blob: 2101e718670d0248110caa4320e51e83c715fad2 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001 Notes on the Generic Block Layer Rewrite in Linux 2.5
2 =====================================================
3
4Notes Written on Jan 15, 2002:
Rob Landley26bbb292007-10-15 11:42:52 +02005 Jens Axboe <jens.axboe@oracle.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -07006 Suparna Bhattacharya <suparna@in.ibm.com>
7
8Last Updated May 2, 2002
9September 2003: Updated I/O Scheduler portions
Nick Piggin6e575592010-08-05 21:08:09 +100010 Nick Piggin <npiggin@kernel.dk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011
12Introduction:
13
14These are some notes describing some aspects of the 2.5 block layer in the
15context of the bio rewrite. The idea is to bring out some of the key
16changes and a glimpse of the rationale behind those changes.
17
18Please mail corrections & suggestions to suparna@in.ibm.com.
19
20Credits:
21---------
22
232.5 bio rewrite:
Rob Landley26bbb292007-10-15 11:42:52 +020024 Jens Axboe <jens.axboe@oracle.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
26Many aspects of the generic block layer redesign were driven by and evolved
27over discussions, prior patches and the collective experience of several
28people. See sections 8 and 9 for a list of some related references.
29
30The following people helped with review comments and inputs for this
31document:
32 Christoph Hellwig <hch@infradead.org>
33 Arjan van de Ven <arjanv@redhat.com>
Adrian Bunkf4b09eb2006-01-03 13:37:51 +010034 Randy Dunlap <rdunlap@xenotime.net>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035 Andre Hedrick <andre@linux-ide.org>
36
37The following people helped with fixes/contributions to the bio patches
38while it was still work-in-progress:
39 David S. Miller <davem@redhat.com>
40
41
42Description of Contents:
43------------------------
44
451. 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
562. 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
613. 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
714. The I/O scheduler
725. Scalability related changes
73 5.1 Granular locking: Removal of io_request_lock
74 5.2 Prepare for transition to 64 bit sector_t
756. Other Changes/Implications
76 6.1 Partition re-mapping handled by the generic block layer
777. A few tips on migration of older drivers
788. A list of prior/related/impacted patches/ideas
799. Other References/Discussion Threads
80
81---------------------------------------------------------------------------
82
83Bio Notes
84--------
85
86Let us discuss the changes in the context of how some overall goals for the
87block layer are addressed.
88
891. Scope for tuning the generic logic to satisfy various requirements
90
91The block layer design supports adaptable abstractions to handle common
92processing with the ability to tune the logic to an appropriate extent
93depending on the nature of the device and the requirements of the caller.
94One of the objectives of the rewrite was to increase the degree of tunability
95and to enable higher level code to utilize underlying device/driver
96capabilities to the maximum extent for better i/o performance. This is
97important especially in the light of ever improving hardware capabilities
98and application/middleware software designed to take advantage of these
99capabilities.
100
1011.1 Tuning based on low level device / driver capabilities
102
103Sophisticated devices with large built-in caches, intelligent i/o scheduling
104optimizations, high memory DMA support, etc may find some of the
105generic processing an overhead, while for less capable devices the
106generic functionality is essential for performance or correctness reasons.
107Knowledge of some of the capabilities or parameters of the device should be
108used at the generic block layer to take the right decisions on
109behalf of the driver.
110
111How is this achieved ?
112
113Tuning at a per-queue level:
114
115i. Per-queue limits/values exported to the generic layer by the driver
116
117Various parameters that the generic i/o scheduler logic uses are set at
118a per-queue level (e.g maximum request size, maximum number of segments in
119a scatter-gather list, hardsect size)
120
121Some parameters that were earlier available as global arrays indexed by
122major/minor are now directly associated with the queue. Some of these may
123move into the block device structure in the future. Some characteristics
124have been incorporated into a queue flags field rather than separate fields
125in themselves. There are blk_queue_xxx functions to set the parameters,
126rather than update the fields directly
127
128Some new queue property settings:
129
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.
133
134 blk_queue_max_sectors(q, max_sectors)
Mike Christie28832e82006-03-08 11:19:51 +0100135 Sets two variables that limit the size of the request.
136
137 - The request queue's max_sectors, which is a soft size in
Paolo Ornati670e9f32006-10-03 22:57:56 +0200138 units of 512 byte sectors, and could be dynamically varied
Mike Christie28832e82006-03-08 11:19:51 +0100139 by the core kernel.
140
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.
144
145 The default for both max_sectors and max_hw_sectors is
146 255. The upper limit of max_sectors is 1024.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147
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)
151
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)
156
157 blk_queue_max_segment_size(q, max_seg_size)
158 Maximum size of a clustered segment, 64kB default.
159
160 blk_queue_hardsect_size(q, hardsect_size)
161 Lowest possible sector size that the hardware can operate
162 on, 512 bytes default.
163
164New queue flags:
165
166 QUEUE_FLAG_CLUSTER (see 3.2.2)
167 QUEUE_FLAG_QUEUED (see 3.2.4)
168
169
170ii. High-mem i/o capabilities are now considered the default
171
172The generic bounce buffer logic, present in 2.4, where the block layer would
173by default copyin/out i/o requests on high-memory buffers to low-memory buffers
174assuming that the driver wouldn't be able to handle it directly, has been
175changed in 2.5. The bounce logic is now applied only for memory ranges
176for which the device cannot handle i/o. A driver can specify this by
177setting the queue bounce limit for the request queue for the device
178(blk_queue_bounce_limit()). This avoids the inefficiencies of the copyin/out
179where a device is capable of handling high memory i/o.
