1. Patchew
  2. linux
  3. View series

[PATCH] mm/zswap: reduce the size of the compression buffer to a single page

Nhat Pham posted 1 patch 1 month, 2 weeks ago 
mm/zswap.c | 9 ++-------
1 file changed, 2 insertions(+), 7 deletions(-)
[PATCH] mm/zswap: reduce the size of the compression buffer to a single page
Posted by Nhat Pham 1 month, 2 weeks ago 
Reduce the compression buffer size from 2 * PAGE_SIZE to only one page,
as the compression output (in the success case) should not exceed the
length of the input.

In the past, Chengming tried to reduce the compression buffer size, but
ran into issues with the LZO algorithm (see [2]). Herbert Xu reported
that the issue has been fixed (see [3]). Now we should have the
guarantee that compressors' output should not exceed one page in the
success case, and the algorithm will just report failure otherwise.

With this patch, we save one page per cpu (per compression algorithm).

[1]: https://lore.kernel.org/linux-mm/20231213-zswap-dstmem-v4-1-f228b059dd89@bytedance.com/
[2]: https://lore.kernel.org/lkml/0000000000000b05cd060d6b5511@google.com/
[3]: https://lore.kernel.org/linux-mm/aKUmyl5gUFCdXGn-@gondor.apana.org.au/

Co-developed-by: Chengming Zhou <chengming.zhou@linux.dev>
Signed-off-by: Chengming Zhou <chengming.zhou@linux.dev>
Signed-off-by: Nhat Pham <nphamcs@gmail.com>
---
 mm/zswap.c | 9 ++-------
 1 file changed, 2 insertions(+), 7 deletions(-)

diff --git a/mm/zswap.c b/mm/zswap.c
index 1f1ac043a2d9..5dd282c5b626 100644
--- a/mm/zswap.c
+++ b/mm/zswap.c
@@ -833,7 +833,7 @@ static int zswap_cpu_comp_prepare(unsigned int cpu, struct hlist_node *node)
 	u8 *buffer = NULL;
 	int ret;
 
-	buffer = kmalloc_node(PAGE_SIZE * 2, GFP_KERNEL, cpu_to_node(cpu));
+	buffer = kmalloc_node(PAGE_SIZE, GFP_KERNEL, cpu_to_node(cpu));
 	if (!buffer) {
 		ret = -ENOMEM;
 		goto fail;
@@ -960,12 +960,7 @@ static bool zswap_compress(struct page *page, struct zswap_entry *entry,
 	sg_init_table(&input, 1);
 	sg_set_page(&input, page, PAGE_SIZE, 0);
 
-	/*
-	 * We need PAGE_SIZE * 2 here since there maybe over-compression case,
-	 * and hardware-accelerators may won't check the dst buffer size, so
-	 * giving the dst buffer with enough length to avoid buffer overflow.
-	 */
-	sg_init_one(&output, dst, PAGE_SIZE * 2);
+	sg_init_one(&output, dst, PAGE_SIZE);
 	acomp_request_set_params(acomp_ctx->req, &input, &output, PAGE_SIZE, dlen);
 
 	/*

base-commit: c0e3b3f33ba7b767368de4afabaf7c1ddfdc3872
-- 
2.47.3
Re: [PATCH] mm/zswap: reduce the size of the compression buffer to a single page
Posted by SeongJae Park 1 month, 2 weeks ago 
On Wed, 20 Aug 2025 11:15:47 -0700 Nhat Pham <nphamcs@gmail.com> wrote:

> Reduce the compression buffer size from 2 * PAGE_SIZE to only one page,
> as the compression output (in the success case) should not exceed the
> length of the input.
> 
> In the past, Chengming tried to reduce the compression buffer size, but
> ran into issues with the LZO algorithm (see [2]). Herbert Xu reported
> that the issue has been fixed (see [3]). Now we should have the
> guarantee that compressors' output should not exceed one page in the
> success case, and the algorithm will just report failure otherwise.
> 
> With this patch, we save one page per cpu (per compression algorithm).
> 
> [1]: https://lore.kernel.org/linux-mm/20231213-zswap-dstmem-v4-1-f228b059dd89@bytedance.com/

This is not mentioned anywhere.  Probably you added this as a reference for the
past work from Chengming?

> [2]: https://lore.kernel.org/lkml/0000000000000b05cd060d6b5511@google.com/
> [3]: https://lore.kernel.org/linux-mm/aKUmyl5gUFCdXGn-@gondor.apana.org.au/
> 
> Co-developed-by: Chengming Zhou <chengming.zhou@linux.dev>
> Signed-off-by: Chengming Zhou <chengming.zhou@linux.dev>
> Signed-off-by: Nhat Pham <nphamcs@gmail.com>

Acked-by: SeongJae Park <sj@kernel.org>

[...]


Thanks,
SJ
Re: [PATCH] mm/zswap: reduce the size of the compression buffer to a single page
Posted by Nhat Pham 1 month, 2 weeks ago 
On Wed, Aug 20, 2025 at 3:47 PM SeongJae Park <sj@kernel.org> wrote:
>
> On Wed, 20 Aug 2025 11:15:47 -0700 Nhat Pham <nphamcs@gmail.com> wrote:
>
> > Reduce the compression buffer size from 2 * PAGE_SIZE to only one page,
> > as the compression output (in the success case) should not exceed the
> > length of the input.
> >
> > In the past, Chengming tried to reduce the compression buffer size, but

Yeah, this should include the reference below. My bad. So it should be:

In the past, Chengming tried to reduce the compression buffer size
(see [1]), but

Andrew, would it be OK if you fix it for me when you put this patch in
your tree :) Or I can also send a new version if that's better. Thank
you!

