About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Documentation / block / biodoc.txt

Custom Search

Based on kernel version 4.0. Page generated on 2015-04-14 21:19 EST.

1		Notes on the Generic Block Layer Rewrite in Linux 2.5
2		=====================================================
4	Notes Written on Jan 15, 2002:
5		Jens Axboe <jens.axboe@oracle.com>
6		Suparna Bhattacharya <suparna@in.ibm.com>
8	Last Updated May 2, 2002
9	September 2003: Updated I/O Scheduler portions
10		Nick Piggin <npiggin@kernel.dk>
12	Introduction:
14	These are some notes describing some aspects of the 2.5 block layer in the
15	context of the bio rewrite. The idea is to bring out some of the key
16	changes and a glimpse of the rationale behind those changes.
18	Please mail corrections & suggestions to suparna@in.ibm.com.
20	Credits:
21	---------
23	2.5 bio rewrite:
24		Jens Axboe <jens.axboe@oracle.com>
26	Many aspects of the generic block layer redesign were driven by and evolved
27	over discussions, prior patches and the collective experience of several
28	people. See sections 8 and 9 for a list of some related references.
30	The following people helped with review comments and inputs for this
31	document:
32		Christoph Hellwig <hch@infradead.org>
33		Arjan van de Ven <arjanv@redhat.com>
34		Randy Dunlap <rdunlap@xenotime.net>
35		Andre Hedrick <andre@linux-ide.org>
37	The following people helped with fixes/contributions to the bio patches
38	while it was still work-in-progress:
39		David S. Miller <davem@redhat.com>
42	Description of Contents:
43	------------------------
45	1. Scope for tuning of logic to various needs
46	  1.1 Tuning based on device or low level driver capabilities
47		- Per-queue parameters
48		- Highmem I/O support
49		- I/O scheduler modularization
50	  1.2 Tuning based on high level requirements/capabilities
51		1.2.1 I/O Barriers
52		1.2.2 Request Priority/Latency
53	  1.3 Direct access/bypass to lower layers for diagnostics and special
54	      device operations
55		1.3.1 Pre-built commands
56	2. New flexible and generic but minimalist i/o structure or descriptor
57	   (instead of using buffer heads at the i/o layer)
58	  2.1 Requirements/Goals addressed
59	  2.2 The bio struct in detail (multi-page io unit)
60	  2.3 Changes in the request structure
61	3. Using bios
62	  3.1 Setup/teardown (allocation, splitting)
63	  3.2 Generic bio helper routines
64	    3.2.1 Traversing segments and completion units in a request
65	    3.2.2 Setting up DMA scatterlists
66	    3.2.3 I/O completion
67	    3.2.4 Implications for drivers that do not interpret bios (don't handle
68	 	  multiple segments)
69	    3.2.5 Request command tagging
70	  3.3 I/O submission
71	4. The I/O scheduler
72	5. Scalability related changes
73	  5.1 Granular locking: Removal of io_request_lock
74	  5.2 Prepare for transition to 64 bit sector_t
75	6. Other Changes/Implications
76	  6.1 Partition re-mapping handled by the generic block layer
77	7. A few tips on migration of older drivers
78	8. A list of prior/related/impacted patches/ideas
79	9. Other References/Discussion Threads
81	---------------------------------------------------------------------------
83	Bio Notes
84	--------
86	Let us discuss the changes in the context of how some overall goals for the
87	block layer are addressed.
89	1. Scope for tuning the generic logic to satisfy various requirements
91	The block layer design supports adaptable abstractions to handle common
92	processing with the ability to tune the logic to an appropriate extent
93	depending on the nature of the device and the requirements of the caller.
94	One of the objectives of the rewrite was to increase the degree of tunability
95	and to enable higher level code to utilize underlying device/driver
96	capabilities to the maximum extent for better i/o performance. This is
97	important especially in the light of ever improving hardware capabilities
98	and application/middleware software designed to take advantage of these
99	capabilities.
101	1.1 Tuning based on low level device / driver capabilities
103	Sophisticated devices with large built-in caches, intelligent i/o scheduling
104	optimizations, high memory DMA support, etc may find some of the
105	generic processing an overhead, while for less capable devices the
106	generic functionality is essential for performance or correctness reasons.
107	Knowledge of some of the capabilities or parameters of the device should be
108	used at the generic block layer to take the right decisions on
109	behalf of the driver.
111	How is this achieved ?
113	Tuning at a per-queue level:
115	i. Per-queue limits/values exported to the generic layer by the driver
117	Various parameters that the generic i/o scheduler logic uses are set at
118	a per-queue level (e.g maximum request size, maximum number of segments in
119	a scatter-gather list, hardsect size)
121	Some parameters that were earlier available as global arrays indexed by
122	major/minor are now directly associated with the queue. Some of these may
123	move into the block device structure in the future. Some characteristics
124	have been incorporated into a queue flags field rather than separate fields
125	in themselves.  There are blk_queue_xxx functions to set the parameters,
126	rather than update the fields directly
128	Some new queue property settings:
130		blk_queue_bounce_limit(q, u64 dma_address)
131			Enable I/O to highmem pages, dma_address being the
132			limit. No highmem default.
134		blk_queue_max_sectors(q, max_sectors)
135			Sets two variables that limit the size of the request.
137			- The request queue's max_sectors, which is a soft size in
138			units of 512 byte sectors, and could be dynamically varied
139			by the core kernel.
141			- The request queue's max_hw_sectors, which is a hard limit
142			and reflects the maximum size request a driver can handle
143			in units of 512 byte sectors.
145			The default for both max_sectors and max_hw_sectors is
146			255. The upper limit of max_sectors is 1024.
148		blk_queue_max_phys_segments(q, max_segments)
149			Maximum physical segments you can handle in a request. 128
150			default (driver limit). (See 3.2.2)
152		blk_queue_max_hw_segments(q, max_segments)
153			Maximum dma segments the hardware can handle in a request. 128
154			default (host adapter limit, after dma remapping).
155			(See 3.2.2)
157		blk_queue_max_segment_size(q, max_seg_size)
158			Maximum size of a clustered segment, 64kB default.
160		blk_queue_hardsect_size(q, hardsect_size)
161			Lowest possible sector size that the hardware can operate
162			on, 512 bytes default.
