[v1] Split netmem from struct page

[RFC 13/19] page_pool: expand scope of is_pp_{netmem,page}() to global

Posted by Byungchul Park 7 months, 1 week ago

Other than skbuff.c might need to check if a page or netmem is for page
pool, for example, page_alloc.c needs to check the page state, whether
it comes from page pool or not for their own purpose.

Expand the scope of is_pp_netmem() and introduce is_pp_page() newly, so
that those who want to check the source can achieve the checking without
accessing page pool member, page->pp_magic, directly.

Signed-off-by: Byungchul Park <byungchul@sk.com>
---
 include/net/page_pool/types.h |  2 ++
 net/core/page_pool.c          | 10 ++++++++++
 net/core/skbuff.c             |  5 -----
 3 files changed, 12 insertions(+), 5 deletions(-)

diff --git a/include/net/page_pool/types.h b/include/net/page_pool/types.h
index 36eb57d73abc6..d3e1a52f01e09 100644
--- a/include/net/page_pool/types.h
+++ b/include/net/page_pool/types.h
@@ -299,4 +299,6 @@ static inline bool is_page_pool_compiled_in(void)
 /* Caller must provide appropriate safe context, e.g. NAPI. */
 void page_pool_update_nid(struct page_pool *pool, int new_nid);
 
+bool is_pp_netmem(netmem_ref netmem);
+bool is_pp_page(struct page *page);
 #endif /* _NET_PAGE_POOL_H */
diff --git a/net/core/page_pool.c b/net/core/page_pool.c
index b61c1038f4c68..9c553e5a1b555 100644
--- a/net/core/page_pool.c
+++ b/net/core/page_pool.c
@@ -1225,3 +1225,13 @@ void net_mp_niov_clear_page_pool(struct netmem_desc *niov)
 
 	page_pool_clear_pp_info(netmem);
 }
+
+bool is_pp_netmem(netmem_ref netmem)
+{
+	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
+}
+
+bool is_pp_page(struct page *page)
+{
+	return is_pp_netmem(page_to_netmem(page));
+}
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 6cbf77bc61fce..11098c204fe3e 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -893,11 +893,6 @@ static void skb_clone_fraglist(struct sk_buff *skb)
 		skb_get(list);
 }
 
-static bool is_pp_netmem(netmem_ref netmem)
-{
-	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
-}
-
 int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb,
 		    unsigned int headroom)
 {
-- 
2.17.1

Re: [RFC 13/19] page_pool: expand scope of is_pp_{netmem,page}() to global

Posted by Toke Høiland-Jørgensen 7 months, 1 week ago

Byungchul Park <byungchul@sk.com> writes:

> Other than skbuff.c might need to check if a page or netmem is for page
> pool, for example, page_alloc.c needs to check the page state, whether
> it comes from page pool or not for their own purpose.
>
> Expand the scope of is_pp_netmem() and introduce is_pp_page() newly, so
> that those who want to check the source can achieve the checking without
> accessing page pool member, page->pp_magic, directly.
>
> Signed-off-by: Byungchul Park <byungchul@sk.com>
> ---
>  include/net/page_pool/types.h |  2 ++
>  net/core/page_pool.c          | 10 ++++++++++
>  net/core/skbuff.c             |  5 -----
>  3 files changed, 12 insertions(+), 5 deletions(-)
>
> diff --git a/include/net/page_pool/types.h b/include/net/page_pool/types.h
> index 36eb57d73abc6..d3e1a52f01e09 100644
> --- a/include/net/page_pool/types.h
> +++ b/include/net/page_pool/types.h
> @@ -299,4 +299,6 @@ static inline bool is_page_pool_compiled_in(void)
>  /* Caller must provide appropriate safe context, e.g. NAPI. */
>  void page_pool_update_nid(struct page_pool *pool, int new_nid);
>  
> +bool is_pp_netmem(netmem_ref netmem);
> +bool is_pp_page(struct page *page);
>  #endif /* _NET_PAGE_POOL_H */
> diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> index b61c1038f4c68..9c553e5a1b555 100644
> --- a/net/core/page_pool.c
> +++ b/net/core/page_pool.c
> @@ -1225,3 +1225,13 @@ void net_mp_niov_clear_page_pool(struct netmem_desc *niov)
>  
>  	page_pool_clear_pp_info(netmem);
>  }
> +
> +bool is_pp_netmem(netmem_ref netmem)
> +{
> +	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
> +}
> +
> +bool is_pp_page(struct page *page)
> +{
> +	return is_pp_netmem(page_to_netmem(page));
> +}
> diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> index 6cbf77bc61fce..11098c204fe3e 100644
> --- a/net/core/skbuff.c
> +++ b/net/core/skbuff.c
> @@ -893,11 +893,6 @@ static void skb_clone_fraglist(struct sk_buff *skb)
>  		skb_get(list);
>  }
>  
> -static bool is_pp_netmem(netmem_ref netmem)
> -{
> -	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
> -}
> -

