1============== 2Page fragments 3============== 4 5A page fragment is an arbitrary-length arbitrary-offset area of memory 6which resides within a 0 or higher order compound page. Multiple 7fragments within that page are individually refcounted, in the page's 8reference counter. 9 10The page_frag functions, page_frag_alloc and page_frag_free, provide a 11simple allocation framework for page fragments. This is used by the 12network stack and network device drivers to provide a backing region of 13memory for use as either an sk_buff->head, or to be used in the "frags" 14portion of skb_shared_info. 15 16In order to make use of the page fragment APIs a backing page fragment 17cache is needed. This provides a central point for the fragment allocation 18and tracks allows multiple calls to make use of a cached page. The 19advantage to doing this is that multiple calls to get_page can be avoided 20which can be expensive at allocation time. However due to the nature of 21this caching it is required that any calls to the cache be protected by 22either a per-cpu limitation, or a per-cpu limitation and forcing interrupts 23to be disabled when executing the fragment allocation. 24 25The network stack uses two separate caches per CPU to handle fragment 26allocation. The netdev_alloc_cache is used by callers making use of the 27netdev_alloc_frag and __netdev_alloc_skb calls. The napi_alloc_cache is 28used by callers of the __napi_alloc_frag and __napi_alloc_skb calls. The 29main difference between these two calls is the context in which they may be 30called. The "netdev" prefixed functions are usable in any context as these 31functions will disable interrupts, while the "napi" prefixed functions are 32only usable within the softirq context. 33 34Many network device drivers use a similar methodology for allocating page 35fragments, but the page fragments are cached at the ring or descriptor 36level. In order to enable these cases it is necessary to provide a generic 37way of tearing down a page cache. For this reason __page_frag_cache_drain 38was implemented. It allows for freeing multiple references from a single 39page via a single call. The advantage to doing this is that it allows for 40cleaning up the multiple references that were added to a page in order to 41avoid calling get_page per allocation. 42 43Alexander Duyck, Nov 29, 2016. 44