> > ran into issues with the LZO algorithm (see [2]). Herbert Xu reported
> > that the issue has been fixed (see [3]). Now we should have the
> > guarantee that compressors' output should not exceed one page in the
> > success case, and the algorithm will just report failure otherwise.
> >
> > With this patch, we save one page per cpu (per compression algorithm).
> >
> > [1]: https://lore.kernel.org/linux-mm/20231213-zswap-dstmem-v4-1-f228b059dd89@bytedance.com/
>
> This is not mentioned anywhere.  Probably you added this as a reference for the
> past work from Chengming?

Thanks for taking a look, SJ!

>
> > [2]: https://lore.kernel.org/lkml/0000000000000b05cd060d6b5511@google.com/
> > [3]: https://lore.kernel.org/linux-mm/aKUmyl5gUFCdXGn-@gondor.apana.org.au/
> >
> > Co-developed-by: Chengming Zhou <chengming.zhou@linux.dev>
> > Signed-off-by: Chengming Zhou <chengming.zhou@linux.dev>
> > Signed-off-by: Nhat Pham <nphamcs@gmail.com>
>
> Acked-by: SeongJae Park <sj@kernel.org>
Re: [PATCH] mm/zswap: reduce the size of the compression buffer to a single page
Posted by Chengming Zhou 1 month, 2 weeks ago 
On 2025/8/21 06:56, Nhat Pham wrote:
> On Wed, Aug 20, 2025 at 3:47 PM SeongJae Park <sj@kernel.org> wrote:
>>
>> On Wed, 20 Aug 2025 11:15:47 -0700 Nhat Pham <nphamcs@gmail.com> wrote:
>>
>>> Reduce the compression buffer size from 2 * PAGE_SIZE to only one page,
>>> as the compression output (in the success case) should not exceed the
>>> length of the input.
>>>
>>> In the past, Chengming tried to reduce the compression buffer size, but
> 
> Yeah, this should include the reference below. My bad. So it should be:
> 
> In the past, Chengming tried to reduce the compression buffer size
> (see [1]), but
> 
> Andrew, would it be OK if you fix it for me when you put this patch in
> your tree :) Or I can also send a new version if that's better. Thank
> you!
> 
>>> ran into issues with the LZO algorithm (see [2]). Herbert Xu reported
>>> that the issue has been fixed (see [3]). Now we should have the
>>> guarantee that compressors' output should not exceed one page in the
>>> success case, and the algorithm will just report failure otherwise.

Great!

Reviewed-by: Chengming Zhou <chengming.zhou@linux.dev>

Thank you!

>>>
>>> With this patch, we save one page per cpu (per compression algorithm).
>>>
>>> [1]: https://lore.kernel.org/linux-mm/20231213-zswap-dstmem-v4-1-f228b059dd89@bytedance.com/
>>
>> This is not mentioned anywhere.  Probably you added this as a reference for the
>> past work from Chengming?
> 
> Thanks for taking a look, SJ!
> 
>>
>>> [2]: https://lore.kernel.org/lkml/0000000000000b05cd060d6b5511@google.com/
>>> [3]: https://lore.kernel.org/linux-mm/aKUmyl5gUFCdXGn-@gondor.apana.org.au/
>>>
>>> Co-developed-by: Chengming Zhou <chengming.zhou@linux.dev>
>>> Signed-off-by: Chengming Zhou <chengming.zhou@linux.dev>
>>> Signed-off-by: Nhat Pham <nphamcs@gmail.com>
>>
>> Acked-by: SeongJae Park <sj@kernel.org>
Re: [PATCH] mm/zswap: reduce the size of the compression buffer to a single page
Posted by Nhat Pham 1 month, 2 weeks ago 
On Wed, Aug 20, 2025 at 11:15 AM Nhat Pham <nphamcs@gmail.com> wrote:
>
> Reduce the compression buffer size from 2 * PAGE_SIZE to only one page,
> as the compression output (in the success case) should not exceed the
> length of the input.
>
> In the past, Chengming tried to reduce the compression buffer size, but
> ran into issues with the LZO algorithm (see [2]). Herbert Xu reported
> that the issue has been fixed (see [3]). Now we should have the
> guarantee that compressors' output should not exceed one page in the
> success case, and the algorithm will just report failure otherwise.
>
> With this patch, we save one page per cpu (per compression algorithm).
>
> [1]: https://lore.kernel.org/linux-mm/20231213-zswap-dstmem-v4-1-f228b059dd89@bytedance.com/
> [2]: https://lore.kernel.org/lkml/0000000000000b05cd060d6b5511@google.com/
> [3]: https://lore.kernel.org/linux-mm/aKUmyl5gUFCdXGn-@gondor.apana.org.au/
>
> Co-developed-by: Chengming Zhou <chengming.zhou@linux.dev>
> Signed-off-by: Chengming Zhou <chengming.zhou@linux.dev>

The syntax changes very slightly, but this is functionally the same
patch as Chengming's. So I'm keeping all of his credit - please let me
know if I miss any proper accreditations. :)

Patchew 2.1-devel-b67e4c2

© 2016 - 2025 Red Hat, Inc.