This has already been moved to mm.h (and the check changed) by commit:

cd3c93167da0 ("page_pool: Move pp_magic check into helper functions")

You should definitely rebase this series on top of that (and the
subsequent ee62ce7a1d90 ("page_pool: Track DMA-mapped pages and unmap
them when destroying the pool")), as these change the semantics of how
page_pool interacts with struct page.

Both of these are in net-next, which Mina already asked you to rebase
on, so I guess you'll pick it up there, put flagging it here just for
completeness :)

-Toke

Re: [RFC 13/19] page_pool: expand scope of is_pp_{netmem,page}() to global

Posted by Byungchul Park 7 months, 1 week ago

On Mon, May 12, 2025 at 02:46:36PM +0200, Toke Høiland-Jørgensen wrote:
> Byungchul Park <byungchul@sk.com> writes:
> 
> > Other than skbuff.c might need to check if a page or netmem is for page
> > pool, for example, page_alloc.c needs to check the page state, whether
> > it comes from page pool or not for their own purpose.
> >
> > Expand the scope of is_pp_netmem() and introduce is_pp_page() newly, so
> > that those who want to check the source can achieve the checking without
> > accessing page pool member, page->pp_magic, directly.
> >
> > Signed-off-by: Byungchul Park <byungchul@sk.com>
> > ---
> >  include/net/page_pool/types.h |  2 ++
> >  net/core/page_pool.c          | 10 ++++++++++
> >  net/core/skbuff.c             |  5 -----
> >  3 files changed, 12 insertions(+), 5 deletions(-)
> >
> > diff --git a/include/net/page_pool/types.h b/include/net/page_pool/types.h
> > index 36eb57d73abc6..d3e1a52f01e09 100644
> > --- a/include/net/page_pool/types.h
> > +++ b/include/net/page_pool/types.h
> > @@ -299,4 +299,6 @@ static inline bool is_page_pool_compiled_in(void)
> >  /* Caller must provide appropriate safe context, e.g. NAPI. */
> >  void page_pool_update_nid(struct page_pool *pool, int new_nid);
> >  
> > +bool is_pp_netmem(netmem_ref netmem);
> > +bool is_pp_page(struct page *page);
> >  #endif /* _NET_PAGE_POOL_H */
> > diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> > index b61c1038f4c68..9c553e5a1b555 100644
> > --- a/net/core/page_pool.c
> > +++ b/net/core/page_pool.c
> > @@ -1225,3 +1225,13 @@ void net_mp_niov_clear_page_pool(struct netmem_desc *niov)
> >  
> >  	page_pool_clear_pp_info(netmem);
> >  }
> > +
> > +bool is_pp_netmem(netmem_ref netmem)
> > +{
> > +	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
> > +}
> > +
> > +bool is_pp_page(struct page *page)
> > +{
> > +	return is_pp_netmem(page_to_netmem(page));
> > +}
> > diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> > index 6cbf77bc61fce..11098c204fe3e 100644
> > --- a/net/core/skbuff.c
> > +++ b/net/core/skbuff.c
> > @@ -893,11 +893,6 @@ static void skb_clone_fraglist(struct sk_buff *skb)
> >  		skb_get(list);
> >  }
> >  
> > -static bool is_pp_netmem(netmem_ref netmem)
> > -{
> > -	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
> > -}
> > -
> 
> This has already been moved to mm.h (and the check changed) by commit:
> 
> cd3c93167da0 ("page_pool: Move pp_magic check into helper functions")
> 
> You should definitely rebase this series on top of that (and the
> subsequent ee62ce7a1d90 ("page_pool: Track DMA-mapped pages and unmap
> them when destroying the pool")), as these change the semantics of how
> page_pool interacts with struct page.
> 
> Both of these are in net-next, which Mina already asked you to rebase

