page_frags.rst 2.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445
  1. .. _page_frags:
  2. ==============
  3. Page fragments
  4. ==============
  5. A page fragment is an arbitrary-length arbitrary-offset area of memory
  6. which resides within a 0 or higher order compound page. Multiple
  7. fragments within that page are individually refcounted, in the page's
  8. reference counter.
  9. The page_frag functions, page_frag_alloc and page_frag_free, provide a
  10. simple allocation framework for page fragments. This is used by the
  11. network stack and network device drivers to provide a backing region of
  12. memory for use as either an sk_buff->head, or to be used in the "frags"
  13. portion of skb_shared_info.
  14. In order to make use of the page fragment APIs a backing page fragment
  15. cache is needed. This provides a central point for the fragment allocation
  16. and tracks allows multiple calls to make use of a cached page. The
  17. advantage to doing this is that multiple calls to get_page can be avoided
  18. which can be expensive at allocation time. However due to the nature of
  19. this caching it is required that any calls to the cache be protected by
  20. either a per-cpu limitation, or a per-cpu limitation and forcing interrupts
  21. to be disabled when executing the fragment allocation.
  22. The network stack uses two separate caches per CPU to handle fragment
  23. allocation. The netdev_alloc_cache is used by callers making use of the
  24. netdev_alloc_frag and __netdev_alloc_skb calls. The napi_alloc_cache is
  25. used by callers of the __napi_alloc_frag and __napi_alloc_skb calls. The
  26. main difference between these two calls is the context in which they may be
  27. called. The "netdev" prefixed functions are usable in any context as these
  28. functions will disable interrupts, while the "napi" prefixed functions are
  29. only usable within the softirq context.
  30. Many network device drivers use a similar methodology for allocating page
  31. fragments, but the page fragments are cached at the ring or descriptor
  32. level. In order to enable these cases it is necessary to provide a generic
  33. way of tearing down a page cache. For this reason __page_frag_cache_drain
  34. was implemented. It allows for freeing multiple references from a single
  35. page via a single call. The advantage to doing this is that it allows for
  36. cleaning up the multiple references that were added to a page in order to
  37. avoid calling get_page per allocation.
  38. Alexander Duyck, Nov 29, 2016.