release/src-rt-6.x/linux/linux-2.6/Documentation/DocBook/libata.tmpl

   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>
  17    <year>2003-2006</year>
  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
  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&lt;-&gt;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
  61   <chapter id="libataDriverApi">
  62      <title>libata Driver API</title>
  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>
  75      <sect1>
  76         <title>struct ata_port_operations</title>
  77
  78         <sect2><title>Disable ATA port</title>
  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).
  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().
  95         </para>
  96
  97         </sect2>
  98
  99         <sect2><title>Post-IDENTIFY device configuration</title>
 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>
 109         <para>
 110         Called by ata_device_add() after ata_dev_identify() determines
 111         a device is present.
 112         </para>
 113         <para>
 114         This entry may be specified as NULL in ata_port_operations.
 115         </para>
 116
 117         </sect2>
 118
 119         <sect2><title>Set PIO/DMA mode</title>
 120         <programlisting>
 121 void (*set_piomode) (struct ata_port *, struct ata_device *);
 122 void (*set_dmamode) (struct ata_port *, struct ata_device *);
 123 void (*post_set_mode) (struct ata_port *);
 124 unsigned int (*mode_filter) (struct ata_port *, struct ata_device *, unsigned int);
 125         </programlisting>
 126
 127         <para>
 128         Hooks called prior to the issue of SET FEATURES - XFER MODE
 129         command.  The optional ->mode_filter() hook is called when libata
 130         has built a mask of the possible modes. This is passed to the
 131         ->mode_filter() function which should return a mask of valid modes
 132         after filtering those unsuitable due to hardware limits. It is not
 133         valid to use this interface to add modes.
 134         </para>
 135         <para>
 136         dev->pio_mode and dev->dma_mode are guaranteed to be valid when
 137         ->set_piomode() and when ->set_dmamode() is called. The timings for
 138         any other drive sharing the cable will also be valid at this point.
 139         That is the library records the decisions for the modes of each
 140         drive on a channel before it attempts to set any of them.
 141         </para>
 142         <para>
 143         ->post_set_mode() is
 144         called unconditionally, after the SET FEATURES - XFER MODE
 145         command completes successfully.
 146         </para>
 147
 148         <para>
 149         ->set_piomode() is always called (if present), but
 150         ->set_dma_mode() is only called if DMA is possible.
 151         </para>
 152
 153         </sect2>
 154
 155         <sect2><title>Taskfile read/write</title>
 156         <programlisting>
 157 void (*tf_load) (struct ata_port *ap, struct ata_taskfile *tf);
 158 void (*tf_read) (struct ata_port *ap, struct ata_taskfile *tf);
 159         </programlisting>
 160
 161         <para>
 162         ->tf_load() is called to load the given taskfile into hardware
 163         registers / DMA buffers.  ->tf_read() is called to read the
 164         hardware registers / DMA buffers, to obtain the current set of
 165         taskfile register values.
 166         Most drivers for taskfile-based hardware (PIO or MMIO) use
 167         ata_tf_load() and ata_tf_read() for these hooks.
 168         </para>
 169
 170         </sect2>
 171
 172         <sect2><title>PIO data read/write</title>
 173         <programlisting>
 174 void (*data_xfer) (struct ata_device *, unsigned char *, unsigned int, int);
 175         </programlisting>
 176
 177         <para>
 178 All bmdma-style drivers must implement this hook.  This is the low-level
 179 operation that actually copies the data bytes during a PIO data
 180 transfer.
 181 Typically the driver
 182 will choose one of ata_pio_data_xfer_noirq(), ata_pio_data_xfer(), or
 183 ata_mmio_data_xfer().
 184         </para>
 185
 186         </sect2>
 187
 188         <sect2><title>ATA command execute</title>
 189         <programlisting>
 190 void (*exec_command)(struct ata_port *ap, struct ata_taskfile *tf);
 191         </programlisting>
 192
 193         <para>
 194         causes an ATA command, previously loaded with
 195         ->tf_load(), to be initiated in hardware.
 196         Most drivers for taskfile-based hardware use ata_exec_command()
 197         for this hook.
 198         </para>
 199
 200         </sect2>
 201
 202         <sect2><title>Per-cmd ATAPI DMA capabilities filter</title>
 203         <programlisting>
 204 int (*check_atapi_dma) (struct ata_queued_cmd *qc);
 205         </programlisting>
 206
 207         <para>
 208 Allow low-level driver to filter ATA PACKET commands, returning a status
 209 indicating whether or not it is OK to use DMA for the supplied PACKET
 210 command.
 211         </para>
 212         <para>
 213         This hook may be specified as NULL, in which case libata will
 214         assume that atapi dma can be supported.
 215         </para>
 216
 217         </sect2>
 218
 219         <sect2><title>Read specific ATA shadow registers</title>
 220         <programlisting>
 221 u8   (*check_status)(struct ata_port *ap);
 222 u8   (*check_altstatus)(struct ata_port *ap);
 223         </programlisting>
 224
 225         <para>
 226         Reads the Status/AltStatus ATA shadow register from
 227         hardware.  On some hardware, reading the Status register has
 228         the side effect of clearing the interrupt condition.
 229         Most drivers for taskfile-based hardware use
 230         ata_check_status() for this hook.
