Skip to contentSkip to conversation

mm/swap: remove plist swap_active_head

PATCH In review Baoquan He 4 participants 22 messages Oct 1, 2025 – Oct 10, 2025

Subsystems (2)
lore
Share & tools
lore

Baoquan He

author
#0 lore raw

cover mm/swap: remove plist swap_active_head

In mm/swapfile.c, there are two plist variables, swap_active_head and swap_avail_head. swap_avail_head contains all available (active, not full) swap_info_structs orderred by priority. While swap_active_head contains all active swap_info_structs (active, nor full and full) orderred by priority. Earlier, it serves three purposes:

1) When swapoff one swap device in the middle, swap devices of priority
lower than the swapped off swap device will be promoted up value one,
e.g, I swapped off zram1 of priority '-3', then zram2 promoted to
have priority '-3', zram3 has priority '-4'. This is done via plist swap_active_head.
- This code has been taken off in
- [PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

# swapon
NAME       TYPE      SIZE USED PRIO
/dev/zram0 partition  16G   0B   -2
/dev/zram1 partition  16G   0B   -3
/dev/zram2 partition  16G   0B   -4
/dev/zram3 partition  16G   0B   -5
# swapoff /dev/zram1
# swapon
NAME       TYPE      SIZE USED PRIO
/dev/zram0 partition  16G   0B   -2
/dev/zram2 partition  16G   0B   -3
/dev/zram3 partition  16G   0B   -4

2) Find a swap device in swap_active_head when swap off.
- This can be done through iterating swap_info[] instead. Change is done in patch 2.

3) Judge if there's any active swap device via __has_usable_swap().
- This can be done by checking total_swap_pages instead. Change is done in patch 1.

Among them, the purpose 1) is the most important, while it has been
taken off in below patchset. So this patchset removing swap_active_head sits on top of it.
- [PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin
- https://lore.kernel.org/all/[email protected]/T/#u

Baoquan He (3):
mm/swapfile.c: remove __has_usable_swap() mm/swapfile.c: use swap_info[] to find the swap device mm/swap: remove unneeded swap_active_head

 include/linux/swap.h |  1 -
 mm/swapfile.c        | 31 ++++++++-----------------------
 2 files changed, 8 insertions(+), 24 deletions(-)
Signature · 1 line
2.41.0
author
#1 lore raw

PATCH 1/3 mm/swapfile.c: remove __has_usable_swap()

Global total_swap_pages can indicate if usable swap deivice is present, use it instead of __has_usable_swap().

---
 mm/swapfile.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)
mm/swapfile.c +1 -6
index 2bd8bd76ea28..5d71c748a2fe 100644
@@ -3836,11 +3836,6 @@ static void free_swap_count_continuations(struct swap_info_struct *si)
 }
 
 #if defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP)
-static bool __has_usable_swap(void)
-{
-	return !plist_head_empty(&swap_active_head);
-}
-
 void __folio_throttle_swaprate(struct folio *folio, gfp_t gfp)
 {
 	struct swap_info_struct *si, *next;
@@ -3848,7 +3843,7 @@ void __folio_throttle_swaprate(struct folio *folio, gfp_t gfp)
 	if (!(gfp & __GFP_IO))
 		return;
 
-	if (!__has_usable_swap())
+	if (!total_swap_pages)
 		return;
 
 	if (!blk_cgroup_congested())
Signature · 1 line
2.41.0
author
#2 lore raw

PATCH 2/3 mm/swapfile.c: use swap_info[] to find the swap device

In reply to Baoquan He

Now, swap_active_head is only used to find a present swap device when trying to swapoff it. In fact, swap_info[] is a short array which is 32 at maximum, and usually the unused one can be reused, so the searching for target mostly only iterates the foremost several used slots. And swapoff is a rarely used operation, efficiency is not so important. Then it's unnecessary to get a plist to make it.

Here go by iterating swap_info[] to find the swap device instead of iterating swap_active_head.

---
 mm/swapfile.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
mm/swapfile.c +3 -1
index 5d71c748a2fe..18b52cc20749 100644
@@ -2641,6 +2641,7 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
 	struct inode *inode;
 	struct filename *pathname;
 	int err, found = 0;
+	unsigned int type;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
@@ -2658,7 +2659,8 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
 
 	mapping = victim->f_mapping;
 	spin_lock(&swap_lock);
-	plist_for_each_entry(p, &swap_active_head, list) {
+	for (type = 0; type < nr_swapfiles; type++) {
+		p = swap_info[type];
 		if (p->flags & SWP_WRITEOK) {
 			if (p->swap_file->f_mapping == mapping) {
 				found = 1;
Signature · 1 line
2.41.0
maintainer
#3 lore raw

Re: [PATCH 2/3] mm/swapfile.c: use swap_info[] to find the swap device

Quoted 29 lines · Baoquan He

Now, swap_active_head is only used to find a present swap device when trying to swapoff it. In fact, swap_info[] is a short array which is 32 at maximum, and usually the unused one can be reused, so the searching for target mostly only iterates the foremost several used slots. And swapoff is a rarely used operation, efficiency is not so important. Then it's unnecessary to get a plist to make it.

Here go by iterating swap_info[] to find the swap device instead of iterating swap_active_head.

---
 mm/swapfile.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
mm/swapfile.c +1 -0
index 5d71c748a2fe..18b52cc20749 100644
@@ -2641,6 +2641,7 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
        struct inode *inode;
        struct filename *pathname;
        int err, found = 0;
+       unsigned int type;
 
        if (!capable(CAP_SYS_ADMIN))
                return -EPERM;

mm/swapfile.c SYSCALL_DEFINE1(swapoff, const char __user *, specialfile) · line 2663

Quoted 6 lines
mm/swapfile.c +2 -1
@@ -2658,7 +2659,8 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
 
        mapping = victim->f_mapping;
        spin_lock(&swap_lock);
-       plist_for_each_entry(p, &swap_active_head, list) {
+       for (type = 0; type < nr_swapfiles; type++) {
+               p = swap_info[type];

After some careful thinking of the swapoff behavior, I now consider this patch buggy.

