]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | <?xml version="1.0" encoding="UTF-8"?> |
2 | <!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook XML V4.1.2//EN" | |
3 | "http://www.oasis-open.org/docbook/xml/4.1.2/docbookx.dtd" []> | |
4 | ||
5 | <book id="libataDevGuide"> | |
6 | <bookinfo> | |
7 | <title>libATA Developer's Guide</title> | |
8 | ||
9 | <authorgroup> | |
10 | <author> | |
11 | <firstname>Jeff</firstname> | |
12 | <surname>Garzik</surname> | |
13 | </author> | |
14 | </authorgroup> | |
15 | ||
16 | <copyright> | |
4f931374 | 17 | <year>2003-2006</year> |
1da177e4 LT |
18 | <holder>Jeff Garzik</holder> |
19 | </copyright> | |
20 | ||
21 | <legalnotice> | |
22 | <para> | |
23 | The contents of this file are subject to the Open | |
24 | Software License version 1.1 that can be found at | |
25 | <ulink url="http://www.opensource.org/licenses/osl-1.1.txt">http://www.opensource.org/licenses/osl-1.1.txt</ulink> and is included herein | |
26 | by reference. | |
27 | </para> | |
28 | ||
29 | <para> | |
30 | Alternatively, the contents of this file may be used under the terms | |
31 | of the GNU General Public License version 2 (the "GPL") as distributed | |
32 | in the kernel source COPYING file, in which case the provisions of | |
33 | the GPL are applicable instead of the above. If you wish to allow | |
34 | the use of your version of this file only under the terms of the | |
35 | GPL and not to allow others to use your version of this file under | |
36 | the OSL, indicate your decision by deleting the provisions above and | |
37 | replace them with the notice and other provisions required by the GPL. | |
38 | If you do not delete the provisions above, a recipient may use your | |
39 | version of this file under either the OSL or the GPL. | |
40 | </para> | |
41 | ||
42 | </legalnotice> | |
43 | </bookinfo> | |
44 | ||
45 | <toc></toc> | |
46 | ||
07dd39b9 JG |
47 | <chapter id="libataIntroduction"> |
48 | <title>Introduction</title> | |
49 | <para> | |
50 | libATA is a library used inside the Linux kernel to support ATA host | |
51 | controllers and devices. libATA provides an ATA driver API, class | |
52 | transports for ATA and ATAPI devices, and SCSI<->ATA translation | |
53 | for ATA devices according to the T10 SAT specification. | |
54 | </para> | |
55 | <para> | |
56 | This Guide documents the libATA driver API, library functions, library | |
57 | internals, and a couple sample ATA low-level drivers. | |
58 | </para> | |
59 | </chapter> | |
60 | ||
1da177e4 LT |
61 | <chapter id="libataDriverApi"> |
62 | <title>libata Driver API</title> | |
92bab26b JG |
63 | <para> |
64 | struct ata_port_operations is defined for every low-level libata | |
65 | hardware driver, and it controls how the low-level driver | |
66 | interfaces with the ATA and SCSI layers. | |
67 | </para> | |
68 | <para> | |
69 | FIS-based drivers will hook into the system with ->qc_prep() and | |
70 | ->qc_issue() high-level hooks. Hardware which behaves in a manner | |
71 | similar to PCI IDE hardware may utilize several generic helpers, | |
72 | defining at a bare minimum the bus I/O addresses of the ATA shadow | |
73 | register blocks. | |
74 | </para> | |
1da177e4 LT |
75 | <sect1> |
76 | <title>struct ata_port_operations</title> | |
77 | ||
92bab26b | 78 | <sect2><title>Disable ATA port</title> |
1da177e4 LT |
79 | <programlisting> |
80 | void (*port_disable) (struct ata_port *); | |
81 | </programlisting> | |
82 | ||
83 | <para> | |
84 | Called from ata_bus_probe() and ata_bus_reset() error paths, | |
85 | as well as when unregistering from the SCSI module (rmmod, hot | |
86 | unplug). | |
8b2af8f0 EF |
87 | This function should do whatever needs to be done to take the |
88 | port out of use. In most cases, ata_port_disable() can be used | |
89 | as this hook. | |
90 | </para> | |
91 | <para> | |
92 | Called from ata_bus_probe() on a failed probe. | |
93 | Called from ata_bus_reset() on a failed bus reset. | |
94 | Called from ata_scsi_release(). | |
1da177e4 LT |
95 | </para> |
96 | ||
92bab26b JG |
97 | </sect2> |
98 | ||
99 | <sect2><title>Post-IDENTIFY device configuration</title> | |
1da177e4 LT |
100 | <programlisting> |
101 | void (*dev_config) (struct ata_port *, struct ata_device *); | |
102 | </programlisting> | |
103 | ||
104 | <para> | |
105 | Called after IDENTIFY [PACKET] DEVICE is issued to each device | |
106 | found. Typically used to apply device-specific fixups prior to | |
107 | issue of SET FEATURES - XFER MODE, and prior to operation. | |
108 | </para> | |
8b2af8f0 | 109 | <para> |
8b2af8f0 EF |
110 | This entry may be specified as NULL in ata_port_operations. |
111 | </para> | |
1da177e4 | 112 | |
92bab26b JG |
113 | </sect2> |
114 | ||
115 | <sect2><title>Set PIO/DMA mode</title> | |
1da177e4 LT |
116 | <programlisting> |
117 | void (*set_piomode) (struct ata_port *, struct ata_device *); | |
118 | void (*set_dmamode) (struct ata_port *, struct ata_device *); | |
5444a6f4 AC |
119 | void (*post_set_mode) (struct ata_port *); |
120 | unsigned int (*mode_filter) (struct ata_port *, struct ata_device *, unsigned int); | |
1da177e4 LT |
121 | </programlisting> |
122 | ||
123 | <para> | |
124 | Hooks called prior to the issue of SET FEATURES - XFER MODE | |
5444a6f4 AC |
125 | command. The optional ->mode_filter() hook is called when libata |
126 | has built a mask of the possible modes. This is passed to the | |
127 | ->mode_filter() function which should return a mask of valid modes | |
128 | after filtering those unsuitable due to hardware limits. It is not | |
129 | valid to use this interface to add modes. | |
130 | </para> | |
131 | <para> | |
132 | dev->pio_mode and dev->dma_mode are guaranteed to be valid when | |
133 | ->set_piomode() and when ->set_dmamode() is called. The timings for | |
134 | any other drive sharing the cable will also be valid at this point. | |
135 | That is the library records the decisions for the modes of each | |
136 | drive on a channel before it attempts to set any of them. | |
137 | </para> | |
138 | <para> | |
139 | ->post_set_mode() is | |
1da177e4 LT |
140 | called unconditionally, after the SET FEATURES - XFER MODE |
141 | command completes successfully. | |
142 | </para> | |
143 | ||
144 | <para> | |
145 | ->set_piomode() is always called (if present), but | |
146 | ->set_dma_mode() is only called if DMA is possible. | |
147 | </para> | |
148 | ||
92bab26b JG |
149 | </sect2> |
150 | ||
151 | <sect2><title>Taskfile read/write</title> | |
1da177e4 | 152 | <programlisting> |
40868c85 SS |
153 | void (*sff_tf_load) (struct ata_port *ap, struct ata_taskfile *tf); |
154 | void (*sff_tf_read) (struct ata_port *ap, struct ata_taskfile *tf); | |
1da177e4 LT |
155 | </programlisting> |
156 | ||
157 | <para> | |
158 | ->tf_load() is called to load the given taskfile into hardware | |
159 | registers / DMA buffers. ->tf_read() is called to read the | |
160 | hardware registers / DMA buffers, to obtain the current set of | |
161 | taskfile register values. | |
8b2af8f0 | 162 | Most drivers for taskfile-based hardware (PIO or MMIO) use |
40868c85 | 163 | ata_sff_tf_load() and ata_sff_tf_read() for these hooks. |
1da177e4 LT |
164 | </para> |
165 | ||
92bab26b JG |
166 | </sect2> |
167 | ||
bf717b11 JG |
168 | <sect2><title>PIO data read/write</title> |
169 | <programlisting> | |
40868c85 | 170 | void (*sff_data_xfer) (struct ata_device *, unsigned char *, unsigned int, int); |
bf717b11 JG |
171 | </programlisting> |
172 | ||
173 | <para> | |
174 | All bmdma-style drivers must implement this hook. This is the low-level | |
175 | operation that actually copies the data bytes during a PIO data | |
176 | transfer. | |
40868c85 SS |
177 | Typically the driver will choose one of ata_sff_data_xfer_noirq(), |
178 | ata_sff_data_xfer(), or ata_sff_data_xfer32(). | |