 231         </para>
 232         <para>
 233         Note that because this is called from ata_device_add(), at
 234         least a dummy function that clears device interrupts must be
 235         provided for all drivers, even if the controller doesn't
 236         actually have a taskfile status register.
 237         </para>
 238
 239         </sect2>
 240
 241         <sect2><title>Select ATA device on bus</title>
 242         <programlisting>
 243 void (*dev_select)(struct ata_port *ap, unsigned int device);
 244         </programlisting>
 245
 246         <para>
 247         Issues the low-level hardware command(s) that causes one of N
 248         hardware devices to be considered 'selected' (active and
 249         available for use) on the ATA bus.  This generally has no
 250         meaning on FIS-based devices.
 251         </para>
 252         <para>
 253         Most drivers for taskfile-based hardware use
 254         ata_std_dev_select() for this hook.  Controllers which do not
 255         support second drives on a port (such as SATA contollers) will
 256         use ata_noop_dev_select().
 257         </para>
 258
 259         </sect2>
 260
 261         <sect2><title>Private tuning method</title>
 262         <programlisting>
 263 void (*set_mode) (struct ata_port *ap);
 264         </programlisting>
 265
 266         <para>
 267         By default libata performs drive and controller tuning in
 268         accordance with the ATA timing rules and also applies blacklists
 269         and cable limits. Some controllers need special handling and have
 270         custom tuning rules, typically raid controllers that use ATA
 271         commands but do not actually do drive timing.
 272         </para>
 273
 274         <warning>
 275         <para>
 276         This hook should not be used to replace the standard controller
 277         tuning logic when a controller has quirks. Replacing the default
 278         tuning logic in that case would bypass handling for drive and
 279         bridge quirks that may be important to data reliability. If a
 280         controller needs to filter the mode selection it should use the
 281         mode_filter hook instead.
 282         </para>
 283         </warning>
 284
 285         </sect2>
 286
 287         <sect2><title>Control PCI IDE BMDMA engine</title>
 288         <programlisting>
 289 void (*bmdma_setup) (struct ata_queued_cmd *qc);
 290 void (*bmdma_start) (struct ata_queued_cmd *qc);
 291 void (*bmdma_stop) (struct ata_port *ap);
 292 u8   (*bmdma_status) (struct ata_port *ap);
 293         </programlisting>
 294
 295         <para>
 296 When setting up an IDE BMDMA transaction, these hooks arm
 297 (->bmdma_setup), fire (->bmdma_start), and halt (->bmdma_stop)
 298 the hardware's DMA engine.  ->bmdma_status is used to read the standard
 299 PCI IDE DMA Status register.
 300         </para>
 301
 302         <para>
 303 These hooks are typically either no-ops, or simply not implemented, in
 304 FIS-based drivers.
 305         </para>
 306         <para>
 307 Most legacy IDE drivers use ata_bmdma_setup() for the bmdma_setup()
 308 hook.  ata_bmdma_setup() will write the pointer to the PRD table to
 309 the IDE PRD Table Address register, enable DMA in the DMA Command
 310 register, and call exec_command() to begin the transfer.
 311         </para>
 312         <para>
 313 Most legacy IDE drivers use ata_bmdma_start() for the bmdma_start()
 314 hook.  ata_bmdma_start() will write the ATA_DMA_START flag to the DMA
 315 Command register.
 316         </para>
 317         <para>
 318 Many legacy IDE drivers use ata_bmdma_stop() for the bmdma_stop()
 319 hook.  ata_bmdma_stop() clears the ATA_DMA_START flag in the DMA
 320 command register.
 321         </para>
 322         <para>
 323 Many legacy IDE drivers use ata_bmdma_status() as the bmdma_status() hook.
 324         </para>
 325
 326         </sect2>
 327
 328         <sect2><title>High-level taskfile hooks</title>
 329         <programlisting>
 330 void (*qc_prep) (struct ata_queued_cmd *qc);
 331 int (*qc_issue) (struct ata_queued_cmd *qc);
 332         </programlisting>
 333
 334         <para>
 335         Higher-level hooks, these two hooks can potentially supercede
 336         several of the above taskfile/DMA engine hooks.  ->qc_prep is
 337         called after the buffers have been DMA-mapped, and is typically
 338         used to populate the hardware's DMA scatter-gather table.
 339         Most drivers use the standard ata_qc_prep() helper function, but
 340         more advanced drivers roll their own.
 341         </para>
 342         <para>
 343         ->qc_issue is used to make a command active, once the hardware
 344         and S/G tables have been prepared.  IDE BMDMA drivers use the
 345         helper function ata_qc_issue_prot() for taskfile protocol-based
 346         dispatch.  More advanced drivers implement their own ->qc_issue.
 347         </para>
 348         <para>
 349         ata_qc_issue_prot() calls ->tf_load(), ->bmdma_setup(), and
 350         ->bmdma_start() as necessary to initiate a transfer.
 351         </para>
 352
 353         </sect2>
 354
 355         <sect2><title>Exception and probe handling (EH)</title>
 356         <programlisting>
 357 void (*eng_timeout) (struct ata_port *ap);
 358 void (*phy_reset) (struct ata_port *ap);
 359         </programlisting>
 360
 361         <para>
 362 Deprecated.  Use ->error_handler() instead.
 363         </para>
 364
 365         <programlisting>
 366 void (*freeze) (struct ata_port *ap);
 367 void (*thaw) (struct ata_port *ap);
 368         </programlisting>
 369
 370         <para>
 371 ata_port_freeze() is called when HSM violations or some other
 372 condition disrupts normal operation of the port.  A frozen port
 373 is not allowed to perform any operation until the port is
 374 thawed, which usually follows a successful reset.