180
181In order to enable high-memory i/o where the device is capable of supporting
182it, the pci dma mapping routines and associated data structures have now been
183modified to accomplish a direct page -> bus translation, without requiring
184a virtual address mapping (unlike the earlier scheme of virtual address
185-> bus translation). So this works uniformly for high-memory pages (which
Matt LaPlante5d3f0832006-11-30 05:21:10 +0100186do not have a corresponding kernel virtual address space mapping) and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187low-memory pages.
188
Paul Bolle395cf962011-08-15 02:02:26 +0200189Note: Please refer to Documentation/DMA-API-HOWTO.txt for a discussion
Randy Dunlap5872fb92009-01-29 16:28:02 -0800190on PCI high mem DMA aspects and mapping of scatter gather lists, and support
191for 64 bit PCI.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192
193Special handling is required only for cases where i/o needs to happen on
194pages at physical memory addresses beyond what the device can support. In these
195cases, a bounce bio representing a buffer from the supported memory range
196is used for performing the i/o with copyin/copyout as needed depending on
197the type of the operation. For example, in case of a read operation, the
198data read has to be copied to the original buffer on i/o completion, so a
199callback routine is set up to do this, while for write, the data is copied
200from the original buffer to the bounce buffer prior to issuing the
201operation. Since an original buffer may be in a high memory area that's not
202mapped in kernel virtual addr, a kmap operation may be required for
203performing the copy, and special care may be needed in the completion path
204as it may not be in irq context. Special care is also required (by way of
205GFP flags) when allocating bounce buffers, to avoid certain highmem
206deadlock possibilities.
207
208It is also possible that a bounce buffer may be allocated from high-memory
209area that's not mapped in kernel virtual addr, but within the range that the
210device can use directly; so the bounce page may need to be kmapped during
211copy operations. [Note: This does not hold in the current implementation,
212though]
213
214There are some situations when pages from high memory may need to
215be kmapped, even if bounce buffers are not necessary. For example a device
216may need to abort DMA operations and revert to PIO for the transfer, in
217which case a virtual mapping of the page is required. For SCSI it is also
218done in some scenarios where the low level driver cannot be trusted to
219handle a single sg entry correctly. The driver is expected to perform the
220kmaps as needed on such occasions using the __bio_kmap_atomic and bio_kmap_irq
221routines as appropriate. A driver could also use the blk_queue_bounce()
222routine on its own to bounce highmem i/o to low memory for specific requests
223if so desired.
224
225iii. The i/o scheduler algorithm itself can be replaced/set as appropriate
226
227As in 2.4, it is possible to plugin a brand new i/o scheduler for a particular
228queue or pick from (copy) existing generic schedulers and replace/override
229certain portions of it. The 2.5 rewrite provides improved modularization
230of the i/o scheduler. There are more pluggable callbacks, e.g for init,
231add request, extract request, which makes it possible to abstract specific
232i/o scheduling algorithm aspects and details outside of the generic loop.
233It also makes it possible to completely hide the implementation details of
234the i/o scheduler from block drivers.
235
236I/O scheduler wrappers are to be used instead of accessing the queue directly.
237See section 4. The I/O scheduler for details.
238
2391.2 Tuning Based on High level code capabilities
240
241i. Application capabilities for raw i/o
242
243This comes from some of the high-performance database/middleware
244requirements where an application prefers to make its own i/o scheduling
245decisions based on an understanding of the access patterns and i/o
246characteristics
247
248ii. High performance filesystems or other higher level kernel code's
249capabilities
250
251Kernel components like filesystems could also take their own i/o scheduling
252decisions for optimizing performance. Journalling filesystems may need
253some control over i/o ordering.
254
255What kind of support exists at the generic block layer for this ?
256
257The flags and rw fields in the bio structure can be used for some tuning
258from above e.g indicating that an i/o is just a readahead request, or for
259marking barrier requests (discussed next), or priority settings (currently
260unused). As far as user applications are concerned they would need an
261additional mechanism either via open flags or ioctls, or some other upper
262level mechanism to communicate such settings to block.
263
2641.2.1 I/O Barriers
265
266There is a way to enforce strict ordering for i/os through barriers.
267All requests before a barrier point must be serviced before the barrier
268request and any other requests arriving after the barrier will not be
269serviced until after the barrier has completed. This is useful for higher
270level control on write ordering, e.g flushing a log of committed updates
271to disk before the corresponding updates themselves.
272
273A flag in the bio structure, BIO_BARRIER is used to identify a barrier i/o.
274The generic i/o scheduler would make sure that it places the barrier request and
275all other requests coming after it after all the previous requests in the
276queue. Barriers may be implemented in different ways depending on the
Tejun Heoff5b8cf2006-01-06 09:58:37 +0100277driver. For more details regarding I/O barriers, please read barrier.txt
278in this directory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279
2801.2.2 Request Priority/Latency
281
282Todo/Under discussion:
283Arjan'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.
292
293
2941.3 Direct Access to Low level Device/Driver Capabilities (Bypass mode)
295 (e.g Diagnostics, Systems Management)
296
297There are situations where high-level code needs to have direct access to
298the low level device capabilities or requires the ability to issue commands
299to the device bypassing some of the intermediate i/o layers.
300These could, for example, be special control commands issued through ioctl
301interfaces, or could be raw read/write commands that stress the drive's
302capabilities for certain kinds of fitness tests. Having direct interfaces at
303multiple levels without having to pass through upper layers makes
304it possible to perform bottom up validation of the i/o path, layer by
305layer, starting from the media.
306
307The normal i/o submission interfaces, e.g submit_bio, could be bypassed
308for specially crafted requests which such ioctl or diagnostics
309interfaces would typically use, and the elevator add_request routine
310can instead be used to directly insert such requests in the queue or preferably
311the blk_do_rq routine can be used to place the request on the queue and
312wait for completion. Alternatively, sometimes the caller might just
313invoke a lower level driver specific interface with the request as a
314parameter.
