i_mutex);

Simon Jeons

2013-03-18 05:16 UTC #6 lore raw

Quoted 1 line · Shaohua Li

Ping! are there any comments for this series?

Could you show me your benchmark and testcase?

Quoted 400 lines · Shaohua Li

Quoted 398 lines

Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.

This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.

We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).

V2 -> V3:
rebase to latest linux-next

V1 -> V2:
1. free cluster is added to a list, which makes searching cluster more efficient
2. only enable the algorithm for SSD.

Signed-off-by Shaohua Li <[email protected]>

---
  include/linux/swap.h |    3
  mm/swapfile.c        |  181 +++++++++++++++++++++++++++++++++++++++++++++++----
  2 files changed, 172 insertions(+), 12 deletions(-)

Index: linux/include/linux/swap.h
===================================================================

include/linux/swap.h +3 -0

@@ -185,6 +185,9 @@ struct swap_info_struct {
  	signed char	next;		/* next type on the swap list */
  	unsigned int	max;		/* extent of the swap_map */
  	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
+	unsigned int *cluster_info;	/* cluster info. Only for SSD */
+	unsigned int free_cluster_head;
+	unsigned int free_cluster_tail;
  	unsigned int lowest_bit;	/* index of first free in swap_map */
  	unsigned int highest_bit;	/* index of last free in swap_map */
  	unsigned int pages;		/* total of usable pages of swap */

Index: linux/mm/swapfile.c
===================================================================

mm/swapfile.c +168 ~~-12~~

@@ -184,6 +184,85 @@ static int wait_for_discard(void *word)
  #define SWAPFILE_CLUSTER	256
  #define LATENCY_LIMIT		256
 
+/*
+ * cluster info is a unsigned int, the highest 8 bits stores flags, the low 24
+ * bits stores next cluster if the cluster is free or cluster counter otherwise
+ */
+#define CLUSTER_FLAG_FREE (1 << 0)
+#define CLUSTER_FLAG_NEXT_NULL (1 << 1)
+#define CLUSTER_NULL (CLUSTER_FLAG_NEXT_NULL << 24)
+#define cluster_flag(info) ((info) >> 24)
+#define cluster_set_flag(info, flag) \
+	do { info = ((info) & 0xffffff) | ((flag) << 24); } while (0)
+#define cluster_count(info) ((info) & 0xffffff)
+#define cluster_set_count(info, c) \
+	do { info = (cluster_flag(info) << 24) | (c); } while (0)
+#define cluster_next(info) ((info) & 0xffffff)
+#define cluster_set_next(info, n) \
+	do { info = (cluster_flag(info) << 24) | (n); } while (0)
+#define cluster_is_free(info) (cluster_flag(info) & CLUSTER_FLAG_FREE)
+
+static inline void inc_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+	if (cluster_is_free(cluster_info[idx])) {
+		VM_BUG_ON(p->free_cluster_head != idx);
+		p->free_cluster_head = cluster_next(cluster_info[idx]);
+		if (p->free_cluster_tail == idx) {
+			p->free_cluster_tail = CLUSTER_NULL;
+			p->free_cluster_head = CLUSTER_NULL;
+		}
+		cluster_set_flag(cluster_info[idx], 0);
+		cluster_set_count(cluster_info[idx], 0);
+	}
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) >= SWAPFILE_CLUSTER);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) + 1);
+}
+
+static inline void dec_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) == 0);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) - 1);
+
+	if (cluster_count(cluster_info[idx]) == 0) {
+		cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+		if (p->free_cluster_head == CLUSTER_NULL) {
+			p->free_cluster_head = idx;
+			p->free_cluster_tail = idx;
+		} else {
+			cluster_set_next(cluster_info[p->free_cluster_tail],
+				idx);
+			p->free_cluster_tail = idx;
+		}
+	}
+}
+
+/*
+ * It's possible scan_swap_map() uses a free cluster in the middle of free
+ * cluster list. Avoiding such abuse to avoid list corruption.
+ */
+static inline bool scan_swap_map_recheck_cluster(struct swap_info_struct *si,
+	unsigned long offset)
+{
+	offset /= SWAPFILE_CLUSTER;
+	return si->free_cluster_head != CLUSTER_NULL &&
+		offset != si->free_cluster_head &&
+		cluster_is_free(si->cluster_info[offset]);
+}
+
  static unsigned long scan_swap_map(struct swap_info_struct *si,
  				   unsigned char usage)
  {

@@ -225,6 +304,24 @@ static unsigned long scan_swap_map(struc
  			si->lowest_alloc = si->max;
  			si->highest_alloc = 0;
  		}
+check_cluster:
+		if (si->free_cluster_head != CLUSTER_NULL) {
+			offset = si->free_cluster_head * SWAPFILE_CLUSTER;
+			last_in_cluster = offset + SWAPFILE_CLUSTER - 1;
+			si->cluster_next = offset;
+			si->cluster_nr = SWAPFILE_CLUSTER - 1;
+			found_free_cluster = 1;
+			goto checks;
+		} else if (si->cluster_info) {
+			/*
+			 * Checking free cluster is fast enough, we can do the
+			 * check every time
+			 */
+			si->cluster_nr = 0;
+			si->lowest_alloc = 0;
+			goto checks;
+		}
+
  		spin_unlock(&si->lock);
 
  		/*

@@ -285,6 +382,8 @@ static unsigned long scan_swap_map(struc
  	}
 
  checks:
+	if (scan_swap_map_recheck_cluster(si, offset))
+		goto check_cluster;
  	if (!(si->flags & SWP_WRITEOK))
  		goto no_page;
  	if (!si->highest_bit)

@@ -317,6 +416,7 @@ checks:
  		si->highest_bit = 0;
  	}
  	si->swap_map[offset] = usage;
+	inc_cluster_info_page(si, si->cluster_info, offset);
  	si->cluster_next = offset + 1;
  	si->flags -= SWP_SCANNING;

@@ -600,6 +700,7 @@ static unsigned char swap_entry_free(str
 
  	/* free if no reference */
  	if (!usage) {
+		dec_cluster_info_page(p, p->cluster_info, offset);
  		if (offset < p->lowest_bit)
  			p->lowest_bit = offset;
  		if (offset > p->highest_bit)

@@ -1497,6 +1598,7 @@ static int setup_swap_extents(struct swa
 
  static void _enable_swap_info(struct swap_info_struct *p, int prio,
  				unsigned char *swap_map,
+				unsigned int *cluster_info,
  				unsigned long *frontswap_map)
  {
  	int i, prev;

@@ -1506,6 +1608,7 @@ static void _enable_swap_info(struct swa
  	else
  		p->prio = --least_priority;
  	p->swap_map = swap_map;
+	p->cluster_info = cluster_info;
  	frontswap_map_set(p, frontswap_map);
  	p->flags |= SWP_WRITEOK;
  	atomic_long_add(p->pages, &nr_swap_pages);

@@ -1527,11 +1630,12 @@ static void _enable_swap_info(struct swa
 
  static void enable_swap_info(struct swap_info_struct *p, int prio,
  				unsigned char *swap_map,
+				unsigned int *cluster_info,
  				unsigned long *frontswap_map)
  {
  	spin_lock(&swap_lock);
  	spin_lock(&p->lock);
-	_enable_swap_info(p, prio, swap_map, frontswap_map);
+	_enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
  	frontswap_init(p->type);
  	spin_unlock(&p->lock);
  	spin_unlock(&swap_lock);

@@ -1541,7 +1645,8 @@ static void reinsert_swap_info(struct sw
  {
  	spin_lock(&swap_lock);
  	spin_lock(&p->lock);
-	_enable_swap_info(p, p->prio, p->swap_map, frontswap_map_get(p));
+	_enable_swap_info(p, p->prio, p->swap_map, p->cluster_info,
+					frontswap_map_get(p));
  	spin_unlock(&p->lock);
  	spin_unlock(&swap_lock);
  }

@@ -1550,6 +1655,7 @@ SYSCALL_DEFINE1(swapoff, const char __us
  {
  	struct swap_info_struct *p = NULL;
  	unsigned char *swap_map;
+	unsigned int *cluster_info;
  	struct file *swap_file, *victim;
  	struct address_space *mapping;
  	struct inode *inode;