164	New queue flags:
166		QUEUE_FLAG_CLUSTER (see 3.2.2)
167		QUEUE_FLAG_QUEUED (see 3.2.4)
170	ii. High-mem i/o capabilities are now considered the default
172	The generic bounce buffer logic, present in 2.4, where the block layer would
173	by default copyin/out i/o requests on high-memory buffers to low-memory buffers
174	assuming that the driver wouldn't be able to handle it directly, has been
175	changed in 2.5. The bounce logic is now applied only for memory ranges
176	for which the device cannot handle i/o. A driver can specify this by
177	setting the queue bounce limit for the request queue for the device
178	(blk_queue_bounce_limit()). This avoids the inefficiencies of the copyin/out
179	where a device is capable of handling high memory i/o.
181	In order to enable high-memory i/o where the device is capable of supporting
182	it, the pci dma mapping routines and associated data structures have now been
183	modified to accomplish a direct page -> bus translation, without requiring
184	a virtual address mapping (unlike the earlier scheme of virtual address
185	-> bus translation). So this works uniformly for high-memory pages (which
186	do not have a corresponding kernel virtual address space mapping) and
187	low-memory pages.
189	Note: Please refer to Documentation/DMA-API-HOWTO.txt for a discussion
190	on PCI high mem DMA aspects and mapping of scatter gather lists, and support
191	for 64 bit PCI.
193	Special handling is required only for cases where i/o needs to happen on
194	pages at physical memory addresses beyond what the device can support. In these
195	cases, a bounce bio representing a buffer from the supported memory range
196	is used for performing the i/o with copyin/copyout as needed depending on
197	the type of the operation.  For example, in case of a read operation, the
198	data read has to be copied to the original buffer on i/o completion, so a
199	callback routine is set up to do this, while for write, the data is copied
200	from the original buffer to the bounce buffer prior to issuing the
201	operation. Since an original buffer may be in a high memory area that's not
202	mapped in kernel virtual addr, a kmap operation may be required for
203	performing the copy, and special care may be needed in the completion path
204	as it may not be in irq context. Special care is also required (by way of
205	GFP flags) when allocating bounce buffers, to avoid certain highmem
206	deadlock possibilities.
208	It is also possible that a bounce buffer may be allocated from high-memory
209	area that's not mapped in kernel virtual addr, but within the range that the
210	device can use directly; so the bounce page may need to be kmapped during
211	copy operations. [Note: This does not hold in the current implementation,
212	though]
214	There are some situations when pages from high memory may need to
215	be kmapped, even if bounce buffers are not necessary. For example a device
216	may need to abort DMA operations and revert to PIO for the transfer, in
217	which case a virtual mapping of the page is required. For SCSI it is also
218	done in some scenarios where the low level driver cannot be trusted to
219	handle a single sg entry correctly. The driver is expected to perform the
220	kmaps as needed on such occasions using the __bio_kmap_atomic and bio_kmap_irq
221	routines as appropriate. A driver could also use the blk_queue_bounce()
222	routine on its own to bounce highmem i/o to low memory for specific requests
223	if so desired.
225	iii. The i/o scheduler algorithm itself can be replaced/set as appropriate
227	As in 2.4, it is possible to plugin a brand new i/o scheduler for a particular
228	queue or pick from (copy) existing generic schedulers and replace/override
229	certain portions of it. The 2.5 rewrite provides improved modularization
230	of the i/o scheduler. There are more pluggable callbacks, e.g for init,
231	add request, extract request, which makes it possible to abstract specific
232	i/o scheduling algorithm aspects and details outside of the generic loop.
233	It also makes it possible to completely hide the implementation details of
234	the i/o scheduler from block drivers.
236	I/O scheduler wrappers are to be used instead of accessing the queue directly.
237	See section 4. The I/O scheduler for details.
239	1.2 Tuning Based on High level code capabilities
241	i. Application capabilities for raw i/o
243	This comes from some of the high-performance database/middleware
244	requirements where an application prefers to make its own i/o scheduling
245	decisions based on an understanding of the access patterns and i/o
246	characteristics
248	ii. High performance filesystems or other higher level kernel code's
249	capabilities
251	Kernel components like filesystems could also take their own i/o scheduling
252	decisions for optimizing performance. Journalling filesystems may need
253	some control over i/o ordering.
255	What kind of support exists at the generic block layer for this ?
257	The flags and rw fields in the bio structure can be used for some tuning
258	from above e.g indicating that an i/o is just a readahead request, or for
259	marking  barrier requests (discussed next), or priority settings (currently
260	unused). As far as user applications are concerned they would need an
261	additional mechanism either via open flags or ioctls, or some other upper
262	level mechanism to communicate such settings to block.
264	1.2.1 I/O Barriers
266	There is a way to enforce strict ordering for i/os through barriers.
267	All requests before a barrier point must be serviced before the barrier
268	request and any other requests arriving after the barrier will not be
269	serviced until after the barrier has completed. This is useful for higher
270	level control on write ordering, e.g flushing a log of committed updates
271	to disk before the corresponding updates themselves.
273	A flag in the bio structure, BIO_BARRIER is used to identify a barrier i/o.
274	The generic i/o scheduler would make sure that it places the barrier request and
275	all other requests coming after it after all the previous requests in the
276	queue. Barriers may be implemented in different ways depending on the
277	driver. For more details regarding I/O barriers, please read barrier.txt
278	in this directory.
280	1.2.2 Request Priority/Latency
282	Todo/Under discussion:
283	Arjan's proposed request priority scheme allows higher levels some broad
284	  control (high/med/low) over the priority  of an i/o request vs other pending
285	  requests in the queue. For example it allows reads for bringing in an
286	  executable page on demand to be given a higher priority over pending write
287	  requests which haven't aged too much on the queue. Potentially this priority
288	  could even be exposed to applications in some manner, providing higher level
289	  tunability. Time based aging avoids starvation of lower priority
290	  requests. Some bits in the bi_rw flags field in the bio structure are
291	  intended to be used for this priority information.
294	1.3 Direct Access to Low level Device/Driver Capabilities (Bypass mode)
295	    (e.g Diagnostics, Systems Management)
297	There are situations where high-level code needs to have direct access to
298	the low level device capabilities or requires the ability to issue commands
299	to the device bypassing some of the intermediate i/o layers.
300	These could, for example, be special control commands issued through ioctl
301	interfaces, or could be raw read/write commands that stress the drive's
302	capabilities for certain kinds of fitness tests. Having direct interfaces at
303	multiple levels without having to pass through upper layers makes
304	it possible to perform bottom up validation of the i/o path, layer by
305	layer, starting from the media.
307	The normal i/o submission interfaces, e.g submit_bio, could be bypassed
308	for specially crafted requests which such ioctl or diagnostics
309	interfaces would typically use, and the elevator add_request routine
310	can instead be used to directly insert such requests in the queue or preferably
311	the blk_do_rq routine can be used to place the request on the queue and
312	wait for completion. Alternatively, sometimes the caller might just
313	invoke a lower level driver specific interface with the request as a
314	parameter.
