* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 5:49 [PATCH] mm, page_pool: introduce a new page type for page pool in page type Byungchul Park
@ 2025-07-21 8:05 ` David Hildenbrand
2025-07-21 8:19 ` Byungchul Park
2025-07-21 11:12 ` Pavel Begunkov
2025-07-22 22:17 ` Mina Almasry
2 siblings, 1 reply; 14+ messages in thread
From: David Hildenbrand @ 2025-07-21 8:05 UTC (permalink / raw)
To: Byungchul Park, linux-mm, netdev
Cc: linux-kernel, kernel_team, harry.yoo, ast, daniel, davem, kuba,
hawk, john.fastabend, sdf, saeedm, leon, tariqt, mbloch,
andrew+netdev, edumazet, pabeni, akpm, lorenzo.stoakes,
Liam.Howlett, vbabka, rppt, surenb, mhocko, horms, jackmanb,
hannes, ziy, ilias.apalodimas, willy, brauner, kas, yuzhao,
usamaarif642, baolin.wang, almasrymina, toke, asml.silence, bpf,
linux-rdma
On 21.07.25 07:49, Byungchul Park wrote:
> Hi,
>
> I focused on converting the existing APIs accessing ->pp_magic field to
> page type APIs. However, yes. Additional works would better be
> considered on top like:
>
> 1. Adjust how to store and retrieve dma index. Maybe network guys
> can work better on top.
>
> 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
>
> Byungchul
>
> ---8<---
> From 7d207a1b3e9f4ff2a72f5b54b09e3ed0c4aaaca3 Mon Sep 17 00:00:00 2001
> From: Byungchul Park <byungchul@sk.com>
> Date: Mon, 21 Jul 2025 14:05:20 +0900
> Subject: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
>
> ->pp_magic field in struct page is current used to identify if a page
> belongs to a page pool. However, page type e.i. PGTY_netpp can be used
> for that purpose.
>
> Use the page type APIs e.g. PageNetpp(), __SetPageNetpp(), and
> __ClearPageNetpp() instead, and remove the existing APIs accessing
> ->pp_magic e.g. page_pool_page_is_pp(), netmem_or_pp_magic(), and
> netmem_clear_pp_magic() since they are totally replaced.
>
> This work was inspired by the following link by Pavel:
>
> [1] https://lore.kernel.org/all/582f41c0-2742-4400-9c81-0d46bf4e8314@gmail.com/
I'm, sure you saw my comment (including my earlier suggestion for using
a page type), in particular around ...
> ---
> .../net/ethernet/mellanox/mlx5/core/en/xdp.c | 2 +-
> include/linux/mm.h | 28 ++-----------------
> include/linux/page-flags.h | 6 ++++
> include/net/netmem.h | 2 +-
> mm/page_alloc.c | 4 +--
> net/core/netmem_priv.h | 16 ++---------
> net/core/page_pool.c | 10 +++++--
> 7 files changed, 24 insertions(+), 44 deletions(-)
>
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> index 5d51600935a6..def274f5c1ca 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> @@ -707,7 +707,7 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
> xdpi = mlx5e_xdpi_fifo_pop(xdpi_fifo);
> page = xdpi.page.page;
>
> - /* No need to check page_pool_page_is_pp() as we
> + /* No need to check PageNetpp() as we
> * know this is a page_pool page.
> */
> page_pool_recycle_direct(pp_page_to_nmdesc(page)->pp,
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index ae50c1641bed..736061749535 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -4135,10 +4135,9 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> * DMA mapping IDs for page_pool
> *
> * When DMA-mapping a page, page_pool allocates an ID (from an xarray) and
> - * stashes it in the upper bits of page->pp_magic. We always want to be able to
> - * unambiguously identify page pool pages (using page_pool_page_is_pp()). Non-PP
> - * pages can have arbitrary kernel pointers stored in the same field as pp_magic
> - * (since it overlaps with page->lru.next), so we must ensure that we cannot
> + * stashes it in the upper bits of page->pp_magic. Non-PP pages can have
> + * arbitrary kernel pointers stored in the same field as pp_magic (since
> + * it overlaps with page->lru.next), so we must ensure that we cannot
> * mistake a valid kernel pointer with any of the values we write into this
> * field.
> *
> @@ -4168,25 +4167,4 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
>
> #define PP_DMA_INDEX_MASK GENMASK(PP_DMA_INDEX_BITS + PP_DMA_INDEX_SHIFT - 1, \
> PP_DMA_INDEX_SHIFT)
> -
> -/* Mask used for checking in page_pool_page_is_pp() below. page->pp_magic is
> - * OR'ed with PP_SIGNATURE after the allocation in order to preserve bit 0 for
> - * the head page of compound page and bit 1 for pfmemalloc page, as well as the
> - * bits used for the DMA index. page_is_pfmemalloc() is checked in
> - * __page_pool_put_page() to avoid recycling the pfmemalloc page.
> - */
> -#define PP_MAGIC_MASK ~(PP_DMA_INDEX_MASK | 0x3UL)
> -
> -#ifdef CONFIG_PAGE_POOL
> -static inline bool page_pool_page_is_pp(const struct page *page)
> -{
> - return (page->pp_magic & PP_MAGIC_MASK) == PP_SIGNATURE;
> -}
> -#else
> -static inline bool page_pool_page_is_pp(const struct page *page)
> -{
> - return false;
> -}
> -#endif
> -
> #endif /* _LINUX_MM_H */
> diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> index 4fe5ee67535b..906ba7c9e372 100644
> --- a/include/linux/page-flags.h
> +++ b/include/linux/page-flags.h
> @@ -957,6 +957,7 @@ enum pagetype {
> PGTY_zsmalloc = 0xf6,
> PGTY_unaccepted = 0xf7,
> PGTY_large_kmalloc = 0xf8,
> + PGTY_netpp = 0xf9,
>
> PGTY_mapcount_underflow = 0xff
> };
> @@ -1101,6 +1102,11 @@ PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
> PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
> FOLIO_TYPE_OPS(large_kmalloc, large_kmalloc)
>
> +/*
> + * Marks page_pool allocated pages.
> + */
> +PAGE_TYPE_OPS(Netpp, netpp, netpp)
> +
> /**
> * PageHuge - Determine if the page belongs to hugetlbfs
> * @page: The page to test.
> diff --git a/include/net/netmem.h b/include/net/netmem.h
> index f7dacc9e75fd..3667334e16e7 100644
> --- a/include/net/netmem.h
> +++ b/include/net/netmem.h
> @@ -298,7 +298,7 @@ static inline struct net_iov *__netmem_clear_lsb(netmem_ref netmem)
> */
> #define pp_page_to_nmdesc(p) \
> ({ \
> - DEBUG_NET_WARN_ON_ONCE(!page_pool_page_is_pp(p)); \
> + DEBUG_NET_WARN_ON_ONCE(!PageNetpp(p)); \
> __pp_page_to_nmdesc(p); \
> })
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 2ef3c07266b3..71c7666e48a9 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -898,7 +898,7 @@ static inline bool page_expected_state(struct page *page,
> #ifdef CONFIG_MEMCG
> page->memcg_data |
> #endif
> - page_pool_page_is_pp(page) |
> + PageNetpp(page) |
> (page->flags & check_flags)))
> return false;
>
> @@ -925,7 +925,7 @@ static const char *page_bad_reason(struct page *page, unsigned long flags)
> if (unlikely(page->memcg_data))
> bad_reason = "page still charged to cgroup";
> #endif
> - if (unlikely(page_pool_page_is_pp(page)))
> + if (unlikely(PageNetpp(page)))
> bad_reason = "page_pool leak";
> return bad_reason;
> }
^ this
This will not work they way you want it once you rebase on top of
linux-next, where we have (from mm/mm-stable)
commit 2dfcd1608f3a96364f10de7fcfe28727c0292e5d
Author: David Hildenbrand <david@redhat.com>
Date: Fri Jul 4 12:24:58 2025 +0200
mm/page_alloc: let page freeing clear any set page type
I commented what to do already.
--
Cheers,
David / dhildenb
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 8:05 ` David Hildenbrand
@ 2025-07-21 8:19 ` Byungchul Park
2025-07-21 8:49 ` David Hildenbrand
0 siblings, 1 reply; 14+ messages in thread
From: Byungchul Park @ 2025-07-21 8:19 UTC (permalink / raw)
To: David Hildenbrand
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, almasrymina, toke,
asml.silence, bpf, linux-rdma
On Mon, Jul 21, 2025 at 10:05:25AM +0200, David Hildenbrand wrote:
> On 21.07.25 07:49, Byungchul Park wrote:
> > Hi,
> >
> > I focused on converting the existing APIs accessing ->pp_magic field to
> > page type APIs. However, yes. Additional works would better be
> > considered on top like:
> >
> > 1. Adjust how to store and retrieve dma index. Maybe network guys
> > can work better on top.
> >
> > 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
> >
> > Byungchul
> >
> > ---8<---
> > From 7d207a1b3e9f4ff2a72f5b54b09e3ed0c4aaaca3 Mon Sep 17 00:00:00 2001
> > From: Byungchul Park <byungchul@sk.com>
> > Date: Mon, 21 Jul 2025 14:05:20 +0900
> > Subject: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
> >
> > ->pp_magic field in struct page is current used to identify if a page
> > belongs to a page pool. However, page type e.i. PGTY_netpp can be used
> > for that purpose.