Is this net-next you are mentioning?  I will rebase on this if so.

   https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/

	Byungchul

> on, so I guess you'll pick it up there, put flagging it here just for
> completeness :)
> 
> -Toke
>

Re: [RFC 13/19] page_pool: expand scope of is_pp_{netmem,page}() to global

Posted by Toke Høiland-Jørgensen 7 months ago

Byungchul Park <byungchul@sk.com> writes:

> On Mon, May 12, 2025 at 02:46:36PM +0200, Toke Høiland-Jørgensen wrote:
>> Byungchul Park <byungchul@sk.com> writes:
>> 
>> > Other than skbuff.c might need to check if a page or netmem is for page
>> > pool, for example, page_alloc.c needs to check the page state, whether
>> > it comes from page pool or not for their own purpose.
>> >
>> > Expand the scope of is_pp_netmem() and introduce is_pp_page() newly, so
>> > that those who want to check the source can achieve the checking without
>> > accessing page pool member, page->pp_magic, directly.
>> >
>> > Signed-off-by: Byungchul Park <byungchul@sk.com>
>> > ---
>> >  include/net/page_pool/types.h |  2 ++
>> >  net/core/page_pool.c          | 10 ++++++++++
>> >  net/core/skbuff.c             |  5 -----
>> >  3 files changed, 12 insertions(+), 5 deletions(-)
>> >
>> > diff --git a/include/net/page_pool/types.h b/include/net/page_pool/types.h
>> > index 36eb57d73abc6..d3e1a52f01e09 100644
>> > --- a/include/net/page_pool/types.h
>> > +++ b/include/net/page_pool/types.h
>> > @@ -299,4 +299,6 @@ static inline bool is_page_pool_compiled_in(void)
>> >  /* Caller must provide appropriate safe context, e.g. NAPI. */
>> >  void page_pool_update_nid(struct page_pool *pool, int new_nid);
>> >  
>> > +bool is_pp_netmem(netmem_ref netmem);
>> > +bool is_pp_page(struct page *page);
>> >  #endif /* _NET_PAGE_POOL_H */
>> > diff --git a/net/core/page_pool.c b/net/core/page_pool.c
>> > index b61c1038f4c68..9c553e5a1b555 100644
>> > --- a/net/core/page_pool.c
>> > +++ b/net/core/page_pool.c
>> > @@ -1225,3 +1225,13 @@ void net_mp_niov_clear_page_pool(struct netmem_desc *niov)
>> >  
>> >  	page_pool_clear_pp_info(netmem);
>> >  }
>> > +
>> > +bool is_pp_netmem(netmem_ref netmem)
>> > +{
>> > +	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
>> > +}
>> > +
>> > +bool is_pp_page(struct page *page)
>> > +{
>> > +	return is_pp_netmem(page_to_netmem(page));
>> > +}
>> > diff --git a/net/core/skbuff.c b/net/core/skbuff.c
>> > index 6cbf77bc61fce..11098c204fe3e 100644
>> > --- a/net/core/skbuff.c
>> > +++ b/net/core/skbuff.c
>> > @@ -893,11 +893,6 @@ static void skb_clone_fraglist(struct sk_buff *skb)
>> >  		skb_get(list);
>> >  }
>> >  
>> > -static bool is_pp_netmem(netmem_ref netmem)
>> > -{
>> > -	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
>> > -}
>> > -
>> 
>> This has already been moved to mm.h (and the check changed) by commit:
>> 
>> cd3c93167da0 ("page_pool: Move pp_magic check into helper functions")
>> 
>> You should definitely rebase this series on top of that (and the
>> subsequent ee62ce7a1d90 ("page_pool: Track DMA-mapped pages and unmap
>> them when destroying the pool")), as these change the semantics of how
>> page_pool interacts with struct page.
>> 
>> Both of these are in net-next, which Mina already asked you to rebase
>
> Is this net-next you are mentioning?  I will rebase on this if so.
>
>    https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/