316	If the request is a means for passing on special information associated with
317	the command, then such information is associated with the request->special
318	field (rather than misuse the request->buffer field which is meant for the
319	request data buffer's virtual mapping).
321	For passing request data, the caller must build up a bio descriptor
322	representing the concerned memory buffer if the underlying driver interprets
323	bio segments or uses the block layer end*request* functions for i/o
324	completion. Alternatively one could directly use the request->buffer field to
325	specify the virtual address of the buffer, if the driver expects buffer
326	addresses passed in this way and ignores bio entries for the request type
327	involved. In the latter case, the driver would modify and manage the
328	request->buffer, request->sector and request->nr_sectors or
329	request->current_nr_sectors fields itself rather than using the block layer
330	end_request or end_that_request_first completion interfaces.
331	(See 2.3 or Documentation/block/request.txt for a brief explanation of
332	the request structure fields)
334	[TBD: end_that_request_last should be usable even in this case;
335	Perhaps an end_that_direct_request_first routine could be implemented to make
336	handling direct requests easier for such drivers; Also for drivers that
337	expect bios, a helper function could be provided for setting up a bio
338	corresponding to a data buffer]
340	<JENS: I dont understand the above, why is end_that_request_first() not
341	usable? Or _last for that matter. I must be missing something>
342	<SUP: What I meant here was that if the request doesn't have a bio, then
343	 end_that_request_first doesn't modify nr_sectors or current_nr_sectors,
344	 and hence can't be used for advancing request state settings on the
345	 completion of partial transfers. The driver has to modify these fields 
346	 directly by hand.
347	 This is because end_that_request_first only iterates over the bio list,
348	 and always returns 0 if there are none associated with the request.
349	 _last works OK in this case, and is not a problem, as I mentioned earlier
350	>
352	1.3.1 Pre-built Commands
354	A request can be created with a pre-built custom command  to be sent directly
355	to the device. The cmd block in the request structure has room for filling
356	in the command bytes. (i.e rq->cmd is now 16 bytes in size, and meant for
357	command pre-building, and the type of the request is now indicated
358	through rq->flags instead of via rq->cmd)
360	The request structure flags can be set up to indicate the type of request
361	in such cases (REQ_PC: direct packet command passed to driver, REQ_BLOCK_PC:
362	packet command issued via blk_do_rq, REQ_SPECIAL: special request).
364	It can help to pre-build device commands for requests in advance.
365	Drivers can now specify a request prepare function (q->prep_rq_fn) that the
366	block layer would invoke to pre-build device commands for a given request,
367	or perform other preparatory processing for the request. This is routine is
368	called by elv_next_request(), i.e. typically just before servicing a request.
369	(The prepare function would not be called for requests that have REQ_DONTPREP
370	enabled)
372	Aside:
373	  Pre-building could possibly even be done early, i.e before placing the
374	  request on the queue, rather than construct the command on the fly in the
375	  driver while servicing the request queue when it may affect latencies in
376	  interrupt context or responsiveness in general. One way to add early
377	  pre-building would be to do it whenever we fail to merge on a request.
378	  Now REQ_NOMERGE is set in the request flags to skip this one in the future,
379	  which means that it will not change before we feed it to the device. So
380	  the pre-builder hook can be invoked there.
383	2. Flexible and generic but minimalist i/o structure/descriptor.
385	2.1 Reason for a new structure and requirements addressed
387	Prior to 2.5, buffer heads were used as the unit of i/o at the generic block
388	layer, and the low level request structure was associated with a chain of
389	buffer heads for a contiguous i/o request. This led to certain inefficiencies
390	when it came to large i/o requests and readv/writev style operations, as it
391	forced such requests to be broken up into small chunks before being passed
392	on to the generic block layer, only to be merged by the i/o scheduler
393	when the underlying device was capable of handling the i/o in one shot.
394	Also, using the buffer head as an i/o structure for i/os that didn't originate
395	from the buffer cache unnecessarily added to the weight of the descriptors
396	which were generated for each such chunk.
398	The following were some of the goals and expectations considered in the
399	redesign of the block i/o data structure in 2.5.
401	i.  Should be appropriate as a descriptor for both raw and buffered i/o  -
402	    avoid cache related fields which are irrelevant in the direct/page i/o path,
403	    or filesystem block size alignment restrictions which may not be relevant
404	    for raw i/o.
405	ii. Ability to represent high-memory buffers (which do not have a virtual
406	    address mapping in kernel address space).
407	iii.Ability to represent large i/os w/o unnecessarily breaking them up (i.e
408	    greater than PAGE_SIZE chunks in one shot)
409	iv. At the same time, ability to retain independent identity of i/os from
410	    different sources or i/o units requiring individual completion (e.g. for
411	    latency reasons)
412	v.  Ability to represent an i/o involving multiple physical memory segments
413	    (including non-page aligned page fragments, as specified via readv/writev)
414	    without unnecessarily breaking it up, if the underlying device is capable of
415	    handling it.
416	vi. Preferably should be based on a memory descriptor structure that can be
417	    passed around different types of subsystems or layers, maybe even
418	    networking, without duplication or extra copies of data/descriptor fields
419	    themselves in the process
420	vii.Ability to handle the possibility of splits/merges as the structure passes
421	    through layered drivers (lvm, md, evms), with minimal overhead.
423	The solution was to define a new structure (bio)  for the block layer,
424	instead of using the buffer head structure (bh) directly, the idea being
425	avoidance of some associated baggage and limitations. The bio structure
426	is uniformly used for all i/o at the block layer ; it forms a part of the
427	bh structure for buffered i/o, and in the case of raw/direct i/o kiobufs are
428	mapped to bio structures.
430	2.2 The bio struct
432	The bio structure uses a vector representation pointing to an array of tuples
433	of <page, offset, len> to describe the i/o buffer, and has various other
434	fields describing i/o parameters and state that needs to be maintained for
435	performing the i/o.
437	Notice that this representation means that a bio has no virtual address
438	mapping at all (unlike buffer heads).
