Swap subsystem does lazy swap slot free with expecting the page
would be swapped out again so we can avoid unnecessary write.
But the problem in in-memory swap(ex, zram) is that it consumes
memory space until vm_swap_full(ie, used half of all of swap device)
condition meet. It could be bad if we use multiple swap device,
small in-memory swap and big storage swap or in-memory swap alone.
This patch makes swap subsystem free swap slot as soon as swap-read
is completed and make the swapcache page dirty so the page should
be written out the swap device to reclaim it.
It means we never lose it.
I tested this patch with kernel compile workload.
1. before
compile time : 9882.42
zram max wasted space by fragmentation: 13471881 byte
memory space consumed by zram: 174227456 byte
the number of slot free notify: 206684
2. after
compile time : 9653.90
zram max wasted space by fragmentation: 11805932 byte
memory space consumed by zram: 154001408 byte
the number of slot free notify: 426972
* changelog from v3
* Rebased on next-20130508
* changelog from v1
* Add more comment
Cc: Hugh Dickins <[email protected]>
Cc: Seth Jennings <[email protected]>
Cc: Nitin Gupta <[email protected]>
Cc: Konrad Rzeszutek Wilk <[email protected]>
Cc: Shaohua Li <[email protected]>
Signed-off-by: Dan Magenheimer <[email protected]>
Signed-off-by: Minchan Kim <[email protected]>
---
mm/page_io.c | 35 +++++++++++++++++++++++++++++++++++
1 file changed, 35 insertions(+)
diff --git a/mm/page_io.c b/mm/page_io.c
index a294076..527db57 100644
--- a/mm/page_io.c
+++ b/mm/page_io.c
@@ -21,6 +21,7 @@
#include <linux/writeback.h>
#include <linux/frontswap.h>
#include <linux/aio.h>
+#include <linux/blkdev.h>
#include <asm/pgtable.h>
static struct bio *get_swap_bio(gfp_t gfp_flags,
@@ -82,8 +83,42 @@ void end_swap_bio_read(struct bio *bio, int err, struct batch_complete *batch)
iminor(bio->bi_bdev->bd_inode),
(unsigned long long)bio->bi_sector);
} else {
+ struct swap_info_struct *sis;
+
SetPageUptodate(page);
+ sis = page_swap_info(page);
+ if (sis->flags & SWP_BLKDEV) {
+ /*
+ * Swap subsystem does lazy swap slot free with
+ * expecting the page would be swapped out again
+ * so we can avoid unnecessary write if the page
+ * isn't redirty.
+ * It's good for real swap storage because we can
+ * reduce unnecessary I/O and enhance wear-leveling
+ * if you use SSD as swap device.
+ * But if you use in-memory swap device(ex, zram),
+ * it causes duplicated copy between uncompressed
+ * data in VM-owned memory and compressed data in
+ * zram-owned memory. So let's free zram-owned memory
+ * and make the VM-owned decompressed page *dirty*
+ * so the page should be swap out somewhere again if
+ * we want to reclaim it, again.
+ */
+ struct gendisk *disk = sis->bdev->bd_disk;
+ if (disk->fops->swap_slot_free_notify) {
+ swp_entry_t entry;
+ unsigned long offset;
+
+ entry.val = page_private(page);
+ offset = swp_offset(entry);
+
+ SetPageDirty(page);
+ disk->fops->swap_slot_free_notify(sis->bdev,
+ offset);
+ }
+ }
}
+
unlock_page(page);
bio_put(bio);
}
--
1.8.2.1
On Thu, May 09, 2013 at 08:41:57AM +0900, Minchan Kim wrote:
Hey Michan,
Just a couple of syntax corrections. The code comment could also
benefit from this.
Otherwise it looks OK to me.
> Swap subsystem does lazy swap slot free with expecting the page
^-a ^- the expectation that
> would be swapped out again so we can avoid unnecessary write.
^--that it
>
> But the problem in in-memory swap(ex, zram) is that it consumes
^^-with
> memory space until vm_swap_full(ie, used half of all of swap device)
> condition meet. It could be bad if we use multiple swap device,
^- 'is' ^^^^^ - 'would' ^^^^^-devices
> small in-memory swap and big storage swap or in-memory swap alone.
^-, ^-,
>
> This patch makes swap subsystem free swap slot as soon as swap-read
> is completed and make the swapcache page dirty so the page should
^-makes ^-'that the'
> be written out the swap device to reclaim it.
> It means we never lose it.
>
> I tested this patch with kernel compile workload.
