xref: /freebsd/sys/dev/isp/DriverManual.txt (revision 81ad626541db97eb356e2c1d4a20eb2a26a766ab)
1/* $FreeBSD$ */
2
3		Driver Theory of Operation Manual
4
51. Introduction
6
7This is a short text document that will describe the background, goals
8for, and current theory of operation for the joint Fibre Channel/SCSI
9HBA driver for QLogic hardware.
10
11Because this driver is an ongoing project, do not expect this manual
12to remain entirely up to date. Like a lot of software engineering, the
13ultimate documentation is the driver source. However, this manual should
14serve as a solid basis for attempting to understand where the driver
15started and what is trying to be accomplished with the current source.
16
17The reader is expected to understand the basics of SCSI and Fibre Channel
18and to be familiar with the range of platforms that Solaris, Linux and
19the variant "BSD" Open Source systems are available on. A glossary and
20a few references will be placed at the end of the document.
21
22There will be references to functions and structures within the body of
23this document. These can be easily found within the source using editor
24tags or grep. There will be few code examples here as the code already
25exists where the reader can easily find it.
26
272. A Brief History for this Driver
28
29This driver originally started as part of work funded by NASA Ames
30Research Center's Numerical Aerodynamic Simulation center ("NAS" for
31short) for the QLogic PCI 1020 and 1040 SCSI Host Adapters as part of my
32work at porting the NetBSD Operating System to the Alpha architectures
33(specifically the AlphaServer 8200 and 8400 platforms).  In short, it
34started just as simple single SCSI HBA driver for just the purpose of
35running off a SCSI disk. This work took place starting in January, 1997.
36
37Because the first implementation was for NetBSD, which runs on a very
38large number of platforms, and because NetBSD supported both systems with
39SBus cards (e.g., Sun SPARC systems) as well as systems with PCI cards,
40and because the QLogic SCSI cards came in both SBus and PCI versions, the
41initial implementation followed the very thoughtful NetBSD design tenet
42of splitting drivers into what are called MI (for Machine Independent)
43and MD (Machine Dependent) portions. The original design therefore was
44from the premise that the driver would drive both SBus and PCI card
45variants. These busses are similar but have quite different constraints,
46and while the QLogic SBus and PCI cards are very similar, there are some
47significant differences.
48
49After this initial goal had been met, there began to be some talk about
50looking into implementing Fibre Channel mass storage at NAS. At this time
51the QLogic 2100 FC/AL HBA was about to become available. After looking at
52the way it was designed I concluded that it was so darned close to being
53just like the SCSI HBAs that it would be insane to *not* leverage off of
54the existing driver. So, we ended up with a driver for NetBSD that drove
55PCI and SBus SCSI cards, and now also drove the QLogic 2100 FC-AL HBA.
56
57After this, ports to non-NetBSD platforms became interesting as well.
58This took the driver out of the interest with NAS and into interested
59support from a number of other places. Since the original NetBSD
60development, the driver has been ported to FreeBSD, OpenBSD, Linux,
61Solaris, and two proprietary systems. Following from the original MI/MD
62design of NetBSD, a rather successful attempt has been made to keep the
63Operating System Platform differences segregated and to a minimum.
64
65Along the way, support for the 2200 as well as full fabric and target
66mode support has been added, and 2300 support as well as an FC-IP stack
67are planned.
68
693. Driver Design Goals
70
71The driver has not started out as one normally would do such an effort.
72Normally you design via top-down methodologies and set an initial goal
73and meet it. This driver has had a design goal that changes from almost
74the very first. This has been an extremely peculiar, if not risque,
75experience. As a consequence, this section of this document contains
76a bit of "reconstruction after the fact" in that the design goals are
77as I perceive them to be now- not necessarily what they started as.
78
79The primary design goal now is to have a driver that can run both the
80SCSI and Fibre Channel SCSI prototocols on multiple OS platforms with
81as little OS platform support code as possible.