 375         </para>
 376
 377         <para>
 378 The optional ->freeze() callback can be used for freezing the port
 379 hardware-wise (e.g. mask interrupt and stop DMA engine).  If a
 380 port cannot be frozen hardware-wise, the interrupt handler
 381 must ack and clear interrupts unconditionally while the port
 382 is frozen.
 383         </para>
 384         <para>
 385 The optional ->thaw() callback is called to perform the opposite of ->freeze():
 386 prepare the port for normal operation once again.  Unmask interrupts,
 387 start DMA engine, etc.
 388         </para>
 389
 390         <programlisting>
 391 void (*error_handler) (struct ata_port *ap);
 392         </programlisting>
 393
 394         <para>
 395 ->error_handler() is a driver's hook into probe, hotplug, and recovery
 396 and other exceptional conditions.  The primary responsibility of an
 397 implementation is to call ata_do_eh() or ata_bmdma_drive_eh() with a set
 398 of EH hooks as arguments:
 399         </para>
 400
 401         <para>
 402 'prereset' hook (may be NULL) is called during an EH reset, before any other actions
 403 are taken.
 404         </para>
 405
 406         <para>
 407 'postreset' hook (may be NULL) is called after the EH reset is performed.  Based on
 408 existing conditions, severity of the problem, and hardware capabilities,
 409         </para>
 410
 411         <para>
 412 Either 'softreset' (may be NULL) or 'hardreset' (may be NULL) will be
 413 called to perform the low-level EH reset.
 414         </para>
 415
 416         <programlisting>
 417 void (*post_internal_cmd) (struct ata_queued_cmd *qc);
 418         </programlisting>
 419
 420         <para>
 421 Perform any hardware-specific actions necessary to finish processing
 422 after executing a probe-time or EH-time command via ata_exec_internal().
 423         </para>
 424
 425         </sect2>
 426
 427         <sect2><title>Hardware interrupt handling</title>
 428         <programlisting>
 429 irqreturn_t (*irq_handler)(int, void *, struct pt_regs *);
 430 void (*irq_clear) (struct ata_port *);
 431         </programlisting>
 432
 433         <para>
 434         ->irq_handler is the interrupt handling routine registered with
 435         the system, by libata.  ->irq_clear is called during probe just
 436         before the interrupt handler is registered, to be sure hardware
 437         is quiet.
 438         </para>
 439         <para>
 440         The second argument, dev_instance, should be cast to a pointer
 441         to struct ata_host_set.
 442         </para>
 443         <para>
 444         Most legacy IDE drivers use ata_interrupt() for the
 445         irq_handler hook, which scans all ports in the host_set,
 446         determines which queued command was active (if any), and calls
 447         ata_host_intr(ap,qc).
 448         </para>
 449         <para>
 450         Most legacy IDE drivers use ata_bmdma_irq_clear() for the
 451         irq_clear() hook, which simply clears the interrupt and error
 452         flags in the DMA status register.
 453         </para>
 454
 455         </sect2>
 456
 457         <sect2><title>SATA phy read/write</title>
 458         <programlisting>
 459 u32 (*scr_read) (struct ata_port *ap, unsigned int sc_reg);
 460 void (*scr_write) (struct ata_port *ap, unsigned int sc_reg,
 461                    u32 val);
 462         </programlisting>
 463
 464         <para>
 465         Read and write standard SATA phy registers.  Currently only used
 466         if ->phy_reset hook called the sata_phy_reset() helper function.
 467         sc_reg is one of SCR_STATUS, SCR_CONTROL, SCR_ERROR, or SCR_ACTIVE.
 468         </para>
 469
 470         </sect2>
 471
 472         <sect2><title>Init and shutdown</title>
 473         <programlisting>
 474 int (*port_start) (struct ata_port *ap);
 475 void (*port_stop) (struct ata_port *ap);
 476 void (*host_stop) (struct ata_host_set *host_set);
 477         </programlisting>
 478
 479         <para>
 480         ->port_start() is called just after the data structures for each
 481         port are initialized.  Typically this is used to alloc per-port
 482         DMA buffers / tables / rings, enable DMA engines, and similar
 483         tasks.  Some drivers also use this entry point as a chance to
 484         allocate driver-private memory for ap->private_data.
 485         </para>
 486         <para>
 487         Many drivers use ata_port_start() as this hook or call
 488         it from their own port_start() hooks.  ata_port_start()
 489         allocates space for a legacy IDE PRD table and returns.
 490         </para>
 491         <para>
 492         ->port_stop() is called after ->host_stop().  It's sole function
 493         is to release DMA/memory resources, now that they are no longer
 494         actively being used.  Many drivers also free driver-private
 495         data from port at this time.
 496         </para>
 497         <para>
 498         Many drivers use ata_port_stop() as this hook, which frees the
 499         PRD table.
 500         </para>
 501         <para>
 502         ->host_stop() is called after all ->port_stop() calls
 503 have completed.  The hook must finalize hardware shutdown, release DMA
 504 and other resources, etc.