On behavior change is that, previously the swapoff is dependent on the swap_active_list and only swapon device is on that list.

Now you change to enumerate every device at swap off and depend on the p->flags & SWP_WRITEOK to filter out the device. Which will go through all 32 swap devices regardless.

mm/swapfile.c SYSCALL_DEFINE1(swapoff, const char __user *, specialfile) · line 2665

Quoted 2 lines
mm/swapfile.c +0 -0
                if (p->flags & SWP_WRITEOK) {
                        if (p->swap_file->f_mapping == mapping) {

Considering the following race:

CPU1:
swapoff( "/dev/sda1")
spin_lock(&swap_lock);

for (type = 0; type < nr_swapfiles; type++) {
     if (p->flags & SWP_WRITEOK) {
         // found the device.
}

....
spin_lock(&p->lock);
del_from_avail_list(p, true);
plist_del(&p->list, &swap_active_head);
atomic_long_sub(p->pages, &nr_swap_pages);
total_swap_pages -= p->pages;
spin_unlock(&p->lock);
spin_unlock(&swap_lock);

// at this point the p->flags still have SWP_WRITEOK. // swap_lock and p->lock are both unlocked.

Now the second CPU swapoff race begins

                                                       CPU2:
                                                       swapoff( "/dev/sda1")

spin_lock(&swap_lock); // can take this lock

                                                       for (type = 0;
type < nr_swapfiles; type++) {
                                                       if (p->flags &
SWP_WRITEOK) {
                                                              // also
found the device. because p->flags are not clear to 0 yet.
                                                       }
                                                       ....

spin_lock(&p->lock); // can also take this lock

del_from_avail_list(p, true);

plist_del(&p->list, &swap_active_head); <=== blow up here

                                                       // remove p not

the in list.

atomic_long_sub(p->pages, &nr_swap_pages);
                                                       // double subtracting ..
total_swap_pages -= p->pages;
                                                       // double
subtracting as well.
                                                       spin_unlock(&p->lock);
                                                       spin_unlock(&swap_lock);

Please confirm or deny if that race is possible.
If it is, consider it a NACK to this patch.

Please consider that even trivial patches can have subtle behavior change and it has non zero cost to reviewers. It can introduce subtle bugs unintentionally as well. In this case, I feel that the gain(removing a list, total less than 1K) is too small, not a very good return on investment.

Chris

author
#4 lore raw

Re: [PATCH 2/3] mm/swapfile.c: use swap_info[] to find the swap device

Quoted 56 lines · Chris Li
Quoted 29 lines · Baoquan He

Now, swap_active_head is only used to find a present swap device when trying to swapoff it. In fact, swap_info[] is a short array which is 32 at maximum, and usually the unused one can be reused, so the searching for target mostly only iterates the foremost several used slots. And swapoff is a rarely used operation, efficiency is not so important. Then it's unnecessary to get a plist to make it.

Here go by iterating swap_info[] to find the swap device instead of iterating swap_active_head.

---
 mm/swapfile.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
mm/swapfile.c +1 -0
index 5d71c748a2fe..18b52cc20749 100644
@@ -2641,6 +2641,7 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
        struct inode *inode;
        struct filename *pathname;
        int err, found = 0;
+       unsigned int type;
 
        if (!capable(CAP_SYS_ADMIN))
                return -EPERM;

mm/swapfile.c SYSCALL_DEFINE1(swapoff, const char __user *, specialfile) · line 2663

Quoted 6 lines
mm/swapfile.c +2 -1
@@ -2658,7 +2659,8 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
 
        mapping = victim->f_mapping;
        spin_lock(&swap_lock);
-       plist_for_each_entry(p, &swap_active_head, list) {
+       for (type = 0; type < nr_swapfiles; type++) {
+               p = swap_info[type];

After some careful thinking of the swapoff behavior, I now consider this patch buggy.

On behavior change is that, previously the swapoff is dependent on the swap_active_list and only swapon device is on that list.

Now you change to enumerate every device at swap off and depend on the p->flags & SWP_WRITEOK to filter out the device. Which will go through all 32 swap devices regardless.

mm/swapfile.c SYSCALL_DEFINE1(swapoff, const char __user *, specialfile) · line 2665

Quoted 2 lines
mm/swapfile.c +0 -0
                if (p->flags & SWP_WRITEOK) {
                        if (p->swap_file->f_mapping == mapping) {

Considering the following race:

CPU1:
swapoff( "/dev/sda1")
spin_lock(&swap_lock);

Quoted 7 lines
for (type = 0; type < nr_swapfiles; type++) {
     if (p->flags & SWP_WRITEOK) {
         // found the device.
}