bf717b11 JG |
179 | </para> |
180 | ||
181 | </sect2> | |
182 | ||
92bab26b | 183 | <sect2><title>ATA command execute</title> |
1da177e4 | 184 | <programlisting> |
40868c85 | 185 | void (*sff_exec_command)(struct ata_port *ap, struct ata_taskfile *tf); |
1da177e4 LT |
186 | </programlisting> |
187 | ||
188 | <para> | |
189 | causes an ATA command, previously loaded with | |
190 | ->tf_load(), to be initiated in hardware. | |
40868c85 | 191 | Most drivers for taskfile-based hardware use ata_sff_exec_command() |
8b2af8f0 | 192 | for this hook. |
1da177e4 LT |
193 | </para> |
194 | ||
92bab26b JG |
195 | </sect2> |
196 | ||
197 | <sect2><title>Per-cmd ATAPI DMA capabilities filter</title> | |
780a87f7 JG |
198 | <programlisting> |
199 | int (*check_atapi_dma) (struct ata_queued_cmd *qc); | |
200 | </programlisting> | |
201 | ||
202 | <para> | |
203 | Allow low-level driver to filter ATA PACKET commands, returning a status | |
204 | indicating whether or not it is OK to use DMA for the supplied PACKET | |
205 | command. | |
206 | </para> | |
8b2af8f0 EF |
207 | <para> |
208 | This hook may be specified as NULL, in which case libata will | |
209 | assume that atapi dma can be supported. | |
210 | </para> | |
780a87f7 | 211 | |
92bab26b JG |
212 | </sect2> |
213 | ||
214 | <sect2><title>Read specific ATA shadow registers</title> | |
1da177e4 | 215 | <programlisting> |
40868c85 SS |
216 | u8 (*sff_check_status)(struct ata_port *ap); |
217 | u8 (*sff_check_altstatus)(struct ata_port *ap); | |
1da177e4 LT |
218 | </programlisting> |
219 | ||
220 | <para> | |
bf717b11 | 221 | Reads the Status/AltStatus ATA shadow register from |
780a87f7 JG |
222 | hardware. On some hardware, reading the Status register has |
223 | the side effect of clearing the interrupt condition. | |
8b2af8f0 | 224 | Most drivers for taskfile-based hardware use |
40868c85 | 225 | ata_sff_check_status() for this hook. |
1da177e4 LT |
226 | </para> |
227 | ||
92bab26b JG |
228 | </sect2> |
229 | ||
230 | <sect2><title>Select ATA device on bus</title> | |
1da177e4 | 231 | <programlisting> |
40868c85 | 232 | void (*sff_dev_select)(struct ata_port *ap, unsigned int device); |
1da177e4 LT |
233 | </programlisting> |
234 | ||
235 | <para> | |
236 | Issues the low-level hardware command(s) that causes one of N | |
237 | hardware devices to be considered 'selected' (active and | |
780a87f7 | 238 | available for use) on the ATA bus. This generally has no |
8b2af8f0 EF |
239 | meaning on FIS-based devices. |
240 | </para> | |
241 | <para> | |
242 | Most drivers for taskfile-based hardware use | |
40868c85 | 243 | ata_sff_dev_select() for this hook. |
1da177e4 LT |
244 | </para> |
245 | ||
92bab26b JG |
246 | </sect2> |
247 | ||
5444a6f4 AC |
248 | <sect2><title>Private tuning method</title> |
249 | <programlisting> | |
250 | void (*set_mode) (struct ata_port *ap); | |
251 | </programlisting> | |
252 | ||
253 | <para> | |
254 | By default libata performs drive and controller tuning in | |
255 | accordance with the ATA timing rules and also applies blacklists | |
256 | and cable limits. Some controllers need special handling and have | |
257 | custom tuning rules, typically raid controllers that use ATA | |
258 | commands but do not actually do drive timing. | |
259 | </para> | |
260 | ||
261 | <warning> | |
262 | <para> | |
263 | This hook should not be used to replace the standard controller | |
264 | tuning logic when a controller has quirks. Replacing the default | |
265 | tuning logic in that case would bypass handling for drive and | |
266 | bridge quirks that may be important to data reliability. If a | |
267 | controller needs to filter the mode selection it should use the | |
268 | mode_filter hook instead. | |
269 | </para> | |
270 | </warning> | |
271 | ||
272 | </sect2> | |
273 | ||
92bab26b | 274 | <sect2><title>Control PCI IDE BMDMA engine</title> |
1da177e4 LT |
275 | <programlisting> |
276 | void (*bmdma_setup) (struct ata_queued_cmd *qc); | |
277 | void (*bmdma_start) (struct ata_queued_cmd *qc); | |
780a87f7 JG |
278 | void (*bmdma_stop) (struct ata_port *ap); |
279 | u8 (*bmdma_status) (struct ata_port *ap); | |
1da177e4 LT |
280 | </programlisting> |
281 | ||
282 | <para> | |
780a87f7 JG |
283 | When setting up an IDE BMDMA transaction, these hooks arm |
284 | (->bmdma_setup), fire (->bmdma_start), and halt (->bmdma_stop) | |
285 | the hardware's DMA engine. ->bmdma_status is used to read the standard | |
286 | PCI IDE DMA Status register. | |
287 | </para> | |
288 | ||
289 | <para> | |
290 | These hooks are typically either no-ops, or simply not implemented, in | |
291 | FIS-based drivers. | |
1da177e4 | 292 | </para> |
8b2af8f0 EF |
293 | <para> |
294 | Most legacy IDE drivers use ata_bmdma_setup() for the bmdma_setup() | |
295 | hook. ata_bmdma_setup() will write the pointer to the PRD table to | |
296 | the IDE PRD Table Address register, enable DMA in the DMA Command | |
297 | register, and call exec_command() to begin the transfer. | |
298 | </para> | |
299 | <para> | |
300 | Most legacy IDE drivers use ata_bmdma_start() for the bmdma_start() | |
301 | hook. ata_bmdma_start() will write the ATA_DMA_START flag to the DMA | |
302 | Command register. | |
303 | </para> | |
304 | <para> | |
305 | Many legacy IDE drivers use ata_bmdma_stop() for the bmdma_stop() | |
306 | hook. ata_bmdma_stop() clears the ATA_DMA_START flag in the DMA | |
307 | command register. | |
308 | </para> | |
309 | <para> | |
310 | Many legacy IDE drivers use ata_bmdma_status() as the bmdma_status() hook. | |
311 | </para> | |
1da177e4 | 312 | |
92bab26b JG |
313 | </sect2> |
314 | ||
315 | <sect2><title>High-level taskfile hooks</title> | |
1da177e4 LT |
316 | <programlisting> |
317 | void (*qc_prep) (struct ata_queued_cmd *qc); | |
318 | int (*qc_issue) (struct ata_queued_cmd *qc); | |
319 | </programlisting> | |
320 | ||
321 | <para> | |
322 | Higher-level hooks, these two hooks can potentially supercede | |
323 | several of the above taskfile/DMA engine hooks. ->qc_prep is | |
324 | called after the buffers have been DMA-mapped, and is typically | |
325 | used to populate the hardware's DMA scatter-gather table. | |
326 | Most drivers use the standard ata_qc_prep() helper function, but | |
327 | more advanced drivers roll their own. | |
328 | </para> | |
329 | <para> | |
330 | ->qc_issue is used to make a command active, once the hardware | |
331 | and S/G tables have been prepared. IDE BMDMA drivers use the | |
332 | helper function ata_qc_issue_prot() for taskfile protocol-based | |
780a87f7 | 333 | dispatch. More advanced drivers implement their own ->qc_issue. |
1da177e4 | 334 | </para> |
8b2af8f0 EF |
335 | <para> |
336 | ata_qc_issue_prot() calls ->tf_load(), ->bmdma_setup(), and | |
337 | ->bmdma_start() as necessary to initiate a transfer. | |
338 | </para> | |
1da177e4 | 339 | |
92bab26b JG |
340 | </sect2> |
341 | ||
bf717b11 | 342 | <sect2><title>Exception and probe handling (EH)</title> |
1da177e4 LT |
343 | <programlisting> |
344 | void (*eng_timeout) (struct ata_port *ap); | |
bf717b11 JG |
345 | void (*phy_reset) (struct ata_port *ap); |
346 | </programlisting> | |
347 | ||
348 | <para> | |
349 | Deprecated. Use ->error_handler() instead. | |
350 | </para> | |
351 | ||
352 | <programlisting> | |
353 | void (*freeze) (struct ata_port *ap); | |
354 | void (*thaw) (struct ata_port *ap); | |
355 | </programlisting> | |
356 | ||
357 | <para> | |
358 | ata_port_freeze() is called when HSM violations or some other | |
359 | condition disrupts normal operation of the port. A frozen port | |
360 | is not allowed to perform any operation until the port is | |
361 | thawed, which usually follows a successful reset. | |
362 | </para> | |
363 | ||
364 | <para> | |
365 | The optional ->freeze() callback can be used for freezing the port | |
