| Message ID | 164329517024.130462.875087745767169286.stgit@localhost.localdomain (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next,v2] page_pool: Refactor page_pool to enable fragmenting after allocation | expand |
On 2022/1/27 22:57, Alexander Duyck wrote: > From: Alexander Duyck <alexanderduyck@fb.com> > > This change is meant to permit a driver to perform "fragmenting" of the > page from within the driver instead of the current model which requires > pre-partitioning the page. The main motivation behind this is to support > use cases where the page will be split up by the driver after DMA instead > of before. > > With this change it becomes possible to start using page pool to replace > some of the existing use cases where multiple references were being used > for a single page, but the number needed was unknown as the size could be > dynamic. > > For example, with this code it would be possible to do something like > the following to handle allocation: > page = page_pool_alloc_pages(); > if (!page) > return NULL; > page_pool_fragment_page(page, DRIVER_PAGECNT_BIAS_MAX); > rx_buf->page = page; > rx_buf->pagecnt_bias = DRIVER_PAGECNT_BIAS_MAX; > > Then we would process a received buffer by handling it with: > rx_buf->pagecnt_bias--; > > Once the page has been fully consumed we could then flush the remaining > instances with: > if (page_pool_defrag_page(page, rx_buf->pagecnt_bias)) > continue; > page_pool_put_defragged_page(pool, page -1, !!budget); page_pool_put_defragged_page(pool, page, -1, !!budget); Also I am not sure exporting the frag count to the driver is a good idea, as the above example seems a little complex, maybe adding the fragmenting after allocation support for a existing driver is a good way to show if the API is really a good one. > > The general idea is that we want to have the ability to allocate a page > with excess fragment count and then trim off the unneeded fragments. > > Signed-off-by: Alexander Duyck <alexanderduyck@fb.com> > --- > > v2: Added page_pool_is_last_frag > Moved comment about CONFIG_PAGE_POOL to page_pool_put_page > Wrapped statements for page_pool_is_last_frag in parenthesis > > include/net/page_pool.h | 82 ++++++++++++++++++++++++++++++----------------- > net/core/page_pool.c | 23 ++++++------- > 2 files changed, 62 insertions(+), 43 deletions(-) > > diff --git a/include/net/page_pool.h b/include/net/page_pool.h > index 79a805542d0f..fbed91469d42 100644 > --- a/include/net/page_pool.h > +++ b/include/net/page_pool.h > @@ -201,21 +201,67 @@ static inline void page_pool_put_page_bulk(struct page_pool *pool, void **data, > } > #endif > > -void page_pool_put_page(struct page_pool *pool, struct page *page, > - unsigned int dma_sync_size, bool allow_direct); > +void page_pool_put_defragged_page(struct page_pool *pool, struct page *page, > + unsigned int dma_sync_size, > + bool allow_direct); > > -/* Same as above but will try to sync the entire area pool->max_len */ > -static inline void page_pool_put_full_page(struct page_pool *pool, > - struct page *page, bool allow_direct) > +static inline void page_pool_fragment_page(struct page *page, long nr) > +{ > + atomic_long_set(&page->pp_frag_count, nr); > +} > + > +static inline long page_pool_defrag_page(struct page *page, long nr) > +{ > + long ret; > + > + /* If nr == pp_frag_count then we are have cleared all remaining > + * references to the page. No need to actually overwrite it, instead > + * we can leave this to be overwritten by the calling function. > + * > + * The main advantage to doing this is that an atomic_read is > + * generally a much cheaper operation than an atomic update, > + * especially when dealing with a page that may be partitioned > + * into only 2 or 3 pieces. > + */ > + if (atomic_long_read(&page->pp_frag_count) == nr) > + return 0; > + > + ret = atomic_long_sub_return(nr, &page->pp_frag_count); > + WARN_ON(ret < 0); > + return ret; > +} > + > +static inline bool page_pool_is_last_frag(struct page_pool *pool, > + struct page *page) > +{ > + /* If fragments aren't enabled or count is 0 we were the last user */ > + return !