 505         This hook may be specified as NULL, in which case it is not called.
 506         </para>
 507
 508         </sect2>
 509
 510      </sect1>
 511   </chapter>
 512
 513   <chapter id="libataEH">
 514         <title>Error handling</title>
 515
 516         <para>
 517         This chapter describes how errors are handled under libata.
 518         Readers are advised to read SCSI EH
 519         (Documentation/scsi/scsi_eh.txt) and ATA exceptions doc first.
 520         </para>
 521
 522         <sect1><title>Origins of commands</title>
 523         <para>
 524         In libata, a command is represented with struct ata_queued_cmd
 525         or qc.  qc's are preallocated during port initialization and
 526         repetitively used for command executions.  Currently only one
 527         qc is allocated per port but yet-to-be-merged NCQ branch
 528         allocates one for each tag and maps each qc to NCQ tag 1-to-1.
 529         </para>
 530         <para>
 531         libata commands can originate from two sources - libata itself
 532         and SCSI midlayer.  libata internal commands are used for
 533         initialization and error handling.  All normal blk requests
 534         and commands for SCSI emulation are passed as SCSI commands
 535         through queuecommand callback of SCSI host template.
 536         </para>
 537         </sect1>
 538
 539         <sect1><title>How commands are issued</title>
 540
 541         <variablelist>
 542
 543         <varlistentry><term>Internal commands</term>
 544         <listitem>
 545         <para>
 546         First, qc is allocated and initialized using
 547         ata_qc_new_init().  Although ata_qc_new_init() doesn't
 548         implement any wait or retry mechanism when qc is not
 549         available, internal commands are currently issued only during
 550         initialization and error recovery, so no other command is
 551         active and allocation is guaranteed to succeed.
 552         </para>
 553         <para>
 554         Once allocated qc's taskfile is initialized for the command to
 555         be executed.  qc currently has two mechanisms to notify
 556         completion.  One is via qc->complete_fn() callback and the
 557         other is completion qc->waiting.  qc->complete_fn() callback
 558         is the asynchronous path used by normal SCSI translated
 559         commands and qc->waiting is the synchronous (issuer sleeps in
 560         process context) path used by internal commands.
 561         </para>
 562         <para>
 563         Once initialization is complete, host_set lock is acquired
 564         and the qc is issued.
 565         </para>
 566         </listitem>
 567         </varlistentry>
 568
 569         <varlistentry><term>SCSI commands</term>
 570         <listitem>
 571         <para>
 572         All libata drivers use ata_scsi_queuecmd() as
 573         hostt->queuecommand callback.  scmds can either be simulated
 574         or translated.  No qc is involved in processing a simulated
 575         scmd.  The result is computed right away and the scmd is
 576         completed.
 577         </para>
 578         <para>
 579         For a translated scmd, ata_qc_new_init() is invoked to
 580         allocate a qc and the scmd is translated into the qc.  SCSI
 581         midlayer's completion notification function pointer is stored
 582         into qc->scsidone.
 583         </para>
 584         <para>
 585         qc->complete_fn() callback is used for completion
 586         notification.  ATA commands use ata_scsi_qc_complete() while
 587         ATAPI commands use atapi_qc_complete().  Both functions end up
 588         calling qc->scsidone to notify upper layer when the qc is
 589         finished.  After translation is completed, the qc is issued
 590         with ata_qc_issue().
 591         </para>
 592         <para>
 593         Note that SCSI midlayer invokes hostt->queuecommand while
 594         holding host_set lock, so all above occur while holding
 595         host_set lock.
 596         </para>
 597         </listitem>
 598         </varlistentry>
 599
 600         </variablelist>
 601         </sect1>
 602
 603         <sect1><title>How commands are processed</title>
 604         <para>
 605         Depending on which protocol and which controller are used,
 606         commands are processed differently.  For the purpose of
 607         discussion, a controller which uses taskfile interface and all
 608         standard callbacks is assumed.
 609         </para>
 610         <para>
 611         Currently 6 ATA command protocols are used.  They can be
 612         sorted into the following four categories according to how
 613         they are processed.
 614         </para>
 615
 616         <variablelist>
 617            <varlistentry><term>ATA NO DATA or DMA</term>
 618            <listitem>
 619            <para>
 620            ATA_PROT_NODATA and ATA_PROT_DMA fall into this category.
 621            These types of commands don't require any software
 622            intervention once issued.  Device will raise interrupt on
 623            completion.
 624            </para>
 625            </listitem>
 626            </varlistentry>
 627
 628            <varlistentry><term>ATA PIO</term>
 629            <listitem>
 630            <para>
 631            ATA_PROT_PIO is in this category.  libata currently
 632            implements PIO with polling.  ATA_NIEN bit is set to turn
 633            off interrupt and pio_task on ata_wq performs polling and
 634            IO.
 635            </para>
 636            </listitem>
 637            </varlistentry>
 638
 639            <varlistentry><term>ATAPI NODATA or DMA</term>
 640            <listitem>
 641            <para>
 642            ATA_PROT_ATAPI_NODATA and ATA_PROT_ATAPI_DMA are in this
 643            category.  packet_task is used to poll BSY bit after
 644            issuing PACKET command.  Once BSY is turned off by the
 645            device, packet_task transfers CDB and hands off processing
 646            to interrupt handler.
 647            </para>
 648            </listitem>
 649            </varlistentry>
 650
 651            <varlistentry><term>ATAPI PIO</term>
 652            <listitem>
 653            <para>
 654            ATA_PROT_ATAPI is in this category.  ATA_NIEN bit is set
 655            and, as in ATAPI NODATA or DMA, packet_task submits cdb.