Yup :)

-Toke

Re: [RFC 13/19] page_pool: expand scope of is_pp_{netmem,page}() to global

Posted by Byungchul Park 7 months, 1 week ago

On Mon, May 12, 2025 at 02:46:36PM +0200, Toke Høiland-Jørgensen wrote:
> Byungchul Park <byungchul@sk.com> writes:
> 
> > Other than skbuff.c might need to check if a page or netmem is for page
> > pool, for example, page_alloc.c needs to check the page state, whether
> > it comes from page pool or not for their own purpose.
> >
> > Expand the scope of is_pp_netmem() and introduce is_pp_page() newly, so
> > that those who want to check the source can achieve the checking without
> > accessing page pool member, page->pp_magic, directly.
> >
> > Signed-off-by: Byungchul Park <byungchul@sk.com>
> > ---
> >  include/net/page_pool/types.h |  2 ++
> >  net/core/page_pool.c          | 10 ++++++++++
> >  net/core/skbuff.c             |  5 -----
> >  3 files changed, 12 insertions(+), 5 deletions(-)
> >
> > diff --git a/include/net/page_pool/types.h b/include/net/page_pool/types.h
> > index 36eb57d73abc6..d3e1a52f01e09 100644
> > --- a/include/net/page_pool/types.h
> > +++ b/include/net/page_pool/types.h
> > @@ -299,4 +299,6 @@ static inline bool is_page_pool_compiled_in(void)
> >  /* Caller must provide appropriate safe context, e.g. NAPI. */
> >  void page_pool_update_nid(struct page_pool *pool, int new_nid);
> >  
> > +bool is_pp_netmem(netmem_ref netmem);
> > +bool is_pp_page(struct page *page);
> >  #endif /* _NET_PAGE_POOL_H */
> > diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> > index b61c1038f4c68..9c553e5a1b555 100644
> > --- a/net/core/page_pool.c
> > +++ b/net/core/page_pool.c
> > @@ -1225,3 +1225,13 @@ void net_mp_niov_clear_page_pool(struct netmem_desc *niov)
> >  
> >  	page_pool_clear_pp_info(netmem);
> >  }
> > +
> > +bool is_pp_netmem(netmem_ref netmem)
> > +{
> > +	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
> > +}
> > +
> > +bool is_pp_page(struct page *page)
> > +{
> > +	return is_pp_netmem(page_to_netmem(page));
> > +}
> > diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> > index 6cbf77bc61fce..11098c204fe3e 100644
> > --- a/net/core/skbuff.c
> > +++ b/net/core/skbuff.c
> > @@ -893,11 +893,6 @@ static void skb_clone_fraglist(struct sk_buff *skb)
> >  		skb_get(list);
> >  }
> >  
> > -static bool is_pp_netmem(netmem_ref netmem)
> > -{
> > -	return (netmem_get_pp_magic(netmem) & ~0x3UL) == PP_SIGNATURE;
> > -}
> > -
> 
> This has already been moved to mm.h (and the check changed) by commit:
> 
> cd3c93167da0 ("page_pool: Move pp_magic check into helper functions")
> 
> You should definitely rebase this series on top of that (and the
> subsequent ee62ce7a1d90 ("page_pool: Track DMA-mapped pages and unmap
> them when destroying the pool")), as these change the semantics of how
> page_pool interacts with struct page.
> 
> Both of these are in net-next, which Mina already asked you to rebase
> on, so I guess you'll pick it up there, put flagging it here just for
> completeness :)

I will not miss it.  Thanks.

	Byungchul
> 
> -Toke
>