> >
> > Use the page type APIs e.g. PageNetpp(), __SetPageNetpp(), and
> > __ClearPageNetpp() instead, and remove the existing APIs accessing
> > ->pp_magic e.g. page_pool_page_is_pp(), netmem_or_pp_magic(), and
> > netmem_clear_pp_magic() since they are totally replaced.
> >
> > This work was inspired by the following link by Pavel:
> >
> > [1] https://lore.kernel.org/all/582f41c0-2742-4400-9c81-0d46bf4e8314@gmail.com/
>
> I'm, sure you saw my comment (including my earlier suggestion for using
> a page type), in particular around ...
Honestly, I was too confused to understand exactly what you meant.
Maybe my bad but I was and I'm still so. Lemme add a question below.
> > ---
> > .../net/ethernet/mellanox/mlx5/core/en/xdp.c | 2 +-
> > include/linux/mm.h | 28 ++-----------------
> > include/linux/page-flags.h | 6 ++++
> > include/net/netmem.h | 2 +-
> > mm/page_alloc.c | 4 +--
> > net/core/netmem_priv.h | 16 ++---------
> > net/core/page_pool.c | 10 +++++--
> > 7 files changed, 24 insertions(+), 44 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > index 5d51600935a6..def274f5c1ca 100644
> > --- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > @@ -707,7 +707,7 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
> > xdpi = mlx5e_xdpi_fifo_pop(xdpi_fifo);
> > page = xdpi.page.page;
> >
> > - /* No need to check page_pool_page_is_pp() as we
> > + /* No need to check PageNetpp() as we
> > * know this is a page_pool page.
> > */
> > page_pool_recycle_direct(pp_page_to_nmdesc(page)->pp,
> > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > index ae50c1641bed..736061749535 100644
> > --- a/include/linux/mm.h
> > +++ b/include/linux/mm.h
> > @@ -4135,10 +4135,9 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> > * DMA mapping IDs for page_pool
> > *
> > * When DMA-mapping a page, page_pool allocates an ID (from an xarray) and
> > - * stashes it in the upper bits of page->pp_magic. We always want to be able to
> > - * unambiguously identify page pool pages (using page_pool_page_is_pp()). Non-PP
> > - * pages can have arbitrary kernel pointers stored in the same field as pp_magic
> > - * (since it overlaps with page->lru.next), so we must ensure that we cannot
> > + * stashes it in the upper bits of page->pp_magic. Non-PP pages can have
> > + * arbitrary kernel pointers stored in the same field as pp_magic (since
> > + * it overlaps with page->lru.next), so we must ensure that we cannot
> > * mistake a valid kernel pointer with any of the values we write into this
> > * field.
> > *
> > @@ -4168,25 +4167,4 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> >
> > #define PP_DMA_INDEX_MASK GENMASK(PP_DMA_INDEX_BITS + PP_DMA_INDEX_SHIFT - 1, \
> > PP_DMA_INDEX_SHIFT)
> > -
> > -/* Mask used for checking in page_pool_page_is_pp() below. page->pp_magic is
> > - * OR'ed with PP_SIGNATURE after the allocation in order to preserve bit 0 for
> > - * the head page of compound page and bit 1 for pfmemalloc page, as well as the
> > - * bits used for the DMA index. page_is_pfmemalloc() is checked in
> > - * __page_pool_put_page() to avoid recycling the pfmemalloc page.
> > - */
> > -#define PP_MAGIC_MASK ~(PP_DMA_INDEX_MASK | 0x3UL)
> > -
> > -#ifdef CONFIG_PAGE_POOL
> > -static inline bool page_pool_page_is_pp(const struct page *page)
> > -{
> > - return (page->pp_magic & PP_MAGIC_MASK) == PP_SIGNATURE;
> > -}
> > -#else
> > -static inline bool page_pool_page_is_pp(const struct page *page)
> > -{
> > - return false;
> > -}
> > -#endif
> > -
> > #endif /* _LINUX_MM_H */
> > diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> > index 4fe5ee67535b..906ba7c9e372 100644
> > --- a/include/linux/page-flags.h
> > +++ b/include/linux/page-flags.h
> > @@ -957,6 +957,7 @@ enum pagetype {
> > PGTY_zsmalloc = 0xf6,
> > PGTY_unaccepted = 0xf7,
> > PGTY_large_kmalloc = 0xf8,
> > + PGTY_netpp = 0xf9,
> >
> > PGTY_mapcount_underflow = 0xff
> > };
> > @@ -1101,6 +1102,11 @@ PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
> > PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
> > FOLIO_TYPE_OPS(large_kmalloc, large_kmalloc)
> >
> > +/*
> > + * Marks page_pool allocated pages.
> > + */
> > +PAGE_TYPE_OPS(Netpp, netpp, netpp)
> > +
> > /**
> > * PageHuge - Determine if the page belongs to hugetlbfs
> > * @page: The page to test.
> > diff --git a/include/net/netmem.h b/include/net/netmem.h
> > index f7dacc9e75fd..3667334e16e7 100644
> > --- a/include/net/netmem.h
> > +++ b/include/net/netmem.h
> > @@ -298,7 +298,7 @@ static inline struct net_iov *__netmem_clear_lsb(netmem_ref netmem)
> > */
> > #define pp_page_to_nmdesc(p) \
> > ({ \
> > - DEBUG_NET_WARN_ON_ONCE(!page_pool_page_is_pp(p)); \
> > + DEBUG_NET_WARN_ON_ONCE(!PageNetpp(p)); \
> > __pp_page_to_nmdesc(p); \
> > })
> >
> > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > index 2ef3c07266b3..71c7666e48a9 100644
> > --- a/mm/page_alloc.c
> > +++ b/mm/page_alloc.c
> > @@ -898,7 +898,7 @@ static inline bool page_expected_state(struct page *page,
> > #ifdef CONFIG_MEMCG
> > page->memcg_data |
> > #endif
> > - page_pool_page_is_pp(page) |
> > + PageNetpp(page) |
> > (page->flags & check_flags)))
> > return false;
> >
> > @@ -925,7 +925,7 @@ static const char *page_bad_reason(struct page *page, unsigned long flags)
> > if (unlikely(page->memcg_data))
> > bad_reason = "page still charged to cgroup";
> > #endif
> > - if (unlikely(page_pool_page_is_pp(page)))
> > + if (unlikely(PageNetpp(page)))
> > bad_reason = "page_pool leak";
> > return bad_reason;
> > }
>
> ^ this
>
> This will not work they way you want it once you rebase on top of
> linux-next, where we have (from mm/mm-stable)
>
> commit 2dfcd1608f3a96364f10de7fcfe28727c0292e5d
I just checked this.
So is it sufficient that I rebase on mm/mm-stable? Or should I wait for
something else? Or should I achieve this in other ways?
Byungchul
> Author: David Hildenbrand <david@redhat.com>
> Date: Fri Jul 4 12:24:58 2025 +0200
>
> mm/page_alloc: let page freeing clear any set page type
>
>
> I commented what to do already.
>
> --
> Cheers,
>
> David / dhildenb
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 8:19 ` Byungchul Park
@ 2025-07-21 8:49 ` David Hildenbrand
2025-07-22 1:04 ` Byungchul Park
2025-07-28 10:57 ` Byungchul Park
0 siblings, 2 replies; 14+ messages in thread
From: David Hildenbrand @ 2025-07-21 8:49 UTC (permalink / raw)
To: Byungchul Park
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, almasrymina, toke,
asml.silence, bpf, linux-rdma
>>
>> This will not work they way you want it once you rebase on top of
>> linux-next, where we have (from mm/mm-stable)
>>
>> commit 2dfcd1608f3a96364f10de7fcfe28727c0292e5d
>
> I just checked this.
>
> So is it sufficient that I rebase on mm/mm-stable? Or should I wait for
> something else? Or should I achieve this in other ways?
Probably best to rebase (+test) to linux-next, where that commit should
be in.
Whatever is in mm-stable is expected to go upstream in the next merge
window (iow, soon), with stable commit ids.
--
Cheers,
David / dhildenb
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 8:49 ` David Hildenbrand
@ 2025-07-22 1:04 ` Byungchul Park
2025-07-28 10:57 ` Byungchul Park
1 sibling, 0 replies; 14+ messages in thread
From: Byungchul Park @ 2025-07-22 1:04 UTC (permalink / raw)
To: David Hildenbrand
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, almasrymina, toke,
asml.silence, bpf, linux-rdma
On Mon, Jul 21, 2025 at 10:49:37AM +0200, David Hildenbrand wrote:
> > > This will not work they way you want it once you rebase on top of
> > > linux-next, where we have (from mm/mm-stable)
> > >
> > > commit 2dfcd1608f3a96364f10de7fcfe28727c0292e5d
> >
> > I just checked this.
> >
> > So is it sufficient that I rebase on mm/mm-stable? Or should I wait for
> > something else? Or should I achieve this in other ways?
>
> Probably best to rebase (+test) to linux-next, where that commit should
> be in.
I will.
Byungchul
> Whatever is in mm-stable is expected to go upstream in the next merge
> window (iow, soon), with stable commit ids.