....
spin_lock(&p->lock);
del_from_avail_list(p, true);

~~~~~~~~~~~~~~~~~~~~~~~~~~~

Quoted 8 lines

plist_del(&p->list, &swap_active_head);
atomic_long_sub(p->pages, &nr_swap_pages);
total_swap_pages -= p->pages;
spin_unlock(&p->lock);
spin_unlock(&swap_lock);

// at this point the p->flags still have SWP_WRITEOK. // swap_lock and p->lock are both unlocked.

No, SWP_WRITEOK is cleared in del_from_avail_list(). At this point, there's no more SWP_WRITEOK in p->flags when both swap_lock and p->lock are lifted.

Quoted 45 lines

Now the second CPU swapoff race begins

                                                       CPU2:
                                                       swapoff( "/dev/sda1")

spin_lock(&swap_lock); // can take this lock

                                                       for (type = 0;
type < nr_swapfiles; type++) {
                                                       if (p->flags &
SWP_WRITEOK) {
                                                              // also
found the device. because p->flags are not clear to 0 yet.
                                                       }
                                                       ....

spin_lock(&p->lock); // can also take this lock

del_from_avail_list(p, true);

plist_del(&p->list, &swap_active_head); <=== blow up here

                                                       // remove p not

the in list.

atomic_long_sub(p->pages, &nr_swap_pages);
                                                       // double subtracting ..
total_swap_pages -= p->pages;
                                                       // double
subtracting as well.
                                                       spin_unlock(&p->lock);
                                                       spin_unlock(&swap_lock);

Please confirm or deny if that race is possible.
If it is, consider it a NACK to this patch.

Please consider that even trivial patches can have subtle behavior change and it has non zero cost to reviewers. It can introduce subtle bugs unintentionally as well. In this case, I feel that the gain(removing a list, total less than 1K) is too small, not a very good return on investment.

Chris

maintainer
#5 lore raw

Re: [PATCH 2/3] mm/swapfile.c: use swap_info[] to find the swap device

Quoted 72 lines · Baoquan He
Quoted 56 lines · Chris Li
Quoted 29 lines · Baoquan He

Now, swap_active_head is only used to find a present swap device when trying to swapoff it. In fact, swap_info[] is a short array which is 32 at maximum, and usually the unused one can be reused, so the searching for target mostly only iterates the foremost several used slots. And swapoff is a rarely used operation, efficiency is not so important. Then it's unnecessary to get a plist to make it.

Here go by iterating swap_info[] to find the swap device instead of iterating swap_active_head.

---
 mm/swapfile.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
mm/swapfile.c +1 -0
index 5d71c748a2fe..18b52cc20749 100644
@@ -2641,6 +2641,7 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
        struct inode *inode;
        struct filename *pathname;
        int err, found = 0;
+       unsigned int type;
 
        if (!capable(CAP_SYS_ADMIN))
                return -EPERM;

mm/swapfile.c SYSCALL_DEFINE1(swapoff, const char __user *, specialfile) · line 2663

Quoted 6 lines
mm/swapfile.c +2 -1
@@ -2658,7 +2659,8 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
 
        mapping = victim->f_mapping;
        spin_lock(&swap_lock);
-       plist_for_each_entry(p, &swap_active_head, list) {
+       for (type = 0; type < nr_swapfiles; type++) {
+               p = swap_info[type];

After some careful thinking of the swapoff behavior, I now consider this patch buggy.

On behavior change is that, previously the swapoff is dependent on the swap_active_list and only swapon device is on that list.

Now you change to enumerate every device at swap off and depend on the p->flags & SWP_WRITEOK to filter out the device. Which will go through all 32 swap devices regardless.

mm/swapfile.c SYSCALL_DEFINE1(swapoff, const char __user *, specialfile) · line 2665

Quoted 2 lines
mm/swapfile.c +0 -0
                if (p->flags & SWP_WRITEOK) {
                        if (p->swap_file->f_mapping == mapping) {

Considering the following race:

CPU1:
swapoff( "/dev/sda1")
spin_lock(&swap_lock);

Quoted 7 lines
for (type = 0; type < nr_swapfiles; type++) {
     if (p->flags & SWP_WRITEOK) {
         // found the device.
}

....
spin_lock(&p->lock);
del_from_avail_list(p, true);

~~~~~~~~~~~~~~~~~~~~~~~~~~~

Quoted 6 lines

plist_del(&p->list, &swap_active_head);
atomic_long_sub(p->pages, &nr_swap_pages);
total_swap_pages -= p->pages;
spin_unlock(&p->lock);
spin_unlock(&swap_lock);

Quoted 6 lines
Quoted 2 lines

// at this point the p->flags still have SWP_WRITEOK. // swap_lock and p->lock are both unlocked.

No, SWP_WRITEOK is cleared in del_from_avail_list(). At this point, there's no more SWP_WRITEOK in p->flags when both swap_lock and p->lock are lifted.

Ah, you are right. I am wrong. I missed the "si->flags &= ~SWP_WRITEOK;" inside the del_from_avail_list().

I read at the end of swapoff it set the si->flags = 0; thinking the SWP_WRITEOK was clear there. It is cleared earlier.

That should be fine. Sorry about the misread.

Quoted 29 lines
Quoted 29 lines

Now the second CPU swapoff race begins

                                                       CPU2:
                                                       swapoff( "/dev/sda1")

spin_lock(&swap_lock); // can take this lock

                                                       for (type = 0;
type < nr_swapfiles; type++) {
                                                       if (p->flags &
SWP_WRITEOK) {
                                                              // also
found the device. because p->flags are not clear to 0 yet.
                                                       }
                                                       ....

spin_lock(&p->lock); // can also take this lock

del_from_avail_list(p, true);

plist_del(&p->list, &swap_active_head); <=== blow up here

                                                       // remove p not

the in list.

atomic_long_sub(p->pages, &nr_swap_pages);
                                                       // double subtracting ..
Quoted 8 lines
Quoted 8 lines
total_swap_pages -= p->pages;
                                                       // double
subtracting as well.
                                                       spin_unlock(&p->lock);
                                                       spin_unlock(&swap_lock);

Please confirm or deny if that race is possible.
If it is, consider it a NACK to this patch.

I withdraw the NACK. Sorry about that.

Chris

author
#6 lore raw

Re: [PATCH 2/3] mm/swapfile.c: use swap_info[] to find the swap device

Quoted 80 lines · Chris Li
Quoted 72 lines · Baoquan He
Quoted 56 lines · Chris Li
Quoted 29 lines · Baoquan He

Now, swap_active_head is only used to find a present swap device when trying to swapoff it. In fact, swap_info[] is a short array which is 32 at maximum, and usually the unused one can be reused, so the searching for target mostly only iterates the foremost several used slots. And swapoff is a rarely used operation, efficiency is not so important. Then it's unnecessary to get a plist to make it.

Here go by iterating swap_info[] to find the swap device instead of iterating swap_active_head.

---
 mm/swapfile.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
mm/swapfile.c +1 -0
index 5d71c748a2fe..18b52cc20749 100644
@@ -2641,6 +2641,7 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
        struct inode *inode;
        struct filename *pathname;
        int err, found = 0;
+       unsigned int type;
 
        if (!capable(CAP_SYS_ADMIN))
                return -EPERM;

mm/swapfile.c SYSCALL_DEFINE1(swapoff, const char __user *, specialfile) · line 2663

Quoted 6 lines
mm/swapfile.c +2 -1
@@ -2658,7 +2659,8 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
 
        mapping = victim->f_mapping;
        spin_lock(&swap_lock);
-       plist_for_each_entry(p, &swap_active_head, list) {
+       for (type = 0; type < nr_swapfiles; type++) {
+               p = swap_info[type];

After some careful thinking of the swapoff behavior, I now consider this patch buggy.

