2 .\" Copyright (c) 2002 Poul-Henning Kamp
3 .\" Copyright (c) 2002 Networks Associates Technology, Inc.
4 .\" All rights reserved.
6 .\" This software was developed for the FreeBSD Project by Poul-Henning Kamp
7 .\" and NAI Labs, the Security Research Division of Network Associates, Inc.
8 .\" under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the
9 .\" DARPA CHATS research program.
11 .\" Redistribution and use in source and binary forms, with or without
12 .\" modification, are permitted provided that the following conditions
14 .\" 1. Redistributions of source code must retain the above copyright
15 .\" notice, this list of conditions and the following disclaimer.
16 .\" 2. Redistributions in binary form must reproduce the above copyright
17 .\" notice, this list of conditions and the following disclaimer in the
18 .\" documentation and/or other materials provided with the distribution.
19 .\" 3. The names of the authors may not be used to endorse or promote
20 .\" products derived from this software without specific prior written
23 .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
24 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
27 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
42 .Nd "modular disk I/O request transformation framework"
47 .Cd options GEOM_CACHE
48 .Cd options GEOM_CONCAT
52 .Cd options GEOM_JOURNAL
53 .Cd options GEOM_LABEL
54 .Cd options GEOM_LINUX_LVM
57 .Cd options GEOM_MIRROR
58 .Cd options GEOM_MULTIPATH
60 .Cd options GEOM_PART_APM
61 .Cd options GEOM_PART_BSD
62 .Cd options GEOM_PART_BSD64
63 .Cd options GEOM_PART_EBR
64 .Cd options GEOM_PART_EBR_COMPAT
65 .Cd options GEOM_PART_GPT
66 .Cd options GEOM_PART_LDM
67 .Cd options GEOM_PART_MBR
68 .Cd options GEOM_PART_PC98
69 .Cd options GEOM_PART_VTOC8
72 .Cd options GEOM_RAID3
73 .Cd options GEOM_SHSEC
74 .Cd options GEOM_STRIPE
75 .Cd options GEOM_SUNLABEL
77 .Cd options GEOM_VIRSTOR
83 framework provides an infrastructure in which
85 can perform transformations on disk I/O requests on their path from
86 the upper kernel to the device drivers and back.
90 context range from the simple geometric
91 displacement performed in typical disk partitioning modules over RAID
92 algorithms and device multipath resolution to full blown cryptographic
93 protection of the stored data.
95 Compared to traditional
96 .Dq "volume management" ,
99 and in some cases all previous implementations in the following ways:
104 It is trivially simple to write a new class
105 of transformation and it will not be given stepchild treatment.
107 someone for some reason wanted to mount IBM MVS diskpacks, a class
108 recognizing and configuring their VTOC information would be a trivial
112 is topologically agnostic.
113 Most volume management implementations
114 have very strict notions of how classes can fit together, very often
115 one fixed hierarchy is provided, for instance, subdisk - plex -
119 Being extensible means that new transformations are treated no differently
120 than existing transformations.
122 Fixed hierarchies are bad because they make it impossible to express
123 the intent efficiently.
124 In the fixed hierarchy above, it is not possible to mirror two
125 physical disks and then partition the mirror into subdisks, instead
126 one is forced to make subdisks on the physical volumes and to mirror
127 these two and two, resulting in a much more complex configuration.
129 on the other hand does not care in which order things are done,
130 the only restriction is that cycles in the graph will not be allowed.
131 .Sh "TERMINOLOGY AND TOPOLOGY"
133 is quite object oriented and consequently the terminology
134 borrows a lot of context and semantics from the OO vocabulary:
138 represented by the data structure
141 particular kind of transformation.
142 Typical examples are MBR disk
143 partition, BSD disklabel, and RAID5 classes.
145 An instance of a class is called a
147 and represented by the data structure
152 will be one geom of class MBR for each disk.
156 represented by the data structure
158 is the front gate at which a geom offers service.