 656            However, after submitting cdb, further processing (data
 657            transfer) is handed off to pio_task.
 658            </para>
 659            </listitem>
 660            </varlistentry>
 661         </variablelist>
 662         </sect1>
 663
 664         <sect1><title>How commands are completed</title>
 665         <para>
 666         Once issued, all qc's are either completed with
 667         ata_qc_complete() or time out.  For commands which are handled
 668         by interrupts, ata_host_intr() invokes ata_qc_complete(), and,
 669         for PIO tasks, pio_task invokes ata_qc_complete().  In error
 670         cases, packet_task may also complete commands.
 671         </para>
 672         <para>
 673         ata_qc_complete() does the following.
 674         </para>
 675
 676         <orderedlist>
 677
 678         <listitem>
 679         <para>
 680         DMA memory is unmapped.
 681         </para>
 682         </listitem>
 683
 684         <listitem>
 685         <para>
 686         ATA_QCFLAG_ACTIVE is clared from qc->flags.
 687         </para>
 688         </listitem>
 689
 690         <listitem>
 691         <para>
 692         qc->complete_fn() callback is invoked.  If the return value of
 693         the callback is not zero.  Completion is short circuited and
 694         ata_qc_complete() returns.
 695         </para>
 696         </listitem>
 697
 698         <listitem>
 699         <para>
 700         __ata_qc_complete() is called, which does
 701            <orderedlist>
 702
 703            <listitem>
 704            <para>
 705            qc->flags is cleared to zero.
 706            </para>
 707            </listitem>
 708
 709            <listitem>
 710            <para>
 711            ap->active_tag and qc->tag are poisoned.
 712            </para>
 713            </listitem>
 714
 715            <listitem>
 716            <para>
 717            qc->waiting is claread &amp; completed (in that order).
 718            </para>
 719            </listitem>
 720
 721            <listitem>
 722            <para>
 723            qc is deallocated by clearing appropriate bit in ap->qactive.
 724            </para>
 725            </listitem>
 726
 727            </orderedlist>
 728         </para>
 729         </listitem>
 730
 731         </orderedlist>
 732
 733         <para>
 734         So, it basically notifies upper layer and deallocates qc.  One
 735         exception is short-circuit path in #3 which is used by
 736         atapi_qc_complete().
 737         </para>
 738         <para>
 739         For all non-ATAPI commands, whether it fails or not, almost
 740         the same code path is taken and very little error handling
 741         takes place.  A qc is completed with success status if it
 742         succeeded, with failed status otherwise.
 743         </para>
 744         <para>
 745         However, failed ATAPI commands require more handling as
 746         REQUEST SENSE is needed to acquire sense data.  If an ATAPI
 747         command fails, ata_qc_complete() is invoked with error status,
 748         which in turn invokes atapi_qc_complete() via
 749         qc->complete_fn() callback.
 750         </para>
 751         <para>
 752         This makes atapi_qc_complete() set scmd->result to
 753         SAM_STAT_CHECK_CONDITION, complete the scmd and return 1.  As
 754         the sense data is empty but scmd->result is CHECK CONDITION,
 755         SCSI midlayer will invoke EH for the scmd, and returning 1
 756         makes ata_qc_complete() to return without deallocating the qc.
 757         This leads us to ata_scsi_error() with partially completed qc.
 758         </para>
 759
 760         </sect1>
 761
 762         <sect1><title>ata_scsi_error()</title>
 763         <para>
 764         ata_scsi_error() is the current transportt->eh_strategy_handler()
 765         for libata.  As discussed above, this will be entered in two
 766         cases - timeout and ATAPI error completion.  This function
 767         calls low level libata driver's eng_timeout() callback, the
 768         standard callback for which is ata_eng_timeout().  It checks
 769         if a qc is active and calls ata_qc_timeout() on the qc if so.
 770         Actual error handling occurs in ata_qc_timeout().
 771         </para>
 772         <para>
 773         If EH is invoked for timeout, ata_qc_timeout() stops BMDMA and
 774         completes the qc.  Note that as we're currently in EH, we
 775         cannot call scsi_done.  As described in SCSI EH doc, a
 776         recovered scmd should be either retried with
 777         scsi_queue_insert() or finished with scsi_finish_command().
 778         Here, we override qc->scsidone with scsi_finish_command() and
 779         calls ata_qc_complete().
 780         </para>
 781         <para>
 782         If EH is invoked due to a failed ATAPI qc, the qc here is
 783         completed but not deallocated.  The purpose of this
 784         half-completion is to use the qc as place holder to make EH
 785         code reach this place.  This is a bit hackish, but it works.
 786         </para>
 787         <para>
 788         Once control reaches here, the qc is deallocated by invoking
 789         __ata_qc_complete() explicitly.  Then, internal qc for REQUEST
 790         SENSE is issued.  Once sense data is acquired, scmd is
 791         finished by directly invoking scsi_finish_command() on the
 792         scmd.  Note that as we already have completed and deallocated
 793         the qc which was associated with the scmd, we don't need
 794         to/cannot call ata_qc_complete() again.