440	struct bio_vec {
441	       struct page     *bv_page;
442	       unsigned short  bv_len;
443	       unsigned short  bv_offset;
444	};
446	/*
447	 * main unit of I/O for the block layer and lower layers (ie drivers)
448	 */
449	struct bio {
450	       struct bio          *bi_next;    /* request queue link */
451	       struct block_device *bi_bdev;	/* target device */
452	       unsigned long       bi_flags;    /* status, command, etc */
453	       unsigned long       bi_rw;       /* low bits: r/w, high: priority */
455	       unsigned int	bi_vcnt;     /* how may bio_vec's */
456	       struct bvec_iter	bi_iter;	/* current index into bio_vec array */
458	       unsigned int	bi_size;     /* total size in bytes */
459	       unsigned short 	bi_phys_segments; /* segments after physaddr coalesce*/
460	       unsigned short	bi_hw_segments; /* segments after DMA remapping */
461	       unsigned int	bi_max;	     /* max bio_vecs we can hold
462	                                        used as index into pool */
463	       struct bio_vec   *bi_io_vec;  /* the actual vec list */
464	       bio_end_io_t	*bi_end_io;  /* bi_end_io (bio) */
465	       atomic_t		bi_cnt;	     /* pin count: free when it hits zero */
466	       void             *bi_private;
467	};
469	With this multipage bio design:
471	- Large i/os can be sent down in one go using a bio_vec list consisting
472	  of an array of <page, offset, len> fragments (similar to the way fragments
473	  are represented in the zero-copy network code)
474	- Splitting of an i/o request across multiple devices (as in the case of
475	  lvm or raid) is achieved by cloning the bio (where the clone points to
476	  the same bi_io_vec array, but with the index and size accordingly modified)
477	- A linked list of bios is used as before for unrelated merges (*) - this
478	  avoids reallocs and makes independent completions easier to handle.
479	- Code that traverses the req list can find all the segments of a bio
480	  by using rq_for_each_segment.  This handles the fact that a request
481	  has multiple bios, each of which can have multiple segments.
482	- Drivers which can't process a large bio in one shot can use the bi_iter
483	  field to keep track of the next bio_vec entry to process.
484	  (e.g a 1MB bio_vec needs to be handled in max 128kB chunks for IDE)
485	  [TBD: Should preferably also have a bi_voffset and bi_vlen to avoid modifying
486	   bi_offset an len fields]
488	(*) unrelated merges -- a request ends up containing two or more bios that
489	    didn't originate from the same place.
491	bi_end_io() i/o callback gets called on i/o completion of the entire bio.
493	At a lower level, drivers build a scatter gather list from the merged bios.
494	The scatter gather list is in the form of an array of <page, offset, len>
495	entries with their corresponding dma address mappings filled in at the
496	appropriate time. As an optimization, contiguous physical pages can be
497	covered by a single entry where <page> refers to the first page and <len>
498	covers the range of pages (up to 16 contiguous pages could be covered this
499	way). There is a helper routine (blk_rq_map_sg) which drivers can use to build
500	the sg list.
502	Note: Right now the only user of bios with more than one page is ll_rw_kio,
503	which in turn means that only raw I/O uses it (direct i/o may not work
504	right now). The intent however is to enable clustering of pages etc to
505	become possible. The pagebuf abstraction layer from SGI also uses multi-page
506	bios, but that is currently not included in the stock development kernels.
507	The same is true of Andrew Morton's work-in-progress multipage bio writeout 
508	and readahead patches.
510	2.3 Changes in the Request Structure
512	The request structure is the structure that gets passed down to low level
513	drivers. The block layer make_request function builds up a request structure,
514	places it on the queue and invokes the drivers request_fn. The driver makes
515	use of block layer helper routine elv_next_request to pull the next request
516	off the queue. Control or diagnostic functions might bypass block and directly
517	invoke underlying driver entry points passing in a specially constructed
518	request structure.
520	Only some relevant fields (mainly those which changed or may be referred
521	to in some of the discussion here) are listed below, not necessarily in
522	the order in which they occur in the structure (see include/linux/blkdev.h)
523	Refer to Documentation/block/request.txt for details about all the request
524	structure fields and a quick reference about the layers which are
525	supposed to use or modify those fields.
527	struct request {
528		struct list_head queuelist;  /* Not meant to be directly accessed by
529						the driver.
530						Used by q->elv_next_request_fn
531						rq->queue is gone
532						*/
533		.
534		.
535		unsigned char cmd[16]; /* prebuilt command data block */
536		unsigned long flags;   /* also includes earlier rq->cmd settings */
537		.
538		.
539		sector_t sector; /* this field is now of type sector_t instead of int
540				    preparation for 64 bit sectors */
541		.
542		.
544		/* Number of scatter-gather DMA addr+len pairs after
545		 * physical address coalescing is performed.
546		 */
547		unsigned short nr_phys_segments;
549		/* Number of scatter-gather addr+len pairs after
550		 * physical and DMA remapping hardware coalescing is performed.
551		 * This is the number of scatter-gather entries the driver
552		 * will actually have to deal with after DMA mapping is done.
553		 */
554		unsigned short nr_hw_segments;
556		/* Various sector counts */
557		unsigned long nr_sectors;  /* no. of sectors left: driver modifiable */
558		unsigned long hard_nr_sectors;  /* block internal copy of above */
559		unsigned int current_nr_sectors; /* no. of sectors left in the
560						   current segment:driver modifiable */
561		unsigned long hard_cur_sectors; /* block internal copy of the above */
562		.
563		.
564		int tag;	/* command tag associated with request */
565		void *special;  /* same as before */
566		char *buffer;   /* valid only for low memory buffers up to
567				 current_nr_sectors */
568		.
569		.
570		struct bio *bio, *biotail;  /* bio list instead of bh */
571		struct request_list *rl;
572	}
574	See the rq_flag_bits definitions for an explanation of the various flags
575	available. Some bits are used by the block layer or i/o scheduler.
577	The behaviour of the various sector counts are almost the same as before,
578	except that since we have multi-segment bios, current_nr_sectors refers
579	to the numbers of sectors in the current segment being processed which could
580	be one of the many segments in the current bio (i.e i/o completion unit).
581	The nr_sectors value refers to the total number of sectors in the whole
582	request that remain to be transferred (no change). The purpose of the
583	hard_xxx values is for block to remember these counts every time it hands
584	over the request to the driver. These values are updated by block on
585	end_that_request_first, i.e. every time the driver completes a part of the
586	transfer and invokes block end*request helpers to mark this. The
587	driver should not modify these values. The block layer sets up the
588	nr_sectors and current_nr_sectors fields (based on the corresponding
589	hard_xxx values and the number of bytes transferred) and updates it on
590	every transfer that invokes end_that_request_first. It does the same for the
591	buffer, bio, bio->bi_iter fields too.
593	The buffer field is just a virtual address mapping of the current segment
594	of the i/o buffer in cases where the buffer resides in low-memory. For high
595	memory i/o, this field is not valid and must not be used by drivers.