@@ -1648,12 +1754,15 @@ SYSCALL_DEFINE1(swapoff, const char __us
  	p->max = 0;
  	swap_map = p->swap_map;
  	p->swap_map = NULL;
+	cluster_info = p->cluster_info;
+	p->cluster_info = NULL;
  	p->flags = 0;
  	frontswap_invalidate_area(type);
  	spin_unlock(&p->lock);
  	spin_unlock(&swap_lock);
  	mutex_unlock(&swapon_mutex);
  	vfree(swap_map);
+	vfree(cluster_info);
  	vfree(frontswap_map_get(p));
  	/* Destroy swap account informatin */
  	swap_cgroup_swapoff(type);

@@ -1966,15 +2075,21 @@ static unsigned long read_swap_header(st
  static int setup_swap_map_and_extents(struct swap_info_struct *p,
  					union swap_header *swap_header,
  					unsigned char *swap_map,
+					unsigned int *cluster_info,
  					unsigned long maxpages,
  					sector_t *span)
  {
  	int i;
  	unsigned int nr_good_pages;
  	int nr_extents;
+	unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);
+	unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
 
  	nr_good_pages = maxpages - 1;	/* omit header page */
 
+	p->free_cluster_head = CLUSTER_NULL;
+	p->free_cluster_tail = CLUSTER_NULL;
+
  	for (i = 0; i < swap_header->info.nr_badpages; i++) {
  		unsigned int page_nr = swap_header->info.badpages[i];
  		if (page_nr == 0 || page_nr > swap_header->info.last_page)

@@ -1982,11 +2097,25 @@ static int setup_swap_map_and_extents(st
  		if (page_nr < maxpages) {
  			swap_map[page_nr] = SWAP_MAP_BAD;
  			nr_good_pages--;
+			/*
+			 * Not mark the cluster free yet, no list
+			 * operation involved
+			 */
+			inc_cluster_info_page(p, cluster_info, page_nr);
  		}
  	}
 
+	/* Not mark the cluster free yet, no list operation involved */
+	for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
+		inc_cluster_info_page(p, cluster_info, i);
+
  	if (nr_good_pages) {
  		swap_map[0] = SWAP_MAP_BAD;
+		/*
+		 * Not mark the cluster free yet, no list
+		 * operation involved
+		 */
+		inc_cluster_info_page(p, cluster_info, 0);
  		p->max = maxpages;
  		p->pages = nr_good_pages;
  		nr_extents = setup_swap_extents(p, span);

@@ -1999,6 +2128,27 @@ static int setup_swap_map_and_extents(st
  		return -EINVAL;
  	}
 
+	if (!cluster_info)
+		return nr_extents;
+
+	for (i = 0; i < nr_clusters; i++) {
+		if (!cluster_count(cluster_info[idx])) {
+			cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+			if (p->free_cluster_head == CLUSTER_NULL) {
+				p->free_cluster_head = idx;
+				p->free_cluster_tail = idx;
+			} else {
+				cluster_set_next(
+					cluster_info[p->free_cluster_tail],
+					idx);
+				p->free_cluster_tail = idx;
+			}
+		}
+		idx++;
+		if (idx == nr_clusters)
+			idx = 0;
+	}
+
  	return nr_extents;
  }

@@ -2016,6 +2166,7 @@ SYSCALL_DEFINE2(swapon, const char __use
  	sector_t span;
  	unsigned long maxpages;
  	unsigned char *swap_map = NULL;
+	unsigned int *cluster_info = NULL;
  	unsigned long *frontswap_map = NULL;
  	struct page *page = NULL;
  	struct inode *inode = NULL;

@@ -2089,13 +2240,24 @@ SYSCALL_DEFINE2(swapon, const char __use
  		error = -ENOMEM;
  		goto bad_swap;
  	}
+	if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
+		p->flags |= SWP_SOLIDSTATE;
+		p->cluster_next = 1 + (random32() % p->highest_bit);
+
+		cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
+			SWAPFILE_CLUSTER) * sizeof(*cluster_info));
+		if (!cluster_info) {
+			error = -ENOMEM;
+			goto bad_swap;
+		}
+	}
 
  	error = swap_cgroup_swapon(p->type, maxpages);
  	if (error)
  		goto bad_swap;
 
  	nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
-		maxpages, &span);
+		cluster_info, maxpages, &span);
  	if (unlikely(nr_extents < 0)) {
  		error = nr_extents;
  		goto bad_swap;

@@ -2104,21 +2266,15 @@ SYSCALL_DEFINE2(swapon, const char __use
  	if (frontswap_enabled)
  		frontswap_map = vzalloc(maxpages / sizeof(long));
 
-	if (p->bdev) {
-		if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
-			p->flags |= SWP_SOLIDSTATE;
-			p->cluster_next = 1 + (random32() % p->highest_bit);
-		}
-		if ((swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
-			p->flags |= SWP_DISCARDABLE;
-	}
+	if (p->bdev && (swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
+		p->flags |= SWP_DISCARDABLE;
 
  	mutex_lock(&swapon_mutex);
  	prio = -1;
  	if (swap_flags & SWAP_FLAG_PREFER)
  		prio =
  		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
-	enable_swap_info(p, prio, swap_map, frontswap_map);
+	enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);

Quoted 15 lines

mm/swapfile.c bad_swap: · line 2310
Quoted 10 lines
mm/swapfile.c +1 -0
  	printk(KERN_INFO "Adding %uk swap on %s.  "
  			"Priority:%d extents:%d across:%lluk %s%s%s\n",
@@ -2148,6 +2304,7 @@ bad_swap:
  	p->flags = 0;
  	spin_unlock(&swap_lock);
  	vfree(swap_map);
+	vfree(cluster_info);
  	if (swap_file) {
  		if (inode && S_ISREG(inode->i_mode)) {
  			mutex_unlock(&inode->i_mutex);
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Shaohua Li

author

2013-03-18 06:41 UTC #7 lore raw

Quoted 4 lines · Simon Jeons

Quoted 1 line · Shaohua Li
Ping! are there any comments for this series?
Could you show me your benchmark and testcase?

I use usemem from ext3-tools. It's a simple multi-thread/process mmap() test.

Simon Jeons

2013-03-18 06:49 UTC #8 lore raw

Quoted 5 lines · Shaohua Li

Quoted 3 lines · Simon Jeons
Quoted 1 line · Shaohua Li
Ping! are there any comments for this series?
Could you show me your benchmark and testcase?
I use usemem from ext3-tools. It's a simple multi-thread/process mmap() test.

Thanks, I know this tool. "scan_swap_map() sometimes uses up to 20~30% CPU time(when cluster is hard to find, the CPU time can be up to 80%)", how you monitor cpu utilization? how can predicate cpu utilization which specific function use?

Rafael Aquini

2013-03-18 21:03 UTC #9 lore raw

In reply to Shaohua Li

Quoted 2 lines · Shaohua Li

Ping! are there any comments for this series?

Did you get my replies from March, 12th? if not ping me and I'll re-submit them.

Rafael

Quoted 399 lines · Shaohua Li

Quoted 398 lines

Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.

This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.

We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).

V2 -> V3:
rebase to latest linux-next

V1 -> V2:
1. free cluster is added to a list, which makes searching cluster more efficient
2. only enable the algorithm for SSD.