>
> --
> Cheers,
>
> David / dhildenb
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 8:49 ` David Hildenbrand
2025-07-22 1:04 ` Byungchul Park
@ 2025-07-28 10:57 ` Byungchul Park
1 sibling, 0 replies; 14+ messages in thread
From: Byungchul Park @ 2025-07-28 10:57 UTC (permalink / raw)
To: David Hildenbrand
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, almasrymina, toke,
asml.silence, bpf, linux-rdma, sfr
On Mon, Jul 21, 2025 at 10:49:37AM +0200, David Hildenbrand wrote:
> > >
> > > This will not work they way you want it once you rebase on top of
> > > linux-next, where we have (from mm/mm-stable)
> > >
> > > commit 2dfcd1608f3a96364f10de7fcfe28727c0292e5d
> >
> > I just checked this.
> >
> > So is it sufficient that I rebase on mm/mm-stable? Or should I wait for
> > something else? Or should I achieve this in other ways?
>
> Probably best to rebase (+test) to linux-next, where that commit should
> be in.
+cc sfr@canb.auug.org.au
Byungchul
>
> Whatever is in mm-stable is expected to go upstream in the next merge
> window (iow, soon), with stable commit ids.
>
> --
> Cheers,
>
> David / dhildenb
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 5:49 [PATCH] mm, page_pool: introduce a new page type for page pool in page type Byungchul Park
2025-07-21 8:05 ` David Hildenbrand
@ 2025-07-21 11:12 ` Pavel Begunkov
2025-07-22 1:03 ` Byungchul Park
2025-07-22 22:20 ` Mina Almasry
2025-07-22 22:17 ` Mina Almasry
2 siblings, 2 replies; 14+ messages in thread
From: Pavel Begunkov @ 2025-07-21 11:12 UTC (permalink / raw)
To: Byungchul Park, linux-mm, netdev
Cc: linux-kernel, kernel_team, harry.yoo, ast, daniel, davem, kuba,
hawk, john.fastabend, sdf, saeedm, leon, tariqt, mbloch,
andrew+netdev, edumazet, pabeni, akpm, david, lorenzo.stoakes,
Liam.Howlett, vbabka, rppt, surenb, mhocko, horms, jackmanb,
hannes, ziy, ilias.apalodimas, willy, brauner, kas, yuzhao,
usamaarif642, baolin.wang, almasrymina, toke, bpf, linux-rdma
On 7/21/25 06:49, Byungchul Park wrote:
> Hi,
>
> I focused on converting the existing APIs accessing ->pp_magic field to
> page type APIs. However, yes. Additional works would better be
> considered on top like:
>
> 1. Adjust how to store and retrieve dma index. Maybe network guys
> can work better on top.
>
> 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
Don't be in a hurry, I've got a branch, but as mentioned before,
it'll be for-6.18. And there will also be more time for testing.
> This work was inspired by the following link by Pavel:
The idea came from David, let's add
Suggested-by: David Hildenbrand <david@redhat.com>
...> -
> static inline bool netmem_is_pp(netmem_ref netmem)
> {
> - return (netmem_get_pp_magic(netmem) & PP_MAGIC_MASK) == PP_SIGNATURE;
> + if (netmem_is_net_iov(netmem))
This needs to return false for tx niovs. Seems like all callers are
gated on ->pp_recycle, so maybe it's fine, but we can at least
check pp. Mina, you've been checking tx doesn't mix with rx, any
opinion on that?
Question to net maintainers, can a ->pp_recycle marked skb contain
not page pool originated pages or a mix?
--
Pavel Begunkov
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 11:12 ` Pavel Begunkov
@ 2025-07-22 1:03 ` Byungchul Park
2025-07-22 22:20 ` Mina Almasry
1 sibling, 0 replies; 14+ messages in thread
From: Byungchul Park @ 2025-07-22 1:03 UTC (permalink / raw)
To: Pavel Begunkov
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm, david,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, almasrymina, toke, bpf,
linux-rdma
On Mon, Jul 21, 2025 at 12:12:39PM +0100, Pavel Begunkov wrote:
> On 7/21/25 06:49, Byungchul Park wrote:
> > Hi,
> >
> > I focused on converting the existing APIs accessing ->pp_magic field to
> > page type APIs. However, yes. Additional works would better be
> > considered on top like:
> >
> > 1. Adjust how to store and retrieve dma index. Maybe network guys
> > can work better on top.
> >
> > 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
>
> Don't be in a hurry, I've got a branch, but as mentioned before,
> it'll be for-6.18. And there will also be more time for testing.
I'm not. I listed the two items above wishing someone to work on it on
top of this patch in the future.
> > This work was inspired by the following link by Pavel:
>
> The idea came from David, let's add
>
> Suggested-by: David Hildenbrand <david@redhat.com>
Okay. I will replace the current one with this.
> ...> -
> > static inline bool netmem_is_pp(netmem_ref netmem)
> > {
> > - return (netmem_get_pp_magic(netmem) & PP_MAGIC_MASK) == PP_SIGNATURE;
> > + if (netmem_is_net_iov(netmem))
>
> This needs to return false for tx niovs. Seems like all callers are
> gated on ->pp_recycle, so maybe it's fine, but we can at least
> check pp. Mina, you've been checking tx doesn't mix with rx, any
> opinion on that?
I will wait for the answer of this before going ahead.
> Question to net maintainers, can a ->pp_recycle marked skb contain
> not page pool originated pages or a mix?
This too.
Byungchul
> --
> Pavel Begunkov
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 11:12 ` Pavel Begunkov
2025-07-22 1:03 ` Byungchul Park
@ 2025-07-22 22:20 ` Mina Almasry
1 sibling, 0 replies; 14+ messages in thread
From: Mina Almasry @ 2025-07-22 22:20 UTC (permalink / raw)
To: Pavel Begunkov
Cc: Byungchul Park, linux-mm, netdev, linux-kernel, kernel_team,
harry.yoo, ast, daniel, davem, kuba, hawk, john.fastabend, sdf,
saeedm, leon, tariqt, mbloch, andrew+netdev, edumazet, pabeni,
akpm, david, lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb,
mhocko, horms, jackmanb, hannes, ziy, ilias.apalodimas, willy,
brauner, kas, yuzhao, usamaarif642, baolin.wang, toke, bpf,
linux-rdma
On Mon, Jul 21, 2025 at 4:11 AM Pavel Begunkov <asml.silence@gmail.com> wrote:
>
> On 7/21/25 06:49, Byungchul Park wrote:
> > Hi,
> >
> > I focused on converting the existing APIs accessing ->pp_magic field to
> > page type APIs. However, yes. Additional works would better be
> > considered on top like:
> >
> > 1. Adjust how to store and retrieve dma index. Maybe network guys
> > can work better on top.
> >
> > 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
>
> Don't be in a hurry, I've got a branch, but as mentioned before,
> it'll be for-6.18. And there will also be more time for testing.
>
> > This work was inspired by the following link by Pavel:
>
> The idea came from David, let's add
>
> Suggested-by: David Hildenbrand <david@redhat.com>
>
> ...> -
> > static inline bool netmem_is_pp(netmem_ref netmem)
> > {
> > - return (netmem_get_pp_magic(netmem) & PP_MAGIC_MASK) == PP_SIGNATURE;
> > + if (netmem_is_net_iov(netmem))
>
> This needs to return false for tx niovs. Seems like all callers are
> gated on ->pp_recycle, so maybe it's fine, but we can at least
> check pp. Mina, you've been checking tx doesn't mix with rx, any
> opinion on that?
>
> Question to net maintainers, can a ->pp_recycle marked skb contain
> not page pool originated pages or a mix?
>
IIRC last I looked at the code ->pp_recycle technically means it could
be a mix. Technically it means "consider this netmem for pp recycling
when the skb is freed." and non-pp netmems don't get recycled to the
pp obviously, because napi_pp_put_page rejects recycling them.
--
Thanks,
Mina
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-21 5:49 [PATCH] mm, page_pool: introduce a new page type for page pool in page type Byungchul Park
2025-07-21 8:05 ` David Hildenbrand
2025-07-21 11:12 ` Pavel Begunkov
@ 2025-07-22 22:17 ` Mina Almasry
2025-07-23 4:46 ` Byungchul Park
2025-07-25 6:50 ` Byungchul Park
2 siblings, 2 replies; 14+ messages in thread
From: Mina Almasry @ 2025-07-22 22:17 UTC (permalink / raw)
To: Byungchul Park
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm, david,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, toke, asml.silence, bpf,
linux-rdma
On Sun, Jul 20, 2025 at 10:49 PM Byungchul Park <byungchul@sk.com> wrote:
>
> Hi,
>
> I focused on converting the existing APIs accessing ->pp_magic field to
> page type APIs. However, yes. Additional works would better be
> considered on top like:
>
> 1. Adjust how to store and retrieve dma index. Maybe network guys
> can work better on top.
>
> 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
>
> Byungchul
>
> ---8<---
> From 7d207a1b3e9f4ff2a72f5b54b09e3ed0c4aaaca3 Mon Sep 17 00:00:00 2001
> From: Byungchul Park <byungchul@sk.com>
> Date: Mon, 21 Jul 2025 14:05:20 +0900
> Subject: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
>
> ->pp_magic field in struct page is current used to identify if a page
> belongs to a page pool. However, page type e.i. PGTY_netpp can be used
> for that purpose.
>
> Use the page type APIs e.g. PageNetpp(), __SetPageNetpp(), and
> __ClearPageNetpp() instead, and remove the existing APIs accessing
> ->pp_magic e.g. page_pool_page_is_pp(), netmem_or_pp_magic(), and
> netmem_clear_pp_magic() since they are totally replaced.