 795         </para>
 796
 797         </sect1>
 798
 799         <sect1><title>Problems with the current EH</title>
 800
 801         <itemizedlist>
 802
 803         <listitem>
 804         <para>
 805         Error representation is too crude.  Currently any and all
 806         error conditions are represented with ATA STATUS and ERROR
 807         registers.  Errors which aren't ATA device errors are treated
 808         as ATA device errors by setting ATA_ERR bit.  Better error
 809         descriptor which can properly represent ATA and other
 810         errors/exceptions is needed.
 811         </para>
 812         </listitem>
 813
 814         <listitem>
 815         <para>
 816         When handling timeouts, no action is taken to make device
 817         forget about the timed out command and ready for new commands.
 818         </para>
 819         </listitem>
 820
 821         <listitem>
 822         <para>
 823         EH handling via ata_scsi_error() is not properly protected
 824         from usual command processing.  On EH entrance, the device is
 825         not in quiescent state.  Timed out commands may succeed or
 826         fail any time.  pio_task and atapi_task may still be running.
 827         </para>
 828         </listitem>
 829
 830         <listitem>
 831         <para>
 832         Too weak error recovery.  Devices / controllers causing HSM
 833         mismatch errors and other errors quite often require reset to
 834         return to known state.  Also, advanced error handling is
 835         necessary to support features like NCQ and hotplug.
 836         </para>
 837         </listitem>
 838
 839         <listitem>
 840         <para>
 841         ATA errors are directly handled in the interrupt handler and
 842         PIO errors in pio_task.  This is problematic for advanced
 843         error handling for the following reasons.
 844         </para>
 845         <para>
 846         First, advanced error handling often requires context and
 847         internal qc execution.
 848         </para>
 849         <para>
 850         Second, even a simple failure (say, CRC error) needs
 851         information gathering and could trigger complex error handling
 852         (say, resetting &amp; reconfiguring).  Having multiple code
 853         paths to gather information, enter EH and trigger actions
 854         makes life painful.
 855         </para>
 856         <para>
 857         Third, scattered EH code makes implementing low level drivers
 858         difficult.  Low level drivers override libata callbacks.  If
 859         EH is scattered over several places, each affected callbacks
 860         should perform its part of error handling.  This can be error
 861         prone and painful.
 862         </para>
 863         </listitem>
 864
 865         </itemizedlist>
 866         </sect1>
 867   </chapter>
 868
 869   <chapter id="libataExt">
 870      <title>libata Library</title>
 871 !Edrivers/ata/libata-core.c
 872   </chapter>
 873
 874   <chapter id="libataInt">
 875      <title>libata Core Internals</title>
 876 !Idrivers/ata/libata-core.c
 877   </chapter>
 878
 879   <chapter id="libataScsiInt">
 880      <title>libata SCSI translation/emulation</title>
 881 !Edrivers/ata/libata-scsi.c
 882 !Idrivers/ata/libata-scsi.c
 883   </chapter>
 884
 885   <chapter id="ataExceptions">
 886      <title>ATA errors and exceptions</title>
 887
 888   <para>
 889   This chapter tries to identify what error/exception conditions exist
 890   for ATA/ATAPI devices and describe how they should be handled in
 891   implementation-neutral way.
 892   </para>
 893
 894   <para>
 895   The term 'error' is used to describe conditions where either an
 896   explicit error condition is reported from device or a command has
 897   timed out.
 898   </para>
 899
 900   <para>
 901   The term 'exception' is either used to describe exceptional
 902   conditions which are not errors (say, power or hotplug events), or
 903   to describe both errors and non-error exceptional conditions.  Where
 904   explicit distinction between error and exception is necessary, the
 905   term 'non-error exception' is used.
 906   </para>
 907
 908   <sect1 id="excat">
 909      <title>Exception categories</title>
 910      <para>
 911      Exceptions are described primarily with respect to legacy
 912      taskfile + bus master IDE interface.  If a controller provides
 913      other better mechanism for error reporting, mapping those into
 914      categories described below shouldn't be difficult.
 915      </para>
 916
 917      <para>
 918      In the following sections, two recovery actions - reset and
 919      reconfiguring transport - are mentioned.  These are described
 920      further in <xref linkend="exrec"/>.
 921      </para>
 922
 923      <sect2 id="excatHSMviolation">
 924         <title>HSM violation</title>
 925         <para>
 926         This error is indicated when STATUS value doesn't match HSM
 927         requirement during issuing or excution any ATA/ATAPI command.
 928         </para>
 929
 930         <itemizedlist>
 931         <title>Examples</title>
 932
 933         <listitem>
 934         <para>
 935         ATA_STATUS doesn't contain !BSY &amp;&amp; DRDY &amp;&amp; !DRQ while trying
 936         to issue a command.
 937         </para>
 938         </listitem>
 939
 940         <listitem>
 941         <para>
 942         !BSY &amp;&amp; !DRQ during PIO data transfer.
 943         </para>
 944         </listitem>
 945
 946         <listitem>
 947         <para>
 948         DRQ on command completion.
 949         </para>
 950         </listitem>
 951
 952         <listitem>
 953         <para>
 954         !BSY &amp;&amp; ERR after CDB tranfer starts but before the
 955         last byte of CDB is transferred.  ATA/ATAPI standard states
 956         that &quot;The device shall not terminate the PACKET command
 957         with an error before the last byte of the command packet has
 958         been written&quot; in the error outputs description of PACKET
 959         command and the state diagram doesn't include such
 960         transitions.