366 | hardware-wise (e.g. mask interrupt and stop DMA engine). If a | |
367 | port cannot be frozen hardware-wise, the interrupt handler | |
368 | must ack and clear interrupts unconditionally while the port | |
369 | is frozen. | |
370 | </para> | |
371 | <para> | |
372 | The optional ->thaw() callback is called to perform the opposite of ->freeze(): | |
373 | prepare the port for normal operation once again. Unmask interrupts, | |
374 | start DMA engine, etc. | |
375 | </para> | |
376 | ||
377 | <programlisting> | |
378 | void (*error_handler) (struct ata_port *ap); | |
379 | </programlisting> | |
380 | ||
381 | <para> | |
382 | ->error_handler() is a driver's hook into probe, hotplug, and recovery | |
383 | and other exceptional conditions. The primary responsibility of an | |
384 | implementation is to call ata_do_eh() or ata_bmdma_drive_eh() with a set | |
385 | of EH hooks as arguments: | |
386 | </para> | |
387 | ||
388 | <para> | |
389 | 'prereset' hook (may be NULL) is called during an EH reset, before any other actions | |
390 | are taken. | |
391 | </para> | |
392 | ||
393 | <para> | |
394 | 'postreset' hook (may be NULL) is called after the EH reset is performed. Based on | |
395 | existing conditions, severity of the problem, and hardware capabilities, | |
396 | </para> | |
397 | ||
398 | <para> | |
399 | Either 'softreset' (may be NULL) or 'hardreset' (may be NULL) will be | |
400 | called to perform the low-level EH reset. | |
401 | </para> | |
402 | ||
403 | <programlisting> | |
404 | void (*post_internal_cmd) (struct ata_queued_cmd *qc); | |
1da177e4 LT |
405 | </programlisting> |
406 | ||
407 | <para> | |
bf717b11 JG |
408 | Perform any hardware-specific actions necessary to finish processing |
409 | after executing a probe-time or EH-time command via ata_exec_internal(). | |
1da177e4 LT |
410 | </para> |
411 | ||
92bab26b JG |
412 | </sect2> |
413 | ||
414 | <sect2><title>Hardware interrupt handling</title> | |
1da177e4 LT |
415 | <programlisting> |
416 | irqreturn_t (*irq_handler)(int, void *, struct pt_regs *); | |
417 | void (*irq_clear) (struct ata_port *); | |
418 | </programlisting> | |
419 | ||
420 | <para> | |
421 | ->irq_handler is the interrupt handling routine registered with | |
422 | the system, by libata. ->irq_clear is called during probe just | |
423 | before the interrupt handler is registered, to be sure hardware | |
424 | is quiet. | |
425 | </para> | |
8b2af8f0 EF |
426 | <para> |
427 | The second argument, dev_instance, should be cast to a pointer | |
428 | to struct ata_host_set. | |
429 | </para> | |
430 | <para> | |
40868c85 | 431 | Most legacy IDE drivers use ata_sff_interrupt() for the |
8b2af8f0 EF |
432 | irq_handler hook, which scans all ports in the host_set, |
433 | determines which queued command was active (if any), and calls | |
40868c85 | 434 | ata_sff_host_intr(ap,qc). |
8b2af8f0 EF |
435 | </para> |
436 | <para> | |
40868c85 | 437 | Most legacy IDE drivers use ata_sff_irq_clear() for the |
8b2af8f0 EF |
438 | irq_clear() hook, which simply clears the interrupt and error |
439 | flags in the DMA status register. | |
440 | </para> | |
1da177e4 | 441 | |
92bab26b JG |
442 | </sect2> |
443 | ||
444 | <sect2><title>SATA phy read/write</title> | |
1da177e4 | 445 | <programlisting> |
110dd8f1 JB |
446 | int (*scr_read) (struct ata_port *ap, unsigned int sc_reg, |
447 | u32 *val); | |
448 | int (*scr_write) (struct ata_port *ap, unsigned int sc_reg, | |
1da177e4 LT |
449 | u32 val); |
450 | </programlisting> | |
451 | ||
452 | <para> | |
453 | Read and write standard SATA phy registers. Currently only used | |
454 | if ->phy_reset hook called the sata_phy_reset() helper function. | |
8b2af8f0 | 455 | sc_reg is one of SCR_STATUS, SCR_CONTROL, SCR_ERROR, or SCR_ACTIVE. |
1da177e4 LT |
456 | </para> |
457 | ||
92bab26b JG |
458 | </sect2> |
459 | ||
460 | <sect2><title>Init and shutdown</title> | |
1da177e4 LT |
461 | <programlisting> |
462 | int (*port_start) (struct ata_port *ap); | |
463 | void (*port_stop) (struct ata_port *ap); | |
464 | void (*host_stop) (struct ata_host_set *host_set); | |
465 | </programlisting> | |
466 | ||
467 | <para> | |
468 | ->port_start() is called just after the data structures for each | |
469 | port are initialized. Typically this is used to alloc per-port | |
470 | DMA buffers / tables / rings, enable DMA engines, and similar | |
8b2af8f0 EF |
471 | tasks. Some drivers also use this entry point as a chance to |
472 | allocate driver-private memory for ap->private_data. | |
473 | </para> | |
474 | <para> | |
475 | Many drivers use ata_port_start() as this hook or call | |
476 | it from their own port_start() hooks. ata_port_start() | |
477 | allocates space for a legacy IDE PRD table and returns. | |
1da177e4 LT |
478 | </para> |
479 | <para> | |
1da177e4 LT |
480 | ->port_stop() is called after ->host_stop(). It's sole function |
481 | is to release DMA/memory resources, now that they are no longer | |
8b2af8f0 EF |
482 | actively being used. Many drivers also free driver-private |
483 | data from port at this time. | |
484 | </para> | |
485 | <para> | |
780a87f7 JG |
486 | ->host_stop() is called after all ->port_stop() calls |
487 | have completed. The hook must finalize hardware shutdown, release DMA | |
488 | and other resources, etc. | |
8b2af8f0 | 489 | This hook may be specified as NULL, in which case it is not called. |
780a87f7 | 490 | </para> |
1da177e4 | 491 | |
92bab26b JG |
492 | </sect2> |
493 | ||
1da177e4 | 494 | </sect1> |
a1213499 JG |
495 | </chapter> |
496 | ||
497 | <chapter id="libataEH"> | |
bfd00722 TH |
498 | <title>Error handling</title> |
499 | ||
500 | <para> | |
501 | This chapter describes how errors are handled under libata. | |
502 | Readers are advised to read SCSI EH | |
503 | (Documentation/scsi/scsi_eh.txt) and ATA exceptions doc first. | |
504 | </para> | |
505 | ||
a1213499 | 506 | <sect1><title>Origins of commands</title> |
bfd00722 TH |
507 | <para> |
508 | In libata, a command is represented with struct ata_queued_cmd | |
509 | or qc. qc's are preallocated during port initialization and | |
510 | repetitively used for command executions. Currently only one | |
511 | qc is allocated per port but yet-to-be-merged NCQ branch | |
512 | allocates one for each tag and maps each qc to NCQ tag 1-to-1. | |
513 | </para> | |
514 | <para> | |
515 | libata commands can originate from two sources - libata itself | |
516 | and SCSI midlayer. libata internal commands are used for | |
517 | initialization and error handling. All normal blk requests | |
518 | and commands for SCSI emulation are passed as SCSI commands | |
519 | through queuecommand callback of SCSI host template. | |
520 | </para> | |
a1213499 | 521 | </sect1> |
bfd00722 | 522 | |
a1213499 | 523 | <sect1><title>How commands are issued</title> |
bfd00722 TH |
524 | |
525 | <variablelist> | |
526 | ||
527 | <varlistentry><term>Internal commands</term> | |
528 | <listitem> | |
529 | <para> | |
530 | First, qc is allocated and initialized using | |
531 | ata_qc_new_init(). Although ata_qc_new_init() doesn't | |
532 | implement any wait or retry mechanism when qc is not | |
533 | available, internal commands are currently issued only during | |
534 | initialization and error recovery, so no other command is | |
535 | active and allocation is guaranteed to succeed. | |
536 | </para> | |
537 | <para> | |
538 | Once allocated qc's taskfile is initialized for the command to | |
539 | be executed. qc currently has two mechanisms to notify | |
540 | completion. One is via qc->complete_fn() callback and the | |
541 | other is completion qc->waiting. qc->complete_fn() callback | |
542 | is the asynchronous path used by normal SCSI translated | |
543 | commands and qc->waiting is the synchronous (issuer sleeps in | |