>
> This work was inspired by the following link by Pavel:
>
> [1] https://lore.kernel.org/all/582f41c0-2742-4400-9c81-0d46bf4e8314@gmail.com/
>
> Suggested-by: Pavel Begunkov <asml.silence@gmail.com>
> Signed-off-by: Byungchul Park <byungchul@sk.com>
> ---
> .../net/ethernet/mellanox/mlx5/core/en/xdp.c | 2 +-
> include/linux/mm.h | 28 ++-----------------
> include/linux/page-flags.h | 6 ++++
> include/net/netmem.h | 2 +-
> mm/page_alloc.c | 4 +--
> net/core/netmem_priv.h | 16 ++---------
> net/core/page_pool.c | 10 +++++--
> 7 files changed, 24 insertions(+), 44 deletions(-)
>
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> index 5d51600935a6..def274f5c1ca 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> @@ -707,7 +707,7 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
> xdpi = mlx5e_xdpi_fifo_pop(xdpi_fifo);
> page = xdpi.page.page;
>
> - /* No need to check page_pool_page_is_pp() as we
> + /* No need to check PageNetpp() as we
> * know this is a page_pool page.
> */
> page_pool_recycle_direct(pp_page_to_nmdesc(page)->pp,
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index ae50c1641bed..736061749535 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -4135,10 +4135,9 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> * DMA mapping IDs for page_pool
> *
> * When DMA-mapping a page, page_pool allocates an ID (from an xarray) and
> - * stashes it in the upper bits of page->pp_magic. We always want to be able to
> - * unambiguously identify page pool pages (using page_pool_page_is_pp()). Non-PP
> - * pages can have arbitrary kernel pointers stored in the same field as pp_magic
> - * (since it overlaps with page->lru.next), so we must ensure that we cannot
> + * stashes it in the upper bits of page->pp_magic. Non-PP pages can have
> + * arbitrary kernel pointers stored in the same field as pp_magic (since
> + * it overlaps with page->lru.next), so we must ensure that we cannot
> * mistake a valid kernel pointer with any of the values we write into this
> * field.
> *
> @@ -4168,25 +4167,4 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
>
> #define PP_DMA_INDEX_MASK GENMASK(PP_DMA_INDEX_BITS + PP_DMA_INDEX_SHIFT - 1, \
> PP_DMA_INDEX_SHIFT)
> -
> -/* Mask used for checking in page_pool_page_is_pp() below. page->pp_magic is
> - * OR'ed with PP_SIGNATURE after the allocation in order to preserve bit 0 for
> - * the head page of compound page and bit 1 for pfmemalloc page, as well as the
> - * bits used for the DMA index. page_is_pfmemalloc() is checked in
> - * __page_pool_put_page() to avoid recycling the pfmemalloc page.
> - */
> -#define PP_MAGIC_MASK ~(PP_DMA_INDEX_MASK | 0x3UL)
> -
> -#ifdef CONFIG_PAGE_POOL
> -static inline bool page_pool_page_is_pp(const struct page *page)
> -{
> - return (page->pp_magic & PP_MAGIC_MASK) == PP_SIGNATURE;
> -}
> -#else
> -static inline bool page_pool_page_is_pp(const struct page *page)
> -{
> - return false;
> -}
> -#endif
> -
> #endif /* _LINUX_MM_H */
> diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> index 4fe5ee67535b..906ba7c9e372 100644
> --- a/include/linux/page-flags.h
> +++ b/include/linux/page-flags.h
> @@ -957,6 +957,7 @@ enum pagetype {
> PGTY_zsmalloc = 0xf6,
> PGTY_unaccepted = 0xf7,
> PGTY_large_kmalloc = 0xf8,
> + PGTY_netpp = 0xf9,
>
> PGTY_mapcount_underflow = 0xff
> };
> @@ -1101,6 +1102,11 @@ PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
> PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
> FOLIO_TYPE_OPS(large_kmalloc, large_kmalloc)
>
> +/*
> + * Marks page_pool allocated pages.
> + */
> +PAGE_TYPE_OPS(Netpp, netpp, netpp)
> +
> /**
> * PageHuge - Determine if the page belongs to hugetlbfs
> * @page: The page to test.
> diff --git a/include/net/netmem.h b/include/net/netmem.h
> index f7dacc9e75fd..3667334e16e7 100644
> --- a/include/net/netmem.h
> +++ b/include/net/netmem.h
> @@ -298,7 +298,7 @@ static inline struct net_iov *__netmem_clear_lsb(netmem_ref netmem)
> */
> #define pp_page_to_nmdesc(p) \
> ({ \
> - DEBUG_NET_WARN_ON_ONCE(!page_pool_page_is_pp(p)); \
> + DEBUG_NET_WARN_ON_ONCE(!PageNetpp(p)); \
> __pp_page_to_nmdesc(p); \
> })
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 2ef3c07266b3..71c7666e48a9 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -898,7 +898,7 @@ static inline bool page_expected_state(struct page *page,
> #ifdef CONFIG_MEMCG
> page->memcg_data |
> #endif
> - page_pool_page_is_pp(page) |
> + PageNetpp(page) |
> (page->flags & check_flags)))
> return false;
>
> @@ -925,7 +925,7 @@ static const char *page_bad_reason(struct page *page, unsigned long flags)
> if (unlikely(page->memcg_data))
> bad_reason = "page still charged to cgroup";
> #endif
> - if (unlikely(page_pool_page_is_pp(page)))
> + if (unlikely(PageNetpp(page)))
> bad_reason = "page_pool leak";
> return bad_reason;
> }
> diff --git a/net/core/netmem_priv.h b/net/core/netmem_priv.h
> index cd95394399b4..39a97703d9ed 100644
> --- a/net/core/netmem_priv.h
> +++ b/net/core/netmem_priv.h
> @@ -8,21 +8,11 @@ static inline unsigned long netmem_get_pp_magic(netmem_ref netmem)
> return __netmem_clear_lsb(netmem)->pp_magic & ~PP_DMA_INDEX_MASK;
> }
>
> -static inline void netmem_or_pp_magic(netmem_ref netmem, unsigned long pp_magic)
> -{
> - __netmem_clear_lsb(netmem)->pp_magic |= pp_magic;
> -}
> -
> -static inline void netmem_clear_pp_magic(netmem_ref netmem)
> -{
> - WARN_ON_ONCE(__netmem_clear_lsb(netmem)->pp_magic & PP_DMA_INDEX_MASK);
> -
> - __netmem_clear_lsb(netmem)->pp_magic = 0;
> -}
> -
> static inline bool netmem_is_pp(netmem_ref netmem)
> {
> - return (netmem_get_pp_magic(netmem) & PP_MAGIC_MASK) == PP_SIGNATURE;
> + if (netmem_is_net_iov(netmem))
> + return true;
As Pavel alludes, this is dubious, and at least it's difficult to
reason about it.
There could be net_iovs that are not attached to pp, and should not be
treated as pp memory. These are in the devmem (and future net_iov) tx
paths.
We need a way to tell if a net_iov is pp or not. A couple of options:
1. We could have it such that if net_iov->pp is set, then the
netmem_is_pp == true, otherwise false.
2. We could implement a page-flags equivalent for net_iov.
Option #1 is simpler and is my preferred. To do that properly, you need to:
1. Make sure everywhere net_iovs are allocated that pp=NULL in the
non-pp case and pp=non NULL in the pp case. those callsites are
net_devmem_bind_dmabuf (devmem rx & tx path), io_zcrx_create_area
(io_uring rx path).
2. Change netmem_is_pp to check net_iov->pp in the net_iov case.
--
Thanks,
Mina
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-22 22:17 ` Mina Almasry
@ 2025-07-23 4:46 ` Byungchul Park
2025-07-24 21:23 ` Mina Almasry
2025-07-25 6:50 ` Byungchul Park
1 sibling, 1 reply; 14+ messages in thread
From: Byungchul Park @ 2025-07-23 4:46 UTC (permalink / raw)
To: Mina Almasry
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm, david,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, toke, asml.silence, bpf,
linux-rdma
On Tue, Jul 22, 2025 at 03:17:15PM -0700, Mina Almasry wrote:
> On Sun, Jul 20, 2025 at 10:49 PM Byungchul Park <byungchul@sk.com> wrote:
> >
> > Hi,
> >
> > I focused on converting the existing APIs accessing ->pp_magic field to
> > page type APIs. However, yes. Additional works would better be
> > considered on top like:
> >
> > 1. Adjust how to store and retrieve dma index. Maybe network guys
> > can work better on top.
> >
> > 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
> >
> > Byungchul
> >
> > ---8<---
> > From 7d207a1b3e9f4ff2a72f5b54b09e3ed0c4aaaca3 Mon Sep 17 00:00:00 2001
> > From: Byungchul Park <byungchul@sk.com>
> > Date: Mon, 21 Jul 2025 14:05:20 +0900
> > Subject: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
> >
> > ->pp_magic field in struct page is current used to identify if a page
> > belongs to a page pool. However, page type e.i. PGTY_netpp can be used
> > for that purpose.
> >
> > Use the page type APIs e.g. PageNetpp(), __SetPageNetpp(), and
> > __ClearPageNetpp() instead, and remove the existing APIs accessing
> > ->pp_magic e.g. page_pool_page_is_pp(), netmem_or_pp_magic(), and
> > netmem_clear_pp_magic() since they are totally replaced.