82
83The intended support targets for SCSI HBAs is to support the single and
84dual channel PCI Ultra2 and PCI Ultra3 cards as well as the older PCI
85Ultra single channel cards and SBus cards.
86
87The intended support targets for Fibre Channel HBAs is the 2100, 2200
88and 2300 PCI cards.
89
90Fibre Channel support should include complete fabric and public loop
91as well as private loop and private loop, direct-attach topologies.
92FC-IP support is also a goal.
93
94For both SCSI and Fibre Channel, simultaneous target/initiator mode support
95is a goal.
96
97Pure, raw, performance is not a primary goal of this design. This design,
98because it has a tremendous amount of code common across multiple
99platforms, will undoubtedly never be able to beat the performance of a
100driver that is specifically designed for a single platform and a single
101card. However, it is a good strong secondary goal to make the performance
102penalties in this design as small as possible.
103
104Another primary aim, which almost need not be stated, is that the
105implementation of platform differences must not clutter up the common
106code with platform specific defines. Instead, some reasonable layering
107semantics are defined such that platform specifics can be kept in the
108platform specific code.
109
1104. QLogic Hardware Architecture
111
112In order to make the design of this driver more intelligible, some
113description of the Qlogic hardware architecture is in order. This will
114not be an exhaustive description of how this card works, but will
115note enough of the important features so that the driver design is
116hopefully clearer.
117
1184.1 Basic QLogic hardware
119
120The QLogic HBA cards all contain a tiny 16-bit RISC-like processor and
121varying sizes of SRAM. Each card contains a Bus Interface Unit (BIU)
122as appropriate for the host bus (SBus or PCI).  The BIUs allow access
123to a set of dual-ranked 16 bit incoming and outgoing mailbox registers
124as well as access to control registers that control the RISC or access
125other portions of the card (e.g., Flash BIOS). The term 'dual-ranked'
126means that at the same host visible address if you write a mailbox
127register, that is a write to an (incoming, to the HBA) mailbox register,
128while a read to the same address reads another (outgoing, to the HBA)
129mailbox register with completely different data. Each HBA also then has
130core and auxiliary logic which either is used to interface to a SCSI bus
131(or to external bus drivers that connect to a SCSI bus), or to connect
132to a Fibre Channel bus.
133
1344.2 Basic Control Interface
135
136There are two principle I/O control mechanisms by which the driver
137communicates with and controls the QLogic HBA. The first mechanism is to
138use the incoming mailbox registers to interrupt and issue commands to
139the RISC processor (with results usually, but not always, ending up in
140the ougtoing mailbox registers). The second mechanism is to establish,
141via mailbox commands, circular request and response queues in system
142memory that are then shared between the QLogic and the driver. The
143request queue is used to queue requests (e.g., I/O requests) for the
144QLogic HBA's RISC engine to copy into the HBA memory and process. The
145result queue is used by the QLogic HBA's RISC engine to place results of
146requests read from the request queue, as well as to place notification
147of asynchronous events (e.g., incoming commands in target mode).
148
149To give a bit more precise scale to the preceding description, the QLogic
150HBA has 8 dual-ranked 16 bit mailbox registers, mostly for out-of-band
151control purposes. The QLogic HBA then utilizes a circular request queue
152of 64 byte fixed size Queue Entries to receive normal initiator mode
153I/O commands (or continue target mode requests). The request queue may
154be up to 256 elements for the QLogic 1020 and 1040 chipsets, but may
155be quite larger for the QLogic 12X0/12160 SCSI and QLogic 2X00 Fibre
156Channel chipsets.
157
158In addition to synchronously initiated usage of mailbox commands by
159the host system, the QLogic may also deliver asynchronous notifications
160solely in outgoing mailbox registers. These asynchronous notifications in
161mailboxes may be things like notification of SCSI Bus resets, or that the
162Fabric Name server has sent a change notification, or even that a specific
163I/O command completed without error (this is called 'Fast Posting'
164and saves the QLogic HBA from having to write a response queue entry).