315
316If the request is a means for passing on special information associated with
317the command, then such information is associated with the request->special
318field (rather than misuse the request->buffer field which is meant for the
319request data buffer's virtual mapping).
320
321For passing request data, the caller must build up a bio descriptor
322representing the concerned memory buffer if the underlying driver interprets
323bio segments or uses the block layer end*request* functions for i/o
324completion. Alternatively one could directly use the request->buffer field to
325specify the virtual address of the buffer, if the driver expects buffer
326addresses passed in this way and ignores bio entries for the request type
327involved. In the latter case, the driver would modify and manage the
328request->buffer, request->sector and request->nr_sectors or
329request->current_nr_sectors fields itself rather than using the block layer
330end_request or end_that_request_first completion interfaces.
331(See 2.3 or Documentation/block/request.txt for a brief explanation of
332the request structure fields)
333
334[TBD: end_that_request_last should be usable even in this case;
335Perhaps an end_that_direct_request_first routine could be implemented to make
336handling direct requests easier for such drivers; Also for drivers that
337expect bios, a helper function could be provided for setting up a bio
338corresponding to a data buffer]
339
340<JENS: I dont understand the above, why is end_that_request_first() not
341usable? 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>
351
3521.3.1 Pre-built Commands
353
354A request can be created with a pre-built custom command to be sent directly
355to the device. The cmd block in the request structure has room for filling
356in the command bytes. (i.e rq->cmd is now 16 bytes in size, and meant for
357command pre-building, and the type of the request is now indicated
358through rq->flags instead of via rq->cmd)
359
360The request structure flags can be set up to indicate the type of request
361in such cases (REQ_PC: direct packet command passed to driver, REQ_BLOCK_PC:
362packet command issued via blk_do_rq, REQ_SPECIAL: special request).
363
364It can help to pre-build device commands for requests in advance.
365Drivers can now specify a request prepare function (q->prep_rq_fn) that the
366block layer would invoke to pre-build device commands for a given request,
367or perform other preparatory processing for the request. This is routine is
368called 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
370enabled)
371
372Aside:
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.
381
382
3832. Flexible and generic but minimalist i/o structure/descriptor.
384
3852.1 Reason for a new structure and requirements addressed
386
387Prior to 2.5, buffer heads were used as the unit of i/o at the generic block
388layer, and the low level request structure was associated with a chain of
389buffer heads for a contiguous i/o request. This led to certain inefficiencies
390when it came to large i/o requests and readv/writev style operations, as it
391forced such requests to be broken up into small chunks before being passed
392on to the generic block layer, only to be merged by the i/o scheduler
393when the underlying device was capable of handling the i/o in one shot.
394Also, using the buffer head as an i/o structure for i/os that didn't originate
Matt LaPlante4ae0edc2006-11-30 04:58:40 +0100395from the buffer cache unnecessarily added to the weight of the descriptors
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396which were generated for each such chunk.
397
398The following were some of the goals and expectations considered in the
399redesign of the block i/o data structure in 2.5.
400
401i. 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.
405ii. Ability to represent high-memory buffers (which do not have a virtual
406 address mapping in kernel address space).
Matt LaPlante4ae0edc2006-11-30 04:58:40 +0100407iii.Ability to represent large i/os w/o unnecessarily breaking them up (i.e
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 greater than PAGE_SIZE chunks in one shot)
409iv. 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)
412v. Ability to represent an i/o involving multiple physical memory segments
413 (including non-page aligned page fragments, as specified via readv/writev)
Matt LaPlante4ae0edc2006-11-30 04:58:40 +0100414 without unnecessarily breaking it up, if the underlying device is capable of
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 handling it.
416vi. 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
420vii.Ability to handle the possibility of splits/merges as the structure passes
421 through layered drivers (lvm, md, evms), with minimal overhead.
422
423The solution was to define a new structure (bio) for the block layer,
424instead of using the buffer head structure (bh) directly, the idea being
425avoidance of some associated baggage and limitations. The bio structure
426is uniformly used for all i/o at the block layer ; it forms a part of the
427bh structure for buffered i/o, and in the case of raw/direct i/o kiobufs are
428mapped to bio structures.
429
4302.2 The bio struct
431
432The bio structure uses a vector representation pointing to an array of tuples
433of <page, offset, len> to describe the i/o buffer, and has various other
434fields describing i/o parameters and state that needs to be maintained for
435performing the i/o.
436
437Notice that this representation means that a bio has no virtual address
438mapping at all (unlike buffer heads).
439
440struct bio_vec {
441 struct page *bv_page;
442 unsigned short bv_len;
443 unsigned short bv_offset;
444};
445
446/*
447 * main unit of I/O for the block layer and lower layers (ie drivers)
448 */
449struct bio {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 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 */
454
455 unsigned int bi_vcnt; /* how may bio_vec's */
Kent Overstreet4f024f32013-10-11 15:44:27 -0700456 struct bvec_iter bi_iter; /* current index into bio_vec array */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457
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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467};
468
469With this multipage bio design:
470
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.
NeilBrown5705f702007-09-25 12:35:59 +0200479- 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.
Kent Overstreet4f024f32013-10-11 15:44:27 -0700482- Drivers which can't process a large bio in one shot can use the bi_iter
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 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]
487
488(*) unrelated merges -- a request ends up containing two or more bios that
489 didn't originate from the same place.
490
491bi_end_io() i/o callback gets called on i/o completion of the entire bio.
492
493At a lower level, drivers build a scatter gather list from the merged bios.