> >
> > This work was inspired by the following link by Pavel:
> >
> > [1] https://lore.kernel.org/all/582f41c0-2742-4400-9c81-0d46bf4e8314@gmail.com/
> >
> > Suggested-by: Pavel Begunkov <asml.silence@gmail.com>
> > Signed-off-by: Byungchul Park <byungchul@sk.com>
> > ---
> > .../net/ethernet/mellanox/mlx5/core/en/xdp.c | 2 +-
> > include/linux/mm.h | 28 ++-----------------
> > include/linux/page-flags.h | 6 ++++
> > include/net/netmem.h | 2 +-
> > mm/page_alloc.c | 4 +--
> > net/core/netmem_priv.h | 16 ++---------
> > net/core/page_pool.c | 10 +++++--
> > 7 files changed, 24 insertions(+), 44 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > index 5d51600935a6..def274f5c1ca 100644
> > --- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > @@ -707,7 +707,7 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
> > xdpi = mlx5e_xdpi_fifo_pop(xdpi_fifo);
> > page = xdpi.page.page;
> >
> > - /* No need to check page_pool_page_is_pp() as we
> > + /* No need to check PageNetpp() as we
> > * know this is a page_pool page.
> > */
> > page_pool_recycle_direct(pp_page_to_nmdesc(page)->pp,
> > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > index ae50c1641bed..736061749535 100644
> > --- a/include/linux/mm.h
> > +++ b/include/linux/mm.h
> > @@ -4135,10 +4135,9 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> > * DMA mapping IDs for page_pool
> > *
> > * When DMA-mapping a page, page_pool allocates an ID (from an xarray) and
> > - * stashes it in the upper bits of page->pp_magic. We always want to be able to
> > - * unambiguously identify page pool pages (using page_pool_page_is_pp()). Non-PP
> > - * pages can have arbitrary kernel pointers stored in the same field as pp_magic
> > - * (since it overlaps with page->lru.next), so we must ensure that we cannot
> > + * stashes it in the upper bits of page->pp_magic. Non-PP pages can have
> > + * arbitrary kernel pointers stored in the same field as pp_magic (since
> > + * it overlaps with page->lru.next), so we must ensure that we cannot
> > * mistake a valid kernel pointer with any of the values we write into this
> > * field.
> > *
> > @@ -4168,25 +4167,4 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> >
> > #define PP_DMA_INDEX_MASK GENMASK(PP_DMA_INDEX_BITS + PP_DMA_INDEX_SHIFT - 1, \
> > PP_DMA_INDEX_SHIFT)
> > -
> > -/* Mask used for checking in page_pool_page_is_pp() below. page->pp_magic is
> > - * OR'ed with PP_SIGNATURE after the allocation in order to preserve bit 0 for
> > - * the head page of compound page and bit 1 for pfmemalloc page, as well as the
> > - * bits used for the DMA index. page_is_pfmemalloc() is checked in
> > - * __page_pool_put_page() to avoid recycling the pfmemalloc page.
> > - */
> > -#define PP_MAGIC_MASK ~(PP_DMA_INDEX_MASK | 0x3UL)
> > -
> > -#ifdef CONFIG_PAGE_POOL
> > -static inline bool page_pool_page_is_pp(const struct page *page)
> > -{
> > - return (page->pp_magic & PP_MAGIC_MASK) == PP_SIGNATURE;
> > -}
> > -#else
> > -static inline bool page_pool_page_is_pp(const struct page *page)
> > -{
> > - return false;
> > -}
> > -#endif
> > -
> > #endif /* _LINUX_MM_H */
> > diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> > index 4fe5ee67535b..906ba7c9e372 100644
> > --- a/include/linux/page-flags.h
> > +++ b/include/linux/page-flags.h
> > @@ -957,6 +957,7 @@ enum pagetype {
> > PGTY_zsmalloc = 0xf6,
> > PGTY_unaccepted = 0xf7,
> > PGTY_large_kmalloc = 0xf8,
> > + PGTY_netpp = 0xf9,
> >
> > PGTY_mapcount_underflow = 0xff
> > };
> > @@ -1101,6 +1102,11 @@ PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
> > PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
> > FOLIO_TYPE_OPS(large_kmalloc, large_kmalloc)
> >
> > +/*
> > + * Marks page_pool allocated pages.
> > + */
> > +PAGE_TYPE_OPS(Netpp, netpp, netpp)
> > +
> > /**
> > * PageHuge - Determine if the page belongs to hugetlbfs
> > * @page: The page to test.
> > diff --git a/include/net/netmem.h b/include/net/netmem.h
> > index f7dacc9e75fd..3667334e16e7 100644
> > --- a/include/net/netmem.h
> > +++ b/include/net/netmem.h
> > @@ -298,7 +298,7 @@ static inline struct net_iov *__netmem_clear_lsb(netmem_ref netmem)
> > */
> > #define pp_page_to_nmdesc(p) \
> > ({ \
> > - DEBUG_NET_WARN_ON_ONCE(!page_pool_page_is_pp(p)); \
> > + DEBUG_NET_WARN_ON_ONCE(!PageNetpp(p)); \
> > __pp_page_to_nmdesc(p); \
> > })
> >
> > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > index 2ef3c07266b3..71c7666e48a9 100644
> > --- a/mm/page_alloc.c
> > +++ b/mm/page_alloc.c
> > @@ -898,7 +898,7 @@ static inline bool page_expected_state(struct page *page,
> > #ifdef CONFIG_MEMCG
> > page->memcg_data |
> > #endif
> > - page_pool_page_is_pp(page) |
> > + PageNetpp(page) |
> > (page->flags & check_flags)))
> > return false;
> >
> > @@ -925,7 +925,7 @@ static const char *page_bad_reason(struct page *page, unsigned long flags)
> > if (unlikely(page->memcg_data))
> > bad_reason = "page still charged to cgroup";
> > #endif
> > - if (unlikely(page_pool_page_is_pp(page)))
> > + if (unlikely(PageNetpp(page)))
> > bad_reason = "page_pool leak";
> > return bad_reason;
> > }
> > diff --git a/net/core/netmem_priv.h b/net/core/netmem_priv.h
> > index cd95394399b4..39a97703d9ed 100644
> > --- a/net/core/netmem_priv.h
> > +++ b/net/core/netmem_priv.h
> > @@ -8,21 +8,11 @@ static inline unsigned long netmem_get_pp_magic(netmem_ref netmem)
> > return __netmem_clear_lsb(netmem)->pp_magic & ~PP_DMA_INDEX_MASK;
> > }
> >
> > -static inline void netmem_or_pp_magic(netmem_ref netmem, unsigned long pp_magic)
> > -{
> > - __netmem_clear_lsb(netmem)->pp_magic |= pp_magic;
> > -}
> > -
> > -static inline void netmem_clear_pp_magic(netmem_ref netmem)
> > -{
> > - WARN_ON_ONCE(__netmem_clear_lsb(netmem)->pp_magic & PP_DMA_INDEX_MASK);
> > -
> > - __netmem_clear_lsb(netmem)->pp_magic = 0;
> > -}
> > -
> > static inline bool netmem_is_pp(netmem_ref netmem)
> > {
> > - return (netmem_get_pp_magic(netmem) & PP_MAGIC_MASK) == PP_SIGNATURE;
> > + if (netmem_is_net_iov(netmem))
> > + return true;
>
> As Pavel alludes, this is dubious, and at least it's difficult to
> reason about it.
>
> There could be net_iovs that are not attached to pp, and should not be
> treated as pp memory. These are in the devmem (and future net_iov) tx
> paths.
>
> We need a way to tell if a net_iov is pp or not. A couple of options:
>
> 1. We could have it such that if net_iov->pp is set, then the
> netmem_is_pp == true, otherwise false.
> 2. We could implement a page-flags equivalent for net_iov.
>
> Option #1 is simpler and is my preferred. To do that properly, you need to:
>
> 1. Make sure everywhere net_iovs are allocated that pp=NULL in the
> non-pp case and pp=non NULL in the pp case. those callsites are
> net_devmem_bind_dmabuf (devmem rx & tx path), io_zcrx_create_area
> (io_uring rx path).
>
> 2. Change netmem_is_pp to check net_iov->pp in the net_iov case.
Good idea, but I'm not sure if I could work on it without consuming your
additional review efforts. Can anyone add net_iov_is_pp() helper?
Or use the page type, Netpp, as an additional way to identify if it's a
pp page for system memory, keeping the current way using ->pp_magic.
So the page type, Netpp, is used for system memory, and ->pp_magic is
used for net_iov. The clean up for ->pp_magic can be done if needed.
Byungchul
> --
> Thanks,
> Mina
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-23 4:46 ` Byungchul Park
@ 2025-07-24 21:23 ` Mina Almasry
2025-07-25 0:26 ` Byungchul Park
0 siblings, 1 reply; 14+ messages in thread
From: Mina Almasry @ 2025-07-24 21:23 UTC (permalink / raw)
To: Byungchul Park
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm, david,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, toke, asml.silence, bpf,
linux-rdma
On Tue, Jul 22, 2025 at 9:46 PM Byungchul Park <byungchul@sk.com> wrote:
>
> On Tue, Jul 22, 2025 at 03:17:15PM -0700, Mina Almasry wrote:
> > On Sun, Jul 20, 2025 at 10:49 PM Byungchul Park <byungchul@sk.com> wrote:
> > >
> > > Hi,
> > >
> > > I focused on converting the existing APIs accessing ->pp_magic field to
> > > page type APIs. However, yes. Additional works would better be
> > > considered on top like:
> > >
> > > 1. Adjust how to store and retrieve dma index. Maybe network guys
> > > can work better on top.