165
166The QLogic HBA is an interrupting card, and when servicing an interrupt
167you really only have to check for either a mailbox interrupt or an
168interrupt notification that the response queue has an entry to
169be dequeued.
170
1714.3 Fibre Channel SCSI out of SCSI
172
173QLogic took the approach in introducing the 2X00 cards to just treat
174FC-AL as a 'fat' SCSI bus (a SCSI bus with more than 15 targets). All
175of the things that you really need to do with Fibre Channel with respect
176to providing FC-4 services on top of a Class 3 connection are performed
177by the RISC engine on the QLogic card itself. This means that from
178an HBA driver point of view, very little needs to change that would
179distinguish addressing a Fibre Channel disk from addressing a plain
180old SCSI disk.
181
182However, in the details it's not *quite* that simple. For example, in
183order to manage Fabric Connections, the HBA driver has to do explicit
184binding of entities it's queried from the name server to specific 'target'
185ids (targets, in this case, being a virtual entity).
186
187Still- the HBA firmware does really nearly all of the tedious management
188of Fibre Channel login state. The corollary to this sometimes is the
189lack of ability to say why a particular login connection to a Fibre
190Channel disk is not working well.
191
192There are clear limits with the QLogic card in managing fabric devices.
193The QLogic manages local loop devices (LoopID or Target 0..126) itself,
194but for the management of fabric devices, it has an absolute limit of
195253 simultaneous connections (256 entries less 3 reserved entries).
196
1975. Driver Architecture
198
1995.1 Driver Assumptions
200
201The first basic assumption for this driver is that the requirements for
202a SCSI HBA driver for any system is that of a 2 or 3 layer model where
203there are SCSI target device drivers (drivers which drive SCSI disks,
204SCSI tapes, and so on), possibly a middle services layer, and a bottom
205layer that manages the transport of SCSI CDB's out a SCSI bus (or across
206Fibre Channel) to a SCSI device. It's assumed that each SCSI command is
207a separate structure (or pointer to a structure) that contains the SCSI
208CDB and a place to store SCSI Status and SCSI Sense Data.
209
210This turns out to be a pretty good assumption. All of the Open Source
211systems (*BSD and Linux) and most of the proprietary systems have this
212kind of structure. This has been the way to manage SCSI subsystems for
213at least ten years.
214
215There are some additional basic assumptions that this driver makes- primarily
216in the arena of basic simple services like memory zeroing, memory copying,
217delay, sleep, microtime functions. It doesn't assume much more than this.
218
2195.2 Overall Driver Architecture
220
221The driver is split into a core (machine independent) module and platform
222and bus specific outer modules (machine dependent).
223
224The core code (in the files isp.c, isp_inline.h, ispvar.h, ispreg.h and
225ispmbox.h) handles:
226
227 + Chipset recognition and reset and firmware download (isp_reset)
228 + Board Initialization (isp_init)
229 + First level interrupt handling (response retrieval) (isp_intr)
230 + A SCSI command queueing entry point (isp_start)
231 + A set of control services accessed either via local requirements within
232   the core module or via an externally visible control entry point
233   (isp_control).
234
235The platform/bus specific modules (and definitions) depend on each
236platform, and they provide both definitions and functions for the core
237module's use.  Generally a platform module set is split into a bus
238dependent module (where configuration is begun from and bus specific
239support functions reside) and relatively thin platform specific layer
240which serves as the interconnect with the rest of this platform's SCSI
241subsystem.
242
243For ease of bus specific access issues, a centralized soft state
244structure is maintained for each HBA instance (struct ispsoftc). This
245soft state structure contains a machine/bus dependent vector (mdvec)
246for functions that read and write hardware registers, set up DMA for the
247request/response queues and fibre channel scratch area, set up and tear
248down DMA mappings for a SCSI command, provide a pointer to firmware to
249load, and other minor things.