(pool->p.flags & PP_FLAG_PAGE_FRAG) || > + (page_pool_defrag_page(page, 1) == 0); > +} > + > +static inline void page_pool_put_page(struct page_pool *pool, > + struct page *page, > + unsigned int dma_sync_size, > + bool allow_direct) > { > /* When page_pool isn't compiled-in, net/core/xdp.c doesn't > * allow registering MEM_TYPE_PAGE_POOL, but shield linker. > */ > #ifdef CONFIG_PAGE_POOL > - page_pool_put_page(pool, page, -1, allow_direct); > + if (!page_pool_is_last_frag(pool, page)) > + return; > + > + page_pool_put_defragged_page(pool, page, dma_sync_size, allow_direct); > #endif > } > > +/* Same as above but will try to sync the entire area pool->max_len */ > +static inline void page_pool_put_full_page(struct page_pool *pool, > + struct page *page, bool allow_direct) > +{ > + page_pool_put_page(pool, page, -1, allow_direct); > +} > + > /* Same as above but the caller must guarantee safe context. e.g NAPI */ > static inline void page_pool_recycle_direct(struct page_pool *pool, > struct page *page) > @@ -243,30 +289,6 @@ static inline void page_pool_set_dma_addr(struct page *page, dma_addr_t addr) > page->dma_addr_upper = upper_32_bits(addr); > } > > -static inline void page_pool_set_frag_count(struct page *page, long nr) > -{ > - atomic_long_set(&page->pp_frag_count, nr); > -} > - > -static inline long page_pool_atomic_sub_frag_count_return(struct page *page, > - long nr) > -{ > - long ret; > - > - /* As suggested by Alexander, atomic_long_read() may cover up the > - * reference count errors, so avoid calling atomic_long_read() in > - * the cases of freeing or draining the page_frags, where we would > - * not expect it to match or that are slowpath anyway. > - */ > - if (__builtin_constant_p(nr) && > - atomic_long_read(&page->pp_frag_count) == nr) > - return 0; > - > - ret = atomic_long_sub_return(nr, &page->pp_frag_count); > - WARN_ON(ret < 0); > - return ret; > -} > - > static inline bool is_page_pool_compiled_in(void) > { > #ifdef CONFIG_PAGE_POOL > diff --git a/net/core/page_pool.c b/net/core/page_pool.c > index bd62c01a2ec3..e25d359d84d9 100644 > --- a/net/core/page_pool.c > +++ b/net/core/page_pool.c > @@ -423,11 +423,6 @@ static __always_inline struct page * > __page_pool_put_page(struct page_pool *pool, struct page *page, > unsigned int dma_sync_size, bool allow_direct) > { > - /* It is not the last user for the page frag case */ > - if (pool->p.flags & PP_FLAG_PAGE_FRAG && > - page_pool_atomic_sub_frag_count_return(page, 1)) > - return NULL; > - > /* This allocator is optimized for the XDP mode that uses > * one-frame-per-page, but have fallbacks that act like the > * regular page allocator APIs. > @@ -471,8 +466,8 @@ __page_pool_put_page(struct page_pool *pool, struct page *page, > return NULL; > } > > -void page_pool_put_page(struct page_pool *pool, struct page *page, > - unsigned int dma_sync_size, bool allow_direct) > +void page_pool_put_defragged_page(struct page_pool *pool, struct page *page, > + unsigned int dma_sync_size, bool allow_direct) > { > page = __page_pool_put_page(pool, page, dma_sync_size, allow_direct); > if (page && !page_pool_recycle_in_ring(pool, page)) { > @@ -480,7 +475,7 @@ void page_pool_put_page(struct page_pool *pool, struct page *page, > page_pool_return_page(pool, page); > } > } > -EXPORT_SYMBOL(page_pool_put_page); > +EXPORT_SYMBOL(page_pool_put_defragged_page); > > /* Caller must not use data area after call, as this function overwrites it */ > void page_pool_put_page_bulk(struct page_pool *pool, void **data, > @@ -491,6 +486,10 @@ void page_pool_put_page_bulk(struct page_pool *pool, void **data, > for (i = 0; i < count; i++) { > struct page *page = virt_to_head_page(data[i]); > > + /* It is not the last user for the page frag case */ > + if (!page_pool_is_last_frag(pool, page)) > + continue; > + > page = __page_pool_put_page(pool, page, -1, false); > /* Approved for bulk recycling in ptr_ring cache */ > if (page) > @@ -526,8 +525,7 @@ static struct page *page_pool_drain_frag(struct page_pool *pool, > long drain_count = BIAS_MAX - pool->frag_users; > > /* Some user is still using the page frag */ > - if (likely(page_pool_atomic_sub_frag_count_return(page, > - drain_count))) > + if (likely(page_pool_defrag_page(page, drain_count))) > return NULL; > > if (page_ref_count(page) == 1 && !page_is_pfmemalloc(page)) { > @@ -548,8 +546,7 @@ static void page_pool_free_frag(struct page_pool *pool) > > pool->frag_page = NULL; > > - if (!page || > - page_pool_atomic_sub_frag_count_return(page, drain_count)) > + if (!page || page_pool_defrag_page(page, drain_count)) > return; > > page_pool_return_page(pool, page); > @@ -588,7 +585,7 @@ struct page *page_pool_alloc_frag(struct page_pool *pool, > pool->frag_users = 1; > *offset = 0; > pool->frag_offset = size; > - page_pool_set_frag_count(page, BIAS_MAX); > + page_pool_fragment_page(page, BIAS_MAX); > return page; > } > > > > . >