494The scatter gather list is in the form of an array of <page, offset, len>
495entries with their corresponding dma address mappings filled in at the
496appropriate time. As an optimization, contiguous physical pages can be
497covered by a single entry where <page> refers to the first page and <len>
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300498covers the range of pages (up to 16 contiguous pages could be covered this
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499way). There is a helper routine (blk_rq_map_sg) which drivers can use to build
500the sg list.
501
502Note: Right now the only user of bios with more than one page is ll_rw_kio,
503which in turn means that only raw I/O uses it (direct i/o may not work
504right now). The intent however is to enable clustering of pages etc to
505become possible. The pagebuf abstraction layer from SGI also uses multi-page
506bios, but that is currently not included in the stock development kernels.
507The same is true of Andrew Morton's work-in-progress multipage bio writeout
508and readahead patches.
509
5102.3 Changes in the Request Structure
511
512The request structure is the structure that gets passed down to low level
513drivers. The block layer make_request function builds up a request structure,
514places it on the queue and invokes the drivers request_fn. The driver makes
515use of block layer helper routine elv_next_request to pull the next request
516off the queue. Control or diagnostic functions might bypass block and directly
517invoke underlying driver entry points passing in a specially constructed
518request structure.
519
520Only some relevant fields (mainly those which changed or may be referred
521to in some of the discussion here) are listed below, not necessarily in
522the order in which they occur in the structure (see include/linux/blkdev.h)
523Refer to Documentation/block/request.txt for details about all the request
524structure fields and a quick reference about the layers which are
525supposed to use or modify those fields.
526
527struct 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 .
543
544 /* Number of scatter-gather DMA addr+len pairs after
545 * physical address coalescing is performed.
546 */
547 unsigned short nr_phys_segments;
548
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;
555
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 */
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300566 char *buffer; /* valid only for low memory buffers up to
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 current_nr_sectors */
568 .
569 .
570 struct bio *bio, *biotail; /* bio list instead of bh */
571 struct request_list *rl;
572}
573
574See the rq_flag_bits definitions for an explanation of the various flags
575available. Some bits are used by the block layer or i/o scheduler.
576
577The behaviour of the various sector counts are almost the same as before,
578except that since we have multi-segment bios, current_nr_sectors refers
579to the numbers of sectors in the current segment being processed which could
580be one of the many segments in the current bio (i.e i/o completion unit).
581The nr_sectors value refers to the total number of sectors in the whole
582request that remain to be transferred (no change). The purpose of the
583hard_xxx values is for block to remember these counts every time it hands
584over the request to the driver. These values are updated by block on
585end_that_request_first, i.e. every time the driver completes a part of the
586transfer and invokes block end*request helpers to mark this. The
587driver should not modify these values. The block layer sets up the
588nr_sectors and current_nr_sectors fields (based on the corresponding
589hard_xxx values and the number of bytes transferred) and updates it on
590every transfer that invokes end_that_request_first. It does the same for the
Kent Overstreet4f024f32013-10-11 15:44:27 -0700591buffer, bio, bio->bi_iter fields too.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
593The buffer field is just a virtual address mapping of the current segment
594of the i/o buffer in cases where the buffer resides in low-memory. For high
595memory i/o, this field is not valid and must not be used by drivers.
596
597Code that sets up its own request structures and passes them down to
598a driver needs to be careful about interoperation with the block layer helper
599functions which the driver uses. (Section 1.3)
600
6013. Using bios
602
6033.1 Setup/Teardown
604
605There are routines for managing the allocation, and reference counting, and
606freeing of bios (bio_alloc, bio_get, bio_put).
607
608This makes use of Ingo Molnar's mempool implementation, which enables
609subsystems like bio to maintain their own reserve memory pools for guaranteed
610deadlock-free allocations during extreme VM load. For example, the VM
611subsystem makes use of the block layer to writeout dirty pages in order to be
612able to free up memory space, a case which needs careful handling. The
613allocation logic draws from the preallocated emergency reserve in situations
614where it cannot allocate through normal means. If the pool is empty and it
615can wait, then it would trigger action that would help free up memory or
616replenish the pool (without deadlocking) and wait for availability in the pool.
617If it is in IRQ context, and hence not in a position to do this, allocation
618could fail if the pool is empty. In general mempool always first tries to
619perform allocation without having to wait, even if it means digging into the
620pool as long it is not less that 50% full.
621
622On a free, memory is released to the pool or directly freed depending on
623the current availability in the pool. The mempool interface lets the
624subsystem specify the routines to be used for normal alloc and free. In the
625case of bio, these routines make use of the standard slab allocator.
626
627The caller of bio_alloc is expected to taken certain steps to avoid
628deadlocks, e.g. avoid trying to allocate more memory from the pool while
629already 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 ]
634
635Memory allocated from the pool should be released back within a limited
636amount of time (in the case of bio, that would be after the i/o is completed).
637This ensures that if part of the pool has been used up, some work (in this
638case i/o) must already be in progress and memory would be available when it
639is over. If allocating from multiple pools in the same code path, the order
640or hierarchy of allocation needs to be consistent, just the way one deals
641with multiple locks.
642
643The bio_alloc routine also needs to allocate the bio_vec_list (bvec_alloc())
644for a non-clone bio. There are the 6 pools setup for different size biovecs,
645so bio_alloc(gfp_mask, nr_iovecs) will allocate a vec_list of the
646given size from these slabs.
647
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648The bio_get() routine may be used to hold an extra reference on a bio prior
649to i/o submission, if the bio fields are likely to be accessed after the
650i/o is issued (since the bio may otherwise get freed in case i/o completion
651happens in the meantime).
652
653The bio_clone() routine may be used to duplicate a bio, where the clone
654shares the bio_vec_list with the original bio (i.e. both point to the
655same bio_vec_list). This would typically be used for splitting i/o requests
656in lvm or md.