Signed-off-by Shaohua Li <[email protected]>

---
 include/linux/swap.h |    3 
 mm/swapfile.c        |  181 +++++++++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 172 insertions(+), 12 deletions(-)

Index: linux/include/linux/swap.h
===================================================================

include/linux/swap.h +3 -0

@@ -185,6 +185,9 @@ struct swap_info_struct {
 	signed char	next;		/* next type on the swap list */
 	unsigned int	max;		/* extent of the swap_map */
 	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
+	unsigned int *cluster_info;	/* cluster info. Only for SSD */
+	unsigned int free_cluster_head;
+	unsigned int free_cluster_tail;
 	unsigned int lowest_bit;	/* index of first free in swap_map */
 	unsigned int highest_bit;	/* index of last free in swap_map */
 	unsigned int pages;		/* total of usable pages of swap */

Index: linux/mm/swapfile.c
===================================================================

mm/swapfile.c +168 ~~-12~~

@@ -184,6 +184,85 @@ static int wait_for_discard(void *word)
 #define SWAPFILE_CLUSTER	256
 #define LATENCY_LIMIT		256
 
+/*
+ * cluster info is a unsigned int, the highest 8 bits stores flags, the low 24
+ * bits stores next cluster if the cluster is free or cluster counter otherwise
+ */
+#define CLUSTER_FLAG_FREE (1 << 0)
+#define CLUSTER_FLAG_NEXT_NULL (1 << 1)
+#define CLUSTER_NULL (CLUSTER_FLAG_NEXT_NULL << 24)
+#define cluster_flag(info) ((info) >> 24)
+#define cluster_set_flag(info, flag) \
+	do { info = ((info) & 0xffffff) | ((flag) << 24); } while (0)
+#define cluster_count(info) ((info) & 0xffffff)
+#define cluster_set_count(info, c) \
+	do { info = (cluster_flag(info) << 24) | (c); } while (0)
+#define cluster_next(info) ((info) & 0xffffff)
+#define cluster_set_next(info, n) \
+	do { info = (cluster_flag(info) << 24) | (n); } while (0)
+#define cluster_is_free(info) (cluster_flag(info) & CLUSTER_FLAG_FREE)
+
+static inline void inc_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+	if (cluster_is_free(cluster_info[idx])) {
+		VM_BUG_ON(p->free_cluster_head != idx);
+		p->free_cluster_head = cluster_next(cluster_info[idx]);
+		if (p->free_cluster_tail == idx) {
+			p->free_cluster_tail = CLUSTER_NULL;
+			p->free_cluster_head = CLUSTER_NULL;
+		}
+		cluster_set_flag(cluster_info[idx], 0);
+		cluster_set_count(cluster_info[idx], 0);
+	}
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) >= SWAPFILE_CLUSTER);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) + 1);
+}
+
+static inline void dec_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) == 0);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) - 1);
+
+	if (cluster_count(cluster_info[idx]) == 0) {
+		cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+		if (p->free_cluster_head == CLUSTER_NULL) {
+			p->free_cluster_head = idx;
+			p->free_cluster_tail = idx;
+		} else {
+			cluster_set_next(cluster_info[p->free_cluster_tail],
+				idx);
+			p->free_cluster_tail = idx;
+		}
+	}
+}
+
+/*
+ * It's possible scan_swap_map() uses a free cluster in the middle of free
+ * cluster list. Avoiding such abuse to avoid list corruption.
+ */
+static inline bool scan_swap_map_recheck_cluster(struct swap_info_struct *si,
+	unsigned long offset)
+{
+	offset /= SWAPFILE_CLUSTER;
+	return si->free_cluster_head != CLUSTER_NULL &&
+		offset != si->free_cluster_head &&
+		cluster_is_free(si->cluster_info[offset]);
+}
+
 static unsigned long scan_swap_map(struct swap_info_struct *si,
 				   unsigned char usage)
 {

@@ -225,6 +304,24 @@ static unsigned long scan_swap_map(struc
 			si->lowest_alloc = si->max;
 			si->highest_alloc = 0;
 		}
+check_cluster:
+		if (si->free_cluster_head != CLUSTER_NULL) {
+			offset = si->free_cluster_head * SWAPFILE_CLUSTER;
+			last_in_cluster = offset + SWAPFILE_CLUSTER - 1;
+			si->cluster_next = offset;
+			si->cluster_nr = SWAPFILE_CLUSTER - 1;
+			found_free_cluster = 1;
+			goto checks;
+		} else if (si->cluster_info) {
+			/*
+			 * Checking free cluster is fast enough, we can do the
+			 * check every time
+			 */
+			si->cluster_nr = 0;
+			si->lowest_alloc = 0;
+			goto checks;
+		}
+
 		spin_unlock(&si->lock);
 
 		/*

@@ -285,6 +382,8 @@ static unsigned long scan_swap_map(struc
 	}
 
 checks:
+	if (scan_swap_map_recheck_cluster(si, offset))
+		goto check_cluster;
 	if (!(si->flags & SWP_WRITEOK))
 		goto no_page;
 	if (!si->highest_bit)

@@ -317,6 +416,7 @@ checks:
 		si->highest_bit = 0;
 	}
 	si->swap_map[offset] = usage;
+	inc_cluster_info_page(si, si->cluster_info, offset);
 	si->cluster_next = offset + 1;
 	si->flags -= SWP_SCANNING;

@@ -600,6 +700,7 @@ static unsigned char swap_entry_free(str
 
 	/* free if no reference */
 	if (!usage) {
+		dec_cluster_info_page(p, p->cluster_info, offset);
 		if (offset < p->lowest_bit)
 			p->lowest_bit = offset;
 		if (offset > p->highest_bit)

@@ -1497,6 +1598,7 @@ static int setup_swap_extents(struct swa
 
 static void _enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	int i, prev;

@@ -1506,6 +1608,7 @@ static void _enable_swap_info(struct swa
 	else
 		p->prio = --least_priority;
 	p->swap_map = swap_map;
+	p->cluster_info = cluster_info;
 	frontswap_map_set(p, frontswap_map);
 	p->flags |= SWP_WRITEOK;
 	atomic_long_add(p->pages, &nr_swap_pages);

@@ -1527,11 +1630,12 @@ static void _enable_swap_info(struct swa
 
 static void enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, prio, swap_map, frontswap_map);
+	_enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 	frontswap_init(p->type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);

@@ -1541,7 +1645,8 @@ static void reinsert_swap_info(struct sw
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, p->prio, p->swap_map, frontswap_map_get(p));
+	_enable_swap_info(p, p->prio, p->swap_map, p->cluster_info,
+					frontswap_map_get(p));
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 }

@@ -1550,6 +1655,7 @@ SYSCALL_DEFINE1(swapoff, const char __us
 {
 	struct swap_info_struct *p = NULL;
 	unsigned char *swap_map;
+	unsigned int *cluster_info;
 	struct file *swap_file, *victim;
 	struct address_space *mapping;
 	struct inode *inode;

@@ -1648,12 +1754,15 @@ SYSCALL_DEFINE1(swapoff, const char __us
 	p->max = 0;
 	swap_map = p->swap_map;
 	p->swap_map = NULL;
+	cluster_info = p->cluster_info;
+	p->cluster_info = NULL;
 	p->flags = 0;
 	frontswap_invalidate_area(type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 	mutex_unlock(&swapon_mutex);
 	vfree(swap_map);
+	vfree(cluster_info);
 	vfree(frontswap_map_get(p));
 	/* Destroy swap account informatin */
 	swap_cgroup_swapoff(type);

@@ -1966,15 +2075,21 @@ static unsigned long read_swap_header(st
 static int setup_swap_map_and_extents(struct swap_info_struct *p,
 					union swap_header *swap_header,
 					unsigned char *swap_map,
+					unsigned int *cluster_info,
 					unsigned long maxpages,
 					sector_t *span)
 {
 	int i;
 	unsigned int nr_good_pages;
 	int nr_extents;
+	unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);
+	unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
 
 	nr_good_pages = maxpages - 1;	/* omit header page */
 
+	p->free_cluster_head = CLUSTER_NULL;
+	p->free_cluster_tail = CLUSTER_NULL;
+
 	for (i = 0; i < swap_header->info.nr_badpages; i++) {
 		unsigned int page_nr = swap_header->info.badpages[i];
 		if (page_nr == 0 || page_nr > swap_header->info.last_page)

@@ -1982,11 +2097,25 @@ static int setup_swap_map_and_extents(st
 		if (page_nr < maxpages) {
 			swap_map[page_nr] = SWAP_MAP_BAD;
 			nr_good_pages--;
+			/*
+			 * Not mark the cluster free yet, no list
+			 * operation involved
+			 */
+			inc_cluster_info_page(p, cluster_info, page_nr);
 		}
 	}
 
+	/* Not mark the cluster free yet, no list operation involved */
+	for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
+		inc_cluster_info_page(p, cluster_info, i);
+
 	if (nr_good_pages) {
 		swap_map[0] = SWAP_MAP_BAD;
+		/*
+		 * Not mark the cluster free yet, no list
+		 * operation involved
+		 */
+		inc_cluster_info_page(p, cluster_info, 0);
 		p->max = maxpages;
 		p->pages = nr_good_pages;
 		nr_extents = setup_swap_extents(p, span);

@@ -1999,6 +2128,27 @@ static int setup_swap_map_and_extents(st
 		return -EINVAL;
 	}
 