On Sat, Jan 29, 2022 at 05:20:37PM +0800, Yunsheng Lin wrote: > On 2022/1/27 22:57, Alexander Duyck wrote: > > From: Alexander Duyck <alexanderduyck@fb.com> > > > > This change is meant to permit a driver to perform "fragmenting" of the > > page from within the driver instead of the current model which requires > > pre-partitioning the page. The main motivation behind this is to support > > use cases where the page will be split up by the driver after DMA instead > > of before. > > > > With this change it becomes possible to start using page pool to replace > > some of the existing use cases where multiple references were being used > > for a single page, but the number needed was unknown as the size could be > > dynamic. > > > > For example, with this code it would be possible to do something like > > the following to handle allocation: > > page = page_pool_alloc_pages(); > > if (!page) > > return NULL; > > page_pool_fragment_page(page, DRIVER_PAGECNT_BIAS_MAX); > > rx_buf->page = page; > > rx_buf->pagecnt_bias = DRIVER_PAGECNT_BIAS_MAX; > > > > Then we would process a received buffer by handling it with: > > rx_buf->pagecnt_bias--; > > > > Once the page has been fully consumed we could then flush the remaining > > instances with: > > if (page_pool_defrag_page(page, rx_buf->pagecnt_bias)) > > continue; > > page_pool_put_defragged_page(pool, page -1, !!budget); > > page_pool_put_defragged_page(pool, page, -1, !!budget); > > Also I am not sure exporting the frag count to the driver is a good > idea, as the above example seems a little complex, maybe adding > the fragmenting after allocation support for a existing driver > is a good way to show if the API is really a good one. This is already kind of exposed since no one limits drivers from calling page_pool_atomic_sub_frag_count_return() right? What this patchset does is allow the drivers to actually use it and release pages without having to atomically decrement all the refcnt bias. And I do get the point that a driver might choose to do the refcounting internally. That was the point all along with the fragment support in page_pool. There's a wide variety of interfaces out there and each one handles buffers differently. What I am missing though is how this works with the current recycling scheme? The driver will still have to to make sure that page_pool_defrag_page(page, 1) == 0 for that to work no? > > > > > > The general idea is that we want to have the ability to allocate a page > > with excess fragment count and then trim off the unneeded fragments. > > > > Signed-off-by: Alexander Duyck <alexanderduyck@fb.com> > > --- > > > > v2: Added page_pool_is_last_frag > > Moved comment about CONFIG_PAGE_POOL to page_pool_put_page > > Wrapped statements for page_pool_is_last_frag in parenthesis > > > > include/net/page_pool.h | 82 ++++++++++++++++++++++++++++++----------------- > > net/core/page_pool.c | 23 ++++++------- > > 2 files changed, 62 insertions(+), 43 deletions(-) > > > > diff --git a/include/net/page_pool.h b/include/net/page_pool.h > > index 79a805542d0f..fbed91469d42 100644 > > --- a/include/net/page_pool.h > > +++ b/include/net/page_pool.h > > @@ -201,21 +201,67 @@ static inline void page_pool_put_page_bulk(struct page_pool *pool, void **data, > > } > > #endif > > > > -void page_pool_put_page(struct page_pool *pool, struct page *page, > > - unsigned int dma_sync_size, bool allow_direct); > > +void page_pool_put_defragged_page(struct page_pool *pool, struct page *page, > > + unsigned int dma_sync_size, > > + bool allow_direct); > > > > -/* Same as above but will try to sync the entire area pool->max_len */ > > -static inline void page_pool_put_full_page(struct page_pool *pool, > > - struct page *page, bool allow_direct) > > +static inline void page_pool_fragment_page(struct page *page, long nr) > > +{ > > + atomic_long_set(&page->pp_frag_count, nr); > > +} > > + > > +static inline long page_pool_defrag_page(struct page *page, long nr) > > +{ > > + long ret; > > + > > + /* If nr == pp_frag_count then we are have cleared all remaining s/are// > > + * references to the page. No need to actually overwrite it, instead > > + * we can leave this to be overwritten by the calling function. > > + * > > + * The main advantage to doing this is that an atomic_read is > > + * generally a much cheaper operation than an atomic update, > > + * especially when dealing with a page that may be partitioned > > + * into only 2 or 3 pieces. > > + */ > > + if (atomic_long_read(&page->pp_frag_count) == nr) > > + return 0; > > + > > + ret = atomic_long_sub_return(nr, &page->pp_frag_count); > > + WARN_ON(ret < 0); > > + return ret; > > +} > > + > > +static inline bool page_pool_is_last_frag(struct page_pool *pool, > > + struct page *page) > > +{ > > + /* If fragments aren't enabled or count is 0 we were the last user */ > > + return !(pool->p.flags & PP_FLAG_PAGE_FRAG) || > > + (page_pool_defrag_page(page, 1) == 0); > > +} > > + > > +static inline void page_pool_put_page(struct page_pool *pool, > > + struct page *page, > > + unsigned int dma_sync_size, > > + bool allow_direct) > > { > > /* When page_pool isn't compiled-in, net/core/xdp.c doesn't > > * allow registering MEM_TYPE_PAGE_POOL, but shield linker. > > */ > > #ifdef CONFIG_PAGE_POOL > > - page_pool_put_page(pool, page, -1, allow_direct); > > + if (!page_pool_is_last_frag(pool, page)) > > + return; > > + > > + page_pool_put_defragged_page(pool, page, dma_sync_size, allow_direct); > > #endif > > } > > > > +/* Same as above but will try to sync the entire area pool->max_len */ > > +static inline void page_pool_put_full_page(struct page_pool *pool, > > + struct page *page, bool allow_direct) > > +{ > > + page_pool_put_page(pool, page, -1, allow_direct); > > +} > > + > > /* Same as above but the caller must guarantee safe context. e.g NAPI */ > > static inline void page_pool_recycle_direct(struct page_pool *pool, > > struct page *page) > > @@ -243,30 +289,6 @@ static inline void page_pool_set_dma_addr(struct page *page, dma_addr_t addr) > > page->dma_addr_upper = upper_32_bits(addr); > > } > > > > -static inline void page_pool_set_frag_count(struct page *page, long nr) > > -{ > > - atomic_long_set(&page->pp_frag_count, nr); > > -} > > - > > -static inline long page_pool_atomic_sub_frag_count_return(struct page *page, > > - long nr) > > -{ > > - long ret; > > - > > - /* As suggested by Alexander, atomic_long_read() may cover up the > > - * reference count errors, so avoid calling atomic_long_read() in > > - * the cases of freeing or draining the page_frags, where we would > > - * not expect it to match or that are slowpath anyway. > > - */ > > - if (__builtin_constant_p(nr) && > > - atomic_long_read(&page->pp_frag_count) == nr) > > - return 0; > > - > > - ret = atomic_long_sub_return(nr, &page->pp_frag_count); > > - WARN_ON(ret < 0); > > - return ret; > > -} > > - > > static inline bool is_page_pool_compiled_in(void) > > { > > #ifdef CONFIG_PAGE_POOL > > diff --git a/net/core/page_pool.c b/net/core/page_pool.c > > index bd62c01a2ec3..e25d359d84d9 100644 > > --- a/net/core/page_pool.c > > +++ b/net/core/page_pool.c > > @@ -423,11 +423,6 @@ static __always_inline struct page * > > __page_pool_put_page(struct page_pool *pool, struct page *page, > > unsigned int dma_sync_size, bool allow_direct) > > { > > - /* It is not the last user for the page frag case */ > > - if (pool->p.flags & PP_FLAG_PAGE_FRAG && > > - page_pool_atomic_sub_frag_count_return(page, 1)) > > - return NULL; > > - > > /* This allocator is optimized for the XDP mode that uses > > * one-frame-per-page, but have fallbacks that act like the > > * regular page allocator APIs. > > @@ -471,8 +466,8 @@ __page_pool_put_page(struct page_pool *pool, struct page *page, > > return NULL; > > } > > > > -void page_pool_put_page(struct page_pool *pool, struct page *page, > > - unsigned int dma_sync_size, bool allow_direct) > > +void page_pool_put_defragged_page(struct page_pool *pool, struct page *page, > > + unsigned int dma_sync_size, bool allow_direct) > > { > > page = __page_pool_put_page(pool, page, dma_sync_size, allow_direct); > > if (page && !page_pool_recycle_in_ring(pool, page)) { > > @@ -480,7 +475,7 @@ void page_pool_put_page(struct page_pool *pool, struct page *page, > > page_pool_return_page(pool, page); > > } > > } > > -EXPORT_SYMBOL(page_pool_put_page); > > +EXPORT_SYMBOL(page_pool_put_defragged_page); > > > > /* Caller must not use data area after call, as this function overwrites it */ > > void page_pool_put_page_bulk(struct page_pool *pool, void **data, > > @@ -491,6 +486,10 @@ void page_pool_put_page_bulk(struct page_pool *pool, void **data, > > for (i = 0; i < count; i++) { > > struct page *page = virt_to_head_page(data[i]); > > > > + /* It is not the last user for the page frag case */ > > + if (!page_pool_is_last_frag(pool, page)) > > + continue; > > + > > page = __page_pool_put_page(pool, page, -1, false); > > /* Approved for bulk recycling in ptr_ring cache */ > > if (page) > > @@ -526,8 +525,7 @@ static struct page *page_pool_drain_frag(struct page_pool *pool, > > long drain_count = BIAS_MAX - pool->frag_users; > > > > /* Some user is still using the page frag */ > > - if (likely(page_pool_atomic_sub_frag_count_return(page, > > - drain_count))) > > + if (likely(page_pool_defrag_page(page, drain_count))) > > return NULL; > > > > if (page_ref_count(page) == 1 && !page_is_pfmemalloc(page)) { > > @@ -548,8 +546,7 @@ static void page_pool_free_frag(struct page_pool *pool) > > > > pool->frag_page = NULL; > > > > - if (!page || > > - page_pool_atomic_sub_frag_count_return(page, drain_count)) > > + if (!page || page_pool_defrag_page(page, drain_count)) > > return; > > > > page_pool_return_page(pool, page); > > @@ -588,7 +585,7 @@ struct page *page_pool_alloc_frag(struct page_pool *pool, > > pool->frag_users = 1; > > *offset = 0; > > pool->frag_offset = size; > > - page_pool_set_frag_count(page, BIAS_MAX); > > + page_pool_fragment_page(page, BIAS_MAX); > > return page; > > } > > > > > > > > . > > Thanks! /Ilias
On Sat, Jan 29, 2022 at 1:52 AM Ilias Apalodimas <ilias.apalodimas@linaro.org> wrote: > > On Sat, Jan 29, 2022 at 05:20:37PM +0800, Yunsheng Lin wrote: > > On 2022/1/27 22:57, Alexander Duyck wrote: > > > From: Alexander Duyck <alexanderduyck@fb.com> > > > > > > This change is meant to permit a driver to perform "fragmenting" of the > > > page from within the driver instead of the current model which requires > > > pre-partitioning the page. The main motivation behind this is to support > > > use cases where the page will be split up by the driver after DMA instead > > > of before. > > > > > > With this change it becomes possible to start using page pool to replace > > > some of the existing use cases where multiple references were being used > > > for a single page, but the number needed was unknown as the size could be > > > dynamic. > > > > > > For example, with this code it would be possible to do something like > > > the following to handle allocation: > > > page = page_pool_alloc_pages(); > > > if (!page) > > > return NULL; > > > page_pool_fragment_page(page, DRIVER_PAGECNT_BIAS_MAX); > > > rx_buf->page = page; > > > rx_buf->pagecnt_bias = DRIVER_PAGECNT_BIAS_MAX; > > > > > > Then we would process a received buffer by handling it with: > > > rx_buf->pagecnt_bias--; > > > > > > Once the page has been fully consumed we could then flush the remaining > > > instances with: > > > if (page_pool_defrag_page(page, rx_buf->pagecnt_bias)) > > > continue; > > > page_pool_put_defragged_page(pool, page -1, !!budget); > > > > page_pool_put_defragged_page(pool, page, -1, !!budget); > > > > Also I am not sure exporting the frag count to the driver is a good > > idea, as the above example seems a little complex, maybe adding > > the fragmenting after allocation support for a existing driver > > is a good way to show if the API is really a good one. > > This is already kind of exposed since no one limits drivers from calling > page_pool_atomic_sub_frag_count_return() right? > What this patchset does is allow the drivers to actually use it and release > pages without having to atomically decrement all the refcnt bias. > > And I do get the point that a driver might choose to do the refcounting > internally. That was the point all along with the fragment support in > page_pool. There's a wide variety of interfaces out there and each one > handles buffers differently. > > What I am missing though is how this works with the current recycling > scheme? The driver will still have to to make sure that > page_pool_defrag_page(page, 1) == 0 for that