597	Code that sets up its own request structures and passes them down to
598	a driver needs to be careful about interoperation with the block layer helper
599	functions which the driver uses. (Section 1.3)
601	3. Using bios
603	3.1 Setup/Teardown
605	There are routines for managing the allocation, and reference counting, and
606	freeing of bios (bio_alloc, bio_get, bio_put).
608	This makes use of Ingo Molnar's mempool implementation, which enables
609	subsystems like bio to maintain their own reserve memory pools for guaranteed
610	deadlock-free allocations during extreme VM load. For example, the VM
611	subsystem makes use of the block layer to writeout dirty pages in order to be
612	able to free up memory space, a case which needs careful handling. The
613	allocation logic draws from the preallocated emergency reserve in situations
614	where it cannot allocate through normal means. If the pool is empty and it
615	can wait, then it would trigger action that would help free up memory or
616	replenish the pool (without deadlocking) and wait for availability in the pool.
617	If it is in IRQ context, and hence not in a position to do this, allocation
618	could fail if the pool is empty. In general mempool always first tries to
619	perform allocation without having to wait, even if it means digging into the
620	pool as long it is not less that 50% full.
622	On a free, memory is released to the pool or directly freed depending on
623	the current availability in the pool. The mempool interface lets the
624	subsystem specify the routines to be used for normal alloc and free. In the
625	case of bio, these routines make use of the standard slab allocator.
627	The caller of bio_alloc is expected to taken certain steps to avoid
628	deadlocks, e.g. avoid trying to allocate more memory from the pool while
629	already holding memory obtained from the pool.
630	[TBD: This is a potential issue, though a rare possibility
631	 in the bounce bio allocation that happens in the current code, since
632	 it ends up allocating a second bio from the same pool while
633	 holding the original bio ]
635	Memory allocated from the pool should be released back within a limited
636	amount of time (in the case of bio, that would be after the i/o is completed).
637	This ensures that if part of the pool has been used up, some work (in this
638	case i/o) must already be in progress and memory would be available when it
639	is over. If allocating from multiple pools in the same code path, the order
640	or hierarchy of allocation needs to be consistent, just the way one deals
641	with multiple locks.
643	The bio_alloc routine also needs to allocate the bio_vec_list (bvec_alloc())
644	for a non-clone bio. There are the 6 pools setup for different size biovecs,
645	so bio_alloc(gfp_mask, nr_iovecs) will allocate a vec_list of the
646	given size from these slabs.
648	The bio_get() routine may be used to hold an extra reference on a bio prior
649	to i/o submission, if the bio fields are likely to be accessed after the
650	i/o is issued (since the bio may otherwise get freed in case i/o completion
651	happens in the meantime).
653	The bio_clone() routine may be used to duplicate a bio, where the clone
654	shares the bio_vec_list with the original bio (i.e. both point to the
655	same bio_vec_list). This would typically be used for splitting i/o requests
656	in lvm or md.
658	3.2 Generic bio helper Routines
660	3.2.1 Traversing segments and completion units in a request
662	The macro rq_for_each_segment() should be used for traversing the bios
663	in the request list (drivers should avoid directly trying to do it
664	themselves). Using these helpers should also make it easier to cope
665	with block changes in the future.
667		struct req_iterator iter;
668		rq_for_each_segment(bio_vec, rq, iter)
669			/* bio_vec is now current segment */
671	I/O completion callbacks are per-bio rather than per-segment, so drivers
672	that traverse bio chains on completion need to keep that in mind. Drivers
673	which don't make a distinction between segments and completion units would
674	need to be reorganized to support multi-segment bios.
676	3.2.2 Setting up DMA scatterlists
678	The blk_rq_map_sg() helper routine would be used for setting up scatter
679	gather lists from a request, so a driver need not do it on its own.
681		nr_segments = blk_rq_map_sg(q, rq, scatterlist);
683	The helper routine provides a level of abstraction which makes it easier
684	to modify the internals of request to scatterlist conversion down the line
685	without breaking drivers. The blk_rq_map_sg routine takes care of several
686	things like collapsing physically contiguous segments (if QUEUE_FLAG_CLUSTER
687	is set) and correct segment accounting to avoid exceeding the limits which
688	the i/o hardware can handle, based on various queue properties.
690	- Prevents a clustered segment from crossing a 4GB mem boundary
691	- Avoids building segments that would exceed the number of physical
692	  memory segments that the driver can handle (phys_segments) and the
693	  number that the underlying hardware can handle at once, accounting for
694	  DMA remapping (hw_segments)  (i.e. IOMMU aware limits).
696	Routines which the low level driver can use to set up the segment limits:
698	blk_queue_max_hw_segments() : Sets an upper limit of the maximum number of
699	hw data segments in a request (i.e. the maximum number of address/length
700	pairs the host adapter can actually hand to the device at once)
702	blk_queue_max_phys_segments() : Sets an upper limit on the maximum number
703	of physical data segments in a request (i.e. the largest sized scatter list
704	a driver could handle)
706	3.2.3 I/O completion
708	The existing generic block layer helper routines end_request,
709	end_that_request_first and end_that_request_last can be used for i/o
710	completion (and setting things up so the rest of the i/o or the next
711	request can be kicked of) as before. With the introduction of multi-page
712	bio support, end_that_request_first requires an additional argument indicating
713	the number of sectors completed.
715	3.2.4 Implications for drivers that do not interpret bios (don't handle
716	 multiple segments)
718	Drivers that do not interpret bios e.g those which do not handle multiple
719	segments and do not support i/o into high memory addresses (require bounce
720	buffers) and expect only virtually mapped buffers, can access the rq->buffer
721	field. As before the driver should use current_nr_sectors to determine the
722	size of remaining data in the current segment (that is the maximum it can
723	transfer in one go unless it interprets segments), and rely on the block layer
724	end_request, or end_that_request_first/last to take care of all accounting
725	and transparent mapping of the next bio segment when a segment boundary
726	is crossed on completion of a transfer. (The end*request* functions should
727	be used if only if the request has come down from block/bio path, not for
728	direct access requests which only specify rq->buffer without a valid rq->bio)
730	3.2.5 Generic request command tagging
732 Tag helpers
734	Block now offers some simple generic functionality to help support command
735	queueing (typically known as tagged command queueing), ie manage more than
736	one outstanding command on a queue at any given time.
738		blk_queue_init_tags(struct request_queue *q, int depth)
740		Initialize internal command tagging structures for a maximum
741		depth of 'depth'.
743		blk_queue_free_tags((struct request_queue *q)
745		Teardown tag info associated with the queue. This will be done
746		automatically by block if blk_queue_cleanup() is called on a queue
747		that is using tagging.
