[PATCH v2] zswap: re-check zswap_is_full after do zswap_shrink

[PATCH v2] zswap: re-check zswap_is_full after do zswap_shrink

[Li Wang]

The '/sys/../zswap/stored_pages:' keep raising in zswap test with
"zswap.max_pool_percent=0" parameter. But theoretically, it should
not compress or store pages any more since there is no space in
compressed pool.

Reproduce steps:
  1. Boot kernel with "zswap.enabled=1"
  2. Set the max_pool_percent to 0
      # echo 0 > /sys/module/zswap/parameters/max_pool_percent
  3. Do memory stress test to see if some pages have been compressed
      # stress --vm 1 --vm-bytes $mem_available"M" --timeout 60s
  4. Watching the 'stored_pages' number increasing or not

The root cause is:
  When zswap_max_pool_percent is setting to 0 via kernel parameter, the
  zswap_is_full() will always return true to do zswap_shrink(). But if
  the shinking is able to reclain a page successful, then proceeds to
  compress/store another page, so the value of stored_pages will keep
  changing.

To solve the issue, this patch adds zswap_is_full() check again after
zswap_shrink() to make sure it's now under the max_pool_percent, and
not to compress/store if reach its limitaion.

Signed-off-by: Li Wang <liwang@redhat.com>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Huang Ying <huang.ying.caritas@gmail.com>
Cc: Yu Zhao <yuzhao@google.com>
---
 mm/zswap.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/mm/zswap.c b/mm/zswap.c
index 61a5c41..fd320c3 100644
--- a/mm/zswap.c
+++ b/mm/zswap.c
@@ -1026,6 +1026,15 @@ static int zswap_frontswap_store(unsigned type, pgoff_t offset,
 			ret = -ENOMEM;
 			goto reject;
 		}
+
+		/* A second zswap_is_full() check after
+		 * zswap_shrink() to make sure it's now
+		 * under the max_pool_percent
+		 */
+		if (zswap_is_full()) {
+			ret = -ENOMEM;
+			goto reject;
+		}
 	}
 
 	/* allocate entry */
-- 
2.9.5

Re: [PATCH v2] zswap: re-check zswap_is_full after do zswap_shrink

[Dan Streetman]

On Wed, May 30, 2018 at 6:39 AM, Li Wang <liwang@redhat.com> wrote:
> The '/sys/../zswap/stored_pages:' keep raising in zswap test with
> "zswap.max_pool_percent=0" parameter. But theoretically, it should
> not compress or store pages any more since there is no space in
> compressed pool.
>
> Reproduce steps:
>   1. Boot kernel with "zswap.enabled=1"
>   2. Set the max_pool_percent to 0
>       # echo 0 > /sys/module/zswap/parameters/max_pool_percent
>   3. Do memory stress test to see if some pages have been compressed
>       # stress --vm 1 --vm-bytes $mem_available"M" --timeout 60s
>   4. Watching the 'stored_pages' number increasing or not
>
> The root cause is:
>   When zswap_max_pool_percent is setting to 0 via kernel parameter, the
>   zswap_is_full() will always return true to do zswap_shrink(). But if
>   the shinking is able to reclain a page successful, then proceeds to
>   compress/store another page, so the value of stored_pages will keep
>   changing.
>
> To solve the issue, this patch adds zswap_is_full() check again after
> zswap_shrink() to make sure it's now under the max_pool_percent, and
> not to compress/store if reach its limitaion.
>
> Signed-off-by: Li Wang <liwang@redhat.com>

Acked-by: Dan Streetman <ddstreet@ieee.org>

> Cc: Seth Jennings <sjenning@redhat.com>
> Cc: Dan Streetman <ddstreet@ieee.org>
> Cc: Huang Ying <huang.ying.caritas@gmail.com>
> Cc: Yu Zhao <yuzhao@google.com>
> ---
>  mm/zswap.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
>
> diff --git a/mm/zswap.c b/mm/zswap.c
> index 61a5c41..fd320c3 100644
> --- a/mm/zswap.c
> +++ b/mm/zswap.c
> @@ -1026,6 +1026,15 @@ static int zswap_frontswap_store(unsigned type, pgoff_t offset,
>                         ret = -ENOMEM;
>                         goto reject;
>                 }
> +
> +               /* A second zswap_is_full() check after
> +                * zswap_shrink() to make sure it's now
> +                * under the max_pool_percent
> +                */
> +               if (zswap_is_full()) {
> +                       ret = -ENOMEM;
> +                       goto reject;
> +               }
>         }
>
>         /* allocate entry */
> --
> 2.9.5
>