250
251The machine dependent outer module must provide functional entry points
252for the core module:
253
254 + A SCSI command completion handoff point (isp_done)
255 + An asynchronous event handler (isp_async)
256 + A logging/printing function (isp_prt)
257
258The machine dependent outer module code must also provide a set of
259abstracting definitions which is what the core module utilizes heavily
260to do its job. These are discussed in detail in the comments in the
261file ispvar.h, but to give a sense of the range of what is required,
262let's illustrate two basic classes of these defines.
263
264The first class are "structure definition/access" class. An
265example of these would be:
266
267	XS_T            Platform SCSI transaction type (i.e., command for HBA)
268	..
269	XS_TGT(xs)      gets the target from an XS_T
270	..
271	XS_TAG_TYPE(xs) which type of tag to use
272	..
273
274The second class are 'functional' class definitions. Some examples of
275this class are:
276
277	MEMZERO(dst, src)                       platform zeroing function
278	..
279	MBOX_WAIT_COMPLETE(struct ispsoftc *)   wait for mailbox cmd to be done
280
281Note that the former is likely to be simple replacement with bzero or
282memset on most systems, while the latter could be quite complex.
283
284This soft state structure also contains different parameter information
285based upon whether this is a SCSI HBA or a Fibre Channel HBA (which is
286filled in by the code module).
287
288In order to clear up what is undoubtedly a seeming confusion of
289interconnects, a description of the typical flow of code that performs
290boards initialization and command transactions may help.
291
2925.3 Initialization Code Flow
293
294Typically a bus specific module for a platform (e.g., one that wants
295to configure a PCI card) is entered via that platform's configuration
296methods. If this module recognizes a card and can utilize or construct the
297space for the HBA instance softc, it does so, and initializes the machine
298dependent vector as well as any other platform specific information that
299can be hidden in or associated with this structure.
300
301Configuration at this point usually involves mapping in board registers
302and registering an interrupt. It's quite possible that the core module's
303isp_intr function is adequate to be the interrupt entry point, but often
304it's more useful have a bus specific wrapper module that calls isp_intr.
305
306After mapping and interrupt registry is done, isp_reset is called.
307Part of the isp_reset call may cause callbacks out to the bus dependent
308module to perform allocation and/or mapping of Request and Response
309queues (as well as a Fibre Channel scratch area if this is a Fibre
310Channel HBA).  The reason this is considered 'bus dependent' is that
311only the bus dependent module may have the information that says how
312one could perform I/O mapping and dependent (e.g., on a Solaris system)
313on the Request and Response queues. Another callback can enable the *use*
314of interrupts should this platform be able to finish configuration in
315interrupt driven mode.
316
317If isp_reset is successful at resetting the QLogic chipset and downloading
318new firmware (if available) and setting it running, isp_init is called. If
319isp_init is successful in doing initial board setups (including reading
320NVRAM from the QLogic card), then this bus specicic module will call the
321platform dependent module that takes the appropriate steps to 'register'
322this HBA with this platform's SCSI subsystem.  Examining either the
323OpenBSD or the NetBSD isp_pci.c or isp_sbus.c files may assist the reader
324here in clarifying some of this.
325
3265.4 Initiator Mode Command Code Flow
327
328A successful execution of isp_init will lead to the driver 'registering'
329itself with this platform's SCSI subsystem. One assumed action for this
330is the registry of a function that the SCSI subsystem for this platform
331will call when it has a SCSI command to run.
332
333The platform specific module function that receives this will do whatever
334it needs to prepare this command for execution in the core module. This
335sounds vague, but it's also very flexible. In principle, this could be
336a complete marshalling/demarshalling of this platform's SCSI command
337structure (should it be impossible to represent in an XS_T). In addition,
338this function can also block commands from running (if, e.g., Fibre
339Channel loop state would preclude successful starting of the command).
340
341When it's ready to do so, the function isp_start is called with this
342command. This core module tries to allocate request queue space for
343this command. It also calls through the machine dependent vector
344function to make sure any DMA mapping for this command is done.