> > >
> > > 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
> > >
> > > Byungchul
> > >
> > > ---8<---
> > > From 7d207a1b3e9f4ff2a72f5b54b09e3ed0c4aaaca3 Mon Sep 17 00:00:00 2001
> > > From: Byungchul Park <byungchul@sk.com>
> > > Date: Mon, 21 Jul 2025 14:05:20 +0900
> > > Subject: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
> > >
> > > ->pp_magic field in struct page is current used to identify if a page
> > > belongs to a page pool. However, page type e.i. PGTY_netpp can be used
> > > for that purpose.
> > >
> > > Use the page type APIs e.g. PageNetpp(), __SetPageNetpp(), and
> > > __ClearPageNetpp() instead, and remove the existing APIs accessing
> > > ->pp_magic e.g. page_pool_page_is_pp(), netmem_or_pp_magic(), and
> > > netmem_clear_pp_magic() since they are totally replaced.
> > >
> > > This work was inspired by the following link by Pavel:
> > >
> > > [1] https://lore.kernel.org/all/582f41c0-2742-4400-9c81-0d46bf4e8314@gmail.com/
> > >
> > > Suggested-by: Pavel Begunkov <asml.silence@gmail.com>
> > > Signed-off-by: Byungchul Park <byungchul@sk.com>
> > > ---
> > > .../net/ethernet/mellanox/mlx5/core/en/xdp.c | 2 +-
> > > include/linux/mm.h | 28 ++-----------------
> > > include/linux/page-flags.h | 6 ++++
> > > include/net/netmem.h | 2 +-
> > > mm/page_alloc.c | 4 +--
> > > net/core/netmem_priv.h | 16 ++---------
> > > net/core/page_pool.c | 10 +++++--
> > > 7 files changed, 24 insertions(+), 44 deletions(-)
> > >
> > > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > > index 5d51600935a6..def274f5c1ca 100644
> > > --- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > > +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > > @@ -707,7 +707,7 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
> > > xdpi = mlx5e_xdpi_fifo_pop(xdpi_fifo);
> > > page = xdpi.page.page;
> > >
> > > - /* No need to check page_pool_page_is_pp() as we
> > > + /* No need to check PageNetpp() as we
> > > * know this is a page_pool page.
> > > */
> > > page_pool_recycle_direct(pp_page_to_nmdesc(page)->pp,
> > > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > > index ae50c1641bed..736061749535 100644
> > > --- a/include/linux/mm.h
> > > +++ b/include/linux/mm.h
> > > @@ -4135,10 +4135,9 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> > > * DMA mapping IDs for page_pool
> > > *
> > > * When DMA-mapping a page, page_pool allocates an ID (from an xarray) and
> > > - * stashes it in the upper bits of page->pp_magic. We always want to be able to
> > > - * unambiguously identify page pool pages (using page_pool_page_is_pp()). Non-PP
> > > - * pages can have arbitrary kernel pointers stored in the same field as pp_magic
> > > - * (since it overlaps with page->lru.next), so we must ensure that we cannot
> > > + * stashes it in the upper bits of page->pp_magic. Non-PP pages can have
> > > + * arbitrary kernel pointers stored in the same field as pp_magic (since
> > > + * it overlaps with page->lru.next), so we must ensure that we cannot
> > > * mistake a valid kernel pointer with any of the values we write into this
> > > * field.
> > > *
> > > @@ -4168,25 +4167,4 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> > >
> > > #define PP_DMA_INDEX_MASK GENMASK(PP_DMA_INDEX_BITS + PP_DMA_INDEX_SHIFT - 1, \
> > > PP_DMA_INDEX_SHIFT)
> > > -
> > > -/* Mask used for checking in page_pool_page_is_pp() below. page->pp_magic is
> > > - * OR'ed with PP_SIGNATURE after the allocation in order to preserve bit 0 for
> > > - * the head page of compound page and bit 1 for pfmemalloc page, as well as the
> > > - * bits used for the DMA index. page_is_pfmemalloc() is checked in
> > > - * __page_pool_put_page() to avoid recycling the pfmemalloc page.
> > > - */
> > > -#define PP_MAGIC_MASK ~(PP_DMA_INDEX_MASK | 0x3UL)
> > > -
> > > -#ifdef CONFIG_PAGE_POOL
> > > -static inline bool page_pool_page_is_pp(const struct page *page)
> > > -{
> > > - return (page->pp_magic & PP_MAGIC_MASK) == PP_SIGNATURE;
> > > -}
> > > -#else
> > > -static inline bool page_pool_page_is_pp(const struct page *page)
> > > -{
> > > - return false;
> > > -}
> > > -#endif
> > > -
> > > #endif /* _LINUX_MM_H */
> > > diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> > > index 4fe5ee67535b..906ba7c9e372 100644
> > > --- a/include/linux/page-flags.h
> > > +++ b/include/linux/page-flags.h
> > > @@ -957,6 +957,7 @@ enum pagetype {
> > > PGTY_zsmalloc = 0xf6,
> > > PGTY_unaccepted = 0xf7,
> > > PGTY_large_kmalloc = 0xf8,
> > > + PGTY_netpp = 0xf9,
> > >
> > > PGTY_mapcount_underflow = 0xff
> > > };
> > > @@ -1101,6 +1102,11 @@ PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
> > > PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
> > > FOLIO_TYPE_OPS(large_kmalloc, large_kmalloc)
> > >
> > > +/*
> > > + * Marks page_pool allocated pages.
> > > + */
> > > +PAGE_TYPE_OPS(Netpp, netpp, netpp)
> > > +
> > > /**
> > > * PageHuge - Determine if the page belongs to hugetlbfs
> > > * @page: The page to test.
> > > diff --git a/include/net/netmem.h b/include/net/netmem.h
> > > index f7dacc9e75fd..3667334e16e7 100644
> > > --- a/include/net/netmem.h
> > > +++ b/include/net/netmem.h
> > > @@ -298,7 +298,7 @@ static inline struct net_iov *__netmem_clear_lsb(netmem_ref netmem)
> > > */
> > > #define pp_page_to_nmdesc(p) \
> > > ({ \
> > > - DEBUG_NET_WARN_ON_ONCE(!page_pool_page_is_pp(p)); \
> > > + DEBUG_NET_WARN_ON_ONCE(!PageNetpp(p)); \
> > > __pp_page_to_nmdesc(p); \
> > > })
> > >
> > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > > index 2ef3c07266b3..71c7666e48a9 100644
> > > --- a/mm/page_alloc.c
> > > +++ b/mm/page_alloc.c
> > > @@ -898,7 +898,7 @@ static inline bool page_expected_state(struct page *page,
> > > #ifdef CONFIG_MEMCG
> > > page->memcg_data |
> > > #endif
> > > - page_pool_page_is_pp(page) |
> > > + PageNetpp(page) |
> > > (page->flags & check_flags)))
> > > return false;
> > >
> > > @@ -925,7 +925,7 @@ static const char *page_bad_reason(struct page *page, unsigned long flags)
> > > if (unlikely(page->memcg_data))
> > > bad_reason = "page still charged to cgroup";
> > > #endif
> > > - if (unlikely(page_pool_page_is_pp(page)))
> > > + if (unlikely(PageNetpp(page)))
> > > bad_reason = "page_pool leak";
> > > return bad_reason;
> > > }
> > > diff --git a/net/core/netmem_priv.h b/net/core/netmem_priv.h
> > > index cd95394399b4..39a97703d9ed 100644
> > > --- a/net/core/netmem_priv.h
> > > +++ b/net/core/netmem_priv.h
> > > @@ -8,21 +8,11 @@ static inline unsigned long netmem_get_pp_magic(netmem_ref netmem)
> > > return __netmem_clear_lsb(netmem)->pp_magic & ~PP_DMA_INDEX_MASK;
> > > }
> > >
> > > -static inline void netmem_or_pp_magic(netmem_ref netmem, unsigned long pp_magic)
> > > -{
> > > - __netmem_clear_lsb(netmem)->pp_magic |= pp_magic;
> > > -}
> > > -
> > > -static inline void netmem_clear_pp_magic(netmem_ref netmem)
> > > -{
> > > - WARN_ON_ONCE(__netmem_clear_lsb(netmem)->pp_magic & PP_DMA_INDEX_MASK);
> > > -
> > > - __netmem_clear_lsb(netmem)->pp_magic = 0;
> > > -}
> > > -
> > > static inline bool netmem_is_pp(netmem_ref netmem)
> > > {
> > > - return (netmem_get_pp_magic(netmem) & PP_MAGIC_MASK) == PP_SIGNATURE;
> > > + if (netmem_is_net_iov(netmem))
> > > + return true;
> >
> > As Pavel alludes, this is dubious, and at least it's difficult to
> > reason about it.
> >
> > There could be net_iovs that are not attached to pp, and should not be
> > treated as pp memory. These are in the devmem (and future net_iov) tx
> > paths.
> >
> > We need a way to tell if a net_iov is pp or not. A couple of options:
> >
> > 1. We could have it such that if net_iov->pp is set, then the
> > netmem_is_pp == true, otherwise false.
> > 2. We could implement a page-flags equivalent for net_iov.