161 a disk-like thing which appears in
165 All providers have three main properties:
173 is the backdoor through which a geom connects to another
174 geom provider and through which I/O requests are sent.
176 The topological relationship between these entities are as follows:
179 A class has zero or more geom instances.
181 A geom has exactly one class it is derived from.
183 A geom has zero or more consumers.
185 A geom has zero or more providers.
187 A consumer can be attached to zero or one providers.
189 A provider can have zero or more consumers attached.
192 All geoms have a rank-number assigned, which is used to detect and
193 prevent loops in the acyclic directed graph.
198 A geom with no attached consumers has rank=1.
200 A geom with attached consumers has a rank one higher than the
201 highest rank of the geoms of the providers its consumers are
204 .Sh "SPECIAL TOPOLOGICAL MANEUVERS"
205 In addition to the straightforward attach, which attaches a consumer
206 to a provider, and detach, which breaks the bond, a number of special
207 topological maneuvers exists to facilitate configuration and to
208 improve the overall flexibility.
211 is a process that happens whenever a new class or new provider
212 is created, and it provides the class a chance to automatically configure an
213 instance on providers which it recognizes as its own.
214 A typical example is the MBR disk-partition class which will look for
215 the MBR table in the first sector and, if found and validated, will
216 instantiate a geom to multiplex according to the contents of the MBR.
218 A new class will be offered to all existing providers in turn and a new
219 provider will be offered to all classes in turn.
221 Exactly what a class does to recognize if it should accept the offered
222 provider is not defined by
224 but the sensible set of options are:
227 Examine specific data structures on the disk.
229 Examine properties like
235 Examine the rank number of the provider's geom.
237 Examine the method name of the provider's geom.
240 is the process by which a provider is removed while
241 it potentially is still being used.
243 When a geom orphans a provider, all future I/O requests will
245 on the provider with an error code set by the geom.
247 consumers attached to the provider will receive notification about
248 the orphanization when the event loop gets around to it, and they
249 can take appropriate action at that time.
251 A geom which came into being as a result of a normal taste operation
252 should self-destruct unless it has a way to keep functioning whilst
253 lacking the orphaned provider.
254 Geoms like disk slicers should therefore self-destruct whereas
255 RAID5 or mirror geoms will be able to continue as long as they do
258 When a provider is orphaned, this does not necessarily result in any
259 immediate change in the topology: any attached consumers are still
260 attached, any opened paths are still open, any outstanding I/O
261 requests are still outstanding.
263 The typical scenario is:
265 .Bl -bullet -offset indent -compact
267 A device driver detects a disk has departed and orphans the provider for it.
269 The geoms on top of the disk receive the orphanization event and
270 orphan all their providers in turn.
271 Providers which are not attached to will typically self-destruct
273 This process continues in a quasi-recursive fashion until all
274 relevant pieces of the tree have heard the bad news.
276 Eventually the buck stops when it reaches geom_dev at the top
281 to stop any more requests from
283 It will sleep until any and all outstanding I/O requests have
285 It will explicitly close (i.e.: zero the access counts), a change
286 which will propagate all the way down through the mesh.
287 It will then detach and destroy its geom.
289 The geom whose provider is now detached will destroy the provider,
290 detach and destroy its consumer and destroy its geom.
292 This process percolates all the way down through the mesh, until
293 the cleanup is complete.
296 While this approach seems byzantine, it does provide the maximum
297 flexibility and robustness in handling disappearing devices.
299 The one absolutely crucial detail to be aware of is that if the
300 device driver does not return all I/O requests, the tree will
303 is a special case of orphanization used to protect
304 against stale metadata.
305 It is probably easiest to understand spoiling by going through
310 on top of which an MBR geom provides
320 and that both the MBR and BSD geoms have
321 autoconfigured based on data structures on the disk media.
322 Now imagine the case where
324 is opened for writing and those
325 data structures are modified or overwritten: now the geoms would
326 be operating on stale metadata unless some notification system
327 can inform them otherwise.