657
6583.2 Generic bio helper Routines
659
6603.2.1 Traversing segments and completion units in a request
661
NeilBrown5705f702007-09-25 12:35:59 +0200662The macro rq_for_each_segment() should be used for traversing the bios
663in the request list (drivers should avoid directly trying to do it
664themselves). Using these helpers should also make it easier to cope
665with block changes in the future.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666
NeilBrown5705f702007-09-25 12:35:59 +0200667 struct req_iterator iter;
668 rq_for_each_segment(bio_vec, rq, iter)
669 /* bio_vec is now current segment */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
671I/O completion callbacks are per-bio rather than per-segment, so drivers
672that traverse bio chains on completion need to keep that in mind. Drivers
673which don't make a distinction between segments and completion units would
674need to be reorganized to support multi-segment bios.
675
6763.2.2 Setting up DMA scatterlists
677
678The blk_rq_map_sg() helper routine would be used for setting up scatter
679gather lists from a request, so a driver need not do it on its own.
680
681 nr_segments = blk_rq_map_sg(q, rq, scatterlist);
682
683The helper routine provides a level of abstraction which makes it easier
684to modify the internals of request to scatterlist conversion down the line
685without breaking drivers. The blk_rq_map_sg routine takes care of several
686things like collapsing physically contiguous segments (if QUEUE_FLAG_CLUSTER
687is set) and correct segment accounting to avoid exceeding the limits which
688the i/o hardware can handle, based on various queue properties.
689
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).
695
696Routines which the low level driver can use to set up the segment limits:
697
698blk_queue_max_hw_segments() : Sets an upper limit of the maximum number of
699hw data segments in a request (i.e. the maximum number of address/length
700pairs the host adapter can actually hand to the device at once)
701
702blk_queue_max_phys_segments() : Sets an upper limit on the maximum number
703of physical data segments in a request (i.e. the largest sized scatter list
704a driver could handle)
705
7063.2.3 I/O completion
707
708The existing generic block layer helper routines end_request,
709end_that_request_first and end_that_request_last can be used for i/o
710completion (and setting things up so the rest of the i/o or the next
711request can be kicked of) as before. With the introduction of multi-page
712bio support, end_that_request_first requires an additional argument indicating
713the number of sectors completed.
714
7153.2.4 Implications for drivers that do not interpret bios (don't handle
716 multiple segments)
717
718Drivers that do not interpret bios e.g those which do not handle multiple
719segments and do not support i/o into high memory addresses (require bounce
720buffers) and expect only virtually mapped buffers, can access the rq->buffer
721field. As before the driver should use current_nr_sectors to determine the
722size of remaining data in the current segment (that is the maximum it can
723transfer in one go unless it interprets segments), and rely on the block layer
724end_request, or end_that_request_first/last to take care of all accounting
725and transparent mapping of the next bio segment when a segment boundary
726is crossed on completion of a transfer. (The end*request* functions should
727be used if only if the request has come down from block/bio path, not for
728direct access requests which only specify rq->buffer without a valid rq->bio)
729
7303.2.5 Generic request command tagging
731
7323.2.5.1 Tag helpers
733
734Block now offers some simple generic functionality to help support command
735queueing (typically known as tagged command queueing), ie manage more than
736one outstanding command on a queue at any given time.
737
Jens Axboe165125e2007-07-24 09:28:11 +0200738 blk_queue_init_tags(struct request_queue *q, int depth)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
740 Initialize internal command tagging structures for a maximum
741 depth of 'depth'.
742
Jens Axboe165125e2007-07-24 09:28:11 +0200743 blk_queue_free_tags((struct request_queue *q)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744
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.
748
749The above are initialization and exit management, the main helpers during
750normal operations are:
751
Jens Axboe165125e2007-07-24 09:28:11 +0200752 blk_queue_start_tag(struct request_queue *q, struct request *rq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
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.
759
Jens Axboe165125e2007-07-24 09:28:11 +0200760 blk_queue_end_tag(struct request_queue *q, struct request *rq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
762 End tagged operation on this request. 'rq' is removed from the internal
763 book keeping structures.
764
765To minimize struct request and queue overhead, the tag helpers utilize some
766of the same request members that are used for normal request queue management.
767This means that a request cannot both be an active tag and be on the queue
768list at the same time. blk_queue_start_tag() will remove the request, but
769the driver must remember to call blk_queue_end_tag() before signalling
770completion of the request to the block layer. This means ending tag
771operations before calling end_that_request_last()! For an example of a user
772of these helpers, see the IDE tagged command queueing support.
773
774Certain hardware conditions may dictate a need to invalidate the block tag
775queue. For instance, on IDE any tagged request error needs to clear both
776the hardware and software block queue and enable the driver to sanely restart
777all the outstanding requests. There's a third helper to do that:
778
Jens Axboe165125e2007-07-24 09:28:11 +0200779 blk_queue_invalidate_tags(struct request_queue *q)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780
Matt LaPlanted6bc8ac2006-10-03 22:54:15 +0200781 Clear the internal block tag queue and re-add all the pending requests
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 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.
785
7863.2.5.2 Tag info
787
788Some block functions exist to query current tag status or to go from a
789tag number to the associated request. These are, in no particular order:
790
791 blk_queue_tagged(q)
792
793 Returns 1 if the queue 'q' is using tagging, 0 if not.
794
795 blk_queue_tag_request(q, tag)
796
797 Returns a pointer to the request associated with tag 'tag'.
798
799 blk_queue_tag_depth(q)
800
801 Return current queue depth.
802
803 blk_queue_tag_queue(q)
804
805 Returns 1 if the queue can accept a new queued command, 0 if we are
806 at the maximum depth already.