749	The above are initialization and exit management, the main helpers during
750	normal operations are:
752		blk_queue_start_tag(struct request_queue *q, struct request *rq)
754		Start tagged operation for this request. A free tag number between
755		0 and 'depth' is assigned to the request (rq->tag holds this number),
756		and 'rq' is added to the internal tag management. If the maximum depth
757		for this queue is already achieved (or if the tag wasn't started for
758		some other reason), 1 is returned. Otherwise 0 is returned.
760		blk_queue_end_tag(struct request_queue *q, struct request *rq)
762		End tagged operation on this request. 'rq' is removed from the internal
763		book keeping structures.
765	To minimize struct request and queue overhead, the tag helpers utilize some
766	of the same request members that are used for normal request queue management.
767	This means that a request cannot both be an active tag and be on the queue
768	list at the same time. blk_queue_start_tag() will remove the request, but
769	the driver must remember to call blk_queue_end_tag() before signalling
770	completion of the request to the block layer. This means ending tag
771	operations before calling end_that_request_last()! For an example of a user
772	of these helpers, see the IDE tagged command queueing support.
774	Certain hardware conditions may dictate a need to invalidate the block tag
775	queue. For instance, on IDE any tagged request error needs to clear both
776	the hardware and software block queue and enable the driver to sanely restart
777	all the outstanding requests. There's a third helper to do that:
779		blk_queue_invalidate_tags(struct request_queue *q)
781		Clear the internal block tag queue and re-add all the pending requests
782		to the request queue. The driver will receive them again on the
783		next request_fn run, just like it did the first time it encountered
784		them.
786 Tag info
788	Some block functions exist to query current tag status or to go from a
789	tag number to the associated request. These are, in no particular order:
791		blk_queue_tagged(q)
793		Returns 1 if the queue 'q' is using tagging, 0 if not.
795		blk_queue_tag_request(q, tag)
797		Returns a pointer to the request associated with tag 'tag'.
799		blk_queue_tag_depth(q)
801		Return current queue depth.
803		blk_queue_tag_queue(q)
805		Returns 1 if the queue can accept a new queued command, 0 if we are
806		at the maximum depth already.
808		blk_queue_rq_tagged(rq)
810		Returns 1 if the request 'rq' is tagged.
812 Internal structure
814	Internally, block manages tags in the blk_queue_tag structure:
816		struct blk_queue_tag {
817			struct request **tag_index;	/* array or pointers to rq */
818			unsigned long *tag_map;		/* bitmap of free tags */
819			struct list_head busy_list;	/* fifo list of busy tags */
820			int busy;			/* queue depth */
821			int max_depth;			/* max queue depth */
822		};
824	Most of the above is simple and straight forward, however busy_list may need
825	a bit of explaining. Normally we don't care too much about request ordering,
826	but in the event of any barrier requests in the tag queue we need to ensure
827	that requests are restarted in the order they were queue. This may happen
828	if the driver needs to use blk_queue_invalidate_tags().
830	3.3 I/O Submission
832	The routine submit_bio() is used to submit a single io. Higher level i/o
833	routines make use of this:
835	(a) Buffered i/o:
836	The routine submit_bh() invokes submit_bio() on a bio corresponding to the
837	bh, allocating the bio if required. ll_rw_block() uses submit_bh() as before.
839	(b) Kiobuf i/o (for raw/direct i/o):
840	The ll_rw_kio() routine breaks up the kiobuf into page sized chunks and
841	maps the array to one or more multi-page bios, issuing submit_bio() to
842	perform the i/o on each of these.
844	The embedded bh array in the kiobuf structure has been removed and no
845	preallocation of bios is done for kiobufs. [The intent is to remove the
846	blocks array as well, but it's currently in there to kludge around direct i/o.]
847	Thus kiobuf allocation has switched back to using kmalloc rather than vmalloc.
849	Todo/Observation:
851	 A single kiobuf structure is assumed to correspond to a contiguous range
852	 of data, so brw_kiovec() invokes ll_rw_kio for each kiobuf in a kiovec.
853	 So right now it wouldn't work for direct i/o on non-contiguous blocks.
854	 This is to be resolved.  The eventual direction is to replace kiobuf
855	 by kvec's.
857	 Badari Pulavarty has a patch to implement direct i/o correctly using
858	 bio and kvec.
861	(c) Page i/o:
862	Todo/Under discussion:
864	 Andrew Morton's multi-page bio patches attempt to issue multi-page
865	 writeouts (and reads) from the page cache, by directly building up
866	 large bios for submission completely bypassing the usage of buffer
867	 heads. This work is still in progress.
869	 Christoph Hellwig had some code that uses bios for page-io (rather than
870	 bh). This isn't included in bio as yet. Christoph was also working on a
871	 design for representing virtual/real extents as an entity and modifying
872	 some of the address space ops interfaces to utilize this abstraction rather
873	 than buffer_heads. (This is somewhat along the lines of the SGI XFS pagebuf
874	 abstraction, but intended to be as lightweight as possible).
876	(d) Direct access i/o:
877	Direct access requests that do not contain bios would be submitted differently
878	as discussed earlier in section 1.3.
880	Aside:
882	  Kvec i/o:
884	  Ben LaHaise's aio code uses a slightly different structure instead
885	  of kiobufs, called a kvec_cb. This contains an array of <page, offset, len>
886	  tuples (very much like the networking code), together with a callback function
887	  and data pointer. This is embedded into a brw_cb structure when passed
888	  to brw_kvec_async().
890	  Now it should be possible to directly map these kvecs to a bio. Just as while
891	  cloning, in this case rather than PRE_BUILT bio_vecs, we set the bi_io_vec
892	  array pointer to point to the veclet array in kvecs.
894	  TBD: In order for this to work, some changes are needed in the way multi-page
895	  bios are handled today. The values of the tuples in such a vector passed in
896	  from higher level code should not be modified by the block layer in the course
897	  of its request processing, since that would make it hard for the higher layer
898	  to continue to use the vector descriptor (kvec) after i/o completes. Instead,
899	  all such transient state should either be maintained in the request structure,
900	  and passed on in some way to the endio completion routine.
903	4. The I/O scheduler
904	I/O scheduler, a.k.a. elevator, is implemented in two layers.  Generic dispatch
905	queue and specific I/O schedulers.  Unless stated otherwise, elevator is used
906	to refer to both parts and I/O scheduler to specific I/O schedulers.
908	Block layer implements generic dispatch queue in block/*.c.
909	The generic dispatch queue is responsible for properly ordering barrier
910	requests, requeueing, handling non-fs requests and all other subtleties.