^-a
>
> 1. before
>
> compile time : 9882.42
> zram max wasted space by fragmentation: 13471881 byte
> memory space consumed by zram: 174227456 byte
> the number of slot free notify: 206684
>
> 2. after
>
> compile time : 9653.90
> zram max wasted space by fragmentation: 11805932 byte
> memory space consumed by zram: 154001408 byte
> the number of slot free notify: 426972
>
> * changelog from v3
> * Rebased on next-20130508
>
> * changelog from v1
> * Add more comment
>
> Cc: Hugh Dickins <[email protected]>
> Cc: Seth Jennings <[email protected]>
> Cc: Nitin Gupta <[email protected]>
> Cc: Konrad Rzeszutek Wilk <[email protected]>
> Cc: Shaohua Li <[email protected]>
> Signed-off-by: Dan Magenheimer <[email protected]>
> Signed-off-by: Minchan Kim <[email protected]>
> ---
> mm/page_io.c | 35 +++++++++++++++++++++++++++++++++++
> 1 file changed, 35 insertions(+)
>
> diff --git a/mm/page_io.c b/mm/page_io.c
> index a294076..527db57 100644
> --- a/mm/page_io.c
> +++ b/mm/page_io.c
> @@ -21,6 +21,7 @@
> #include <linux/writeback.h>
> #include <linux/frontswap.h>
> #include <linux/aio.h>
> +#include <linux/blkdev.h>
> #include <asm/pgtable.h>
>
> static struct bio *get_swap_bio(gfp_t gfp_flags,
> @@ -82,8 +83,42 @@ void end_swap_bio_read(struct bio *bio, int err, struct batch_complete *batch)
> iminor(bio->bi_bdev->bd_inode),
> (unsigned long long)bio->bi_sector);
> } else {
> + struct swap_info_struct *sis;
> +
> SetPageUptodate(page);
> + sis = page_swap_info(page);
> + if (sis->flags & SWP_BLKDEV) {
> + /*
> + * Swap subsystem does lazy swap slot free with
> + * expecting the page would be swapped out again
> + * so we can avoid unnecessary write if the page
> + * isn't redirty.
> + * It's good for real swap storage because we can
> + * reduce unnecessary I/O and enhance wear-leveling
> + * if you use SSD as swap device.
> + * But if you use in-memory swap device(ex, zram),
> + * it causes duplicated copy between uncompressed
> + * data in VM-owned memory and compressed data in
> + * zram-owned memory. So let's free zram-owned memory
> + * and make the VM-owned decompressed page *dirty*
> + * so the page should be swap out somewhere again if
> + * we want to reclaim it, again.
> + */
> + struct gendisk *disk = sis->bdev->bd_disk;
> + if (disk->fops->swap_slot_free_notify) {
> + swp_entry_t entry;
> + unsigned long offset;
> +
> + entry.val = page_private(page);
> + offset = swp_offset(entry);
> +
> + SetPageDirty(page);
> + disk->fops->swap_slot_free_notify(sis->bdev,
> + offset);
> + }
> + }
> }
> +
> unlock_page(page);
> bio_put(bio);
> }
> --
> 1.8.2.1
>
Hi Konrad,
On Thu, May 09, 2013 at 04:15:42PM -0400, Konrad Rzeszutek Wilk wrote:
> On Thu, May 09, 2013 at 08:41:57AM +0900, Minchan Kim wrote:
>
> Hey Michan,
^-n
It's a only thing I can know better than other native speakers. :)
> Just a couple of syntax corrections. The code comment could also
> benefit from this.
>
> Otherwise it looks OK to me.
>
> > Swap subsystem does lazy swap slot free with expecting the page
> ^-a ^- the expectation that
> > would be swapped out again so we can avoid unnecessary write.
> ^--that it
> >
> > But the problem in in-memory swap(ex, zram) is that it consumes
> ^^-with
> > memory space until vm_swap_full(ie, used half of all of swap device)
> > condition meet. It could be bad if we use multiple swap device,
> ^- 'is' ^^^^^ - 'would' ^^^^^-devices
> > small in-memory swap and big storage swap or in-memory swap alone.
> ^-, ^-,
> >
> > This patch makes swap subsystem free swap slot as soon as swap-read
> > is completed and make the swapcache page dirty so the page should
> ^-makes ^-'that the'
> > be written out the swap device to reclaim it.
> > It means we never lose it.
> >
> > I tested this patch with kernel compile workload.
> ^-a
Thanks for the correct whole sentence!
But Andrew alreay correted it with his style.
Although he was done, I'm giving a million thanks to you.
Surely, Thanks Andrew, too.
--
Kind regards,
Minchan Kim
On Fri, May 10, 2013 at 4:44 AM, Minchan Kim <[email protected]> wrote:
> Hi Konrad,
>
> On Thu, May 09, 2013 at 04:15:42PM -0400, Konrad Rzeszutek Wilk wrote:
>> On Thu, May 09, 2013 at 08:41:57AM +0900, Minchan Kim wrote:
>>
>> Hey Michan,
> ^-n
>
> It's a only thing I can know better than other native speakers. :)
I keep on misspelling your name. I am really sorry about that.
>
>
>> Just a couple of syntax corrections. The code comment could also
>> benefit from this.
>>
>> Otherwise it looks OK to me.
>>
>> > Swap subsystem does lazy swap slot free with expecting the page
>> ^-a ^- the expectation that
>> > would be swapped out again so we can avoid unnecessary write.
>> ^--that it
>> >
>> > But the problem in in-memory swap(ex, zram) is that it consumes
>> ^^-with
>> > memory space until vm_swap_full(ie, used half of all of swap device)
>> > condition meet. It could be bad if we use multiple swap device,
>> ^- 'is' ^^^^^ - 'would' ^^^^^-devices
>> > small in-memory swap and big storage swap or in-memory swap alone.
>> ^-, ^-,
>> >
>> > This patch makes swap subsystem free swap slot as soon as swap-read
>> > is completed and make the swapcache page dirty so the page should
>> ^-makes ^-'that the'
>> > be written out the swap device to reclaim it.
>> > It means we never lose it.
>> >
>> > I tested this patch with kernel compile workload.
>> ^-a
>
> Thanks for the correct whole sentence!
> But Andrew alreay correted it with his style.
<nods> I saw his email a couple of hours ago.
> Although he was done, I'm giving a million thanks to you.
> Surely, Thanks Andrew, too.
>
> --
> Kind regards,
> Minchan Kim
>