807
808 blk_queue_rq_tagged(rq)
809
810 Returns 1 if the request 'rq' is tagged.
811
8123.2.5.2 Internal structure
813
814Internally, block manages tags in the blk_queue_tag structure:
815
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 };
823
824Most of the above is simple and straight forward, however busy_list may need
825a bit of explaining. Normally we don't care too much about request ordering,
826but in the event of any barrier requests in the tag queue we need to ensure
827that requests are restarted in the order they were queue. This may happen
828if the driver needs to use blk_queue_invalidate_tags().
829
830Tagging also defines a new request flag, REQ_QUEUED. This is set whenever
831a request is currently tagged. You should not use this flag directly,
832blk_rq_tagged(rq) is the portable way to do so.
833
8343.3 I/O Submission
835
836The routine submit_bio() is used to submit a single io. Higher level i/o
837routines make use of this:
838
839(a) Buffered i/o:
840The routine submit_bh() invokes submit_bio() on a bio corresponding to the
841bh, allocating the bio if required. ll_rw_block() uses submit_bh() as before.
842
843(b) Kiobuf i/o (for raw/direct i/o):
844The ll_rw_kio() routine breaks up the kiobuf into page sized chunks and
845maps the array to one or more multi-page bios, issuing submit_bio() to
846perform the i/o on each of these.
847
848The embedded bh array in the kiobuf structure has been removed and no
849preallocation of bios is done for kiobufs. [The intent is to remove the
850blocks array as well, but it's currently in there to kludge around direct i/o.]
851Thus kiobuf allocation has switched back to using kmalloc rather than vmalloc.
852
853Todo/Observation:
854
855 A single kiobuf structure is assumed to correspond to a contiguous range
856 of data, so brw_kiovec() invokes ll_rw_kio for each kiobuf in a kiovec.
857 So right now it wouldn't work for direct i/o on non-contiguous blocks.
858 This is to be resolved. The eventual direction is to replace kiobuf
859 by kvec's.
860
861 Badari Pulavarty has a patch to implement direct i/o correctly using
862 bio and kvec.
863
864
865(c) Page i/o:
866Todo/Under discussion:
867
868 Andrew Morton's multi-page bio patches attempt to issue multi-page
869 writeouts (and reads) from the page cache, by directly building up
870 large bios for submission completely bypassing the usage of buffer
871 heads. This work is still in progress.
872
873 Christoph Hellwig had some code that uses bios for page-io (rather than
874 bh). This isn't included in bio as yet. Christoph was also working on a
875 design for representing virtual/real extents as an entity and modifying
876 some of the address space ops interfaces to utilize this abstraction rather
877 than buffer_heads. (This is somewhat along the lines of the SGI XFS pagebuf
878 abstraction, but intended to be as lightweight as possible).
879
880(d) Direct access i/o:
881Direct access requests that do not contain bios would be submitted differently
882as discussed earlier in section 1.3.
883
884Aside:
885
886 Kvec i/o:
887
Matt LaPlante53cb4722006-10-03 22:55:17 +0200888 Ben LaHaise's aio code uses a slightly different structure instead
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 of kiobufs, called a kvec_cb. This contains an array of <page, offset, len>
890 tuples (very much like the networking code), together with a callback function
891 and data pointer. This is embedded into a brw_cb structure when passed
892 to brw_kvec_async().
893
894 Now it should be possible to directly map these kvecs to a bio. Just as while
895 cloning, in this case rather than PRE_BUILT bio_vecs, we set the bi_io_vec
896 array pointer to point to the veclet array in kvecs.
897
898 TBD: In order for this to work, some changes are needed in the way multi-page
899 bios are handled today. The values of the tuples in such a vector passed in
900 from higher level code should not be modified by the block layer in the course
901 of its request processing, since that would make it hard for the higher layer
902 to continue to use the vector descriptor (kvec) after i/o completes. Instead,
903 all such transient state should either be maintained in the request structure,
904 and passed on in some way to the endio completion routine.
905
906
9074. The I/O scheduler
Tejun Heo4c9f7832005-10-20 16:47:40 +0200908I/O scheduler, a.k.a. elevator, is implemented in two layers. Generic dispatch
909queue and specific I/O schedulers. Unless stated otherwise, elevator is used
910to refer to both parts and I/O scheduler to specific I/O schedulers.
911
Nikanth Karthikesan42364692008-11-24 10:46:29 +0100912Block layer implements generic dispatch queue in block/*.c.
Tejun Heo4c9f7832005-10-20 16:47:40 +0200913The generic dispatch queue is responsible for properly ordering barrier
914requests, requeueing, handling non-fs requests and all other subtleties.
915
916Specific I/O schedulers are responsible for ordering normal filesystem
917requests. They can also choose to delay certain requests to improve
918throughput or whatever purpose. As the plural form indicates, there are
919multiple I/O schedulers. They can be built as modules but at least one should
920be built inside the kernel. Each queue can choose different one and can also
921change to another one dynamically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922
923A block layer call to the i/o scheduler follows the convention elv_xxx(). This
Nikanth Karthikesan42364692008-11-24 10:46:29 +0100924calls elevator_xxx_fn in the elevator switch (block/elevator.c). Oh, xxx
925and xxx might not match exactly, but use your imagination. If an elevator
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926doesn't implement a function, the switch does nothing or some minimal house
927keeping work.
928
9294.1. I/O scheduler API
930
931The functions an elevator may implement are: (* are mandatory)
932elevator_merge_fn called to query requests for merge with a bio
933
Tejun Heo4c9f7832005-10-20 16:47:40 +0200934elevator_merge_req_fn called when two requests get merged. the one
935 which gets merged into the other one will be
936 never seen by I/O scheduler again. IOW, after
937 being merged, the request is gone.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938
939elevator_merged_fn called when a request in the scheduler has been
940 involved in a merge. It is used in the deadline
941 scheduler for example, to reposition the request
942 if its sorting order has changed.