544 | process context) path used by internal commands. | |
545 | </para> | |
546 | <para> | |
547 | Once initialization is complete, host_set lock is acquired | |
548 | and the qc is issued. | |
549 | </para> | |
550 | </listitem> | |
551 | </varlistentry> | |
552 | ||
553 | <varlistentry><term>SCSI commands</term> | |
554 | <listitem> | |
555 | <para> | |
556 | All libata drivers use ata_scsi_queuecmd() as | |
557 | hostt->queuecommand callback. scmds can either be simulated | |
558 | or translated. No qc is involved in processing a simulated | |
559 | scmd. The result is computed right away and the scmd is | |
560 | completed. | |
561 | </para> | |
562 | <para> | |
563 | For a translated scmd, ata_qc_new_init() is invoked to | |
564 | allocate a qc and the scmd is translated into the qc. SCSI | |
565 | midlayer's completion notification function pointer is stored | |
566 | into qc->scsidone. | |
567 | </para> | |
568 | <para> | |
569 | qc->complete_fn() callback is used for completion | |
570 | notification. ATA commands use ata_scsi_qc_complete() while | |
571 | ATAPI commands use atapi_qc_complete(). Both functions end up | |
572 | calling qc->scsidone to notify upper layer when the qc is | |
573 | finished. After translation is completed, the qc is issued | |
574 | with ata_qc_issue(). | |
575 | </para> | |
576 | <para> | |
577 | Note that SCSI midlayer invokes hostt->queuecommand while | |
578 | holding host_set lock, so all above occur while holding | |
579 | host_set lock. | |
580 | </para> | |
581 | </listitem> | |
582 | </varlistentry> | |
583 | ||
584 | </variablelist> | |
a1213499 | 585 | </sect1> |
bfd00722 | 586 | |
a1213499 | 587 | <sect1><title>How commands are processed</title> |
bfd00722 TH |
588 | <para> |
589 | Depending on which protocol and which controller are used, | |
590 | commands are processed differently. For the purpose of | |
591 | discussion, a controller which uses taskfile interface and all | |
592 | standard callbacks is assumed. | |
593 | </para> | |
594 | <para> | |
595 | Currently 6 ATA command protocols are used. They can be | |
596 | sorted into the following four categories according to how | |
597 | they are processed. | |
598 | </para> | |
599 | ||
600 | <variablelist> | |
601 | <varlistentry><term>ATA NO DATA or DMA</term> | |
602 | <listitem> | |
603 | <para> | |
604 | ATA_PROT_NODATA and ATA_PROT_DMA fall into this category. | |
605 | These types of commands don't require any software | |
606 | intervention once issued. Device will raise interrupt on | |
607 | completion. | |
608 | </para> | |
609 | </listitem> | |
610 | </varlistentry> | |
611 | ||
612 | <varlistentry><term>ATA PIO</term> | |
613 | <listitem> | |
614 | <para> | |
615 | ATA_PROT_PIO is in this category. libata currently | |
616 | implements PIO with polling. ATA_NIEN bit is set to turn | |
617 | off interrupt and pio_task on ata_wq performs polling and | |
618 | IO. | |
619 | </para> | |
620 | </listitem> | |
621 | </varlistentry> | |
622 | ||
623 | <varlistentry><term>ATAPI NODATA or DMA</term> | |
624 | <listitem> | |
625 | <para> | |
626 | ATA_PROT_ATAPI_NODATA and ATA_PROT_ATAPI_DMA are in this | |
627 | category. packet_task is used to poll BSY bit after | |
628 | issuing PACKET command. Once BSY is turned off by the | |
629 | device, packet_task transfers CDB and hands off processing | |
630 | to interrupt handler. | |
631 | </para> | |
632 | </listitem> | |
633 | </varlistentry> | |
634 | ||
635 | <varlistentry><term>ATAPI PIO</term> | |
636 | <listitem> | |
637 | <para> | |
638 | ATA_PROT_ATAPI is in this category. ATA_NIEN bit is set | |
639 | and, as in ATAPI NODATA or DMA, packet_task submits cdb. | |
640 | However, after submitting cdb, further processing (data | |
641 | transfer) is handed off to pio_task. | |
642 | </para> | |
643 | </listitem> | |
644 | </varlistentry> | |
645 | </variablelist> | |
a1213499 | 646 | </sect1> |
bfd00722 | 647 | |
a1213499 | 648 | <sect1><title>How commands are completed</title> |
bfd00722 TH |
649 | <para> |
650 | Once issued, all qc's are either completed with | |
651 | ata_qc_complete() or time out. For commands which are handled | |
652 | by interrupts, ata_host_intr() invokes ata_qc_complete(), and, | |
653 | for PIO tasks, pio_task invokes ata_qc_complete(). In error | |
654 | cases, packet_task may also complete commands. | |
655 | </para> | |
656 | <para> | |
657 | ata_qc_complete() does the following. | |
658 | </para> | |
659 | ||
660 | <orderedlist> | |
661 | ||
662 | <listitem> | |
663 | <para> | |
664 | DMA memory is unmapped. | |
665 | </para> | |
666 | </listitem> | |
667 | ||
668 | <listitem> | |
669 | <para> | |
670 | ATA_QCFLAG_ACTIVE is clared from qc->flags. | |
671 | </para> | |
672 | </listitem> | |
673 | ||
674 | <listitem> | |
675 | <para> | |
676 | qc->complete_fn() callback is invoked. If the return value of | |
677 | the callback is not zero. Completion is short circuited and | |
678 | ata_qc_complete() returns. | |
679 | </para> | |
680 | </listitem> | |
681 | ||
682 | <listitem> | |
683 | <para> | |
684 | __ata_qc_complete() is called, which does | |
685 | <orderedlist> | |
686 | ||
687 | <listitem> | |
688 | <para> | |
689 | qc->flags is cleared to zero. | |
690 | </para> | |
691 | </listitem> | |
692 | ||
693 | <listitem> | |
694 | <para> | |
695 | ap->active_tag and qc->tag are poisoned. | |
696 | </para> | |
697 | </listitem> | |
698 | ||
699 | <listitem> | |
700 | <para> | |
701 | qc->waiting is claread & completed (in that order). | |
702 | </para> | |
703 | </listitem> | |
704 | ||
705 | <listitem> | |
706 | <para> | |
707 | qc is deallocated by clearing appropriate bit in ap->qactive. | |
708 | </para> | |
709 | </listitem> | |
710 | ||
711 | </orderedlist> | |
712 | </para> | |
713 | </listitem> | |
714 | ||
715 | </orderedlist> | |
716 | ||
717 | <para> | |
718 | So, it basically notifies upper layer and deallocates qc. One | |
719 | exception is short-circuit path in #3 which is used by | |
720 | atapi_qc_complete(). | |
721 | </para> | |
722 | <para> | |
723 | For all non-ATAPI commands, whether it fails or not, almost | |
724 | the same code path is taken and very little error handling | |
725 | takes place. A qc is completed with success status if it | |
726 | succeeded, with failed status otherwise. | |
727 | </para> | |
728 | <para> | |
729 | However, failed ATAPI commands require more handling as | |
730 | REQUEST SENSE is needed to acquire sense data. If an ATAPI | |
731 | command fails, ata_qc_complete() is invoked with error status, | |
732 | which in turn invokes atapi_qc_complete() via | |
733 | qc->complete_fn() callback. | |
734 | </para> | |
735 | <para> | |
736 | This makes atapi_qc_complete() set scmd->result to | |
737 | SAM_STAT_CHECK_CONDITION, complete the scmd and return 1. As | |
738 | the sense data is empty but scmd->result is CHECK CONDITION, | |
739 | SCSI midlayer will invoke EH for the scmd, and returning 1 | |
740 | makes ata_qc_complete() to return without deallocating the qc. | |
741 | This leads us to ata_scsi_error() with partially completed qc. | |
742 | </para> | |
743 | ||
a1213499 | 744 | </sect1> |
bfd00722 | 745 | |
a1213499 | 746 | <sect1><title>ata_scsi_error()</title> |
bfd00722 | 747 | <para> |
9227c33d | 748 | ata_scsi_error() is the current transportt->eh_strategy_handler() |
bfd00722 TH |
749 | for libata. As discussed above, this will be entered in two |
750 | cases - timeout and ATAPI error completion. This function | |
751 | calls low level libata driver's eng_timeout() callback, the | |
752 | standard callback for which is ata_eng_timeout(). It checks | |
753 | if a qc is active and calls ata_qc_timeout() on the qc if so. | |
754 | Actual error handling occurs in ata_qc_timeout(). | |
755 | </para> | |