to work no? The general idea here is that we are getting away from doing in-driver recycling and instead letting page pool take care of all that. That was the original idea behind page pool, however the original implementation was limited to a single use per page only. So most of the legacy code out there is having to use the page_ref_count == 1 or page_ref_count == bias trick in order to determine if it can recycle the page. The page pool already takes care of the page recycling by returning the pages to the pool when page_ref_count == 1, what we get by adding the frag count is the ability for the drivers to drop the need to perform their own ref count tricks and instead offloads that to the kernel so when page_pool_defrag_page(page, 1) == 0 it can then go immediately into the checks for page_ref_count == 1 and just recycle the page into the page pool. > > > > > > > > > > The general idea is that we want to have the ability to allocate a page > > > with excess fragment count and then trim off the unneeded fragments. > > > > > > Signed-off-by: Alexander Duyck <alexanderduyck@fb.com> > > > --- > > > > > > v2: Added page_pool_is_last_frag > > > Moved comment about CONFIG_PAGE_POOL to page_pool_put_page > > > Wrapped statements for page_pool_is_last_frag in parenthesis > > > > > > include/net/page_pool.h | 82 ++++++++++++++++++++++++++++++----------------- > > > net/core/page_pool.c | 23 ++++++------- > > > 2 files changed, 62 insertions(+), 43 deletions(-) > > > > > > diff --git a/include/net/page_pool.h b/include/net/page_pool.h > > > index 79a805542d0f..fbed91469d42 100644 > > > --- a/include/net/page_pool.h > > > +++ b/include/net/page_pool.h > > > @@ -201,21 +201,67 @@ static inline void page_pool_put_page_bulk(struct page_pool *pool, void **data, > > > } > > > #endif > > > > > > -void page_pool_put_page(struct page_pool *pool, struct page *page, > > > - unsigned int dma_sync_size, bool allow_direct); > > > +void page_pool_put_defragged_page(struct page_pool *pool, struct page *page, > > > + unsigned int dma_sync_size, > > > + bool allow_direct); > > > > > > -/* Same as above but will try to sync the entire area pool->max_len */ > > > -static inline void page_pool_put_full_page(struct page_pool *pool, > > > - struct page *page, bool allow_direct) > > > +static inline void page_pool_fragment_page(struct page *page, long nr) > > > +{ > > > + atomic_long_set(&page->pp_frag_count, nr); > > > +} > > > + > > > +static inline long page_pool_defrag_page(struct page *page, long nr) > > > +{ > > > + long ret; > > > + > > > + /* If nr == pp_frag_count then we are have cleared all remaining > > s/are// Will fix for v3. Thanks, Alex
Hi Alexander, On Sat, 29 Jan 2022 at 20:35, Alexander Duyck <alexander.duyck@gmail.com> wrote: > [...] > > > > > > Also I am not sure exporting the frag count to the driver is a good > > > idea, as the above example seems a little complex, maybe adding > > > the fragmenting after allocation support for a existing driver > > > is a good way to show if the API is really a good one. > > > > This is already kind of exposed since no one limits drivers from calling > > page_pool_atomic_sub_frag_count_return() right? > > What this patchset does is allow the drivers to actually use it and release > > pages without having to atomically decrement all the refcnt bias. > > > > And I do get the point that a driver might choose to do the refcounting > > internally. That was the point all along with the fragment support in > > page_pool. There's a wide variety of interfaces out there and each one > > handles buffers differently. > > > > What I am missing though is how this works with the current recycling > > scheme? The driver will still have to to make sure that > > page_pool_defrag_page(page, 1) == 0 for that to work no? > > The general idea here is that we are getting away from doing in-driver > recycling and instead letting page pool take care of all that. That > was the original idea behind page pool, however the original > implementation was limited to a single use per page only. > > So most of the legacy code out there is having to use the > page_ref_count == 1 or page_ref_count == bias trick in order to > determine if it can recycle the page. The page pool already takes care > of the page recycling by returning the pages to the pool when > page_ref_count == 1, what we get by adding the frag count is the > ability for the drivers to drop the need to perform their own ref > count tricks and instead offloads that to the kernel so when > page_pool_defrag_page(page, 1) == 0 it can then go immediately into > the checks for page_ref_count == 1 and just recycle the page into the > page pool. Excellent, that's what I assumed tbh Thanks! Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org> > > > > > > > > > > > > > > > The general idea is that we want to have the ability to allocate a page > > > > with excess fragment count and then trim off the unneeded fragments. > > > > > > > > Signed-off-by: Alexander Duyck <alexanderduyck@fb.com> > > > > --- > > > > > > > > v2: Added page_pool_is_last_frag > > > > Moved comment about CONFIG_PAGE_POOL to page_pool_put_page > > > > Wrapped statements for page_pool_is_last_frag in parenthesis > > > > > > > > include/net/page_pool.h | 82 ++++++++++++++++++++++++++++++----------------- > > > > net/core/page_pool.c | 23 ++++++------- > > > > 2 files changed, 62 insertions(+), 43 deletions(-) > > > > > > > > diff --git a/include/net/page_pool.h b/include/net/page_pool.h > > > > index 79a805542d0f..fbed91469d42 100644 > > > > --- a/include/net/page_pool.h > > > > +++ b/include/net/page_pool.h > > > > @@ -201,21 +201,67 @@ static inline void page_pool_put_page_bulk(struct page_pool *pool, void **data, > > > > } > > > > #endif > > > > > > > > -void page_pool_put_page(struct page_pool *pool, struct page *page, > > > > - unsigned int dma_sync_size, bool allow_direct); > > > > +void page_pool_put_defragged_page(struct page_pool *pool, struct page *page, > > > > + unsigned int dma_sync_size, > > > > + bool allow_direct); > > > > > > > > -/* Same as above but will try to sync the entire area pool->max_len */ > > > > -static inline void page_pool_put_full_page(struct page_pool *pool, > > > > - struct page *page, bool allow_direct) > > > > +static inline void page_pool_fragment_page(struct page *page, long nr) > > > > +{ > > > > + atomic_long_set(&page->pp_frag_count, nr); > > > > +} > > > > + > > > > +static inline long page_pool_defrag_page(struct page *page, long nr) > > > > +{ > > > > + long ret; > > > > + > > > > + /* If nr == pp_frag_count then we are have cleared all remaining > > > > s/are// > > Will fix for v3. > > Thanks, > > Alex
diff --git a/include/net/page_pool.h b/include/net/page_pool.h index 79a805542d0f..fbed91469d42 100644 --- a/include/net/page_pool.h +++ b/include/net/page_pool.h @@ -201,21 +201,67 @@ static inline void page_pool_put_page_bulk(struct page_pool *pool, void **data, } #endif -void page_pool_put_page(struct page_pool *pool, struct page *page, - unsigned int dma_sync_size, bool allow_direct); +void page_pool_put_defragged_page(struct page_pool *pool, struct page *page, + unsigned int dma_sync_size, + bool allow_direct); -/* Same as above but will try to sync the entire area pool->max_len */ -static inline void page_pool_put_full_page(struct page_pool *pool, - struct page *page, bool allow_direct) +static inline void page_pool_fragment_page(struct page *page, long nr) +{ + atomic_long_set(&page->pp_frag_count, nr); +} + +static inline long page_pool_defrag_page(struct page *page, long nr) +{ + long ret; + + /* If nr == pp_frag_count then we are have cleared all remaining + * references to the page. No need to actually overwrite it, instead + * we can leave this to be overwritten by the calling function. + * + * The main advantage to doing this is that an atomic_read is + * generally a much cheaper operation than an atomic update, + * especially when dealing with a page that may be partitioned + * into only 2 or 3 pieces. + */ + if (atomic_long_read(&page->pp_frag_count) == nr) + return 0; + + ret = atomic_long_sub_return(nr, &page->pp_frag_count); + WARN_ON(ret < 0); + return ret; +} + +static inline bool page_pool_is_last_frag(struct page_pool *pool, + struct page *page) +{ + /* If fragments aren't enabled or count is 0 we were the last user */ + return !