On behavior change is that, previously the swapoff is dependent on the swap_active_list and only swapon device is on that list.

Now you change to enumerate every device at swap off and depend on the p->flags & SWP_WRITEOK to filter out the device. Which will go through all 32 swap devices regardless.

mm/swapfile.c SYSCALL_DEFINE1(swapoff, const char __user *, specialfile) · line 2665

Quoted 2 lines
mm/swapfile.c +0 -0
                if (p->flags & SWP_WRITEOK) {
                        if (p->swap_file->f_mapping == mapping) {

Considering the following race:

CPU1:
swapoff( "/dev/sda1")
spin_lock(&swap_lock);

Quoted 7 lines
for (type = 0; type < nr_swapfiles; type++) {
     if (p->flags & SWP_WRITEOK) {
         // found the device.
}

....
spin_lock(&p->lock);
del_from_avail_list(p, true);

~~~~~~~~~~~~~~~~~~~~~~~~~~~

Quoted 6 lines

plist_del(&p->list, &swap_active_head);
atomic_long_sub(p->pages, &nr_swap_pages);
total_swap_pages -= p->pages;
spin_unlock(&p->lock);
spin_unlock(&swap_lock);

Quoted 6 lines
Quoted 2 lines

// at this point the p->flags still have SWP_WRITEOK. // swap_lock and p->lock are both unlocked.

No, SWP_WRITEOK is cleared in del_from_avail_list(). At this point, there's no more SWP_WRITEOK in p->flags when both swap_lock and p->lock are lifted.

Quoted 7 lines

Ah, you are right. I am wrong. I missed the "si->flags &= ~SWP_WRITEOK;" inside the del_from_avail_list().

I read at the end of swapoff it set the si->flags = 0; thinking the SWP_WRITEOK was clear there. It is cleared earlier.

That should be fine. Sorry about the misread.

It's OK, thanks for careful reviewing and checking.

author
#7 lore raw

PATCH 3/3 mm/swap: remove unneeded swap_active_head

In reply to Baoquan He

There's no user of swap_active_head, remove it now.

---
 include/linux/swap.h |  1 -
 mm/swapfile.c        | 20 ++++----------------
 2 files changed, 4 insertions(+), 17 deletions(-)
include/linux/swap.h +0 -1
index 5b7a39b20f58..dfc0cc9fc166 100644
@@ -296,7 +296,6 @@ struct swap_info_struct {
 	struct percpu_ref users;	/* indicate and keep swap device valid. */
 	unsigned long	flags;		/* SWP_USED etc: see above */
 	signed short	prio;		/* swap priority of this type */
-	struct plist_node list;		/* entry in swap_active_head */
 	signed char	type;		/* strange name for an index */
 	unsigned int	max;		/* extent of the swap_map */
 	unsigned char *swap_map;	/* vmalloc'ed array of usage counts */
mm/swapfile.c +4 -16
index 18b52cc20749..80b34dc86a95 100644
@@ -85,17 +85,10 @@ static const char Bad_offset[] = "Bad swap offset entry ";
 static const char Unused_offset[] = "Unused swap offset entry ";
 
 /*
- * all active swap_info_structs
- * protected with swap_lock, and ordered by priority.
- */
-static PLIST_HEAD(swap_active_head);
-
-/*
- * all available (active, not full) swap_info_structs
- * protected with swap_avail_lock, ordered by priority.
- * This is used by folio_alloc_swap() instead of swap_active_head
- * because swap_active_head includes all swap_info_structs,
- * but folio_alloc_swap() doesn't need to look at full ones.
+ * All available (active, not full) swap_info_structs protected with
+ * swap_avail_lock, ordered by priority.
+ * This is used by folio_alloc_swap() because folio_alloc_swap()
+ * doesn't need to look at full ones.
  * This uses its own lock instead of swap_lock because when a
  * swap_info_struct changes between not-full/full, it needs to
  * add/remove itself to/from this list, but the swap_info_struct->lock
@@ -2539,7 +2532,6 @@ static void setup_swap_info(struct swap_info_struct *si, int prio,
 	 * the plist prio is negated because plist ordering is
 	 * low-to-high, while swap ordering is high-to-low
 	 */
-	si->list.prio = -si->prio;
 	si->avail_list.prio = -si->prio;
 	si->swap_map = swap_map;
 	si->cluster_info = cluster_info;
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
 	assert_spin_locked(&swap_lock);
 
-	plist_add(&si->list, &swap_active_head);
-
 	/* Add back to available list */
 	add_to_avail_list(si, true);
 }
@@ -2682,7 +2672,6 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
 	}
 	spin_lock(&p->lock);
 	del_from_avail_list(p, true);
-	plist_del(&p->list, &swap_active_head);
 	atomic_long_sub(p->pages, &nr_swap_pages);
 	total_swap_pages -= p->pages;
 	spin_unlock(&p->lock);
@@ -2958,7 +2947,6 @@ static struct swap_info_struct *alloc_swap_info(void)
 		 */
 	}
 	p->swap_extent_root = RB_ROOT;
-	plist_node_init(&p->list, 0);
 	plist_node_init(&p->avail_list, 0);
 	p->flags = SWP_USED;
 	spin_unlock(&swap_lock);
Signature · 1 line
2.41.0
maintainer
#8 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

Quoted 2 lines · Baoquan He

There's no user of swap_active_head, remove it now.

Removing swap_active_head for 32 swap devices is less than 1K memory saving? Only the power off code uses it now so the reduction of complexity is very limited as well. I consider this series payout is pretty small but has considerable review overhead.

Another consideration is that swap.tiers is possible to use the swap_active_head to find swap devices belonging to a tier, because the swap tier is a range of priority. We can revisit it after the initial implementation for swap.tiers.

Let's hold off the removal of swap_active_head a bit.

Chris

Quoted 81 lines
---
 include/linux/swap.h |  1 -
 mm/swapfile.c        | 20 ++++----------------
 2 files changed, 4 insertions(+), 17 deletions(-)
include/linux/swap.h +0 -1
index 5b7a39b20f58..dfc0cc9fc166 100644
@@ -296,7 +296,6 @@ struct swap_info_struct {
        struct percpu_ref users;        /* indicate and keep swap device valid. */
        unsigned long   flags;          /* SWP_USED etc: see above */
        signed short    prio;           /* swap priority of this type */
-       struct plist_node list;         /* entry in swap_active_head */
        signed char     type;           /* strange name for an index */
        unsigned int    max;            /* extent of the swap_map */
        unsigned char *swap_map;        /* vmalloc'ed array of usage counts */
mm/swapfile.c +4 -17
index 18b52cc20749..80b34dc86a95 100644
@@ -85,17 +85,10 @@ static const char Bad_offset[] = "Bad swap offset entry ";
 static const char Unused_offset[] = "Unused swap offset entry ";
 
 /*
- * all active swap_info_structs
- * protected with swap_lock, and ordered by priority.