> >
> > Option #1 is simpler and is my preferred. To do that properly, you need to:
> >
> > 1. Make sure everywhere net_iovs are allocated that pp=NULL in the
> > non-pp case and pp=non NULL in the pp case. those callsites are
> > net_devmem_bind_dmabuf (devmem rx & tx path), io_zcrx_create_area
> > (io_uring rx path).
> >
> > 2. Change netmem_is_pp to check net_iov->pp in the net_iov case.
>
> Good idea, but I'm not sure if I could work on it without consuming your
> additional review efforts. Can anyone add net_iov_is_pp() helper?
>
Things did indeed get busy for me with work work the past week and I
still need to look at your merged netmem desc series, but I'm happy to
review whenever I can.
> Or use the page type, Netpp, as an additional way to identify if it's a
> pp page for system memory, keeping the current way using ->pp_magic.
> So the page type, Netpp, is used for system memory, and ->pp_magic is
> used for net_iov. The clean up for ->pp_magic can be done if needed.
>
IMO I would like to avoid deviations like this, especially since
->pp_magic is in the netmem_desc struct that is now shared between
page and net_iov. I'd rather both use pp_magic or both not, but that
may just be me.
--
Thanks,
Mina
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-24 21:23 ` Mina Almasry
@ 2025-07-25 0:26 ` Byungchul Park
0 siblings, 0 replies; 14+ messages in thread
From: Byungchul Park @ 2025-07-25 0:26 UTC (permalink / raw)
To: Mina Almasry
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm, david,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, toke, asml.silence, bpf,
linux-rdma
On Thu, Jul 24, 2025 at 02:23:05PM -0700, Mina Almasry wrote:
> On Tue, Jul 22, 2025 at 9:46 PM Byungchul Park <byungchul@sk.com> wrote:
> >
> > On Tue, Jul 22, 2025 at 03:17:15PM -0700, Mina Almasry wrote:
> > > On Sun, Jul 20, 2025 at 10:49 PM Byungchul Park <byungchul@sk.com> wrote:
> > > >
> > > > Hi,
> > > >
> > > > I focused on converting the existing APIs accessing ->pp_magic field to
> > > > page type APIs. However, yes. Additional works would better be
> > > > considered on top like:
> > > >
> > > > 1. Adjust how to store and retrieve dma index. Maybe network guys
> > > > can work better on top.
> > > >
> > > > 2. Move the sanity check for page pool in mm/page_alloc.c to on free.
> > > >
> > > > Byungchul
> > > >
> > > > ---8<---
> > > > From 7d207a1b3e9f4ff2a72f5b54b09e3ed0c4aaaca3 Mon Sep 17 00:00:00 2001
> > > > From: Byungchul Park <byungchul@sk.com>
> > > > Date: Mon, 21 Jul 2025 14:05:20 +0900
> > > > Subject: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
> > > >
> > > > ->pp_magic field in struct page is current used to identify if a page
> > > > belongs to a page pool. However, page type e.i. PGTY_netpp can be used
> > > > for that purpose.
> > > >
> > > > Use the page type APIs e.g. PageNetpp(), __SetPageNetpp(), and
> > > > __ClearPageNetpp() instead, and remove the existing APIs accessing
> > > > ->pp_magic e.g. page_pool_page_is_pp(), netmem_or_pp_magic(), and
> > > > netmem_clear_pp_magic() since they are totally replaced.
> > > >
> > > > This work was inspired by the following link by Pavel:
> > > >
> > > > [1] https://lore.kernel.org/all/582f41c0-2742-4400-9c81-0d46bf4e8314@gmail.com/
> > > >
> > > > Suggested-by: Pavel Begunkov <asml.silence@gmail.com>
> > > > Signed-off-by: Byungchul Park <byungchul@sk.com>
> > > > ---
> > > > .../net/ethernet/mellanox/mlx5/core/en/xdp.c | 2 +-
> > > > include/linux/mm.h | 28 ++-----------------
> > > > include/linux/page-flags.h | 6 ++++
> > > > include/net/netmem.h | 2 +-
> > > > mm/page_alloc.c | 4 +--
> > > > net/core/netmem_priv.h | 16 ++---------
> > > > net/core/page_pool.c | 10 +++++--
> > > > 7 files changed, 24 insertions(+), 44 deletions(-)
> > > >
> > > > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > > > index 5d51600935a6..def274f5c1ca 100644
> > > > --- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > > > +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
> > > > @@ -707,7 +707,7 @@ static void mlx5e_free_xdpsq_desc(struct mlx5e_xdpsq *sq,
> > > > xdpi = mlx5e_xdpi_fifo_pop(xdpi_fifo);
> > > > page = xdpi.page.page;
> > > >
> > > > - /* No need to check page_pool_page_is_pp() as we
> > > > + /* No need to check PageNetpp() as we
> > > > * know this is a page_pool page.
> > > > */
> > > > page_pool_recycle_direct(pp_page_to_nmdesc(page)->pp,
> > > > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > > > index ae50c1641bed..736061749535 100644
> > > > --- a/include/linux/mm.h
> > > > +++ b/include/linux/mm.h
> > > > @@ -4135,10 +4135,9 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> > > > * DMA mapping IDs for page_pool
> > > > *
> > > > * When DMA-mapping a page, page_pool allocates an ID (from an xarray) and
> > > > - * stashes it in the upper bits of page->pp_magic. We always want to be able to
> > > > - * unambiguously identify page pool pages (using page_pool_page_is_pp()). Non-PP
> > > > - * pages can have arbitrary kernel pointers stored in the same field as pp_magic
> > > > - * (since it overlaps with page->lru.next), so we must ensure that we cannot
> > > > + * stashes it in the upper bits of page->pp_magic. Non-PP pages can have
> > > > + * arbitrary kernel pointers stored in the same field as pp_magic (since
> > > > + * it overlaps with page->lru.next), so we must ensure that we cannot
> > > > * mistake a valid kernel pointer with any of the values we write into this
> > > > * field.
> > > > *
> > > > @@ -4168,25 +4167,4 @@ int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
> > > >
> > > > #define PP_DMA_INDEX_MASK GENMASK(PP_DMA_INDEX_BITS + PP_DMA_INDEX_SHIFT - 1, \
> > > > PP_DMA_INDEX_SHIFT)
> > > > -
> > > > -/* Mask used for checking in page_pool_page_is_pp() below. page->pp_magic is
> > > > - * OR'ed with PP_SIGNATURE after the allocation in order to preserve bit 0 for
> > > > - * the head page of compound page and bit 1 for pfmemalloc page, as well as the
> > > > - * bits used for the DMA index. page_is_pfmemalloc() is checked in
> > > > - * __page_pool_put_page() to avoid recycling the pfmemalloc page.
> > > > - */
> > > > -#define PP_MAGIC_MASK ~(PP_DMA_INDEX_MASK | 0x3UL)
> > > > -
> > > > -#ifdef CONFIG_PAGE_POOL
> > > > -static inline bool page_pool_page_is_pp(const struct page *page)
> > > > -{
> > > > - return (page->pp_magic & PP_MAGIC_MASK) == PP_SIGNATURE;
> > > > -}
> > > > -#else
> > > > -static inline bool page_pool_page_is_pp(const struct page *page)
> > > > -{
> > > > - return false;
> > > > -}
> > > > -#endif
> > > > -
> > > > #endif /* _LINUX_MM_H */
> > > > diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> > > > index 4fe5ee67535b..906ba7c9e372 100644
> > > > --- a/include/linux/page-flags.h
> > > > +++ b/include/linux/page-flags.h
> > > > @@ -957,6 +957,7 @@ enum pagetype {
> > > > PGTY_zsmalloc = 0xf6,
> > > > PGTY_unaccepted = 0xf7,
> > > > PGTY_large_kmalloc = 0xf8,
> > > > + PGTY_netpp = 0xf9,
> > > >
> > > > PGTY_mapcount_underflow = 0xff
> > > > };
> > > > @@ -1101,6 +1102,11 @@ PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
> > > > PAGE_TYPE_OPS(Unaccepted, unaccepted, unaccepted)
> > > > FOLIO_TYPE_OPS(large_kmalloc, large_kmalloc)
> > > >
> > > > +/*
> > > > + * Marks page_pool allocated pages.
> > > > + */
> > > > +PAGE_TYPE_OPS(Netpp, netpp, netpp)
> > > > +
> > > > /**
> > > > * PageHuge - Determine if the page belongs to hugetlbfs
> > > > * @page: The page to test.