 961         </para>
 962         </listitem>
 963
 964         </itemizedlist>
 965
 966         <para>
 967         In these cases, HSM is violated and not much information
 968         regarding the error can be acquired from STATUS or ERROR
 969         register.  IOW, this error can be anything - driver bug,
 970         faulty device, controller and/or cable.
 971         </para>
 972
 973         <para>
 974         As HSM is violated, reset is necessary to restore known state.
 975         Reconfiguring transport for lower speed might be helpful too
 976         as transmission errors sometimes cause this kind of errors.
 977         </para>
 978      </sect2>
 979
 980      <sect2 id="excatDevErr">
 981         <title>ATA/ATAPI device error (non-NCQ / non-CHECK CONDITION)</title>
 982
 983         <para>
 984         These are errors detected and reported by ATA/ATAPI devices
 985         indicating device problems.  For this type of errors, STATUS
 986         and ERROR register values are valid and describe error
 987         condition.  Note that some of ATA bus errors are detected by
 988         ATA/ATAPI devices and reported using the same mechanism as
 989         device errors.  Those cases are described later in this
 990         section.
 991         </para>
 992
 993         <para>
 994         For ATA commands, this type of errors are indicated by !BSY
 995         &amp;&amp; ERR during command execution and on completion.
 996         </para>
 997
 998         <para>For ATAPI commands,</para>
 999
1000         <itemizedlist>
1001
1002         <listitem>
1003         <para>
1004         !BSY &amp;&amp; ERR &amp;&amp; ABRT right after issuing PACKET
1005         indicates that PACKET command is not supported and falls in
1006         this category.
1007         </para>
1008         </listitem>
1009
1010         <listitem>
1011         <para>
1012         !BSY &amp;&amp; ERR(==CHK) &amp;&amp; !ABRT after the last
1013         byte of CDB is transferred indicates CHECK CONDITION and
1014         doesn't fall in this category.
1015         </para>
1016         </listitem>
1017
1018         <listitem>
1019         <para>
1020         !BSY &amp;&amp; ERR(==CHK) &amp;&amp; ABRT after the last byte
1021         of CDB is transferred *probably* indicates CHECK CONDITION and
1022         doesn't fall in this category.
1023         </para>
1024         </listitem>
1025
1026         </itemizedlist>
1027
1028         <para>
1029         Of errors detected as above, the followings are not ATA/ATAPI
1030         device errors but ATA bus errors and should be handled
1031         according to <xref linkend="excatATAbusErr"/>.
1032         </para>
1033
1034         <variablelist>
1035
1036            <varlistentry>
1037            <term>CRC error during data transfer</term>
1038            <listitem>
1039            <para>
1040            This is indicated by ICRC bit in the ERROR register and
1041            means that corruption occurred during data transfer.  Upto
1042            ATA/ATAPI-7, the standard specifies that this bit is only
1043            applicable to UDMA transfers but ATA/ATAPI-8 draft revision
1044            1f says that the bit may be applicable to multiword DMA and
1045            PIO.
1046            </para>
1047            </listitem>
1048            </varlistentry>
1049
1050            <varlistentry>
1051            <term>ABRT error during data transfer or on completion</term>
1052            <listitem>
1053            <para>
1054            Upto ATA/ATAPI-7, the standard specifies that ABRT could be
1055            set on ICRC errors and on cases where a device is not able
1056            to complete a command.  Combined with the fact that MWDMA
1057            and PIO transfer errors aren't allowed to use ICRC bit upto
1058            ATA/ATAPI-7, it seems to imply that ABRT bit alone could
1059            indicate tranfer errors.
1060            </para>
1061            <para>
1062            However, ATA/ATAPI-8 draft revision 1f removes the part
1063            that ICRC errors can turn on ABRT.  So, this is kind of
1064            gray area.  Some heuristics are needed here.
1065            </para>
1066            </listitem>
1067            </varlistentry>
1068
1069         </variablelist>
1070
1071         <para>
1072         ATA/ATAPI device errors can be further categorized as follows.
1073         </para>
1074
1075         <variablelist>
1076
1077            <varlistentry>
1078            <term>Media errors</term>
1079            <listitem>
1080            <para>
1081            This is indicated by UNC bit in the ERROR register.  ATA
1082            devices reports UNC error only after certain number of
1083            retries cannot recover the data, so there's nothing much
1084            else to do other than notifying upper layer.
1085            </para>
1086            <para>
1087            READ and WRITE commands report CHS or LBA of the first
1088            failed sector but ATA/ATAPI standard specifies that the
1089            amount of transferred data on error completion is
1090            indeterminate, so we cannot assume that sectors preceding
1091            the failed sector have been transferred and thus cannot
1092            complete those sectors successfully as SCSI does.
1093            </para>
1094            </listitem>
1095            </varlistentry>
1096
1097            <varlistentry>
1098            <term>Media changed / media change requested error</term>
1099            <listitem>
1100            <para>
1101            &lt;&lt;TODO: fill here&gt;&gt;
1102            </para>
1103            </listitem>
1104            </varlistentry>
1105
1106            <varlistentry><term>Address error</term>
1107            <listitem>
1108            <para>
1109            This is indicated by IDNF bit in the ERROR register.
1110            Report to upper layer.
1111            </para>
1112            </listitem>
1113            </varlistentry>
1114
1115            <varlistentry><term>Other errors</term>
1116            <listitem>
1117            <para>
1118            This can be invalid command or parameter indicated by ABRT
1119            ERROR bit or some other error condition.  Note that ABRT
1120            bit can indicate a lot of things including ICRC and Address
1121            errors.  Heuristics needed.