+	if (!cluster_info)
+		return nr_extents;
+
+	for (i = 0; i < nr_clusters; i++) {
+		if (!cluster_count(cluster_info[idx])) {
+			cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+			if (p->free_cluster_head == CLUSTER_NULL) {
+				p->free_cluster_head = idx;
+				p->free_cluster_tail = idx;
+			} else {
+				cluster_set_next(
+					cluster_info[p->free_cluster_tail],
+					idx);
+				p->free_cluster_tail = idx;
+			}
+		}
+		idx++;
+		if (idx == nr_clusters)
+			idx = 0;
+	}
+
 	return nr_extents;
 }

@@ -2016,6 +2166,7 @@ SYSCALL_DEFINE2(swapon, const char __use
 	sector_t span;
 	unsigned long maxpages;
 	unsigned char *swap_map = NULL;
+	unsigned int *cluster_info = NULL;
 	unsigned long *frontswap_map = NULL;
 	struct page *page = NULL;
 	struct inode *inode = NULL;

@@ -2089,13 +2240,24 @@ SYSCALL_DEFINE2(swapon, const char __use
 		error = -ENOMEM;
 		goto bad_swap;
 	}
+	if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
+		p->flags |= SWP_SOLIDSTATE;
+		p->cluster_next = 1 + (random32() % p->highest_bit);
+
+		cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
+			SWAPFILE_CLUSTER) * sizeof(*cluster_info));
+		if (!cluster_info) {
+			error = -ENOMEM;
+			goto bad_swap;
+		}
+	}
 
 	error = swap_cgroup_swapon(p->type, maxpages);
 	if (error)
 		goto bad_swap;
 
 	nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
-		maxpages, &span);
+		cluster_info, maxpages, &span);
 	if (unlikely(nr_extents < 0)) {
 		error = nr_extents;
 		goto bad_swap;

@@ -2104,21 +2266,15 @@ SYSCALL_DEFINE2(swapon, const char __use
 	if (frontswap_enabled)
 		frontswap_map = vzalloc(maxpages / sizeof(long));
 
-	if (p->bdev) {
-		if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
-			p->flags |= SWP_SOLIDSTATE;
-			p->cluster_next = 1 + (random32() % p->highest_bit);
-		}
-		if ((swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
-			p->flags |= SWP_DISCARDABLE;
-	}
+	if (p->bdev && (swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
+		p->flags |= SWP_DISCARDABLE;
 
 	mutex_lock(&swapon_mutex);
 	prio = -1;
 	if (swap_flags & SWAP_FLAG_PREFER)
 		prio =
 		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
-	enable_swap_info(p, prio, swap_map, frontswap_map);
+	enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);

Quoted 16 lines

mm/swapfile.c bad_swap: · line 2310
Quoted 10 lines
mm/swapfile.c +1 -0
 	printk(KERN_INFO "Adding %uk swap on %s.  "
 			"Priority:%d extents:%d across:%lluk %s%s%s\n",
@@ -2148,6 +2304,7 @@ bad_swap:
 	p->flags = 0;
 	spin_unlock(&swap_lock);
 	vfree(swap_map);
+	vfree(cluster_info);
 	if (swap_file) {
 		if (inode && S_ISREG(inode->i_mode)) {
 			mutex_unlock(&inode->i_mutex);
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Shaohua Li

author

2013-03-19 01:32 UTC #10 lore raw

Quoted 4 lines · Rafael Aquini

Quoted 2 lines · Shaohua Li
Ping! are there any comments for this series?
Did you get my replies from March, 12th? if not ping me and I'll re-submit them.

I did and looks you are happy with them, and thanks for testing. I'm checking if there are anything I need address from others.

Thanks,
Shaohua

Quoted 399 lines

Quoted 399 lines · Shaohua Li

Quoted 398 lines

Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.

This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.

We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).

V2 -> V3:
rebase to latest linux-next

V1 -> V2:
1. free cluster is added to a list, which makes searching cluster more efficient
2. only enable the algorithm for SSD.

Signed-off-by Shaohua Li <[email protected]>

---
 include/linux/swap.h |    3 
 mm/swapfile.c        |  181 +++++++++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 172 insertions(+), 12 deletions(-)

Index: linux/include/linux/swap.h
===================================================================

include/linux/swap.h +3 -0

@@ -185,6 +185,9 @@ struct swap_info_struct {
 	signed char	next;		/* next type on the swap list */
 	unsigned int	max;		/* extent of the swap_map */
 	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
+	unsigned int *cluster_info;	/* cluster info. Only for SSD */
+	unsigned int free_cluster_head;
+	unsigned int free_cluster_tail;
 	unsigned int lowest_bit;	/* index of first free in swap_map */
 	unsigned int highest_bit;	/* index of last free in swap_map */
 	unsigned int pages;		/* total of usable pages of swap */

Index: linux/mm/swapfile.c
===================================================================

mm/swapfile.c +168 ~~-12~~

@@ -184,6 +184,85 @@ static int wait_for_discard(void *word)
 #define SWAPFILE_CLUSTER	256
 #define LATENCY_LIMIT		256
 
+/*
+ * cluster info is a unsigned int, the highest 8 bits stores flags, the low 24
+ * bits stores next cluster if the cluster is free or cluster counter otherwise
+ */
+#define CLUSTER_FLAG_FREE (1 << 0)
+#define CLUSTER_FLAG_NEXT_NULL (1 << 1)
+#define CLUSTER_NULL (CLUSTER_FLAG_NEXT_NULL << 24)
+#define cluster_flag(info) ((info) >> 24)
+#define cluster_set_flag(info, flag) \
+	do { info = ((info) & 0xffffff) | ((flag) << 24); } while (0)
+#define cluster_count(info) ((info) & 0xffffff)
+#define cluster_set_count(info, c) \
+	do { info = (cluster_flag(info) << 24) | (c); } while (0)
+#define cluster_next(info) ((info) & 0xffffff)
+#define cluster_set_next(info, n) \
+	do { info = (cluster_flag(info) << 24) | (n); } while (0)
+#define cluster_is_free(info) (cluster_flag(info) & CLUSTER_FLAG_FREE)
+
+static inline void inc_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+	if (cluster_is_free(cluster_info[idx])) {
+		VM_BUG_ON(p->free_cluster_head != idx);
+		p->free_cluster_head = cluster_next(cluster_info[idx]);
+		if (p->free_cluster_tail == idx) {
+			p->free_cluster_tail = CLUSTER_NULL;
+			p->free_cluster_head = CLUSTER_NULL;
+		}
+		cluster_set_flag(cluster_info[idx], 0);
+		cluster_set_count(cluster_info[idx], 0);
+	}
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) >= SWAPFILE_CLUSTER);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) + 1);
+}
+
+static inline void dec_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) == 0);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) - 1);
+
+	if (cluster_count(cluster_info[idx]) == 0) {
+		cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+		if (p->free_cluster_head == CLUSTER_NULL) {
+			p->free_cluster_head = idx;
+			p->free_cluster_tail = idx;
+		} else {
+			cluster_set_next(cluster_info[p->free_cluster_tail],
+				idx);
+			p->free_cluster_tail = idx;
+		}
+	}
+}
+
+/*
+ * It's possible scan_swap_map() uses a free cluster in the middle of free
+ * cluster list. Avoiding such abuse to avoid list corruption.
+ */
+static inline bool scan_swap_map_recheck_cluster(struct swap_info_struct *si,
+	unsigned long offset)
+{
+	offset /= SWAPFILE_CLUSTER;
+	return si->free_cluster_head != CLUSTER_NULL &&
+		offset != si->free_cluster_head &&
+		cluster_is_free(si->cluster_info[offset]);
+}
+
 static unsigned long scan_swap_map(struct swap_info_struct *si,
 				   unsigned char usage)
 {

@@ -225,6 +304,24 @@ static unsigned long scan_swap_map(struc
 			si->lowest_alloc = si->max;
 			si->highest_alloc = 0;
 		}
+check_cluster:
+		if (si->free_cluster_head != CLUSTER_NULL) {
+			offset = si->free_cluster_head * SWAPFILE_CLUSTER;
+			last_in_cluster = offset + SWAPFILE_CLUSTER - 1;
+			si->cluster_next = offset;
+			si->cluster_nr = SWAPFILE_CLUSTER - 1;
+			found_free_cluster = 1;
+			goto checks;
+		} else if (si->cluster_info) {
+			/*
+			 * Checking free cluster is fast enough, we can do the
+			 * check every time
+			 */
+			si->cluster_nr = 0;
+			si->lowest_alloc = 0;
+			goto checks;
+		}
+
 		spin_unlock(&si->lock);
 