> > > > diff --git a/include/net/netmem.h b/include/net/netmem.h
> > > > index f7dacc9e75fd..3667334e16e7 100644
> > > > --- a/include/net/netmem.h
> > > > +++ b/include/net/netmem.h
> > > > @@ -298,7 +298,7 @@ static inline struct net_iov *__netmem_clear_lsb(netmem_ref netmem)
> > > > */
> > > > #define pp_page_to_nmdesc(p) \
> > > > ({ \
> > > > - DEBUG_NET_WARN_ON_ONCE(!page_pool_page_is_pp(p)); \
> > > > + DEBUG_NET_WARN_ON_ONCE(!PageNetpp(p)); \
> > > > __pp_page_to_nmdesc(p); \
> > > > })
> > > >
> > > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > > > index 2ef3c07266b3..71c7666e48a9 100644
> > > > --- a/mm/page_alloc.c
> > > > +++ b/mm/page_alloc.c
> > > > @@ -898,7 +898,7 @@ static inline bool page_expected_state(struct page *page,
> > > > #ifdef CONFIG_MEMCG
> > > > page->memcg_data |
> > > > #endif
> > > > - page_pool_page_is_pp(page) |
> > > > + PageNetpp(page) |
> > > > (page->flags & check_flags)))
> > > > return false;
> > > >
> > > > @@ -925,7 +925,7 @@ static const char *page_bad_reason(struct page *page, unsigned long flags)
> > > > if (unlikely(page->memcg_data))
> > > > bad_reason = "page still charged to cgroup";
> > > > #endif
> > > > - if (unlikely(page_pool_page_is_pp(page)))
> > > > + if (unlikely(PageNetpp(page)))
> > > > bad_reason = "page_pool leak";
> > > > return bad_reason;
> > > > }
> > > > diff --git a/net/core/netmem_priv.h b/net/core/netmem_priv.h
> > > > index cd95394399b4..39a97703d9ed 100644
> > > > --- a/net/core/netmem_priv.h
> > > > +++ b/net/core/netmem_priv.h
> > > > @@ -8,21 +8,11 @@ static inline unsigned long netmem_get_pp_magic(netmem_ref netmem)
> > > > return __netmem_clear_lsb(netmem)->pp_magic & ~PP_DMA_INDEX_MASK;
> > > > }
> > > >
> > > > -static inline void netmem_or_pp_magic(netmem_ref netmem, unsigned long pp_magic)
> > > > -{
> > > > - __netmem_clear_lsb(netmem)->pp_magic |= pp_magic;
> > > > -}
> > > > -
> > > > -static inline void netmem_clear_pp_magic(netmem_ref netmem)
> > > > -{
> > > > - WARN_ON_ONCE(__netmem_clear_lsb(netmem)->pp_magic & PP_DMA_INDEX_MASK);
> > > > -
> > > > - __netmem_clear_lsb(netmem)->pp_magic = 0;
> > > > -}
> > > > -
> > > > static inline bool netmem_is_pp(netmem_ref netmem)
> > > > {
> > > > - return (netmem_get_pp_magic(netmem) & PP_MAGIC_MASK) == PP_SIGNATURE;
> > > > + if (netmem_is_net_iov(netmem))
> > > > + return true;
> > >
> > > As Pavel alludes, this is dubious, and at least it's difficult to
> > > reason about it.
> > >
> > > There could be net_iovs that are not attached to pp, and should not be
> > > treated as pp memory. These are in the devmem (and future net_iov) tx
> > > paths.
> > >
> > > We need a way to tell if a net_iov is pp or not. A couple of options:
> > >
> > > 1. We could have it such that if net_iov->pp is set, then the
> > > netmem_is_pp == true, otherwise false.
> > > 2. We could implement a page-flags equivalent for net_iov.
> > >
> > > Option #1 is simpler and is my preferred. To do that properly, you need to:
> > >
> > > 1. Make sure everywhere net_iovs are allocated that pp=NULL in the
> > > non-pp case and pp=non NULL in the pp case. those callsites are
> > > net_devmem_bind_dmabuf (devmem rx & tx path), io_zcrx_create_area
> > > (io_uring rx path).
> > >
> > > 2. Change netmem_is_pp to check net_iov->pp in the net_iov case.
> >
> > Good idea, but I'm not sure if I could work on it without consuming your
> > additional review efforts. Can anyone add net_iov_is_pp() helper?
> >
>
> Things did indeed get busy for me with work work the past week and I
> still need to look at your merged netmem desc series, but I'm happy to
> review whenever I can.
>
> > Or use the page type, Netpp, as an additional way to identify if it's a
> > pp page for system memory, keeping the current way using ->pp_magic.
> > So the page type, Netpp, is used for system memory, and ->pp_magic is
> > used for net_iov. The clean up for ->pp_magic can be done if needed.
> >
>
> IMO I would like to avoid deviations like this, especially since
> ->pp_magic is in the netmem_desc struct that is now shared between
> page and net_iov. I'd rather both use pp_magic or both not, but that
> may just be me.
Totally Agree. Lemme try your option #1 then.
Byungchul
>
>
> --
> Thanks,
> Mina
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH] mm, page_pool: introduce a new page type for page pool in page type
2025-07-22 22:17 ` Mina Almasry
2025-07-23 4:46 ` Byungchul Park
@ 2025-07-25 6:50 ` Byungchul Park
1 sibling, 0 replies; 14+ messages in thread
From: Byungchul Park @ 2025-07-25 6:50 UTC (permalink / raw)
To: Mina Almasry
Cc: linux-mm, netdev, linux-kernel, kernel_team, harry.yoo, ast,
daniel, davem, kuba, hawk, john.fastabend, sdf, saeedm, leon,
tariqt, mbloch, andrew+netdev, edumazet, pabeni, akpm, david,
lorenzo.stoakes, Liam.Howlett, vbabka, rppt, surenb, mhocko,
horms, jackmanb, hannes, ziy, ilias.apalodimas, willy, brauner,
kas, yuzhao, usamaarif642, baolin.wang, toke, asml.silence, bpf,
linux-rdma
On Tue, Jul 22, 2025 at 03:17:15PM -0700, Mina Almasry wrote:
> On Sun, Jul 20, 2025 at 10:49 PM Byungchul Park <byungchul@sk.com> wrote:
> > diff --git a/net/core/netmem_priv.h b/net/core/netmem_priv.h
> > index cd95394399b4..39a97703d9ed 100644
> > --- a/net/core/netmem_priv.h
> > +++ b/net/core/netmem_priv.h
> > @@ -8,21 +8,11 @@ static inline unsigned long netmem_get_pp_magic(netmem_ref netmem)
> > return __netmem_clear_lsb(netmem)->pp_magic & ~PP_DMA_INDEX_MASK;
> > }
> >
> > -static inline void netmem_or_pp_magic(netmem_ref netmem, unsigned long pp_magic)
> > -{
> > - __netmem_clear_lsb(netmem)->pp_magic |= pp_magic;
> > -}
> > -
> > -static inline void netmem_clear_pp_magic(netmem_ref netmem)
> > -{
> > - WARN_ON_ONCE(__netmem_clear_lsb(netmem)->pp_magic & PP_DMA_INDEX_MASK);
> > -
> > - __netmem_clear_lsb(netmem)->pp_magic = 0;
> > -}
> > -
> > static inline bool netmem_is_pp(netmem_ref netmem)
> > {
> > - return (netmem_get_pp_magic(netmem) & PP_MAGIC_MASK) == PP_SIGNATURE;
> > + if (netmem_is_net_iov(netmem))
> > + return true;
>
> As Pavel alludes, this is dubious, and at least it's difficult to
> reason about it.
>
> There could be net_iovs that are not attached to pp, and should not be
> treated as pp memory. These are in the devmem (and future net_iov) tx
> paths.
>
> We need a way to tell if a net_iov is pp or not. A couple of options:
>
> 1. We could have it such that if net_iov->pp is set, then the
> netmem_is_pp == true, otherwise false.
> 2. We could implement a page-flags equivalent for net_iov.
>
> Option #1 is simpler and is my preferred. To do that properly, you need to:
>
> 1. Make sure everywhere net_iovs are allocated that pp=NULL in the
> non-pp case and pp=non NULL in the pp case. those callsites are
> net_devmem_bind_dmabuf (devmem rx & tx path), io_zcrx_create_area
A few seconds reviewing the code, fortunately netmem_set_pp(pool) and
netmem_or_pp_magic(PP_SIGNATURE) are always called paired, and
netmem_set_pp(NULL) and netmem_clear_pp_magic() are always called paired
too.
And there's no code to directly assign a value to ->pp and ->pp_magic,
except in net_devmem_alloc_dmabuf() but that is also safe because always
followed by page_pool_set_pp_info().
Even though I think it's already equivalent between checking
'->pp != NULL' and '->pp_magic == PP_SIGNATURE' with the current code,
more consideration for better code should be always welcome.
As you mentioned, at net_devmem_bind_dmabuf() and io_zcrx_create_area(),
it'd better initialize ->pp and ->pp_magic like:
--
diff --git a/io_uring/zcrx.c b/io_uring/zcrx.c
index 00d0064b22a5..8f2051b2c505 100644
--- a/io_uring/zcrx.c
+++ b/io_uring/zcrx.c
@@ -430,6 +430,7 @@ static int io_zcrx_create_area(struct io_zcrx_ifq *ifq,
area->freelist[i] = i;
atomic_set(&area->user_refs[i], 0);
niov->type = NET_IOV_IOURING;
+ page_pool_clear_pp_info(net_iov_to_netmem(niov));
}
area->free_count = nr_iovs;
diff --git a/net/core/devmem.c b/net/core/devmem.c
index b3a62ca0df65..5d017c9f4986 100644
--- a/net/core/devmem.c
+++ b/net/core/devmem.c
@@ -285,6 +285,7 @@ net_devmem_bind_dmabuf(struct net_device *dev,
niov = &owner->area.niovs[i];
niov->type = NET_IOV_DMABUF;
niov->owner = &owner->area;
+ page_pool_clear_pp_info(net_iov_to_netmem(niov));
page_pool_set_dma_addr_netmem(net_iov_to_netmem(niov),
net_devmem_get_dma_addr(niov));
if (direction == DMA_TO_DEVICE)
--
Do you think it works for using ->pp to check if a niov is pp?
Byungchul
^ permalink raw reply related [flat|nested] 14+ messages in thread