1.. SPDX-License-Identifier: GPL-2.0 2 3============= 4Page Pool API 5============= 6 7The page_pool allocator is optimized for the XDP mode that uses one frame 8per-page, but it can fallback on the regular page allocator APIs. 9 10Basic use involves replacing alloc_pages() calls with the 11page_pool_alloc_pages() call. Drivers should use page_pool_dev_alloc_pages() 12replacing dev_alloc_pages(). 13 14API keeps track of in-flight pages, in order to let API user know 15when it is safe to free a page_pool object. Thus, API users 16must call page_pool_put_page() to free the page, or attach 17the page to a page_pool-aware objects like skbs marked with 18skb_mark_for_recycle(). 19 20API user must call page_pool_put_page() once on a page, as it 21will either recycle the page, or in case of refcnt > 1, it will 22release the DMA mapping and in-flight state accounting. 23 24Architecture overview 25===================== 26 27.. code-block:: none 28 29 +------------------+ 30 | Driver | 31 +------------------+ 32 ^ 33 | 34 | 35 | 36 v 37 +--------------------------------------------+ 38 | request memory | 39 +--------------------------------------------+ 40 ^ ^ 41 | | 42 | Pool empty | Pool has entries 43 | | 44 v v 45 +-----------------------+ +------------------------+ 46 | alloc (and map) pages | | get page from cache | 47 +-----------------------+ +------------------------+ 48 ^ ^ 49 | | 50 | cache available | No entries, refill 51 | | from ptr-ring 52 | | 53 v v 54 +-----------------+ +------------------+ 55 | Fast cache | | ptr-ring cache | 56 +-----------------+ +------------------+ 57 58API interface 59============= 60The number of pools created **must** match the number of hardware queues 61unless hardware restrictions make that impossible. This would otherwise beat the 62purpose of page pool, which is allocate pages fast from cache without locking. 63This lockless guarantee naturally comes from running under a NAPI softirq. 64The protection doesn't strictly have to be NAPI, any guarantee that allocating 65a page will cause no race conditions is enough. 66 67* page_pool_create(): Create a pool. 68 * flags: PP_FLAG_DMA_MAP, PP_FLAG_DMA_SYNC_DEV 69 * order: 2^order pages on allocation 70 * pool_size: size of the ptr_ring 71 * nid: preferred NUMA node for allocation 72 * dev: struct device. Used on DMA operations 73 * dma_dir: DMA direction 74 * max_len: max DMA sync memory size 75 * offset: DMA address offset 76 77* page_pool_put_page(): The outcome of this depends on the page refcnt. If the 78 driver bumps the refcnt > 1 this will unmap the page. If the page refcnt is 1 79 the allocator owns the page and will try to recycle it in one of the pool 80 caches. If PP_FLAG_DMA_SYNC_DEV is set, the page will be synced for_device 81 using dma_sync_single_range_for_device(). 82 83* page_pool_put_full_page(): Similar to page_pool_put_page(), but will DMA sync 84 for the entire memory area configured in area pool->max_len. 85 86* page_pool_recycle_direct(): Similar to page_pool_put_full_page() but caller 87 must guarantee safe context (e.g NAPI), since it will recycle the page 88 directly into the pool fast cache. 89 90* page_pool_dev_alloc_pages(): Get a page from the page allocator or page_pool 91 caches. 92 93* page_pool_get_dma_addr(): Retrieve the stored DMA address. 94 95* page_pool_get_dma_dir(): Retrieve the stored DMA direction. 96 97* page_pool_put_page_bulk(): Tries to refill a number of pages into the 98 ptr_ring cache holding ptr_ring producer lock. If the ptr_ring is full, 99 page_pool_put_page_bulk() will release leftover pages to the page allocator. 