- */
-static PLIST_HEAD(swap_active_head);
-
-/*
- * all available (active, not full) swap_info_structs
- * protected with swap_avail_lock, ordered by priority.
- * This is used by folio_alloc_swap() instead of swap_active_head
- * because swap_active_head includes all swap_info_structs,
- * but folio_alloc_swap() doesn't need to look at full ones.
+ * All available (active, not full) swap_info_structs protected with
+ * swap_avail_lock, ordered by priority.
+ * This is used by folio_alloc_swap() because folio_alloc_swap()
+ * doesn't need to look at full ones.
  * This uses its own lock instead of swap_lock because when a
  * swap_info_struct changes between not-full/full, it needs to
  * add/remove itself to/from this list, but the swap_info_struct->lock
@@ -2539,7 +2532,6 @@ static void setup_swap_info(struct swap_info_struct *si, int prio,
         * the plist prio is negated because plist ordering is
         * low-to-high, while swap ordering is high-to-low
         */
-       si->list.prio = -si->prio;
        si->avail_list.prio = -si->prio;
        si->swap_map = swap_map;
        si->cluster_info = cluster_info;
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
        assert_spin_locked(&swap_lock);
 
-       plist_add(&si->list, &swap_active_head);
-
        /* Add back to available list */
        add_to_avail_list(si, true);
 }
@@ -2682,7 +2672,6 @@ SYSCALL_DEFINE1(swapoff, const char __user *, specialfile)
        }
        spin_lock(&p->lock);
        del_from_avail_list(p, true);
-       plist_del(&p->list, &swap_active_head);
        atomic_long_sub(p->pages, &nr_swap_pages);
        total_swap_pages -= p->pages;
        spin_unlock(&p->lock);
@@ -2958,7 +2947,6 @@ static struct swap_info_struct *alloc_swap_info(void)
                 */
        }
        p->swap_extent_root = RB_ROOT;
-       plist_node_init(&p->list, 0);
        plist_node_init(&p->avail_list, 0);
        p->flags = SWP_USED;
        spin_unlock(&swap_lock);
--

2.41.0

author
#9 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

Quoted 9 lines · Chris Li
Quoted 2 lines · Baoquan He

There's no user of swap_active_head, remove it now.

Removing swap_active_head for 32 swap devices is less than 1K memory saving? Only the power off code uses it now so the reduction of complexity is very limited as well. I consider this series payout is pretty small but has considerable review overhead.

Quoted 6 lines

Another consideration is that swap.tiers is possible to use the swap_active_head to find swap devices belonging to a tier, because the swap tier is a range of priority. We can revisit it after the initial implementation for swap.tiers.

Let's hold off the removal of swap_active_head a bit.

OK, if it has a potential user, agree we can hold it off for now. Thanks for careful reviewing.

maintainer
#10 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

In reply to Baoquan He

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2557

Quoted 10 lines · On Wed, 1 Oct 2025 12:34:36 +0800 Baoquan He <[email protected]> wrote:

There's no user of swap_active_head, remove it now.

...

mm/swapfile.c +0 -2
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
 	assert_spin_locked(&swap_lock);
 
-	plist_add(&si->list, &swap_active_head);
-

There's a large comment in _enable_swap_info() which needs to be removed or updated?

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2550

Quoted 3 lines
mm/swapfile.c +0 -0
 	/* Add back to available list */
 	add_to_avail_list(si, true);
 }

I fixed the rejects and my swapfile.c still has a couple of references to swap_active_head. Can we please have a redo&resend against mm-new?

author
#11 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

Quoted 8 lines · Andrew Morton
Quoted 6 lines · On Wed, 1 Oct 2025 12:34:36 +0800 Baoquan He <[email protected]> wrote:

There's no user of swap_active_head, remove it now.

...

(quoted diff) +0 -0
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
Quoted 7 lines

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2557

Quoted 4 lines
mm/swapfile.c +0 -2
 	assert_spin_locked(&swap_lock);
 
-	plist_add(&si->list, &swap_active_head);
-

There's a large comment in _enable_swap_info() which needs to be removed or updated?

This patchset depends on below patchset:
[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

That large comment in _enable_swap_info() has been removed in "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin"

Quoted 7 lines

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2550

Quoted 4 lines
mm/swapfile.c +0 -0
 	/* Add back to available list */
 	add_to_avail_list(si, true);
 }

I fixed the rejects and my swapfile.c still has a couple of references to swap_active_head. Can we please have a redo&resend against mm-new?

not sure if it's caused by the dependency on "[PATCH v3 0/2] mm/swapfile.c:
select the swap device with default priority round robin". I will rebase it on mm-new to see.

By the way, Chris worried the plist swap_active_head could be reused by the ongoing swap-tier work and suggested to hold off this patchset till swap-tier work is clear. Now seems it won't impact swap-tier, I will send a v2 against mm-new.

Thanks a lot.

maintainer
#12 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

Hi Baoquan,

Quoted 5 lines · Baoquan He

By the way, Chris worried the plist swap_active_head could be reused by the ongoing swap-tier work and suggested to hold off this patchset till swap-tier work is clear. Now seems it won't impact swap-tier, I will send a v2 against mm-new.

Sorry I haven't made it clear earlier. I withdraw the NACK because that is my misunderstanding of the code behavior. It does not mean I clear the swap tiers will not use swap_active_head yet, I actually haven't done that investigation.

I still suggest deferring this series which deletes the "swap_active_head". Wait until YoungPark's swap tiers to be sent out on the list. Let's wait for about 1 month or so, we should have a much better understanding if the swap tiers series needs to use the "swap_active_head". We are not in a hurry to remove it, right? The earliest this can go into the official tree is 6.19 anyway. We still have some time.

YoungPark, do you know if you want to use "swap_active_head" in your pending patch series? Even better send your pending RFC patches to the list and Baoquan can collaborate on it.