756 | <para> | |
757 | If EH is invoked for timeout, ata_qc_timeout() stops BMDMA and | |
758 | completes the qc. Note that as we're currently in EH, we | |
759 | cannot call scsi_done. As described in SCSI EH doc, a | |
760 | recovered scmd should be either retried with | |
761 | scsi_queue_insert() or finished with scsi_finish_command(). | |
762 | Here, we override qc->scsidone with scsi_finish_command() and | |
763 | calls ata_qc_complete(). | |
764 | </para> | |
765 | <para> | |
766 | If EH is invoked due to a failed ATAPI qc, the qc here is | |
767 | completed but not deallocated. The purpose of this | |
768 | half-completion is to use the qc as place holder to make EH | |
769 | code reach this place. This is a bit hackish, but it works. | |
770 | </para> | |
771 | <para> | |
772 | Once control reaches here, the qc is deallocated by invoking | |
773 | __ata_qc_complete() explicitly. Then, internal qc for REQUEST | |
774 | SENSE is issued. Once sense data is acquired, scmd is | |
775 | finished by directly invoking scsi_finish_command() on the | |
776 | scmd. Note that as we already have completed and deallocated | |
777 | the qc which was associated with the scmd, we don't need | |
778 | to/cannot call ata_qc_complete() again. | |
779 | </para> | |
780 | ||
a1213499 | 781 | </sect1> |
bfd00722 | 782 | |
a1213499 | 783 | <sect1><title>Problems with the current EH</title> |
bfd00722 TH |
784 | |
785 | <itemizedlist> | |
786 | ||
787 | <listitem> | |
788 | <para> | |
789 | Error representation is too crude. Currently any and all | |
790 | error conditions are represented with ATA STATUS and ERROR | |
791 | registers. Errors which aren't ATA device errors are treated | |
792 | as ATA device errors by setting ATA_ERR bit. Better error | |
793 | descriptor which can properly represent ATA and other | |
794 | errors/exceptions is needed. | |
795 | </para> | |
796 | </listitem> | |
797 | ||
798 | <listitem> | |
799 | <para> | |
800 | When handling timeouts, no action is taken to make device | |
801 | forget about the timed out command and ready for new commands. | |
802 | </para> | |
803 | </listitem> | |
804 | ||
805 | <listitem> | |
806 | <para> | |
807 | EH handling via ata_scsi_error() is not properly protected | |
808 | from usual command processing. On EH entrance, the device is | |
809 | not in quiescent state. Timed out commands may succeed or | |
810 | fail any time. pio_task and atapi_task may still be running. | |
811 | </para> | |
812 | </listitem> | |
813 | ||
814 | <listitem> | |
815 | <para> | |
816 | Too weak error recovery. Devices / controllers causing HSM | |
817 | mismatch errors and other errors quite often require reset to | |
818 | return to known state. Also, advanced error handling is | |
819 | necessary to support features like NCQ and hotplug. | |
820 | </para> | |
821 | </listitem> | |
822 | ||
823 | <listitem> | |
824 | <para> | |
825 | ATA errors are directly handled in the interrupt handler and | |
826 | PIO errors in pio_task. This is problematic for advanced | |
827 | error handling for the following reasons. | |
828 | </para> | |
829 | <para> | |
830 | First, advanced error handling often requires context and | |
831 | internal qc execution. | |
832 | </para> | |
833 | <para> | |
834 | Second, even a simple failure (say, CRC error) needs | |
835 | information gathering and could trigger complex error handling | |
836 | (say, resetting & reconfiguring). Having multiple code | |
837 | paths to gather information, enter EH and trigger actions | |
838 | makes life painful. | |
839 | </para> | |
840 | <para> | |
841 | Third, scattered EH code makes implementing low level drivers | |
842 | difficult. Low level drivers override libata callbacks. If | |
843 | EH is scattered over several places, each affected callbacks | |
844 | should perform its part of error handling. This can be error | |
845 | prone and painful. | |
846 | </para> | |
847 | </listitem> | |
848 | ||
849 | </itemizedlist> | |
a1213499 | 850 | </sect1> |
1da177e4 LT |
851 | </chapter> |
852 | ||
853 | <chapter id="libataExt"> | |
854 | <title>libata Library</title> | |
58707cce | 855 | !Edrivers/ata/libata-core.c |
1da177e4 LT |
856 | </chapter> |
857 | ||
858 | <chapter id="libataInt"> | |
859 | <title>libata Core Internals</title> | |
58707cce | 860 | !Idrivers/ata/libata-core.c |
1da177e4 LT |
861 | </chapter> |
862 | ||
863 | <chapter id="libataScsiInt"> | |
864 | <title>libata SCSI translation/emulation</title> | |
58707cce HK |
865 | !Edrivers/ata/libata-scsi.c |
866 | !Idrivers/ata/libata-scsi.c | |
1da177e4 LT |
867 | </chapter> |
868 | ||
fe998aa7 | 869 | <chapter id="ataExceptions"> |
7db51fbe | 870 | <title>ATA errors and exceptions</title> |
fe998aa7 TH |
871 | |
872 | <para> | |
873 | This chapter tries to identify what error/exception conditions exist | |
874 | for ATA/ATAPI devices and describe how they should be handled in | |
875 | implementation-neutral way. | |
876 | </para> | |
877 | ||
878 | <para> | |
879 | The term 'error' is used to describe conditions where either an | |
880 | explicit error condition is reported from device or a command has | |
881 | timed out. | |
882 | </para> | |
883 | ||
884 | <para> | |
885 | The term 'exception' is either used to describe exceptional | |
886 | conditions which are not errors (say, power or hotplug events), or | |
887 | to describe both errors and non-error exceptional conditions. Where | |
888 | explicit distinction between error and exception is necessary, the | |
889 | term 'non-error exception' is used. | |
890 | </para> | |
891 | ||
892 | <sect1 id="excat"> | |
893 | <title>Exception categories</title> | |
894 | <para> | |
895 | Exceptions are described primarily with respect to legacy | |
896 | taskfile + bus master IDE interface. If a controller provides | |
897 | other better mechanism for error reporting, mapping those into | |
898 | categories described below shouldn't be difficult. | |
899 | </para> | |
900 | ||
901 | <para> | |
902 | In the following sections, two recovery actions - reset and | |
903 | reconfiguring transport - are mentioned. These are described | |
904 | further in <xref linkend="exrec"/>. | |
905 | </para> | |
906 | ||
907 | <sect2 id="excatHSMviolation"> | |
908 | <title>HSM violation</title> | |
909 | <para> | |
910 | This error is indicated when STATUS value doesn't match HSM | |
911 | requirement during issuing or excution any ATA/ATAPI command. | |
912 | </para> | |
913 | ||
914 | <itemizedlist> | |
915 | <title>Examples</title> | |
916 | ||
917 | <listitem> | |
918 | <para> | |
919 | ATA_STATUS doesn't contain !BSY && DRDY && !DRQ while trying | |
920 | to issue a command. | |
921 | </para> | |
922 | </listitem> | |
923 | ||
924 | <listitem> | |
925 | <para> | |
926 | !BSY && !DRQ during PIO data transfer. | |
927 | </para> | |
928 | </listitem> | |
929 | ||
930 | <listitem> | |
931 | <para> | |
932 | DRQ on command completion. | |
933 | </para> | |
934 | </listitem> | |
935 | ||
936 | <listitem> | |
937 | <para> | |
938 | !BSY && ERR after CDB tranfer starts but before the | |
939 | last byte of CDB is transferred. ATA/ATAPI standard states | |
940 | that "The device shall not terminate the PACKET command | |
941 | with an error before the last byte of the command packet has | |
942 | been written" in the error outputs description of PACKET | |
943 | command and the state diagram doesn't include such | |
944 | transitions. | |
945 | </para> | |
946 | </listitem> | |
947 | ||
948 | </itemizedlist> | |
949 | ||
950 | <para> | |
951 | In these cases, HSM is violated and not much information | |
952 | regarding the error can be acquired from STATUS or ERROR | |
953 | register. IOW, this error can be anything - driver bug, | |
954 | faulty device, controller and/or cable. | |
955 | </para> | |