943
Jens Axboe126ec9a62006-12-20 11:06:15 +0100944elevator_allow_merge_fn called whenever the block layer determines
945 that a bio can be merged into an existing
946 request safely. The io scheduler may still
947 want to stop a merge at this point if it
948 results in some sort of conflict internally,
949 this hook allows it to do that.
950
Nikanth Karthikesan75989092009-01-27 09:29:24 +0100951elevator_dispatch_fn* fills the dispatch queue with ready requests.
Tejun Heo4c9f7832005-10-20 16:47:40 +0200952 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.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957
Nikanth Karthikesan75989092009-01-27 09:29:24 +0100958elevator_add_req_fn* called to add a new request into the scheduler
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960elevator_former_req_fn
961elevator_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.
964
Tejun Heo4c9f7832005-10-20 16:47:40 +0200965elevator_completed_req_fn called when a request is completed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
967elevator_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!!
971
972elevator_set_req_fn
973elevator_put_req_fn Must be used to allocate and free any elevator
Tejun Heo4c9f7832005-10-20 16:47:40 +0200974 specific storage for a request.
975
976elevator_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.
980elevator_deactivate_req_fn Called when device driver decides to delay
981 a request by requeueing it.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
Nikanth Karthikesan75989092009-01-27 09:29:24 +0100983elevator_init_fn*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984elevator_exit_fn Allocate and free any elevator specific storage
985 for a queue.
986
Tejun Heo4c9f7832005-10-20 16:47:40 +02009874.2 Request flows seen by I/O schedulers
Matt LaPlante53cb4722006-10-03 22:55:17 +0200988All requests seen by I/O schedulers strictly follow one of the following three
Tejun Heo4c9f7832005-10-20 16:47:40 +0200989flows.
990
991 set_req_fn ->
992
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]
997
998 -> put_req_fn
999
10004.3 I/O scheduler implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001The generic i/o scheduler algorithm attempts to sort/merge/batch requests for
1002optimal disk scan and request servicing performance (based on generic
1003principles and device capabilities), optimized for:
1004i. improved throughput
1005ii. improved latency
1006iii. better utilization of h/w & CPU time
1007
1008Characteristics:
1009
1010i. Binary tree
1011AS and deadline i/o schedulers use red black binary trees for disk position
1012sorting and searching, and a fifo linked list for time-based searching. This
Matt LaPlante5d3f0832006-11-30 05:21:10 +01001013gives good scalability and good availability of information. Requests are
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014almost always dispatched in disk sort order, so a cache is kept of the next
1015request in sort order to prevent binary tree lookups.
1016
1017This arrangement is not a generic block layer characteristic however, so
1018elevators may implement queues as they please.
1019
Tejun Heo4c9f7832005-10-20 16:47:40 +02001020ii. Merge hash
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021AS and deadline use a hash table indexed by the last sector of a request. This
1022enables merging code to quickly look up "back merge" candidates, even when
1023multiple I/O streams are being performed at once on one disk.
1024
1025"Front merges", a new request being merged at the front of an existing request,
1026are far less common than "back merges" due to the nature of most I/O patterns.
1027Front merges are handled by the binary trees in AS and deadline schedulers.
1028
Tejun Heo4c9f7832005-10-20 16:47:40 +02001029iii. Plugging the queue to batch requests in anticipation of opportunities for
1030 merge/sort optimizations
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032Plugging is an approach that the current i/o scheduling algorithm resorts to so
1033that it collects up enough requests in the queue to be able to take
1034advantage of the sorting/merging logic in the elevator. If the
1035queue is empty when a request comes in, then it plugs the request queue
Jens Axboe329007c2009-04-08 11:38:50 +02001036(sort of like plugging the bath tub of a vessel to get fluid to build up)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037till it fills up with a few more requests, before starting to service
1038the requests. This provides an opportunity to merge/sort the requests before
1039passing them down to the device. There are various conditions when the queue is
1040unplugged (to open up the flow again), either through a scheduled task or
1041could be on demand. For example wait_on_buffer sets the unplugging going
Jens Axboe329007c2009-04-08 11:38:50 +02001042through sync_buffer() running blk_run_address_space(mapping). Or the caller
1043can do it explicity through blk_unplug(bdev). So in the read case,
1044the queue gets explicitly unplugged as part of waiting for completion on that
1045buffer. For page driven IO, the address space ->sync_page() takes care of
1046doing the blk_run_address_space().
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047
1048Aside:
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.
1056
Tejun Heo4c9f7832005-10-20 16:47:40 +020010574.4 I/O contexts
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058I/O contexts provide a dynamically allocated per process data area. They may
1059be used in I/O schedulers, and in the block layer (could be used for IO statis,
Ben Collins1d193f42005-11-15 00:09:21 -08001060priorities for example). See *io_context in block/ll_rw_blk.c, and as-iosched.c
1061for an example of usage in an i/o scheduler.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
1063
10645. Scalability related changes
1065
10665.1 Granular Locking: io_request_lock replaced by a per-queue lock
1067
1068The global io_request_lock has been removed as of 2.5, to avoid
1069the scalability bottleneck it was causing, and has been replaced by more
1070granular locking. The request queue structure has a pointer to the
1071lock to be used for that queue. As a result, locking can now be
1072per-queue, with a provision for sharing a lock across queues if
1073necessary (e.g the scsi layer sets the queue lock pointers to the
1074corresponding adapter lock, which results in a per host locking
1075granularity). The locking semantics are the same, i.e. locking is
1076still imposed by the block layer, grabbing the lock before
1077request_fn execution which it means that lots of older drivers
1078should still be SMP safe. Drivers are free to drop the queue
1079lock themselves, if required. Drivers that explicitly used the
1080io_request_lock for serialization need to be modified accordingly.