Chris

reviewer
#13 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

Quoted 9 lines · Chris Li

Hi Baoquan,

Quoted 5 lines · Baoquan He

By the way, Chris worried the plist swap_active_head could be reused by the ongoing swap-tier work and suggested to hold off this patchset till swap-tier work is clear. Now seems it won't impact swap-tier, I will send a v2 against mm-new.

Quoted 7 lines

Sorry I haven't made it clear earlier. I withdraw the NACK because that is my misunderstanding of the code behavior. It does not mean I clear the swap tiers will not use swap_active_head yet, I actually haven't done that investigation.

I still suggest deferring this series which deletes the "swap_active_head". Wait until YoungPark's swap tiers to be sent out

Thanks Chris.
"swap_active_head" is used on my current working patch.

Quoted 1 line

on the list. Let's wait for about 1 month or so, we should have a much

Yes, I am on the construct writing swap tier RFC.
it takes a few weeks to share on LKML.

Quoted 5 lines

better understanding if the swap tiers series needs to use the "swap_active_head". We are not in a hurry to remove it, right? The earliest this can go into the official tree is 6.19 anyway. We still have some time.
YoungPark, do you know if you want to use "swap_active_head" in your

On current version of my patch, I use "swap_active_head" on swap_tier assignment logic when traversing current swapfile. It is efficient to use "swap_active_head" on traversing the swapfile in priority order.

I think about more on this part.

Quoted 2 lines

pending patch series? Even better send your pending RFC patches to the list and Baoquan can collaborate on it.

As mentioned above, share it a few weeks later.

Thanks,
Youngjun Park

maintainer
#14 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

In reply to Baoquan He
Quoted 11 lines · On Thu, 9 Oct 2025 15:47:06 +0800 Baoquan He <[email protected]> wrote:
Quoted 10 lines · Andrew Morton
Quoted 8 lines · On Wed, 1 Oct 2025 12:34:36 +0800 Baoquan He <[email protected]> wrote:

There's no user of swap_active_head, remove it now.

...

mm/swapfile.c +0 -0
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
 	assert_spin_locked(&swap_lock);
 
Quoted 8 lines
Quoted 5 lines

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2557

Quoted 2 lines
mm/swapfile.c +0 -2
-	plist_add(&si->list, &swap_active_head);
-

There's a large comment in _enable_swap_info() which needs to be removed or updated?

This patchset depends on below patchset:
[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

Oh, OK, sorry, missed that.

(When I'm working on the post-merge-window patch pile (700+ emails) I go in reverse-time order, so highlighting dependencies like this helps!)

Quoted 4 lines

By the way, Chris worried the plist swap_active_head could be reused by the ongoing swap-tier work and suggested to hold off this patchset till swap-tier work is clear. Now seems it won't impact swap-tier, I will send a v2 against mm-new.

OK.

Is Baoquan's "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series still considered good to merge?

author
#15 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

Quoted 19 lines · Andrew Morton
Quoted 17 lines · On Thu, 9 Oct 2025 15:47:06 +0800 Baoquan He <[email protected]> wrote:
Quoted 15 lines · Andrew Morton

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2557

Quoted 10 lines · On Wed, 1 Oct 2025 12:34:36 +0800 Baoquan He <[email protected]> wrote:

There's no user of swap_active_head, remove it now.

...

mm/swapfile.c +0 -2
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
 	assert_spin_locked(&swap_lock);
 
-	plist_add(&si->list, &swap_active_head);
-

There's a large comment in _enable_swap_info() which needs to be removed or updated?

Quoted 7 lines
Quoted 2 lines

This patchset depends on below patchset:
[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

Oh, OK, sorry, missed that.

(When I'm working on the post-merge-window patch pile (700+ emails) I go in reverse-time order, so highlighting dependencies like this helps!)

I see, will mention the dependency at the beginning of cover letter to highlight this.

Quoted 10 lines
Quoted 5 lines

By the way, Chris worried the plist swap_active_head could be reused by the ongoing swap-tier work and suggested to hold off this patchset till swap-tier work is clear. Now seems it won't impact swap-tier, I will send a v2 against mm-new.

OK.

Is Baoquan's "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series still considered good to merge?

Up to now no objection received to the patchset and got ACK from Chris for "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin".

maintainer
#16 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

Quoted 24 lines · Baoquan He
Quoted 22 lines · Andrew Morton
Quoted 19 lines · On Thu, 9 Oct 2025 15:47:06 +0800 Baoquan He <[email protected]> wrote:
Quoted 8 lines · Andrew Morton
Quoted 6 lines · On Wed, 1 Oct 2025 12:34:36 +0800 Baoquan He <[email protected]> wrote:

There's no user of swap_active_head, remove it now.

...

(quoted diff) +0 -0
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
Quoted 7 lines

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2557

Quoted 4 lines
mm/swapfile.c +0 -2
        assert_spin_locked(&swap_lock);
 
-       plist_add(&si->list, &swap_active_head);
-

There's a large comment in _enable_swap_info() which needs to be removed or updated?

This patchset depends on below patchset:
[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

Quoted 7 lines
Quoted 4 lines

Oh, OK, sorry, missed that.

(When I'm working on the post-merge-window patch pile (700+ emails) I go in reverse-time order, so highlighting dependencies like this helps!)

I see, will mention the dependency at the beginning of cover letter to highlight this.

Right. Please hold off this series for swap tiers for a month or so.

Quoted 5 lines
Quoted 5 lines
Quoted 4 lines

By the way, Chris worried the plist swap_active_head could be reused by the ongoing swap-tier work and suggested to hold off this patchset till swap-tier work is clear. Now seems it won't impact swap-tier, I will send a v2 against mm-new.

Quoted 8 lines
Quoted 4 lines

OK.