Re: [PATCH v2] zswap: re-check zswap_is_full after do zswap_shrink

[Li Wang]

On 30 May 2018 at 20:53, Dan Streetman <ddstreet@ieee.org> wrote:
> On Wed, May 30, 2018 at 6:39 AM, Li Wang <liwang@redhat.com> wrote:
>> The '/sys/../zswap/stored_pages:' keep raising in zswap test with
>> "zswap.max_pool_percent=0" parameter. But theoretically, it should
>> not compress or store pages any more since there is no space in
>> compressed pool.
>>
>> Reproduce steps:
>>   1. Boot kernel with "zswap.enabled=1"
>>   2. Set the max_pool_percent to 0
>>       # echo 0 > /sys/module/zswap/parameters/max_pool_percent
>>   3. Do memory stress test to see if some pages have been compressed
>>       # stress --vm 1 --vm-bytes $mem_available"M" --timeout 60s
>>   4. Watching the 'stored_pages' number increasing or not
>>
>> The root cause is:
>>   When zswap_max_pool_percent is setting to 0 via kernel parameter, the
>>   zswap_is_full() will always return true to do zswap_shrink(). But if
>>   the shinking is able to reclain a page successful, then proceeds to
>>   compress/store another page, so the value of stored_pages will keep
>>   changing.
>>
>> To solve the issue, this patch adds zswap_is_full() check again after
>> zswap_shrink() to make sure it's now under the max_pool_percent, and
>> not to compress/store if reach its limitaion.
>>
>> Signed-off-by: Li Wang <liwang@redhat.com>
>
> Acked-by: Dan Streetman <ddstreet@ieee.org>

ping~

Any possible to merge this in kernel-4.18-rcX? My zswap test always
fails on the upstream kernel.


-- 
Regards,
Li Wang
Email: wangli.ahau@gmail.com

Re: [PATCH v2] zswap: re-check zswap_is_full after do zswap_shrink

[Dan Streetman]

On Mon, Jun 25, 2018 at 4:08 AM Li Wang <wangli.ahau@gmail.com> wrote:
>
> On 30 May 2018 at 20:53, Dan Streetman <ddstreet@ieee.org> wrote:
> > On Wed, May 30, 2018 at 6:39 AM, Li Wang <liwang@redhat.com> wrote:
> >> The '/sys/../zswap/stored_pages:' keep raising in zswap test with
> >> "zswap.max_pool_percent=0" parameter. But theoretically, it should
> >> not compress or store pages any more since there is no space in
> >> compressed pool.
> >>
> >> Reproduce steps:
> >>   1. Boot kernel with "zswap.enabled=1"
> >>   2. Set the max_pool_percent to 0
> >>       # echo 0 > /sys/module/zswap/parameters/max_pool_percent
> >>   3. Do memory stress test to see if some pages have been compressed
> >>       # stress --vm 1 --vm-bytes $mem_available"M" --timeout 60s
> >>   4. Watching the 'stored_pages' number increasing or not
> >>
> >> The root cause is:
> >>   When zswap_max_pool_percent is setting to 0 via kernel parameter, the
> >>   zswap_is_full() will always return true to do zswap_shrink(). But if
> >>   the shinking is able to reclain a page successful, then proceeds to
> >>   compress/store another page, so the value of stored_pages will keep
> >>   changing.
> >>
> >> To solve the issue, this patch adds zswap_is_full() check again after
> >> zswap_shrink() to make sure it's now under the max_pool_percent, and
> >> not to compress/store if reach its limitaion.
> >>
> >> Signed-off-by: Li Wang <liwang@redhat.com>
> >
> > Acked-by: Dan Streetman <ddstreet@ieee.org>
>
> ping~
>
> Any possible to merge this in kernel-4.18-rcX? My zswap test always
> fails on the upstream kernel.

cc'ing Andrew as he may have missed this.

>
>
> --
> Regards,
> Li Wang
> Email: wangli.ahau@gmail.com