345
346Now, DMA mapping here is possibly a misnomer, as more than just
347DMA mapping can be done in this bus dependent function. This is
348also the place where any endian byte-swizzling will be done. At any
349rate, this function is called last because the process of establishing
350DMA addresses for any command may in fact consume more Request Queue
351entries than there are currently available. If the mapping and other
352functions are successful, the QLogic mailbox inbox pointer register
353is updated to indicate to the QLogic that it has a new request to
354read.
355
356If this function is unsuccessful, policy as to what to do at this point is
357left to the machine dependent platform function which called isp_start. In
358some platforms, temporary resource shortages can be handled by the main
359SCSI subsystem. In other platforms, the machine dependent code has to
360handle this.
361
362In order to keep track of commands that are in progress, the soft state
363structure contains an array of 'handles' that are associated with each
364active command. When you send a command to the QLogic firmware, a portion
365of the Request Queue entry can contain a non-zero handle identifier so
366that at a later point in time in reading either a Response Queue entry
367or from a Fast Posting mailbox completion interrupt, you can take this
368handle to find the command you were waiting on. It should be noted that
369this is probably one of the most dangerous areas of this driver. Corrupted
370handles will lead to system panics.
371
372At some later point in time an interrupt will occur. Eventually,
373isp_intr will be called. This core module will determine what the cause
374of the interrupt is, and if it is for a completing command. That is,
375it'll determine the handle and fetch the pointer to the command out of
376storage within the soft state structure. Skipping over a lot of details,
377the machine dependent code supplied function isp_done is called with the
378pointer to the completing command. This would then be the glue layer that
379informs the SCSI subsystem for this platform that a command is complete.
380
3815.5 Asynchronous Events
382
383Interrupts occur for events other than commands (mailbox or request queue
384started commands) completing. These are called Asynchronous Mailbox
385interrupts. When some external event causes the SCSI bus to be reset,
386or when a Fibre Channel loop changes state (e.g., a LIP is observed),
387this generates such an asynchronous event.
388
389Each platform module has to provide an isp_async entry point that will
390handle a set of these. This isp_async entry point also handles things
391which aren't properly async events but are simply natural outgrowths
392of code flow for another core function (see discussion on fabric device
393management below).
394
3955.6 Target Mode Code Flow
396
397This section could use a lot of expansion, but this covers the basics.
398
399The QLogic cards, when operating in target mode, follow a code flow that is
400essentially the inverse of that for intiator mode describe above. In this
401scenario, an interrupt occurs, and present on the Response Queue is a
402queue entry element defining a new command arriving from an initiator.
403
404This is passed to possibly external target mode handler. This driver
405provides some handling for this in a core module, but also leaves
406things open enough that a completely different target mode handler
407may accept this incoming queue entry.
408
409The external target mode handler then turns around forms up a response
410to this 'response' that just arrived which is then placed on the Request
411Queue and handled very much like an initiator mode command (i.e., calling
412the bus dependent DMA mapping function). If this entry completes the
413command, no more need occur. But often this handles only part of the
414requested command, so the QLogic firmware will rewrite the response
415to the initial 'response' again onto the Response Queue, whereupon the
416target mode handler will respond to that, and so on until the command
417is completely handled.
418
419Because almost no platform provides basic SCSI Subsystem target mode
420support, this design has been left extremely open ended, and as such
421it's a bit hard to describe in more detail than this.
422
4235.7 Locking Assumptions
424
425The observant reader by now is likely to have asked the question, "but what
426about locking? Or interrupt masking" by now.
427
428The basic assumption about this is that the core module does not know
429anything directly about locking or interrupt masking. It may assume that
430upon entry (e.g., via isp_start, isp_control, isp_intr) that appropriate
431locking and interrupt masking has been done.
432
433The platform dependent code may also therefore assume that if it is
434called (e.g., isp_done or isp_async) that any locking or masking that
435was in place upon the entry to the core module is still there. It is up
436to the platform dependent code to worry about avoiding any lock nesting
437issues. As an example of this, the Linux implementation simply queues
438up commands completed via the callout to isp_done, which it then pushes
439out to the SCSI subsystem after a return from it's calling isp_intr is
440executed (and locks dropped appropriately, as well as avoidance of deep
441interrupt stacks).