 		/*

@@ -285,6 +382,8 @@ static unsigned long scan_swap_map(struc
 	}
 
 checks:
+	if (scan_swap_map_recheck_cluster(si, offset))
+		goto check_cluster;
 	if (!(si->flags & SWP_WRITEOK))
 		goto no_page;
 	if (!si->highest_bit)

@@ -317,6 +416,7 @@ checks:
 		si->highest_bit = 0;
 	}
 	si->swap_map[offset] = usage;
+	inc_cluster_info_page(si, si->cluster_info, offset);
 	si->cluster_next = offset + 1;
 	si->flags -= SWP_SCANNING;

@@ -600,6 +700,7 @@ static unsigned char swap_entry_free(str
 
 	/* free if no reference */
 	if (!usage) {
+		dec_cluster_info_page(p, p->cluster_info, offset);
 		if (offset < p->lowest_bit)
 			p->lowest_bit = offset;
 		if (offset > p->highest_bit)

@@ -1497,6 +1598,7 @@ static int setup_swap_extents(struct swa
 
 static void _enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	int i, prev;

@@ -1506,6 +1608,7 @@ static void _enable_swap_info(struct swa
 	else
 		p->prio = --least_priority;
 	p->swap_map = swap_map;
+	p->cluster_info = cluster_info;
 	frontswap_map_set(p, frontswap_map);
 	p->flags |= SWP_WRITEOK;
 	atomic_long_add(p->pages, &nr_swap_pages);

@@ -1527,11 +1630,12 @@ static void _enable_swap_info(struct swa
 
 static void enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, prio, swap_map, frontswap_map);
+	_enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 	frontswap_init(p->type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);

@@ -1541,7 +1645,8 @@ static void reinsert_swap_info(struct sw
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, p->prio, p->swap_map, frontswap_map_get(p));
+	_enable_swap_info(p, p->prio, p->swap_map, p->cluster_info,
+					frontswap_map_get(p));
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 }

@@ -1550,6 +1655,7 @@ SYSCALL_DEFINE1(swapoff, const char __us
 {
 	struct swap_info_struct *p = NULL;
 	unsigned char *swap_map;
+	unsigned int *cluster_info;
 	struct file *swap_file, *victim;
 	struct address_space *mapping;
 	struct inode *inode;

@@ -1648,12 +1754,15 @@ SYSCALL_DEFINE1(swapoff, const char __us
 	p->max = 0;
 	swap_map = p->swap_map;
 	p->swap_map = NULL;
+	cluster_info = p->cluster_info;
+	p->cluster_info = NULL;
 	p->flags = 0;
 	frontswap_invalidate_area(type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 	mutex_unlock(&swapon_mutex);
 	vfree(swap_map);
+	vfree(cluster_info);
 	vfree(frontswap_map_get(p));
 	/* Destroy swap account informatin */
 	swap_cgroup_swapoff(type);

@@ -1966,15 +2075,21 @@ static unsigned long read_swap_header(st
 static int setup_swap_map_and_extents(struct swap_info_struct *p,
 					union swap_header *swap_header,
 					unsigned char *swap_map,
+					unsigned int *cluster_info,
 					unsigned long maxpages,
 					sector_t *span)
 {
 	int i;
 	unsigned int nr_good_pages;
 	int nr_extents;
+	unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);
+	unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
 
 	nr_good_pages = maxpages - 1;	/* omit header page */
 
+	p->free_cluster_head = CLUSTER_NULL;
+	p->free_cluster_tail = CLUSTER_NULL;
+
 	for (i = 0; i < swap_header->info.nr_badpages; i++) {
 		unsigned int page_nr = swap_header->info.badpages[i];
 		if (page_nr == 0 || page_nr > swap_header->info.last_page)

@@ -1982,11 +2097,25 @@ static int setup_swap_map_and_extents(st
 		if (page_nr < maxpages) {
 			swap_map[page_nr] = SWAP_MAP_BAD;
 			nr_good_pages--;
+			/*
+			 * Not mark the cluster free yet, no list
+			 * operation involved
+			 */
+			inc_cluster_info_page(p, cluster_info, page_nr);
 		}
 	}
 
+	/* Not mark the cluster free yet, no list operation involved */
+	for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
+		inc_cluster_info_page(p, cluster_info, i);
+
 	if (nr_good_pages) {
 		swap_map[0] = SWAP_MAP_BAD;
+		/*
+		 * Not mark the cluster free yet, no list
+		 * operation involved
+		 */
+		inc_cluster_info_page(p, cluster_info, 0);
 		p->max = maxpages;
 		p->pages = nr_good_pages;
 		nr_extents = setup_swap_extents(p, span);

@@ -1999,6 +2128,27 @@ static int setup_swap_map_and_extents(st
 		return -EINVAL;
 	}
 
+	if (!cluster_info)
+		return nr_extents;
+
+	for (i = 0; i < nr_clusters; i++) {
+		if (!cluster_count(cluster_info[idx])) {
+			cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+			if (p->free_cluster_head == CLUSTER_NULL) {
+				p->free_cluster_head = idx;
+				p->free_cluster_tail = idx;
+			} else {
+				cluster_set_next(
+					cluster_info[p->free_cluster_tail],
+					idx);
+				p->free_cluster_tail = idx;
+			}
+		}
+		idx++;
+		if (idx == nr_clusters)
+			idx = 0;
+	}
+
 	return nr_extents;
 }

@@ -2016,6 +2166,7 @@ SYSCALL_DEFINE2(swapon, const char __use
 	sector_t span;
 	unsigned long maxpages;
 	unsigned char *swap_map = NULL;
+	unsigned int *cluster_info = NULL;
 	unsigned long *frontswap_map = NULL;
 	struct page *page = NULL;
 	struct inode *inode = NULL;

@@ -2089,13 +2240,24 @@ SYSCALL_DEFINE2(swapon, const char __use
 		error = -ENOMEM;
 		goto bad_swap;
 	}
+	if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
+		p->flags |= SWP_SOLIDSTATE;
+		p->cluster_next = 1 + (random32() % p->highest_bit);
+
+		cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
+			SWAPFILE_CLUSTER) * sizeof(*cluster_info));
+		if (!cluster_info) {
+			error = -ENOMEM;
+			goto bad_swap;
+		}
+	}
 
 	error = swap_cgroup_swapon(p->type, maxpages);
 	if (error)
 		goto bad_swap;
 
 	nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
-		maxpages, &span);
+		cluster_info, maxpages, &span);
 	if (unlikely(nr_extents < 0)) {
 		error = nr_extents;
 		goto bad_swap;

@@ -2104,21 +2266,15 @@ SYSCALL_DEFINE2(swapon, const char __use
 	if (frontswap_enabled)
 		frontswap_map = vzalloc(maxpages / sizeof(long));
 
-	if (p->bdev) {
-		if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
-			p->flags |= SWP_SOLIDSTATE;
-			p->cluster_next = 1 + (random32() % p->highest_bit);
-		}
-		if ((swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
-			p->flags |= SWP_DISCARDABLE;
-	}
+	if (p->bdev && (swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
+		p->flags |= SWP_DISCARDABLE;
 
 	mutex_lock(&swapon_mutex);
 	prio = -1;
 	if (swap_flags & SWAP_FLAG_PREFER)
 		prio =
 		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
-	enable_swap_info(p, prio, swap_map, frontswap_map);
+	enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);

Quoted 16 lines

Quoted 8 lines
Quoted 8 lines
mm/swapfile.c +1 -0
 	printk(KERN_INFO "Adding %uk swap on %s.  "
 			"Priority:%d extents:%d across:%lluk %s%s%s\n",
@@ -2148,6 +2304,7 @@ bad_swap:
 	p->flags = 0;
 	spin_unlock(&swap_lock);
 	vfree(swap_map);
+	vfree(cluster_info);
 	if (swap_file) {
Quoted 8 lines
mm/swapfile.c bad_swap: · line 2310
Quoted 2 lines
mm/swapfile.c +0 -0
 		if (inode && S_ISREG(inode->i_mode)) {
 			mutex_unlock(&inode->i_mutex);
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Hugh Dickins