100 page_pool_put_page_bulk() is suitable to be run inside the driver NAPI tx 101 completion loop for the XDP_REDIRECT use case. 102 Please note the caller must not use data area after running 103 page_pool_put_page_bulk(), as this function overwrites it. 104 105* page_pool_get_stats(): Retrieve statistics about the page_pool. This API 106 is only available if the kernel has been configured with 107 ``CONFIG_PAGE_POOL_STATS=y``. A pointer to a caller allocated ``struct 108 page_pool_stats`` structure is passed to this API which is filled in. The 109 caller can then report those stats to the user (perhaps via ethtool, 110 debugfs, etc.). See below for an example usage of this API. 111 112Stats API and structures 113------------------------ 114If the kernel is configured with ``CONFIG_PAGE_POOL_STATS=y``, the API 115``page_pool_get_stats()`` and structures described below are available. It 116takes a pointer to a ``struct page_pool`` and a pointer to a ``struct 117page_pool_stats`` allocated by the caller. 118 119The API will fill in the provided ``struct page_pool_stats`` with 120statistics about the page_pool. 121 122The stats structure has the following fields:: 123 124 struct page_pool_stats { 125 struct page_pool_alloc_stats alloc_stats; 126 struct page_pool_recycle_stats recycle_stats; 127 }; 128 129 130The ``struct page_pool_alloc_stats`` has the following fields: 131 * ``fast``: successful fast path allocations 132 * ``slow``: slow path order-0 allocations 133 * ``slow_high_order``: slow path high order allocations 134 * ``empty``: ptr ring is empty, so a slow path allocation was forced. 135 * ``refill``: an allocation which triggered a refill of the cache 136 * ``waive``: pages obtained from the ptr ring that cannot be added to 137 the cache due to a NUMA mismatch. 138 139The ``struct page_pool_recycle_stats`` has the following fields: 140 * ``cached``: recycling placed page in the page pool cache 141 * ``cache_full``: page pool cache was full 142 * ``ring``: page placed into the ptr ring 143 * ``ring_full``: page released from page pool because the ptr ring was full 144 * ``released_refcnt``: page released (and not recycled) because refcnt > 1 145 146Coding examples 147=============== 148 149Registration 150------------ 151 152.. code-block:: c 153 154 /* Page pool registration */ 155 struct page_pool_params pp_params = { 0 }; 156 struct xdp_rxq_info xdp_rxq; 157 int err; 158 159 pp_params.order = 0; 160 /* internal DMA mapping in page_pool */ 161 pp_params.flags = PP_FLAG_DMA_MAP; 162 pp_params.pool_size = DESC_NUM; 163 pp_params.nid = NUMA_NO_NODE; 164 pp_params.dev = priv->dev; 165 pp_params.napi = napi; /* only if locking is tied to NAPI */ 166 pp_params.dma_dir = xdp_prog ? DMA_BIDIRECTIONAL : DMA_FROM_DEVICE; 167 page_pool = page_pool_create(&pp_params); 168 169 err = xdp_rxq_info_reg(&xdp_rxq, ndev, 0); 170 if (err) 171 goto err_out; 172 173 err = xdp_rxq_info_reg_mem_model(&xdp_rxq, MEM_TYPE_PAGE_POOL, page_pool); 174 if (err) 175 goto err_out; 176 177NAPI poller 178----------- 179 180 181.. code-block:: c 182 183 /* NAPI Rx poller */ 184 enum dma_data_direction dma_dir; 185 186 dma_dir = page_pool_get_dma_dir(dring->page_pool); 187 while (done < budget) { 188 if (some error) 189 page_pool_recycle_direct(page_pool, page); 190 if (packet_is_xdp) { 191 if XDP_DROP: 192 page_pool_recycle_direct(page_pool, page); 193 } else (packet_is_skb) { 194 skb_mark_for_recycle(skb); 195 new_page = page_pool_dev_alloc_pages(page_pool); 196 } 197 } 198 199Stats 200----- 201 202.. code-block:: c 203 204 #ifdef CONFIG_PAGE_POOL_STATS 205 /* retrieve stats */ 206 struct page_pool_stats stats = { 0 }; 207 if (page_pool_get_stats(page_pool, &stats)) { 208 /* perhaps the driver reports statistics with ethool */ 209 ethtool_print_allocation_stats(&stats.alloc_stats); 210 ethtool_print_recycle_stats(&stats.recycle_stats); 211 } 212 #endif 213 214Driver unload 215------------- 216 217.. code-block:: c 218 219 /* Driver unload */ 220 page_pool_put_full_page(page_pool, page, false); 221 xdp_rxq_info_unreg(&xdp_rxq); 222