Is Baoquan's "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series still considered good to merge?

Up to now no objection received to the patchset and got ACK from Chris for "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin".

Ack that series is good to merge.

Chris

maintainer
#17 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

In reply to Andrew Morton
Quoted 20 lines · Andrew Morton
Quoted 17 lines · On Thu, 9 Oct 2025 15:47:06 +0800 Baoquan He <[email protected]> wrote:
Quoted 15 lines · Andrew Morton

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2557

Quoted 10 lines · On Wed, 1 Oct 2025 12:34:36 +0800 Baoquan He <[email protected]> wrote:

There's no user of swap_active_head, remove it now.

...

mm/swapfile.c +0 -2
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
  assert_spin_locked(&swap_lock);
 
- plist_add(&si->list, &swap_active_head);
-

There's a large comment in _enable_swap_info() which needs to be removed or updated?

Quoted 4 lines
Quoted 2 lines

This patchset depends on below patchset:
[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

Oh, OK, sorry, missed that.

Yes, that round robin series is consider good to merge.

Quoted 3 lines

(When I'm working on the post-merge-window patch pile (700+ emails) I go in reverse-time order, so highlighting dependencies like this helps!)

Understand, that is why I emphasize to Baoquan to highlight the dependency patch series as well.

Quoted 4 lines
Quoted 4 lines

By the way, Chris worried the plist swap_active_head could be reused by the ongoing swap-tier work and suggested to hold off this patchset till swap-tier work is clear. Now seems it won't impact swap-tier, I will send a v2 against mm-new.

I will wish this swap_active_head patch series to hold off for a month or so to make sure swap tiers won't use it.

Quoted 5 lines

OK.

Is Baoquan's "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series still considered good to merge?

That series is still considered good to merge. It has 18% performance gain by switching the complex per node id swap file priority to simple round robin. Let me reply to that patch email thread.

Chris

maintainer
#18 lore raw

Re: [PATCH 3/3] mm/swap: remove unneeded swap_active_head

In reply to Andrew Morton

Hi Andrew,

Please drop/ignore this patch series that removes the swap_active_head. I just received the patch from YoungJun. Just as I worried, YoungJun is indeed using the swap_active_head in his swap tiers patch series. Let's not remove it.

I copy/paste from YoungJun's patches here. It has confirmed usage of that list.

+
+ p = plist_first_entry(&swap_active_head, struct swap_info_struct, list);
+ for_each_active_tier(tier) {
+ plist_for_each_entry_continue(p, &swap_active_head, list) {
+ if (p->tier_off_mask != DEFAULT_OFF_MASK)

Chris

Quoted 37 lines · Andrew Morton
Quoted 11 lines · On Thu, 9 Oct 2025 15:47:06 +0800 Baoquan He <[email protected]> wrote:
Quoted 10 lines · Andrew Morton
Quoted 8 lines · On Wed, 1 Oct 2025 12:34:36 +0800 Baoquan He <[email protected]> wrote:

There's no user of swap_active_head, remove it now.

...

mm/swapfile.c +0 -0
@@ -2553,8 +2545,6 @@ static void _enable_swap_info(struct swap_info_struct *si)
 
  assert_spin_locked(&swap_lock);
 
Quoted 8 lines
Quoted 5 lines

mm/swapfile.c static void _enable_swap_info(struct swap_info_struct *si) · line 2557

Quoted 2 lines
mm/swapfile.c +0 -2
- plist_add(&si->list, &swap_active_head);
-

There's a large comment in _enable_swap_info() which needs to be removed or updated?

This patchset depends on below patchset:
[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

Oh, OK, sorry, missed that.

(When I'm working on the post-merge-window patch pile (700+ emails) I go in reverse-time order, so highlighting dependencies like this helps!)

Quoted 4 lines

By the way, Chris worried the plist swap_active_head could be reused by the ongoing swap-tier work and suggested to hold off this patchset till swap-tier work is clear. Now seems it won't impact swap-tier, I will send a v2 against mm-new.

OK.

Is Baoquan's "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series still considered good to merge?

maintainer
#19 lore raw
In reply to Baoquan He
Quoted 1 line · Baoquan He

Can you please confirm or deny this series is depend on the "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series.

If it is depend on that series, you need to declare that in the beginning, so Andrew will not try to apply it without the depended series.

Quoted 3 lines

In mm/swapfile.c, there are two plist variables, swap_active_head and swap_avail_head. swap_avail_head contains all available (active, not full) swap_info_structs orderred by priority. While swap_active_head contains

Nit: spelling : orderred -> ordered

Quoted 1 line

all active swap_info_structs (active, nor full and full) orderred by

Same here, ordered.

Quoted 9 lines

priority. Earlier, it serves three purposes:

1) When swapoff one swap device in the middle, swap devices of priority
lower than the swapped off swap device will be promoted up value one,
e.g, I swapped off zram1 of priority '-3', then zram2 promoted to
have priority '-3', zram3 has priority '-4'. This is done via plist swap_active_head.
- This code has been taken off in
- [PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

See above, if there is patch dependency, declare it early.

Quoted 6 lines
# swapon
NAME       TYPE      SIZE USED PRIO
/dev/zram0 partition  16G   0B   -2
/dev/zram1 partition  16G   0B   -3
/dev/zram2 partition  16G   0B   -4
/dev/zram3 partition  16G   0B   -5

I am a bit confused. If the "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series already change the default priority to -1. How does it get the -2 -3 -4 -5? You seems describe the behavior before that revert series. Can you please clarify.

Quoted 6 lines
# swapoff /dev/zram1
# swapon
NAME       TYPE      SIZE USED PRIO
/dev/zram0 partition  16G   0B   -2
/dev/zram2 partition  16G   0B   -3
/dev/zram3 partition  16G   0B   -4

Is that the behavior before your reverting node id for swapfile series? Assume it has patch dependency on the earlier series. Wouldn't better you describe the behavior after the round robin series and how you can do better?

Quoted 3 lines

2) Find a swap device in swap_active_head when swap off.