956 | ||
957 | <para> | |
958 | As HSM is violated, reset is necessary to restore known state. | |
959 | Reconfiguring transport for lower speed might be helpful too | |
960 | as transmission errors sometimes cause this kind of errors. | |
961 | </para> | |
962 | </sect2> | |
963 | ||
964 | <sect2 id="excatDevErr"> | |
965 | <title>ATA/ATAPI device error (non-NCQ / non-CHECK CONDITION)</title> | |
966 | ||
967 | <para> | |
968 | These are errors detected and reported by ATA/ATAPI devices | |
969 | indicating device problems. For this type of errors, STATUS | |
970 | and ERROR register values are valid and describe error | |
971 | condition. Note that some of ATA bus errors are detected by | |
972 | ATA/ATAPI devices and reported using the same mechanism as | |
973 | device errors. Those cases are described later in this | |
974 | section. | |
975 | </para> | |
976 | ||
977 | <para> | |
978 | For ATA commands, this type of errors are indicated by !BSY | |
979 | && ERR during command execution and on completion. | |
980 | </para> | |
981 | ||
982 | <para>For ATAPI commands,</para> | |
983 | ||
984 | <itemizedlist> | |
985 | ||
986 | <listitem> | |
987 | <para> | |
988 | !BSY && ERR && ABRT right after issuing PACKET | |
989 | indicates that PACKET command is not supported and falls in | |
990 | this category. | |
991 | </para> | |
992 | </listitem> | |
993 | ||
994 | <listitem> | |
995 | <para> | |
996 | !BSY && ERR(==CHK) && !ABRT after the last | |
997 | byte of CDB is transferred indicates CHECK CONDITION and | |
998 | doesn't fall in this category. | |
999 | </para> | |
1000 | </listitem> | |
1001 | ||
1002 | <listitem> | |
1003 | <para> | |
1004 | !BSY && ERR(==CHK) && ABRT after the last byte | |
1005 | of CDB is transferred *probably* indicates CHECK CONDITION and | |
1006 | doesn't fall in this category. | |
1007 | </para> | |
1008 | </listitem> | |
1009 | ||
1010 | </itemizedlist> | |
1011 | ||
1012 | <para> | |
1013 | Of errors detected as above, the followings are not ATA/ATAPI | |
1014 | device errors but ATA bus errors and should be handled | |
1015 | according to <xref linkend="excatATAbusErr"/>. | |
1016 | </para> | |
1017 | ||
1018 | <variablelist> | |
1019 | ||
1020 | <varlistentry> | |
1021 | <term>CRC error during data transfer</term> | |
1022 | <listitem> | |
1023 | <para> | |
1024 | This is indicated by ICRC bit in the ERROR register and | |
1025 | means that corruption occurred during data transfer. Upto | |
1026 | ATA/ATAPI-7, the standard specifies that this bit is only | |
1027 | applicable to UDMA transfers but ATA/ATAPI-8 draft revision | |
1028 | 1f says that the bit may be applicable to multiword DMA and | |
1029 | PIO. | |
1030 | </para> | |
1031 | </listitem> | |
1032 | </varlistentry> | |
1033 | ||
1034 | <varlistentry> | |
1035 | <term>ABRT error during data transfer or on completion</term> | |
1036 | <listitem> | |
1037 | <para> | |
1038 | Upto ATA/ATAPI-7, the standard specifies that ABRT could be | |
1039 | set on ICRC errors and on cases where a device is not able | |
1040 | to complete a command. Combined with the fact that MWDMA | |
1041 | and PIO transfer errors aren't allowed to use ICRC bit upto | |
1042 | ATA/ATAPI-7, it seems to imply that ABRT bit alone could | |
1043 | indicate tranfer errors. | |
1044 | </para> | |
1045 | <para> | |
1046 | However, ATA/ATAPI-8 draft revision 1f removes the part | |
1047 | that ICRC errors can turn on ABRT. So, this is kind of | |
1048 | gray area. Some heuristics are needed here. | |
1049 | </para> | |
1050 | </listitem> | |
1051 | </varlistentry> | |
1052 | ||
1053 | </variablelist> | |
1054 | ||
1055 | <para> | |
1056 | ATA/ATAPI device errors can be further categorized as follows. | |
1057 | </para> | |
1058 | ||
1059 | <variablelist> | |
1060 | ||
1061 | <varlistentry> | |
1062 | <term>Media errors</term> | |
1063 | <listitem> | |
1064 | <para> | |
1065 | This is indicated by UNC bit in the ERROR register. ATA | |
1066 | devices reports UNC error only after certain number of | |
1067 | retries cannot recover the data, so there's nothing much | |
1068 | else to do other than notifying upper layer. | |
1069 | </para> | |
1070 | <para> | |
1071 | READ and WRITE commands report CHS or LBA of the first | |
1072 | failed sector but ATA/ATAPI standard specifies that the | |
1073 | amount of transferred data on error completion is | |
1074 | indeterminate, so we cannot assume that sectors preceding | |
1075 | the failed sector have been transferred and thus cannot | |
1076 | complete those sectors successfully as SCSI does. | |
1077 | </para> | |
1078 | </listitem> | |
1079 | </varlistentry> | |
1080 | ||
1081 | <varlistentry> | |
1082 | <term>Media changed / media change requested error</term> | |
1083 | <listitem> | |
1084 | <para> | |
1085 | <<TODO: fill here>> | |
1086 | </para> | |
1087 | </listitem> | |
1088 | </varlistentry> | |
1089 | ||
1090 | <varlistentry><term>Address error</term> | |
1091 | <listitem> | |
1092 | <para> | |
1093 | This is indicated by IDNF bit in the ERROR register. | |
1094 | Report to upper layer. | |
1095 | </para> | |
1096 | </listitem> | |
1097 | </varlistentry> | |
1098 | ||
1099 | <varlistentry><term>Other errors</term> | |
1100 | <listitem> | |
1101 | <para> | |
1102 | This can be invalid command or parameter indicated by ABRT | |
1103 | ERROR bit or some other error condition. Note that ABRT | |
1104 | bit can indicate a lot of things including ICRC and Address | |
1105 | errors. Heuristics needed. | |
1106 | </para> | |
1107 | </listitem> | |
1108 | </varlistentry> | |
1109 | ||
1110 | </variablelist> | |
1111 | ||
1112 | <para> | |
1113 | Depending on commands, not all STATUS/ERROR bits are | |
1114 | applicable. These non-applicable bits are marked with | |
1115 | "na" in the output descriptions but upto ATA/ATAPI-7 | |
1116 | no definition of "na" can be found. However, | |
1117 | ATA/ATAPI-8 draft revision 1f describes "N/A" as | |
1118 | follows. | |
1119 | </para> | |
1120 | ||
1121 | <blockquote> | |
1122 | <variablelist> | |
1123 | <varlistentry><term>3.2.3.3a N/A</term> | |
1124 | <listitem> | |
1125 | <para> | |
1126 | A keyword the indicates a field has no defined value in | |
1127 | this standard and should not be checked by the host or | |
1128 | device. N/A fields should be cleared to zero. | |
1129 | </para> | |
1130 | </listitem> | |
1131 | </varlistentry> | |
1132 | </variablelist> | |
1133 | </blockquote> | |
1134 | ||
1135 | <para> | |
1136 | So, it seems reasonable to assume that "na" bits are | |
1137 | cleared to zero by devices and thus need no explicit masking. | |
1138 | </para> | |
1139 | ||
1140 | </sect2> | |
1141 | ||
1142 | <sect2 id="excatATAPIcc"> | |
1143 | <title>ATAPI device CHECK CONDITION</title> | |
1144 | ||
1145 | <para> | |
1146 | ATAPI device CHECK CONDITION error is indicated by set CHK bit | |
1147 | (ERR bit) in the STATUS register after the last byte of CDB is | |
1148 | transferred for a PACKET command. For this kind of errors, | |
1149 | sense data should be acquired to gather information regarding | |
1150 | the errors. REQUEST SENSE packet command should be used to | |
1151 | acquire sense data. | |
1152 | </para> | |
1153 | ||
1154 | <para> | |
1155 | Once sense data is acquired, this type of errors can be | |
1156 | handled similary to other SCSI errors. Note that sense data | |
1157 | may indicate ATA bus error (e.g. Sense Key 04h HARDWARE ERROR | |
1158 | && ASC/ASCQ 47h/00h SCSI PARITY ERROR). In such | |
1159 | cases, the error should be considered as an ATA bus error and | |
1160 | handled according to <xref linkend="excatATAbusErr"/>. | |
1161 | </para> | |
1162 | ||
1163 | </sect2> | |
1164 | ||
1165 | <sect2 id="excatNCQerr"> | |
1166 | <title>ATA device error (NCQ)</title> | |
1167 | ||
1168 | <para> | |