912	Specific I/O schedulers are responsible for ordering normal filesystem
913	requests.  They can also choose to delay certain requests to improve
914	throughput or whatever purpose.  As the plural form indicates, there are
915	multiple I/O schedulers.  They can be built as modules but at least one should
916	be built inside the kernel.  Each queue can choose different one and can also
917	change to another one dynamically.
919	A block layer call to the i/o scheduler follows the convention elv_xxx(). This
920	calls elevator_xxx_fn in the elevator switch (block/elevator.c). Oh, xxx
921	and xxx might not match exactly, but use your imagination. If an elevator
922	doesn't implement a function, the switch does nothing or some minimal house
923	keeping work.
925	4.1. I/O scheduler API
927	The functions an elevator may implement are: (* are mandatory)
928	elevator_merge_fn		called to query requests for merge with a bio
930	elevator_merge_req_fn		called when two requests get merged. the one
931					which gets merged into the other one will be
932					never seen by I/O scheduler again. IOW, after
933					being merged, the request is gone.
935	elevator_merged_fn		called when a request in the scheduler has been
936					involved in a merge. It is used in the deadline
937					scheduler for example, to reposition the request
938					if its sorting order has changed.
940	elevator_allow_merge_fn		called whenever the block layer determines
941					that a bio can be merged into an existing
942					request safely. The io scheduler may still
943					want to stop a merge at this point if it
944					results in some sort of conflict internally,
945					this hook allows it to do that. Note however
946					that two *requests* can still be merged at later
947					time. Currently the io scheduler has no way to
948					prevent that. It can only learn about the fact
949					from elevator_merge_req_fn callback.
951	elevator_dispatch_fn*		fills the dispatch queue with ready requests.
952					I/O schedulers are free to postpone requests by
953					not filling the dispatch queue unless @force
954					is non-zero.  Once dispatched, I/O schedulers
955					are not allowed to manipulate the requests -
956					they belong to generic dispatch queue.
958	elevator_add_req_fn*		called to add a new request into the scheduler
960	elevator_former_req_fn
961	elevator_latter_req_fn		These return the request before or after the
962					one specified in disk sort order. Used by the
963					block layer to find merge possibilities.
965	elevator_completed_req_fn	called when a request is completed.
967	elevator_may_queue_fn		returns true if the scheduler wants to allow the
968					current context to queue a new request even if
969					it is over the queue limit. This must be used
970					very carefully!!
972	elevator_set_req_fn
973	elevator_put_req_fn		Must be used to allocate and free any elevator
974					specific storage for a request.
976	elevator_activate_req_fn	Called when device driver first sees a request.
977					I/O schedulers can use this callback to
978					determine when actual execution of a request
979					starts.
980	elevator_deactivate_req_fn	Called when device driver decides to delay
981					a request by requeueing it.
983	elevator_init_fn*
984	elevator_exit_fn		Allocate and free any elevator specific storage
985					for a queue.
987	4.2 Request flows seen by I/O schedulers
988	All requests seen by I/O schedulers strictly follow one of the following three
989	flows.
991	 set_req_fn ->
993	 i.   add_req_fn -> (merged_fn ->)* -> dispatch_fn -> activate_req_fn ->
994	      (deactivate_req_fn -> activate_req_fn ->)* -> completed_req_fn
995	 ii.  add_req_fn -> (merged_fn ->)* -> merge_req_fn
996	 iii. [none]
998	 -> put_req_fn
1000	4.3 I/O scheduler implementation
1001	The generic i/o scheduler algorithm attempts to sort/merge/batch requests for
1002	optimal disk scan and request servicing performance (based on generic
1003	principles and device capabilities), optimized for:
1004	i.   improved throughput
1005	ii.  improved latency
1006	iii. better utilization of h/w & CPU time
1008	Characteristics:
1010	i. Binary tree
1011	AS and deadline i/o schedulers use red black binary trees for disk position
1012	sorting and searching, and a fifo linked list for time-based searching. This
1013	gives good scalability and good availability of information. Requests are
1014	almost always dispatched in disk sort order, so a cache is kept of the next
1015	request in sort order to prevent binary tree lookups.
1017	This arrangement is not a generic block layer characteristic however, so
1018	elevators may implement queues as they please.
1020	ii. Merge hash
1021	AS and deadline use a hash table indexed by the last sector of a request. This
1022	enables merging code to quickly look up "back merge" candidates, even when
1023	multiple I/O streams are being performed at once on one disk.
1025	"Front merges", a new request being merged at the front of an existing request,
1026	are far less common than "back merges" due to the nature of most I/O patterns.
1027	Front merges are handled by the binary trees in AS and deadline schedulers.
1029	iii. Plugging the queue to batch requests in anticipation of opportunities for
1030	     merge/sort optimizations
1032	Plugging is an approach that the current i/o scheduling algorithm resorts to so
1033	that it collects up enough requests in the queue to be able to take
1034	advantage of the sorting/merging logic in the elevator. If the
1035	queue is empty when a request comes in, then it plugs the request queue
1036	(sort of like plugging the bath tub of a vessel to get fluid to build up)
1037	till it fills up with a few more requests, before starting to service
1038	the requests. This provides an opportunity to merge/sort the requests before
1039	passing them down to the device. There are various conditions when the queue is
1040	unplugged (to open up the flow again), either through a scheduled task or
1041	could be on demand. For example wait_on_buffer sets the unplugging going
1042	through sync_buffer() running blk_run_address_space(mapping). Or the caller
1043	can do it explicity through blk_unplug(bdev). So in the read case,
1044	the queue gets explicitly unplugged as part of waiting for completion on that
1045	buffer. For page driven IO, the address space ->sync_page() takes care of
1046	doing the blk_run_address_space().
1048	Aside:
1049	  This is kind of controversial territory, as it's not clear if plugging is
1050	  always the right thing to do. Devices typically have their own queues,
1051	  and allowing a big queue to build up in software, while letting the device be
1052	  idle for a while may not always make sense. The trick is to handle the fine
1053	  balance between when to plug and when to open up. Also now that we have
1054	  multi-page bios being queued in one shot, we may not need to wait to merge
1055	  a big request from the broken up pieces coming by.
1057	4.4 I/O contexts
1058	I/O contexts provide a dynamically allocated per process data area. They may
1059	be used in I/O schedulers, and in the block layer (could be used for IO statis,
1060	priorities for example). See *io_context in block/ll_rw_blk.c, and as-iosched.c
1061	for an example of usage in an i/o scheduler.