Acked-by on v3 1/4

2013-03-19 20:51 UTC #11 lore raw

In reply to Shaohua Li

Quoted 34 lines · Shaohua Li

I'm using a fast SSD to do swap. scan_swap_map() sometimes uses up to 20~30% CPU time (when cluster is hard to find, the CPU time can be up to 80%), which becomes a bottleneck. scan_swap_map() scans a byte array to search a 256 page cluster, which is very slow.
Here I introduced a simple algorithm to search cluster. Since we only care about 256 pages cluster, we can just use a counter to track if a cluster is free. Every 256 pages use one int to store the counter. If the counter of a cluster is 0, the cluster is free. All free clusters will be added to a list, so searching cluster is very efficient. With this, scap_swap_map() overhead disappears.
Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.
This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.
We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).
The patch slightly changes which cluster is choosen. It always adds free cluster to list tail. This can help wear leveling for low end SSD too. And if no cluster found, the scan_swap_map() will do search from the end of last cluster. So if no cluster found, the scan_swap_map() will do search from the end of last free cluster, which is random. For SSD, this isn't a problem at all.
Another downside is the cluster must be aligned to 256 pages, which will reduce the chance to find a cluster. I would expect this isn't a big problem for SSD because of the non-seek penality. (And this is the reason I only enable the algorithm for SSD).

Quoted 8 lines

V2 -> V3:
rebase to latest linux-next
V1 -> V2:
1. free cluster is added to a list, which makes searching cluster more efficient
2. only enable the algorithm for SSD.
Signed-off-by Shaohua Li <[email protected]>

Acked-by Hugh Dickins <[email protected]>

Though I wouldn't particularly like this, if it did not lead on to your 3/4 addressing discard.

I find it a bit confusing that we now have these two different clustering strategies in scan_swap_map(), one for SSD and one for the rest; and it's not immediately obvious what's used for what.

But I respect your preference not to interfere with the HDD allocation right now (even to the extent of not indenting it into "else" blocks), since that's not what's driving you and not what you're testing so much.

I hope that maybe later on we could try deleting the older strategy, and see if any ill effect is actually noticed. If I didn't hope that, perhaps I'd be asking you to rename cluster_info to ssd_cluster_info.

I've a little more to say on that against 4/4, where I've paid more attention to the final result than to these intervening steps.

Quoted 360 lines

---
 include/linux/swap.h |    3 
 mm/swapfile.c        |  181 +++++++++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 172 insertions(+), 12 deletions(-)

Index: linux/include/linux/swap.h
===================================================================

include/linux/swap.h +3 -0

@@ -185,6 +185,9 @@ struct swap_info_struct {
 	signed char	next;		/* next type on the swap list */
 	unsigned int	max;		/* extent of the swap_map */
 	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
+	unsigned int *cluster_info;	/* cluster info. Only for SSD */
+	unsigned int free_cluster_head;
+	unsigned int free_cluster_tail;
 	unsigned int lowest_bit;	/* index of first free in swap_map */
 	unsigned int highest_bit;	/* index of last free in swap_map */
 	unsigned int pages;		/* total of usable pages of swap */

Index: linux/mm/swapfile.c
===================================================================

mm/swapfile.c +168 ~~-12~~

@@ -184,6 +184,85 @@ static int wait_for_discard(void *word)
 #define SWAPFILE_CLUSTER	256
 #define LATENCY_LIMIT		256
 
+/*
+ * cluster info is a unsigned int, the highest 8 bits stores flags, the low 24
+ * bits stores next cluster if the cluster is free or cluster counter otherwise
+ */
+#define CLUSTER_FLAG_FREE (1 << 0)
+#define CLUSTER_FLAG_NEXT_NULL (1 << 1)
+#define CLUSTER_NULL (CLUSTER_FLAG_NEXT_NULL << 24)
+#define cluster_flag(info) ((info) >> 24)
+#define cluster_set_flag(info, flag) \
+	do { info = ((info) & 0xffffff) | ((flag) << 24); } while (0)
+#define cluster_count(info) ((info) & 0xffffff)
+#define cluster_set_count(info, c) \
+	do { info = (cluster_flag(info) << 24) | (c); } while (0)
+#define cluster_next(info) ((info) & 0xffffff)
+#define cluster_set_next(info, n) \
+	do { info = (cluster_flag(info) << 24) | (n); } while (0)
+#define cluster_is_free(info) (cluster_flag(info) & CLUSTER_FLAG_FREE)
+
+static inline void inc_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+	if (cluster_is_free(cluster_info[idx])) {
+		VM_BUG_ON(p->free_cluster_head != idx);
+		p->free_cluster_head = cluster_next(cluster_info[idx]);
+		if (p->free_cluster_tail == idx) {
+			p->free_cluster_tail = CLUSTER_NULL;
+			p->free_cluster_head = CLUSTER_NULL;
+		}
+		cluster_set_flag(cluster_info[idx], 0);
+		cluster_set_count(cluster_info[idx], 0);
+	}
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) >= SWAPFILE_CLUSTER);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) + 1);
+}
+
+static inline void dec_cluster_info_page(struct swap_info_struct *p,
+	unsigned int *cluster_info, unsigned long page_nr)
+{
+	unsigned long idx = page_nr / SWAPFILE_CLUSTER;
+
+	if (!cluster_info)
+		return;
+
+	VM_BUG_ON(cluster_count(cluster_info[idx]) == 0);
+	cluster_set_count(cluster_info[idx],
+		cluster_count(cluster_info[idx]) - 1);
+
+	if (cluster_count(cluster_info[idx]) == 0) {
+		cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+		if (p->free_cluster_head == CLUSTER_NULL) {
+			p->free_cluster_head = idx;
+			p->free_cluster_tail = idx;
+		} else {
+			cluster_set_next(cluster_info[p->free_cluster_tail],
+				idx);
+			p->free_cluster_tail = idx;
+		}
+	}
+}
+
+/*
+ * It's possible scan_swap_map() uses a free cluster in the middle of free
+ * cluster list. Avoiding such abuse to avoid list corruption.
+ */
+static inline bool scan_swap_map_recheck_cluster(struct swap_info_struct *si,
+	unsigned long offset)
+{
+	offset /= SWAPFILE_CLUSTER;
+	return si->free_cluster_head != CLUSTER_NULL &&
+		offset != si->free_cluster_head &&
+		cluster_is_free(si->cluster_info[offset]);
+}
+
 static unsigned long scan_swap_map(struct swap_info_struct *si,
 				   unsigned char usage)
 {

@@ -225,6 +304,24 @@ static unsigned long scan_swap_map(struc
 			si->lowest_alloc = si->max;
 			si->highest_alloc = 0;
 		}
+check_cluster:
+		if (si->free_cluster_head != CLUSTER_NULL) {
+			offset = si->free_cluster_head * SWAPFILE_CLUSTER;
+			last_in_cluster = offset + SWAPFILE_CLUSTER - 1;
+			si->cluster_next = offset;
+			si->cluster_nr = SWAPFILE_CLUSTER - 1;
+			found_free_cluster = 1;
+			goto checks;
+		} else if (si->cluster_info) {
+			/*
+			 * Checking free cluster is fast enough, we can do the
+			 * check every time
+			 */
+			si->cluster_nr = 0;
+			si->lowest_alloc = 0;
+			goto checks;
+		}
+
 		spin_unlock(&si->lock);
 
 		/*

@@ -285,6 +382,8 @@ static unsigned long scan_swap_map(struc
 	}
 
 checks:
+	if (scan_swap_map_recheck_cluster(si, offset))
+		goto check_cluster;
 	if (!(si->flags & SWP_WRITEOK))
 		goto no_page;
 	if (!si->highest_bit)

@@ -317,6 +416,7 @@ checks:
 		si->highest_bit = 0;
 	}
 	si->swap_map[offset] = usage;
+	inc_cluster_info_page(si, si->cluster_info, offset);
 	si->cluster_next = offset + 1;
 	si->flags -= SWP_SCANNING;

@@ -600,6 +700,7 @@ static unsigned char swap_entry_free(str
 
 	/* free if no reference */
 	if (!usage) {
+		dec_cluster_info_page(p, p->cluster_info, offset);
 		if (offset < p->lowest_bit)
 			p->lowest_bit = offset;
 		if (offset > p->highest_bit)