442
443Recent changes in the design have now eased what had been an original
444requirement that the while in the core module no locks or interrupt
445masking could be dropped. It's now up to each platform to figure out how
446to implement this. This is principally used in the execution of mailbox
447commands (which are principally used for Loop and Fabric management via
448the isp_control function).
449
4505.8 SCSI Specifics
451
452The driver core or platform dependent architecture issues that are specific
453to SCSI are few. There is a basic assumption that the QLogic firmware
454supported Automatic Request sense will work- there is no particular provision
455for disabling it's usage on a per-command basis.
456
4575.9 Fibre Channel Specifics
458
459Fibre Channel presents an interesting challenge here. The QLogic firmware
460architecture for dealing with Fibre Channel as just a 'fat' SCSI bus
461is fine on the face of it, but there are some subtle and not so subtle
462problems here.
463
4645.9.1 Firmware State
465
466Part of the initialization (isp_init) for Fibre Channel HBAs involves
467sending a command (Initialize Control Block) that establishes Node
468and Port WWNs as well as topology preferences. After this occurs,
469the QLogic firmware tries to traverese through serveral states:
470
471	FW_CONFIG_WAIT
472	FW_WAIT_AL_PA
473	FW_WAIT_LOGIN
474	FW_READY
475	FW_LOSS_OF_SYNC
476	FW_ERROR
477	FW_REINIT
478	FW_NON_PART
479
480It starts with FW_CONFIG_WAIT, attempts to get an AL_PA (if on an FC-AL
481loop instead of being connected as an N-port), waits to log into all
482FC-AL loop entities and then hopefully transitions to FW_READY state.
483
484Clearly, no command should be attempted prior to FW_READY state is
485achieved. The core internal function isp_fclink_test (reachable via
486isp_control with the ISPCTL_FCLINK_TEST function code). This function
487also determines connection topology (i.e., whether we're attached to a
488fabric or not).
489
4905.9.2. Loop State Transitions- From Nil to Ready
491
492Once the firmware has transitioned to a ready state, then the state of the
493connection to either arbitrated loop or to a fabric has to be ascertained,
494and the identity of all loop members (and fabric members validated).
495
496This can be very complicated, and it isn't made easy in that the QLogic
497firmware manages PLOGI and PRLI to devices that are on a local loop, but
498it is the driver that must manage PLOGI/PRLI with devices on the fabric.
499
500In order to manage this state an eight level staging of current "Loop"
501(where "Loop" is taken to mean FC-AL or N- or F-port connections) states
502in the following ascending order:
503
504	LOOP_NIL
505	LOOP_LIP_RCVD
506	LOOP_PDB_RCVD
507	LOOP_SCANNING_FABRIC
508	LOOP_FSCAN_DONE
509	LOOP_SCANNING_LOOP
510	LOOP_LSCAN_DONE
511	LOOP_SYNCING_PDB
512	LOOP_READY
513
514When the core code initializes the QLogic firmware, it sets the loop
515state to LOOP_NIL. The first 'LIP Received' asynchronous event sets state
516to LOOP_LIP_RCVD. This should be followed by a "Port Database Changed"
517asynchronous event which will set the state to LOOP_PDB_RCVD. Each of
518these states, when entered, causes an isp_async event call to the
519machine dependent layers with the ISPASYNC_CHANGE_NOTIFY code.
520
521After the state of LOOP_PDB_RCVD is reached, the internal core function
522isp_scan_fabric (reachable via isp_control(..ISPCTL_SCAN_FABRIC)) will,
523if the connection is to a fabric, use Simple Name Server mailbox mediated
524commands to dump the entire fabric contents. For each new entity, an
525isp_async event will be generated that says a Fabric device has arrived
526(ISPASYNC_FABRIC_DEV). The function that isp_async must perform in this
527step is to insert possibly remove devices that it wants to have the
528QLogic firmware log into (at LOOP_SYNCING_PDB state level)).