1064	5. Scalability related changes
1066	5.1 Granular Locking: io_request_lock replaced by a per-queue lock
1068	The global io_request_lock has been removed as of 2.5, to avoid
1069	the scalability bottleneck it was causing, and has been replaced by more
1070	granular locking. The request queue structure has a pointer to the
1071	lock to be used for that queue. As a result, locking can now be
1072	per-queue, with a provision for sharing a lock across queues if
1073	necessary (e.g the scsi layer sets the queue lock pointers to the
1074	corresponding adapter lock, which results in a per host locking
1075	granularity). The locking semantics are the same, i.e. locking is
1076	still imposed by the block layer, grabbing the lock before
1077	request_fn execution which it means that lots of older drivers
1078	should still be SMP safe. Drivers are free to drop the queue
1079	lock themselves, if required. Drivers that explicitly used the
1080	io_request_lock for serialization need to be modified accordingly.
1081	Usually it's as easy as adding a global lock:
1083		static DEFINE_SPINLOCK(my_driver_lock);
1085	and passing the address to that lock to blk_init_queue().
1087	5.2 64 bit sector numbers (sector_t prepares for 64 bit support)
1089	The sector number used in the bio structure has been changed to sector_t,
1090	which could be defined as 64 bit in preparation for 64 bit sector support.
1092	6. Other Changes/Implications
1094	6.1 Partition re-mapping handled by the generic block layer
1096	In 2.5 some of the gendisk/partition related code has been reorganized.
1097	Now the generic block layer performs partition-remapping early and thus
1098	provides drivers with a sector number relative to whole device, rather than
1099	having to take partition number into account in order to arrive at the true
1100	sector number. The routine blk_partition_remap() is invoked by
1101	generic_make_request even before invoking the queue specific make_request_fn,
1102	so the i/o scheduler also gets to operate on whole disk sector numbers. This
1103	should typically not require changes to block drivers, it just never gets
1104	to invoke its own partition sector offset calculations since all bios
1105	sent are offset from the beginning of the device.
1108	7. A Few Tips on Migration of older drivers
1110	Old-style drivers that just use CURRENT and ignores clustered requests,
1111	may not need much change.  The generic layer will automatically handle
1112	clustered requests, multi-page bios, etc for the driver.
1114	For a low performance driver or hardware that is PIO driven or just doesn't
1115	support scatter-gather changes should be minimal too.
1117	The following are some points to keep in mind when converting old drivers
1118	to bio.
1120	Drivers should use elv_next_request to pick up requests and are no longer
1121	supposed to handle looping directly over the request list.
1122	(struct request->queue has been removed)
1124	Now end_that_request_first takes an additional number_of_sectors argument.
1125	It used to handle always just the first buffer_head in a request, now
1126	it will loop and handle as many sectors (on a bio-segment granularity)
1127	as specified.
1129	Now bh->b_end_io is replaced by bio->bi_end_io, but most of the time the
1130	right thing to use is bio_endio(bio, uptodate) instead.
1132	If the driver is dropping the io_request_lock from its request_fn strategy,
1133	then it just needs to replace that with q->queue_lock instead.
1135	As described in Sec 1.1, drivers can set max sector size, max segment size
1136	etc per queue now. Drivers that used to define their own merge functions i
1137	to handle things like this can now just use the blk_queue_* functions at
1138	blk_init_queue time.
1140	Drivers no longer have to map a {partition, sector offset} into the
1141	correct absolute location anymore, this is done by the block layer, so
1142	where a driver received a request ala this before:
1144		rq->rq_dev = mk_kdev(3, 5);	/* /dev/hda5 */
1145		rq->sector = 0;			/* first sector on hda5 */
1147	  it will now see
1149		rq->rq_dev = mk_kdev(3, 0);	/* /dev/hda */
1150		rq->sector = 123128;		/* offset from start of disk */
1152	As mentioned, there is no virtual mapping of a bio. For DMA, this is
1153	not a problem as the driver probably never will need a virtual mapping.
1154	Instead it needs a bus mapping (dma_map_page for a single segment or
1155	use dma_map_sg for scatter gather) to be able to ship it to the driver. For
1156	PIO drivers (or drivers that need to revert to PIO transfer once in a
1157	while (IDE for example)), where the CPU is doing the actual data
1158	transfer a virtual mapping is needed. If the driver supports highmem I/O,
1159	(Sec 1.1, (ii) ) it needs to use __bio_kmap_atomic and bio_kmap_irq to
1160	temporarily map a bio into the virtual address space.
1163	8. Prior/Related/Impacted patches
1165	8.1. Earlier kiobuf patches (sct/axboe/chait/hch/mkp)
1166	- orig kiobuf & raw i/o patches (now in 2.4 tree)
1167	- direct kiobuf based i/o to devices (no intermediate bh's)
1168	- page i/o using kiobuf
1169	- kiobuf splitting for lvm (mkp)
1170	- elevator support for kiobuf request merging (axboe)
1171	8.2. Zero-copy networking (Dave Miller)
1172	8.3. SGI XFS - pagebuf patches - use of kiobufs
1173	8.4. Multi-page pioent patch for bio (Christoph Hellwig)
1174	8.5. Direct i/o implementation (Andrea Arcangeli) since 2.4.10-pre11
1175	8.6. Async i/o implementation patch (Ben LaHaise)
1176	8.7. EVMS layering design (IBM EVMS team)
1177	8.8. Larger page cache size patch (Ben LaHaise) and
1178	     Large page size (Daniel Phillips)
1179	    => larger contiguous physical memory buffers
1180	8.9. VM reservations patch (Ben LaHaise)
1181	8.10. Write clustering patches ? (Marcelo/Quintela/Riel ?)
1182	8.11. Block device in page cache patch (Andrea Archangeli) - now in 2.4.10+
1183	8.12. Multiple block-size transfers for faster raw i/o (Shailabh Nagar,
1184	      Badari)
1185	8.13  Priority based i/o scheduler - prepatches (Arjan van de Ven)
1186	8.14  IDE Taskfile i/o patch (Andre Hedrick)
1187	8.15  Multi-page writeout and readahead patches (Andrew Morton)
1188	8.16  Direct i/o patches for 2.5 using kvec and bio (Badari Pulavarthy)
1190	9. Other References:
1192	9.1 The Splice I/O Model - Larry McVoy (and subsequent discussions on lkml,
1193	and Linus' comments - Jan 2001)
1194	9.2 Discussions about kiobuf and bh design on lkml between sct, linus, alan
1195	et al - Feb-March 2001 (many of the initial thoughts that led to bio were
1196	brought up in this discussion thread)
1197	9.3 Discussions on mempool on lkml - Dec 2001.
Hide Line Numbers
About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Information is copyright its respective author. All material is available from the Linux Kernel Source distributed under a GPL License. This page is provided as a free service by mjmwired.net.