(pool->p.flags & PP_FLAG_PAGE_FRAG) || + (page_pool_defrag_page(page, 1) == 0); +} + +static inline void page_pool_put_page(struct page_pool *pool, + struct page *page, + unsigned int dma_sync_size, + bool allow_direct) { /* When page_pool isn't compiled-in, net/core/xdp.c doesn't * allow registering MEM_TYPE_PAGE_POOL, but shield linker. */ #ifdef CONFIG_PAGE_POOL - page_pool_put_page(pool, page, -1, allow_direct); + if (!page_pool_is_last_frag(pool, page)) + return; + + page_pool_put_defragged_page(pool, page, dma_sync_size, allow_direct); #endif } +/* Same as above but will try to sync the entire area pool->max_len */ +static inline void page_pool_put_full_page(struct page_pool *pool, + struct page *page, bool allow_direct) +{ + page_pool_put_page(pool, page, -1, allow_direct); +} + /* Same as above but the caller must guarantee safe context. e.g NAPI */ static inline void page_pool_recycle_direct(struct page_pool *pool, struct page *page) @@ -243,30 +289,6 @@ static inline void page_pool_set_dma_addr(struct page *page, dma_addr_t addr) page->dma_addr_upper = upper_32_bits(addr); } -static inline void page_pool_set_frag_count(struct page *page, long nr) -{ - atomic_long_set(&page->pp_frag_count, nr); -} - -static inline long page_pool_atomic_sub_frag_count_return(struct page *page, - long nr) -{ - long ret; - - /* As suggested by Alexander, atomic_long_read() may cover up the - * reference count errors, so avoid calling atomic_long_read() in - * the cases of freeing or draining the page_frags, where we would - * not expect it to match or that are slowpath anyway. - */ - if (__builtin_constant_p(nr) && - atomic_long_read(&page->pp_frag_count) == nr) - return 0; - - ret = atomic_long_sub_return(nr, &page->pp_frag_count); - WARN_ON(ret < 0); - return ret; -} - static inline bool is_page_pool_compiled_in(void) { #ifdef CONFIG_PAGE_POOL diff --git a/net/core/page_pool.c b/net/core/page_pool.c index bd62c01a2ec3..e25d359d84d9 100644 --- a/net/core/page_pool.c +++ b/net/core/page_pool.c @@ -423,11 +423,6 @@ static __always_inline struct page * __page_pool_put_page(struct page_pool *pool, struct page *page, unsigned int dma_sync_size, bool allow_direct) { - /* It is not the last user for the page frag case */ - if (pool->p.flags & PP_FLAG_PAGE_FRAG && - page_pool_atomic_sub_frag_count_return(page, 1)) - return NULL; - /* This allocator is optimized for the XDP mode that uses * one-frame-per-page, but have fallbacks that act like the * regular page allocator APIs. @@ -471,8 +466,8 @@ __page_pool_put_page(struct page_pool *pool, struct page *page, return NULL; } -void page_pool_put_page(struct page_pool *pool, struct page *page, - unsigned int dma_sync_size, bool allow_direct) +void page_pool_put_defragged_page(struct page_pool *pool, struct page *page, + unsigned int dma_sync_size, bool allow_direct) { page = __page_pool_put_page(pool, page, dma_sync_size, allow_direct); if (page && !page_pool_recycle_in_ring(pool, page)) { @@ -480,7 +475,7 @@ void page_pool_put_page(struct page_pool *pool, struct page *page, page_pool_return_page(pool, page); } } -EXPORT_SYMBOL(page_pool_put_page); +EXPORT_SYMBOL(page_pool_put_defragged_page); /* Caller must not use data area after call, as this function overwrites it */ void page_pool_put_page_bulk(struct page_pool *pool, void **data, @@ -491,6 +486,10 @@ void page_pool_put_page_bulk(struct page_pool *pool, void **data, for (i = 0; i < count; i++) { struct page *page = virt_to_head_page(data[i]); + /* It is not the last user for the page frag case */ + if (!page_pool_is_last_frag(pool, page)) + continue; + page = __page_pool_put_page(pool, page, -1, false); /* Approved for bulk recycling in ptr_ring cache */ if (page) @@ -526,8 +525,7 @@ static struct page *page_pool_drain_frag(struct page_pool *pool, long drain_count = BIAS_MAX - pool->frag_users; /* Some user is still using the page frag */ - if (likely(page_pool_atomic_sub_frag_count_return(page, - drain_count))) + if (likely(page_pool_defrag_page(page, drain_count))) return NULL; if (page_ref_count(page) == 1 && !page_is_pfmemalloc(page)) { @@ -548,8 +546,7 @@ static void page_pool_free_frag(struct page_pool *pool) pool->frag_page = NULL; - if (!page || - page_pool_atomic_sub_frag_count_return(page, drain_count)) + if (!page || page_pool_defrag_page(page, drain_count)) return; page_pool_return_page(pool, page); @@ -588,7 +585,7 @@ struct page *page_pool_alloc_frag(struct page_pool *pool, pool->frag_users = 1; *offset = 0; pool->frag_offset = size; - page_pool_set_frag_count(page, BIAS_MAX); + page_pool_fragment_page(page, BIAS_MAX); return page; }