329 To avoid this situation, when the open of
332 all attached consumers are told about this and geoms like
333 MBR and BSD will self-destruct as a result.
336 is closed, it will be offered for tasting again
337 and, if the data structures for MBR and BSD are still there, new
338 geoms will instantiate themselves anew.
340 Now for the fine print:
342 If any of the paths through the MBR or BSD module were open, they
343 would have opened downwards with an exclusive bit thus rendering it
346 for writing in that case.
348 the requested exclusive bit would render it impossible to open a
349 path through the MBR geom while
353 From this it also follows that changing the size of open geoms can
354 only be done with their cooperation.
356 Finally: the spoiling only happens when the write count goes from
357 zero to non-zero and the retasting happens only when the write count goes
358 from non-zero to zero.
360 is the process where the administrator issues instructions
361 for a particular class to instantiate itself.
363 ways to express intent in this case - a particular provider may be
364 specified with a level of override forcing, for instance, a BSD
365 disklabel module to attach to a provider which was not found palatable
366 during the TASTE operation.
368 Finally, I/O is the reason we even do this: it concerns itself with
369 sending I/O requests through the graph.
370 .It Em "I/O REQUESTS" ,
373 originate at a consumer,
374 are scheduled on its attached provider and, when processed, are returned
376 It is important to realize that the
378 which enters through the provider of a particular geom does not
380 come out on the other side
382 Even simple transformations like MBR and BSD will clone the
384 modify the clone, and schedule the clone on their
386 Note that cloning the
388 does not involve cloning the
389 actual data area specified in the I/O request.
391 In total, four different I/O requests exist in
393 read, write, delete, and
396 Read and write are self explanatory.
398 Delete indicates that a certain range of data is no longer used
399 and that it can be erased or freed as the underlying technology
401 Technologies like flash adaptation layers can arrange to erase
402 the relevant blocks before they will become reassigned and
403 cryptographic devices may want to fill random bits into the
404 range to reduce the amount of data available for attack.
406 It is important to recognize that a delete indication is not a
407 request and consequently there is no guarantee that the data actually
408 will be erased or made unavailable unless guaranteed by specific
412 semantics are required, a
413 geom should be pushed which converts delete indications into (a
414 sequence of) write requests.
417 supports inspection and manipulation
418 of out-of-band attributes on a particular provider or path.
419 Attributes are named by
421 strings and they will be discussed in
422 a separate section below.
425 (Stay tuned while the author rests his brain and fingers: more to come.)
427 Several flags are provided for tracing
429 operations and unlocking
430 protection mechanisms via the
431 .Va kern.geom.debugflags
433 All of these flags are off by default, and great care should be taken in
435 .Bl -tag -width indent
436 .It 0x01 Pq Dv G_T_TOPOLOGY
437 Provide tracing of topology change events.
438 .It 0x02 Pq Dv G_T_BIO
439 Provide tracing of buffer I/O requests.
440 .It 0x04 Pq Dv G_T_ACCESS
441 Provide tracing of access check controls.
443 .It 0x10 (allow foot shooting)
444 Allow writing to Rank 1 providers.
445 This would, for example, allow the super-user to overwrite the MBR on the root
446 disk or write random sectors elsewhere to a mounted disk.
447 The implications are obvious.
448 .It 0x40 Pq Dv G_F_DISKIOCTL
449 This is unused at this time.
450 .It 0x80 Pq Dv G_F_CTLDUMP
451 Dump contents of gctl requests.
455 .Xr DECLARE_GEOM_CLASS 9 ,
465 .Xr g_provider_by_name 9
467 This software was developed for the
470 .An Poul-Henning Kamp
471 and NAI Labs, the Security Research Division of Network Associates, Inc.\&
472 under DARPA/SPAWAR contract N66001-01-C-8035
475 DARPA CHATS research program.
477 The first precursor for
479 was a gruesome hack to Minix 1.2 and was
481 An earlier attempt to implement a less general scheme
486 .An Poul-Henning Kamp Aq Mt phk@FreeBSD.org