1081Usually it's as easy as adding a global lock:
1082
Robert P. J. Dayc0d1f292008-04-21 22:44:50 +00001083 static DEFINE_SPINLOCK(my_driver_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084
1085and passing the address to that lock to blk_init_queue().
1086
10875.2 64 bit sector numbers (sector_t prepares for 64 bit support)
1088
1089The sector number used in the bio structure has been changed to sector_t,
1090which could be defined as 64 bit in preparation for 64 bit sector support.
1091
10926. Other Changes/Implications
1093
10946.1 Partition re-mapping handled by the generic block layer
1095
1096In 2.5 some of the gendisk/partition related code has been reorganized.
1097Now the generic block layer performs partition-remapping early and thus
1098provides drivers with a sector number relative to whole device, rather than
1099having to take partition number into account in order to arrive at the true
1100sector number. The routine blk_partition_remap() is invoked by
1101generic_make_request even before invoking the queue specific make_request_fn,
1102so the i/o scheduler also gets to operate on whole disk sector numbers. This
1103should typically not require changes to block drivers, it just never gets
1104to invoke its own partition sector offset calculations since all bios
1105sent are offset from the beginning of the device.
1106
1107
11087. A Few Tips on Migration of older drivers
1109
1110Old-style drivers that just use CURRENT and ignores clustered requests,
1111may not need much change. The generic layer will automatically handle
1112clustered requests, multi-page bios, etc for the driver.
1113
1114For a low performance driver or hardware that is PIO driven or just doesn't
1115support scatter-gather changes should be minimal too.
1116
1117The following are some points to keep in mind when converting old drivers
1118to bio.
1119
1120Drivers should use elv_next_request to pick up requests and are no longer
1121supposed to handle looping directly over the request list.
1122(struct request->queue has been removed)
1123
1124Now end_that_request_first takes an additional number_of_sectors argument.
1125It used to handle always just the first buffer_head in a request, now
1126it will loop and handle as many sectors (on a bio-segment granularity)
1127as specified.
1128
1129Now bh->b_end_io is replaced by bio->bi_end_io, but most of the time the
1130right thing to use is bio_endio(bio, uptodate) instead.
1131
1132If the driver is dropping the io_request_lock from its request_fn strategy,
1133then it just needs to replace that with q->queue_lock instead.
1134
1135As described in Sec 1.1, drivers can set max sector size, max segment size
1136etc per queue now. Drivers that used to define their own merge functions i
1137to handle things like this can now just use the blk_queue_* functions at
1138blk_init_queue time.
1139
1140Drivers no longer have to map a {partition, sector offset} into the
1141correct absolute location anymore, this is done by the block layer, so
1142where a driver received a request ala this before:
1143
1144 rq->rq_dev = mk_kdev(3, 5); /* /dev/hda5 */
1145 rq->sector = 0; /* first sector on hda5 */
1146
1147 it will now see
1148
1149 rq->rq_dev = mk_kdev(3, 0); /* /dev/hda */
1150 rq->sector = 123128; /* offset from start of disk */
1151
1152As mentioned, there is no virtual mapping of a bio. For DMA, this is
1153not a problem as the driver probably never will need a virtual mapping.
FUJITA Tomonoric2282ad2010-03-08 09:11:07 +01001154Instead it needs a bus mapping (dma_map_page for a single segment or
1155use dma_map_sg for scatter gather) to be able to ship it to the driver. For
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156PIO drivers (or drivers that need to revert to PIO transfer once in a
1157while (IDE for example)), where the CPU is doing the actual data
1158transfer 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
1160temporarily map a bio into the virtual address space.
1161
1162
11638. Prior/Related/Impacted patches
1164
11658.1. 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)
11718.2. Zero-copy networking (Dave Miller)
11728.3. SGI XFS - pagebuf patches - use of kiobufs
11738.4. Multi-page pioent patch for bio (Christoph Hellwig)
11748.5. Direct i/o implementation (Andrea Arcangeli) since 2.4.10-pre11
11758.6. Async i/o implementation patch (Ben LaHaise)
11768.7. EVMS layering design (IBM EVMS team)
11778.8. Larger page cache size patch (Ben LaHaise) and
1178 Large page size (Daniel Phillips)
1179 => larger contiguous physical memory buffers
11808.9. VM reservations patch (Ben LaHaise)
11818.10. Write clustering patches ? (Marcelo/Quintela/Riel ?)
11828.11. Block device in page cache patch (Andrea Archangeli) - now in 2.4.10+
11838.12. Multiple block-size transfers for faster raw i/o (Shailabh Nagar,
1184 Badari)
11858.13 Priority based i/o scheduler - prepatches (Arjan van de Ven)
11868.14 IDE Taskfile i/o patch (Andre Hedrick)
11878.15 Multi-page writeout and readahead patches (Andrew Morton)
11888.16 Direct i/o patches for 2.5 using kvec and bio (Badari Pulavarthy)
1189
11909. Other References:
1191
11929.1 The Splice I/O Model - Larry McVoy (and subsequent discussions on lkml,
1193and Linus' comments - Jan 2001)
11949.2 Discussions about kiobuf and bh design on lkml between sct, linus, alan
1195et al - Feb-March 2001 (many of the initial thoughts that led to bio were
Matt LaPlantefff92892006-10-03 22:47:42 +02001196brought up in this discussion thread)
Linus Torvalds1da177e2005-04-16 15:20:36 -070011979.3 Discussions on mempool on lkml - Dec 2001.
1198