1169 | NCQ command error is indicated by cleared BSY and set ERR bit | |
1170 | during NCQ command phase (one or more NCQ commands | |
1171 | outstanding). Although STATUS and ERROR registers will | |
1172 | contain valid values describing the error, READ LOG EXT is | |
1173 | required to clear the error condition, determine which command | |
1174 | has failed and acquire more information. | |
1175 | </para> | |
1176 | ||
1177 | <para> | |
1178 | READ LOG EXT Log Page 10h reports which tag has failed and | |
1179 | taskfile register values describing the error. With this | |
1180 | information the failed command can be handled as a normal ATA | |
1181 | command error as in <xref linkend="excatDevErr"/> and all | |
1182 | other in-flight commands must be retried. Note that this | |
1183 | retry should not be counted - it's likely that commands | |
1184 | retried this way would have completed normally if it were not | |
1185 | for the failed command. | |
1186 | </para> | |
1187 | ||
1188 | <para> | |
1189 | Note that ATA bus errors can be reported as ATA device NCQ | |
1190 | errors. This should be handled as described in <xref | |
1191 | linkend="excatATAbusErr"/>. | |
1192 | </para> | |
1193 | ||
1194 | <para> | |
1195 | If READ LOG EXT Log Page 10h fails or reports NQ, we're | |
1196 | thoroughly screwed. This condition should be treated | |
1197 | according to <xref linkend="excatHSMviolation"/>. | |
1198 | </para> | |
1199 | ||
1200 | </sect2> | |
1201 | ||
1202 | <sect2 id="excatATAbusErr"> | |
1203 | <title>ATA bus error</title> | |
1204 | ||
1205 | <para> | |
1206 | ATA bus error means that data corruption occurred during | |
1207 | transmission over ATA bus (SATA or PATA). This type of errors | |
1208 | can be indicated by | |
1209 | </para> | |
1210 | ||
1211 | <itemizedlist> | |
1212 | ||
1213 | <listitem> | |
1214 | <para> | |
1215 | ICRC or ABRT error as described in <xref linkend="excatDevErr"/>. | |
1216 | </para> | |
1217 | </listitem> | |
1218 | ||
1219 | <listitem> | |
1220 | <para> | |
1221 | Controller-specific error completion with error information | |
1222 | indicating transmission error. | |
1223 | </para> | |
1224 | </listitem> | |
1225 | ||
1226 | <listitem> | |
1227 | <para> | |
1228 | On some controllers, command timeout. In this case, there may | |
1229 | be a mechanism to determine that the timeout is due to | |
1230 | transmission error. | |
1231 | </para> | |
1232 | </listitem> | |
1233 | ||
1234 | <listitem> | |
1235 | <para> | |
1236 | Unknown/random errors, timeouts and all sorts of weirdities. | |
1237 | </para> | |
1238 | </listitem> | |
1239 | ||
1240 | </itemizedlist> | |
1241 | ||
1242 | <para> | |
1243 | As described above, transmission errors can cause wide variety | |
1244 | of symptoms ranging from device ICRC error to random device | |
1245 | lockup, and, for many cases, there is no way to tell if an | |
1246 | error condition is due to transmission error or not; | |
1247 | therefore, it's necessary to employ some kind of heuristic | |
1248 | when dealing with errors and timeouts. For example, | |
1249 | encountering repetitive ABRT errors for known supported | |
1250 | command is likely to indicate ATA bus error. | |
1251 | </para> | |
1252 | ||
1253 | <para> | |
1254 | Once it's determined that ATA bus errors have possibly | |
1255 | occurred, lowering ATA bus transmission speed is one of | |
1256 | actions which may alleviate the problem. See <xref | |
1257 | linkend="exrecReconf"/> for more information. | |
1258 | </para> | |
1259 | ||
1260 | </sect2> | |
1261 | ||
1262 | <sect2 id="excatPCIbusErr"> | |
1263 | <title>PCI bus error</title> | |
1264 | ||
1265 | <para> | |
1266 | Data corruption or other failures during transmission over PCI | |
1267 | (or other system bus). For standard BMDMA, this is indicated | |
1268 | by Error bit in the BMDMA Status register. This type of | |
1269 | errors must be logged as it indicates something is very wrong | |
1270 | with the system. Resetting host controller is recommended. | |
1271 | </para> | |
1272 | ||
1273 | </sect2> | |
1274 | ||
1275 | <sect2 id="excatLateCompletion"> | |
1276 | <title>Late completion</title> | |
1277 | ||
1278 | <para> | |
1279 | This occurs when timeout occurs and the timeout handler finds | |
1280 | out that the timed out command has completed successfully or | |
1281 | with error. This is usually caused by lost interrupts. This | |
1282 | type of errors must be logged. Resetting host controller is | |
1283 | recommended. | |
1284 | </para> | |
1285 | ||
1286 | </sect2> | |
1287 | ||
1288 | <sect2 id="excatUnknown"> | |
1289 | <title>Unknown error (timeout)</title> | |
1290 | ||
1291 | <para> | |
1292 | This is when timeout occurs and the command is still | |
1293 | processing or the host and device are in unknown state. When | |
1294 | this occurs, HSM could be in any valid or invalid state. To | |
1295 | bring the device to known state and make it forget about the | |
1296 | timed out command, resetting is necessary. The timed out | |
1297 | command may be retried. | |
1298 | </para> | |
1299 | ||
1300 | <para> | |
1301 | Timeouts can also be caused by transmission errors. Refer to | |
1302 | <xref linkend="excatATAbusErr"/> for more details. | |
1303 | </para> | |
1304 | ||
1305 | </sect2> | |
1306 | ||
1307 | <sect2 id="excatHoplugPM"> | |
1308 | <title>Hotplug and power management exceptions</title> | |
1309 | ||
1310 | <para> | |
1311 | <<TODO: fill here>> | |
1312 | </para> | |
1313 | ||
1314 | </sect2> | |
1315 | ||
1316 | </sect1> | |
1317 | ||
1318 | <sect1 id="exrec"> | |
1319 | <title>EH recovery actions</title> | |
1320 | ||
1321 | <para> | |
1322 | This section discusses several important recovery actions. | |
1323 | </para> | |
1324 | ||
1325 | <sect2 id="exrecClr"> | |
1326 | <title>Clearing error condition</title> | |
1327 | ||
1328 | <para> | |
1329 | Many controllers require its error registers to be cleared by | |
1330 | error handler. Different controllers may have different | |
1331 | requirements. | |
1332 | </para> | |
1333 | ||
1334 | <para> | |
1335 | For SATA, it's strongly recommended to clear at least SError | |
1336 | register during error handling. | |
1337 | </para> | |
1338 | </sect2> | |
1339 | ||
1340 | <sect2 id="exrecRst"> | |
1341 | <title>Reset</title> | |
1342 | ||
1343 | <para> | |
1344 | During EH, resetting is necessary in the following cases. | |
1345 | </para> | |
1346 | ||
1347 | <itemizedlist> | |
1348 | ||
1349 | <listitem> | |
1350 | <para> | |
1351 | HSM is in unknown or invalid state | |
1352 | </para> | |
1353 | </listitem> | |
1354 | ||
1355 | <listitem> | |
1356 | <para> | |
1357 | HBA is in unknown or invalid state | |
1358 | </para> | |
1359 | </listitem> | |
1360 | ||
1361 | <listitem> | |
1362 | <para> | |
1363 | EH needs to make HBA/device forget about in-flight commands | |
1364 | </para> | |
1365 | </listitem> | |
1366 | ||
1367 | <listitem> | |
1368 | <para> | |
1369 | HBA/device behaves weirdly | |
1370 | </para> | |
1371 | </listitem> | |
1372 | ||
1373 | </itemizedlist> | |
1374 | ||
1375 | <para> | |
1376 | Resetting during EH might be a good idea regardless of error | |
1377 | condition to improve EH robustness. Whether to reset both or | |
1378 | either one of HBA and device depends on situation but the | |
1379 | following scheme is recommended. | |
1380 | </para> | |
1381 | ||
1382 | <itemizedlist> | |
1383 | ||
1384 | <listitem> | |
1385 | <para> | |
1386 | When it's known that HBA is in ready state but ATA/ATAPI | |
670e9f34 | 1387 | device is in unknown state, reset only device. |
fe998aa7 TH |
1388 | </para> |
1389 | </listitem> | |
1390 | ||
1391 | <listitem> | |
1392 | <para> | |
1393 | If HBA is in unknown state, reset both HBA and device. | |
1394 | </para> | |
1395 | </listitem> | |
1396 | ||