@@ -1497,6 +1598,7 @@ static int setup_swap_extents(struct swa
 
 static void _enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	int i, prev;

@@ -1506,6 +1608,7 @@ static void _enable_swap_info(struct swa
 	else
 		p->prio = --least_priority;
 	p->swap_map = swap_map;
+	p->cluster_info = cluster_info;
 	frontswap_map_set(p, frontswap_map);
 	p->flags |= SWP_WRITEOK;
 	atomic_long_add(p->pages, &nr_swap_pages);

@@ -1527,11 +1630,12 @@ static void _enable_swap_info(struct swa
 
 static void enable_swap_info(struct swap_info_struct *p, int prio,
 				unsigned char *swap_map,
+				unsigned int *cluster_info,
 				unsigned long *frontswap_map)
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, prio, swap_map, frontswap_map);
+	_enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 	frontswap_init(p->type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);

@@ -1541,7 +1645,8 @@ static void reinsert_swap_info(struct sw
 {
 	spin_lock(&swap_lock);
 	spin_lock(&p->lock);
-	_enable_swap_info(p, p->prio, p->swap_map, frontswap_map_get(p));
+	_enable_swap_info(p, p->prio, p->swap_map, p->cluster_info,
+					frontswap_map_get(p));
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 }

@@ -1550,6 +1655,7 @@ SYSCALL_DEFINE1(swapoff, const char __us
 {
 	struct swap_info_struct *p = NULL;
 	unsigned char *swap_map;
+	unsigned int *cluster_info;
 	struct file *swap_file, *victim;
 	struct address_space *mapping;
 	struct inode *inode;

@@ -1648,12 +1754,15 @@ SYSCALL_DEFINE1(swapoff, const char __us
 	p->max = 0;
 	swap_map = p->swap_map;
 	p->swap_map = NULL;
+	cluster_info = p->cluster_info;
+	p->cluster_info = NULL;
 	p->flags = 0;
 	frontswap_invalidate_area(type);
 	spin_unlock(&p->lock);
 	spin_unlock(&swap_lock);
 	mutex_unlock(&swapon_mutex);
 	vfree(swap_map);
+	vfree(cluster_info);
 	vfree(frontswap_map_get(p));
 	/* Destroy swap account informatin */
 	swap_cgroup_swapoff(type);

@@ -1966,15 +2075,21 @@ static unsigned long read_swap_header(st
 static int setup_swap_map_and_extents(struct swap_info_struct *p,
 					union swap_header *swap_header,
 					unsigned char *swap_map,
+					unsigned int *cluster_info,
 					unsigned long maxpages,
 					sector_t *span)
 {
 	int i;
 	unsigned int nr_good_pages;
 	int nr_extents;
+	unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);
+	unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
 
 	nr_good_pages = maxpages - 1;	/* omit header page */
 
+	p->free_cluster_head = CLUSTER_NULL;
+	p->free_cluster_tail = CLUSTER_NULL;
+
 	for (i = 0; i < swap_header->info.nr_badpages; i++) {
 		unsigned int page_nr = swap_header->info.badpages[i];
 		if (page_nr == 0 || page_nr > swap_header->info.last_page)

@@ -1982,11 +2097,25 @@ static int setup_swap_map_and_extents(st
 		if (page_nr < maxpages) {
 			swap_map[page_nr] = SWAP_MAP_BAD;
 			nr_good_pages--;
+			/*
+			 * Not mark the cluster free yet, no list
+			 * operation involved
+			 */
+			inc_cluster_info_page(p, cluster_info, page_nr);
 		}
 	}
 
+	/* Not mark the cluster free yet, no list operation involved */
+	for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
+		inc_cluster_info_page(p, cluster_info, i);
+
 	if (nr_good_pages) {
 		swap_map[0] = SWAP_MAP_BAD;
+		/*
+		 * Not mark the cluster free yet, no list
+		 * operation involved
+		 */
+		inc_cluster_info_page(p, cluster_info, 0);
 		p->max = maxpages;
 		p->pages = nr_good_pages;
 		nr_extents = setup_swap_extents(p, span);

@@ -1999,6 +2128,27 @@ static int setup_swap_map_and_extents(st
 		return -EINVAL;
 	}
 
+	if (!cluster_info)
+		return nr_extents;
+
+	for (i = 0; i < nr_clusters; i++) {
+		if (!cluster_count(cluster_info[idx])) {
+			cluster_set_flag(cluster_info[idx], CLUSTER_FLAG_FREE);
+			if (p->free_cluster_head == CLUSTER_NULL) {
+				p->free_cluster_head = idx;
+				p->free_cluster_tail = idx;
+			} else {
+				cluster_set_next(
+					cluster_info[p->free_cluster_tail],
+					idx);
+				p->free_cluster_tail = idx;
+			}
+		}
+		idx++;
+		if (idx == nr_clusters)
+			idx = 0;
+	}
+
 	return nr_extents;
 }

@@ -2016,6 +2166,7 @@ SYSCALL_DEFINE2(swapon, const char __use
 	sector_t span;
 	unsigned long maxpages;
 	unsigned char *swap_map = NULL;
+	unsigned int *cluster_info = NULL;
 	unsigned long *frontswap_map = NULL;
 	struct page *page = NULL;
 	struct inode *inode = NULL;

@@ -2089,13 +2240,24 @@ SYSCALL_DEFINE2(swapon, const char __use
 		error = -ENOMEM;
 		goto bad_swap;
 	}
+	if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
+		p->flags |= SWP_SOLIDSTATE;
+		p->cluster_next = 1 + (random32() % p->highest_bit);
+
+		cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
+			SWAPFILE_CLUSTER) * sizeof(*cluster_info));
+		if (!cluster_info) {
+			error = -ENOMEM;
+			goto bad_swap;
+		}
+	}
 
 	error = swap_cgroup_swapon(p->type, maxpages);
 	if (error)
 		goto bad_swap;
 
 	nr_extents = setup_swap_map_and_extents(p, swap_header, swap_map,
-		maxpages, &span);
+		cluster_info, maxpages, &span);
 	if (unlikely(nr_extents < 0)) {
 		error = nr_extents;
 		goto bad_swap;

@@ -2104,21 +2266,15 @@ SYSCALL_DEFINE2(swapon, const char __use
 	if (frontswap_enabled)
 		frontswap_map = vzalloc(maxpages / sizeof(long));
 
-	if (p->bdev) {
-		if (blk_queue_nonrot(bdev_get_queue(p->bdev))) {
-			p->flags |= SWP_SOLIDSTATE;
-			p->cluster_next = 1 + (random32() % p->highest_bit);
-		}
-		if ((swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
-			p->flags |= SWP_DISCARDABLE;
-	}
+	if (p->bdev && (swap_flags & SWAP_FLAG_DISCARD) && discard_swap(p) == 0)
+		p->flags |= SWP_DISCARDABLE;
 
 	mutex_lock(&swapon_mutex);
 	prio = -1;
 	if (swap_flags & SWAP_FLAG_PREFER)
 		prio =
 		  (swap_flags & SWAP_FLAG_PRIO_MASK) >> SWAP_FLAG_PRIO_SHIFT;
-	enable_swap_info(p, prio, swap_map, frontswap_map);
+	enable_swap_info(p, prio, swap_map, cluster_info, frontswap_map);
 
 	printk(KERN_INFO "Adding %uk swap on %s.  "
 			"Priority:%d extents:%d across:%lluk %s%s%s\n",

mm/swapfile.c bad_swap: · line 2310

Quoted 7 lines

mm/swapfile.c +1 -0

@@ -2148,6 +2304,7 @@ bad_swap:
 	p->flags = 0;
 	spin_unlock(&swap_lock);
 	vfree(swap_map);
+	vfree(cluster_info);
 	if (swap_file) {
 		if (inode && S_ISREG(inode->i_mode)) {
 			mutex_unlock(&inode->i_mutex);

Andrew Morton

maintainer

2013-03-20 20:58 UTC #12 lore raw

Quoted 3 lines · On Tue, 19 Mar 2013 13:50:57 -0700 (PDT) Hugh Dickins <[email protected]> wrote:

I find it a bit confusing that we now have these two different clustering strategies in scan_swap_map(), one for SSD and one for the rest; and it's not immediately obvious what's used for what.