529
530After this has occurred, the state LOOP_FSCAN_DONE is set, and then the
531internal function isp_scan_loop (isp_control(...ISPCTL_SCAN_LOOP)) can
532be called which will then scan for any local (FC-AL) entries by asking
533for each possible local loop id the QLogic firmware for a Port Database
534entry. It's at this level some entries cached locally are purged
535or shifting loopids are managed (see section 5.9.4).
536
537The final step after this is to call the internal function isp_pdb_sync
538(isp_control(..ISPCTL_PDB_SYNC)). The purpose of this function is to
539then perform the PLOGI/PRLI functions for fabric devices. The next state
540entered after this is LOOP_READY, which means that the driver is ready
541to process commands to send to Fibre Channel devices.
542
5435.9.3 Fibre Channel variants of Initiator Mode Code Flow
544
545The code flow in isp_start for Fibre Channel devices is the same as it is
546for SCSI devices, but with a notable exception.
547
548Maintained within the fibre channel specific portion of the driver soft
549state structure is a distillation of the existing population of both
550local loop and fabric devices. Because Loop IDs can shift on a local
551loop but we wish to retain a 'constant' Target ID (see 5.9.4), this
552is indexed directly via the Target ID for the command (XS_TGT(xs)).
553
554If there is a valid entry for this Target ID, the command is started
555(with the stored 'Loop ID'). If not the command is completed with
556the error that is just like a SCSI Selection Timeout error.
557
558This code is currently somewhat in transition. Some platforms to
559do firmware and loop state management (as described above) at this
560point. Other platforms manage this from the machine dependent layers. The
561important function to watch in this respect is isp_fc_runstate (in
562isp_inline.h).
563
5645.9.4 "Target" in Fibre Channel is a fixed virtual construct
565
566Very few systems can cope with the notion that "Target" for a disk
567device can change while you're using it. But one of the properties of
568for arbitrated loop is that the physical bus address for a loop member
569(the AL_PA) can change depending on when and how things are inserted in
570the loop.
571
572To illustrate this, let's take an example. Let's say you start with a
573loop that has 5 disks in it. At boot time, the system will likely find
574them and see them in this order:
575
576disk#   Loop ID         Target ID
577disk0   0               0
578disk1   1               1
579disk2   2               2
580disk3   3               3
581disk4   4               4
582
583The driver uses 'Loop ID' when it forms requests to send a comamnd to
584each disk. However, it reports to NetBSD that things exist as 'Target
585ID'. As you can see here, there is perfect correspondence between disk,
586Loop ID and Target ID.
587
588Let's say you add a new disk between disk2 and disk3 while things are
589running. You don't really often see this, but you *could* see this where
590the loop has to renegotiate, and you end up with:
591
592disk#   Loop ID         Target ID
593disk0   0               0
594disk1   1               1
595disk2   2               2
596diskN   3               ?
597disk3   4               ?
598disk4   5               ?
599
600Clearly, you don't want disk3 and disk4's "Target ID" to change while you're
601running since currently mounted filesystems will get trashed.
602
603What the driver is supposed to do (this is the function of isp_scan_loop),
604is regenerate things such that the following then occurs:
605
606disk#   Loop ID         Target ID
607disk0   0               0
608disk1   1               1
609disk2   2               2
610diskN   3               5
611disk3   4               3
612disk4   5               4
613
614So, "Target" is a virtual entity that is maintained while you're running.
615
6166. Glossary
617
618HBA - Host Bus Adapter
619
620SCSI - Small Computer
621
6227. References
623
624Various URLs of interest:
625
626http://www.netbsd.org		-	NetBSD's Web Page
627http://www.openbsd.org		-	OpenBSD's Web Page
628https://www.freebsd.org		-	FreeBSD's Web Page
629
630http://www.t10.org		-	ANSI SCSI Commitee's Web Page
631					(SCSI Specs)
632http://www.t11.org		-	NCITS Device Interface Web Page
633					(Fibre Channel Specs)
634
635