- This can be done through iterating swap_info[] instead. Change is done in patch 2.

Can you please clarify what is the benefit of remove plist swap_active_head? e.g. Does it simply the code or make swap code perform better.

Quoted 9 lines

3) Judge if there's any active swap device via __has_usable_swap().
- This can be done by checking total_swap_pages instead. Change is done in patch 1.

Among them, the purpose 1) is the most important, while it has been
taken off in below patchset. So this patchset removing swap_active_head sits on top of it.
- [PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin
- https://lore.kernel.org/all/[email protected]/T/#u

It seems you mean to say this patch series is depend on that PATCH v3 series.

Chris

Quoted 13 lines

Baoquan He (3):
mm/swapfile.c: remove __has_usable_swap() mm/swapfile.c: use swap_info[] to find the swap device mm/swap: remove unneeded swap_active_head

 include/linux/swap.h |  1 -
 mm/swapfile.c        | 31 ++++++++-----------------------
 2 files changed, 8 insertions(+), 24 deletions(-)

--
2.41.0

author
#20 lore raw
Quoted 6 lines · Chris Li
Quoted 1 line · Baoquan He

Can you please confirm or deny this series is depend on the "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series.

Thanks a lot for reviewing this carefully. This series depends on patchset [PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series.

Quoted 4 lines

If it is depend on that series, you need to declare that in the beginning, so Andrew will not try to apply it without the depended series.

Sorry, I should have told that clearly at the beginning.

Quoted 6 lines
Quoted 4 lines

In mm/swapfile.c, there are two plist variables, swap_active_head and swap_avail_head. swap_avail_head contains all available (active, not full) swap_info_structs orderred by priority. While swap_active_head contains

Nit: spelling : orderred -> ordered

Thanks.

Quoted 15 lines
Quoted 2 lines

all active swap_info_structs (active, nor full and full) orderred by

Same here, ordered.

Quoted 9 lines

priority. Earlier, it serves three purposes:

1) When swapoff one swap device in the middle, swap devices of priority
lower than the swapped off swap device will be promoted up value one,
e.g, I swapped off zram1 of priority '-3', then zram2 promoted to
have priority '-3', zram3 has priority '-4'. This is done via plist swap_active_head.
- This code has been taken off in
- [PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin

Quoted 1 line

See above, if there is patch dependency, declare it early.

Agree, will do in future posting.

Quoted 8 lines
Quoted 7 lines
# swapon
NAME       TYPE      SIZE USED PRIO
/dev/zram0 partition  16G   0B   -2
/dev/zram1 partition  16G   0B   -3
/dev/zram2 partition  16G   0B   -4
/dev/zram3 partition  16G   0B   -5
Quoted 5 lines

I am a bit confused. If the "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin" series already change the default priority to -1. How does it get the -2 -3 -4 -5? You seems describe the behavior before that revert series. Can you please clarify.

In the given example, It demonstrates every device will get +1 priority promotion after the swapped off device.

I just intended to tell the purposes of swap_active_head in the current kernel, since my patches haven't been merged. I also had a feeling this description is messy before posting, diddn't expect it could cause these confusion.

I should have mentioned this series depends on "[PATCH v3 0/2] mm/swapfile.c:
select the swap device with default priority round robin" series, then I don't need to mention this purpose of swap_active_head, that could make the describing easier.

Quoted 8 lines
Quoted 7 lines
# swapoff /dev/zram1
# swapon
NAME       TYPE      SIZE USED PRIO
/dev/zram0 partition  16G   0B   -2
/dev/zram2 partition  16G   0B   -3
/dev/zram3 partition  16G   0B   -4
Quoted 4 lines

Is that the behavior before your reverting node id for swapfile series? Assume it has patch dependency on the earlier series. Wouldn't better you describe the behavior after the round robin series and how you can do better?

Right, I agree.

Quoted 7 lines
Quoted 4 lines

2) Find a swap device in swap_active_head when swap off.
- This can be done through iterating swap_info[] instead. Change is done in patch 2.

Can you please clarify what is the benefit of remove plist swap_active_head? e.g. Does it simply the code or make swap code perform better.

After series "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin", plist swap_active_head is only used in swapoff api to scan it to find a swap device according to the device name. It can be replaced with scanning swap_info[], and there's no efficiency degradation. And we can remove field "struct plist_node list;" in struct swap_info_struct, and the relevant handling code which is unnecessary.

Quoted 5 lines
Quoted 5 lines

3) Judge if there's any active swap device via __has_usable_swap().
- This can be done by checking total_swap_pages instead. Change is done in patch 1.

Quoted 7 lines
Quoted 5 lines

Among them, the purpose 1) is the most important, while it has been
taken off in below patchset. So this patchset removing swap_active_head sits on top of it.
- [PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin
- https://lore.kernel.org/all/[email protected]/T/#u

It seems you mean to say this patch series is depend on that PATCH v3 series.

Yeah, I should have mentioned this at the beginning to make the log clearer.

maintainer
#21 lore raw
Quoted 10 lines · Baoquan He
Quoted 2 lines

Can you please clarify what is the benefit of remove plist swap_active_head? e.g. Does it simply the code or make swap code perform better.

After series "[PATCH v3 0/2] mm/swapfile.c: select the swap device with default priority round robin", plist swap_active_head is only used in swapoff api to scan it to find a swap device according to the device name. It can be replaced with scanning swap_info[], and there's no efficiency degradation. And we can remove field "struct plist_node list;" in struct swap_info_struct, and the relevant handling code which is unnecessary.

So the benefit is removing a list. Thank you for confirming that.

BTW, I think I found a race bug in your patch 2/3, I replied in the patch email thread.
Let me know if that race is valid or not.

Chris