1397 | </itemizedlist> | |
1398 | ||
1399 | <para> | |
1400 | HBA resetting is implementation specific. For a controller | |
1401 | complying to taskfile/BMDMA PCI IDE, stopping active DMA | |
1402 | transaction may be sufficient iff BMDMA state is the only HBA | |
1403 | context. But even mostly taskfile/BMDMA PCI IDE complying | |
1404 | controllers may have implementation specific requirements and | |
1405 | mechanism to reset themselves. This must be addressed by | |
1406 | specific drivers. | |
1407 | </para> | |
1408 | ||
1409 | <para> | |
1410 | OTOH, ATA/ATAPI standard describes in detail ways to reset | |
1411 | ATA/ATAPI devices. | |
1412 | </para> | |
1413 | ||
1414 | <variablelist> | |
1415 | ||
1416 | <varlistentry><term>PATA hardware reset</term> | |
1417 | <listitem> | |
1418 | <para> | |
1419 | This is hardware initiated device reset signalled with | |
1420 | asserted PATA RESET- signal. There is no standard way to | |
1421 | initiate hardware reset from software although some | |
1422 | hardware provides registers that allow driver to directly | |
1423 | tweak the RESET- signal. | |
1424 | </para> | |
1425 | </listitem> | |
1426 | </varlistentry> | |
1427 | ||
1428 | <varlistentry><term>Software reset</term> | |
1429 | <listitem> | |
1430 | <para> | |
1431 | This is achieved by turning CONTROL SRST bit on for at | |
1432 | least 5us. Both PATA and SATA support it but, in case of | |
1433 | SATA, this may require controller-specific support as the | |
1434 | second Register FIS to clear SRST should be transmitted | |
1435 | while BSY bit is still set. Note that on PATA, this resets | |
1436 | both master and slave devices on a channel. | |
1437 | </para> | |
1438 | </listitem> | |
1439 | </varlistentry> | |
1440 | ||
1441 | <varlistentry><term>EXECUTE DEVICE DIAGNOSTIC command</term> | |
1442 | <listitem> | |
1443 | <para> | |
1444 | Although ATA/ATAPI standard doesn't describe exactly, EDD | |
1445 | implies some level of resetting, possibly similar level | |
1446 | with software reset. Host-side EDD protocol can be handled | |
1447 | with normal command processing and most SATA controllers | |
1448 | should be able to handle EDD's just like other commands. | |
1449 | As in software reset, EDD affects both devices on a PATA | |
1450 | bus. | |
1451 | </para> | |
1452 | <para> | |
1453 | Although EDD does reset devices, this doesn't suit error | |
1454 | handling as EDD cannot be issued while BSY is set and it's | |
1455 | unclear how it will act when device is in unknown/weird | |
1456 | state. | |
1457 | </para> | |
1458 | </listitem> | |
1459 | </varlistentry> | |
1460 | ||
1461 | <varlistentry><term>ATAPI DEVICE RESET command</term> | |
1462 | <listitem> | |
1463 | <para> | |
1464 | This is very similar to software reset except that reset | |
1465 | can be restricted to the selected device without affecting | |
1466 | the other device sharing the cable. | |
1467 | </para> | |
1468 | </listitem> | |
1469 | </varlistentry> | |
1470 | ||
1471 | <varlistentry><term>SATA phy reset</term> | |
1472 | <listitem> | |
1473 | <para> | |
1474 | This is the preferred way of resetting a SATA device. In | |
1475 | effect, it's identical to PATA hardware reset. Note that | |
1476 | this can be done with the standard SCR Control register. | |
1477 | As such, it's usually easier to implement than software | |
1478 | reset. | |
1479 | </para> | |
1480 | </listitem> | |
1481 | </varlistentry> | |
1482 | ||
1483 | </variablelist> | |
1484 | ||
1485 | <para> | |
1486 | One more thing to consider when resetting devices is that | |
1487 | resetting clears certain configuration parameters and they | |
1488 | need to be set to their previous or newly adjusted values | |
1489 | after reset. | |
1490 | </para> | |
1491 | ||
1492 | <para> | |
1493 | Parameters affected are. | |
1494 | </para> | |
1495 | ||
1496 | <itemizedlist> | |
1497 | ||
1498 | <listitem> | |
1499 | <para> | |
1500 | CHS set up with INITIALIZE DEVICE PARAMETERS (seldomly used) | |
1501 | </para> | |
1502 | </listitem> | |
1503 | ||
1504 | <listitem> | |
1505 | <para> | |
1506 | Parameters set with SET FEATURES including transfer mode setting | |
1507 | </para> | |
1508 | </listitem> | |
1509 | ||
1510 | <listitem> | |
1511 | <para> | |
1512 | Block count set with SET MULTIPLE MODE | |
1513 | </para> | |
1514 | </listitem> | |
1515 | ||
1516 | <listitem> | |
1517 | <para> | |
1518 | Other parameters (SET MAX, MEDIA LOCK...) | |
1519 | </para> | |
1520 | </listitem> | |
1521 | ||
1522 | </itemizedlist> | |
1523 | ||
1524 | <para> | |
1525 | ATA/ATAPI standard specifies that some parameters must be | |
1526 | maintained across hardware or software reset, but doesn't | |
1527 | strictly specify all of them. Always reconfiguring needed | |
1528 | parameters after reset is required for robustness. Note that | |
1529 | this also applies when resuming from deep sleep (power-off). | |
1530 | </para> | |
1531 | ||
1532 | <para> | |
1533 | Also, ATA/ATAPI standard requires that IDENTIFY DEVICE / | |
1534 | IDENTIFY PACKET DEVICE is issued after any configuration | |
1535 | parameter is updated or a hardware reset and the result used | |
1536 | for further operation. OS driver is required to implement | |
1537 | revalidation mechanism to support this. | |
1538 | </para> | |
1539 | ||
1540 | </sect2> | |
1541 | ||
1542 | <sect2 id="exrecReconf"> | |
1543 | <title>Reconfigure transport</title> | |
1544 | ||
1545 | <para> | |
1546 | For both PATA and SATA, a lot of corners are cut for cheap | |
1547 | connectors, cables or controllers and it's quite common to see | |
1548 | high transmission error rate. This can be mitigated by | |
1549 | lowering transmission speed. | |
1550 | </para> | |
1551 | ||
1552 | <para> | |
1553 | The following is a possible scheme Jeff Garzik suggested. | |
1554 | </para> | |
1555 | ||
1556 | <blockquote> | |
1557 | <para> | |
1558 | If more than $N (3?) transmission errors happen in 15 minutes, | |
1559 | </para> | |
1560 | <itemizedlist> | |
1561 | <listitem> | |
1562 | <para> | |
1563 | if SATA, decrease SATA PHY speed. if speed cannot be decreased, | |
1564 | </para> | |
1565 | </listitem> | |
1566 | <listitem> | |
1567 | <para> | |
1568 | decrease UDMA xfer speed. if at UDMA0, switch to PIO4, | |
1569 | </para> | |
1570 | </listitem> | |
1571 | <listitem> | |
1572 | <para> | |
1573 | decrease PIO xfer speed. if at PIO3, complain, but continue | |
1574 | </para> | |
1575 | </listitem> | |
1576 | </itemizedlist> | |
1577 | </blockquote> | |
1578 | ||
1579 | </sect2> | |
1580 | ||
1581 | </sect1> | |
1582 | ||
1583 | </chapter> | |
1584 | ||
1da177e4 LT |
1585 | <chapter id="PiixInt"> |
1586 | <title>ata_piix Internals</title> | |
58707cce | 1587 | !Idrivers/ata/ata_piix.c |
1da177e4 LT |
1588 | </chapter> |
1589 | ||
1590 | <chapter id="SILInt"> | |
1591 | <title>sata_sil Internals</title> | |
58707cce | 1592 | !Idrivers/ata/sata_sil.c |
1da177e4 LT |
1593 | </chapter> |
1594 | ||
0cba632b JG |
1595 | <chapter id="libataThanks"> |
1596 | <title>Thanks</title> | |
1597 | <para> | |
1598 | The bulk of the ATA knowledge comes thanks to long conversations with | |
1599 | Andre Hedrick (www.linux-ide.org), and long hours pondering the ATA | |
1600 | and SCSI specifications. | |
1601 | </para> | |
1602 | <para> | |
1603 | Thanks to Alan Cox for pointing out similarities | |
1604 | between SATA and SCSI, and in general for motivation to hack on | |
1605 | libata. | |
1606 | </para> | |
1607 | <para> | |
1608 | libata's device detection | |
1609 | method, ata_pio_devchk, and in general all the early probing was | |
1610 | based on extensive study of Hale Landis's probe/reset code in his | |
1611 | ATADRVR driver (www.ata-atapi.com). | |
1612 | </para> | |
1613 | </chapter> | |
1614 | ||
1da177e4 | 1615 | </book> |