Yes, having two separation allocation paths is bad and we should work to avoid it, please. Sooner rather than later (which sometimes never comes).

We have a few theories about how the SSD code will worsen things for rotating disks. But have those theories been tested? Any performance results? If regressions *are* observed, what is the feasibility of fixing them up?

Shaohua Li

author

2013-03-21 02:02 UTC #13 lore raw

Quoted 6 lines · Andrew Morton

Quoted 3 lines · On Tue, 19 Mar 2013 13:50:57 -0700 (PDT) Hugh Dickins <[email protected]> wrote:
I find it a bit confusing that we now have these two different clustering strategies in scan_swap_map(), one for SSD and one for the rest; and it's not immediately obvious what's used for what.

Quoted 8 lines

Yes, having two separation allocation paths is bad and we should work to avoid it, please. Sooner rather than later (which sometimes never comes).
We have a few theories about how the SSD code will worsen things for rotating disks. But have those theories been tested? Any performance results? If regressions *are* observed, what is the feasibility of fixing them up?

The problem is I don't know which workload is proper to measure the cluster change impact for rotating disks. That would only happen when swap is fragmented but not too much. I'd assume the impact isn't big, but who knows. If you have something I can test, I'm happy to do.

Will address other issues soon.

Thanks,
Shaohua

Andrew Morton

maintainer

2013-03-20 20:37 UTC #14 lore raw

In reply to Shaohua Li

Quoted 38 lines · On Thu, 21 Feb 2013 10:17:10 +0800 Shaohua Li <[email protected]> wrote:

I'm using a fast SSD to do swap. scan_swap_map() sometimes uses up to 20~30% CPU time (when cluster is hard to find, the CPU time can be up to 80%), which becomes a bottleneck. scan_swap_map() scans a byte array to search a 256 page cluster, which is very slow.
Here I introduced a simple algorithm to search cluster. Since we only care about 256 pages cluster, we can just use a counter to track if a cluster is free. Every 256 pages use one int to store the counter. If the counter of a cluster is 0, the cluster is free. All free clusters will be added to a list, so searching cluster is very efficient. With this, scap_swap_map() overhead disappears.
Since searching cluster with a list is easy, we can easily implement a per-cpu cluster algorithm to do block allocation, which can make swapout more efficient. This is in my TODO list.
This might help low end SD card swap too. Because if the cluster is aligned, SD firmware can do flash erase more efficiently.
We only enable the algorithm for SSD. Hard disk swap isn't fast enough and has downside with the algorithm which might introduce regression (see below).
The patch slightly changes which cluster is choosen. It always adds free cluster to list tail. This can help wear leveling for low end SSD too. And if no cluster found, the scan_swap_map() will do search from the end of last cluster. So if no cluster found, the scan_swap_map() will do search from the end of last free cluster, which is random. For SSD, this isn't a problem at all.
Another downside is the cluster must be aligned to 256 pages, which will reduce the chance to find a cluster. I would expect this isn't a big problem for SSD because of the non-seek penality. (And this is the reason I only enable the algorithm for SSD).
...

mm/swapfile.c static int wait_for_discard(void *word) · line 193

Quoted 5 lines

mm/swapfile.c +3 -0

...
 
+#define CLUSTER_FLAG_FREE (1 << 0)
+#define CLUSTER_FLAG_NEXT_NULL (1 << 1)
+#define CLUSTER_NULL (CLUSTER_FLAG_NEXT_NULL << 24)

Some code comments describing the above wouldn't hurt.

mm/swapfile.c static int wait_for_discard(void *word) · line 203

Quoted 10 lines

mm/swapfile.c +10 -0

+#define cluster_flag(info) ((info) >> 24)
+#define cluster_set_flag(info, flag) \
+	do { info = ((info) & 0xffffff) | ((flag) << 24); } while (0)
+#define cluster_count(info) ((info) & 0xffffff)
+#define cluster_set_count(info, c) \
+	do { info = (cluster_flag(info) << 24) | (c); } while (0)
+#define cluster_next(info) ((info) & 0xffffff)
+#define cluster_set_next(info, n) \
+	do { info = (cluster_flag(info) << 24) | (n); } while (0)
+#define cluster_is_free(info) (cluster_flag(info) & CLUSTER_FLAG_FREE)

All the above can and should be implemented in C, please. Doing so will

a) improve readability.

b) increase the likelihood of them being documented (why does this happen?)

c) provide additional typesafety and

d) fix potential bugs when the "function" is passed an expression with side-effects. Three instances of this problem were added here.

Quoted 4 lines

...
static int setup_swap_map_and_extents(struct swap_info_struct *p,

mm/swapfile.c static unsigned long read_swap_header(st · line 2085

Quoted 10 lines

mm/swapfile.c +2 -0

 					union swap_header *swap_header,
 					unsigned char *swap_map,
+					unsigned int *cluster_info,
 					unsigned long maxpages,
 					sector_t *span)
 {
 	int i;
 	unsigned int nr_good_pages;
 	int nr_extents;
+	unsigned long nr_clusters = DIV_ROUND_UP(maxpages, SWAPFILE_CLUSTER);

Rounding up seems wrong. Will nr_clusters be off-by-one if maxpages is not a multiple of SWAPFILE_CLUSTER?

Quoted 11 lines

mm/swapfile.c +4 -0

+	unsigned long idx = p->cluster_next / SWAPFILE_CLUSTER;
 
 	nr_good_pages = maxpages - 1;	/* omit header page */
 
+	p->free_cluster_head = CLUSTER_NULL;
+	p->free_cluster_tail = CLUSTER_NULL;
+
 	for (i = 0; i < swap_header->info.nr_badpages; i++) {
 		unsigned int page_nr = swap_header->info.badpages[i];
 		if (page_nr == 0 || page_nr > swap_header->info.last_page)

mm/swapfile.c static int setup_swap_map_and_extents(st · line 2101

Quoted 5 lines

mm/swapfile.c +2 -0

@@ -1982,11 +2097,25 @@ static int setup_swap_map_and_extents(st
 		if (page_nr < maxpages) {
 			swap_map[page_nr] = SWAP_MAP_BAD;
 			nr_good_pages--;
+			/*
+			 * Not mark the cluster free yet, no list

s/Not/Don't/

mm/swapfile.c static int setup_swap_map_and_extents(st · line 2108

Quoted 7 lines

mm/swapfile.c +4 -0

+			 * operation involved
+			 */
+			inc_cluster_info_page(p, cluster_info, page_nr);
 		}
 	}
 
+	/* Not mark the cluster free yet, no list operation involved */

s/Not/Won't/

Quoted 17 lines

mm/swapfile.c +8 -0

+	for (i = maxpages; i < round_up(maxpages, SWAPFILE_CLUSTER); i++)
+		inc_cluster_info_page(p, cluster_info, i);
+
 	if (nr_good_pages) {
 		swap_map[0] = SWAP_MAP_BAD;
+		/*
+		 * Not mark the cluster free yet, no list
+		 * operation involved
+		 */
+		inc_cluster_info_page(p, cluster_info, 0);
 		p->max = maxpages;
 		p->pages = nr_good_pages;
 		nr_extents = setup_swap_extents(p, span);
 
...

mm/swapfile.c SYSCALL_DEFINE2(swapon, const char __use · line 2245

Quoted 6 lines

mm/swapfile.c +3 -0

@@ -2089,13 +2240,24 @@ SYSCALL_DEFINE2(swapon, const char __use
 		error = -ENOMEM;
 		goto bad_swap;
 	}
+	if (p->bdev && blk_queue_nonrot(bdev_get_queue(p->bdev))) {
+		p->flags |= SWP_SOLIDSTATE;
+		p->cluster_next = 1 + (random32() % p->highest_bit);

The random32() usage was unchangelogged and uncommented. A comment would be better, please. Explain the concept, not the code.

Also, random32 was removed. Use prandom_u32().

mm/swapfile.c SYSCALL_DEFINE2(swapon, const char __use · line 2254

Quoted 10 lines

mm/swapfile.c +7 -0

+		cluster_info = vzalloc(DIV_ROUND_UP(maxpages,
+			SWAPFILE_CLUSTER) * sizeof(*cluster_info));
+		if (!cluster_info) {
+			error = -ENOMEM;
+			goto bad_swap;
+		}
+	}
 
...