1122            </para>
1123            </listitem>
1124            </varlistentry>
1125
1126         </variablelist>
1127
1128         <para>
1129         Depending on commands, not all STATUS/ERROR bits are
1130         applicable.  These non-applicable bits are marked with
1131         &quot;na&quot; in the output descriptions but upto ATA/ATAPI-7
1132         no definition of &quot;na&quot; can be found.  However,
1133         ATA/ATAPI-8 draft revision 1f describes &quot;N/A&quot; as
1134         follows.
1135         </para>
1136
1137         <blockquote>
1138         <variablelist>
1139            <varlistentry><term>3.2.3.3a N/A</term>
1140            <listitem>
1141            <para>
1142            A keyword the indicates a field has no defined value in
1143            this standard and should not be checked by the host or
1144            device. N/A fields should be cleared to zero.
1145            </para>
1146            </listitem>
1147            </varlistentry>
1148         </variablelist>
1149         </blockquote>
1150
1151         <para>
1152         So, it seems reasonable to assume that &quot;na&quot; bits are
1153         cleared to zero by devices and thus need no explicit masking.
1154         </para>
1155
1156      </sect2>
1157
1158      <sect2 id="excatATAPIcc">
1159         <title>ATAPI device CHECK CONDITION</title>
1160
1161         <para>
1162         ATAPI device CHECK CONDITION error is indicated by set CHK bit
1163         (ERR bit) in the STATUS register after the last byte of CDB is
1164         transferred for a PACKET command.  For this kind of errors,
1165         sense data should be acquired to gather information regarding
1166         the errors.  REQUEST SENSE packet command should be used to
1167         acquire sense data.
1168         </para>
1169
1170         <para>
1171         Once sense data is acquired, this type of errors can be
1172         handled similary to other SCSI errors.  Note that sense data
1173         may indicate ATA bus error (e.g. Sense Key 04h HARDWARE ERROR
1174         &amp;&amp; ASC/ASCQ 47h/00h SCSI PARITY ERROR).  In such
1175         cases, the error should be considered as an ATA bus error and
1176         handled according to <xref linkend="excatATAbusErr"/>.
1177         </para>
1178
1179      </sect2>
1180
1181      <sect2 id="excatNCQerr">
1182         <title>ATA device error (NCQ)</title>
1183
1184         <para>
1185         NCQ command error is indicated by cleared BSY and set ERR bit
1186         during NCQ command phase (one or more NCQ commands
1187         outstanding).  Although STATUS and ERROR registers will
1188         contain valid values describing the error, READ LOG EXT is
1189         required to clear the error condition, determine which command
1190         has failed and acquire more information.
1191         </para>
1192
1193         <para>
1194         READ LOG EXT Log Page 10h reports which tag has failed and
1195         taskfile register values describing the error.  With this
1196         information the failed command can be handled as a normal ATA
1197         command error as in <xref linkend="excatDevErr"/> and all
1198         other in-flight commands must be retried.  Note that this
1199         retry should not be counted - it's likely that commands
1200         retried this way would have completed normally if it were not
1201         for the failed command.
1202         </para>
1203
1204         <para>
1205         Note that ATA bus errors can be reported as ATA device NCQ
1206         errors.  This should be handled as described in <xref
1207         linkend="excatATAbusErr"/>.
1208         </para>
1209
1210         <para>
1211         If READ LOG EXT Log Page 10h fails or reports NQ, we're
1212         thoroughly screwed.  This condition should be treated
1213         according to <xref linkend="excatHSMviolation"/>.
1214         </para>
1215
1216      </sect2>
1217
1218      <sect2 id="excatATAbusErr">
1219         <title>ATA bus error</title>
1220
1221         <para>
1222         ATA bus error means that data corruption occurred during
1223         transmission over ATA bus (SATA or PATA).  This type of errors
1224         can be indicated by
1225         </para>
1226
1227         <itemizedlist>
1228
1229         <listitem>
1230         <para>
1231         ICRC or ABRT error as described in <xref linkend="excatDevErr"/>.
1232         </para>
1233         </listitem>
1234
1235         <listitem>
1236         <para>
1237         Controller-specific error completion with error information
1238         indicating transmission error.
1239         </para>
1240         </listitem>
1241
1242         <listitem>
1243         <para>
1244         On some controllers, command timeout.  In this case, there may
1245         be a mechanism to determine that the timeout is due to
1246         transmission error.
1247         </para>
1248         </listitem>
1249
1250         <listitem>
1251         <para>
1252         Unknown/random errors, timeouts and all sorts of weirdities.
1253         </para>
1254         </listitem>
1255
1256         </itemizedlist>
1257
1258         <para>
1259         As described above, transmission errors can cause wide variety
1260         of symptoms ranging from device ICRC error to random device
1261         lockup, and, for many cases, there is no way to tell if an
1262         error condition is due to transmission error or not;
1263         therefore, it's necessary to employ some kind of heuristic
1264         when dealing with errors and timeouts.  For example,
1265         encountering repetitive ABRT errors for known supported
1266         command is likely to indicate ATA bus error.
1267         </para>
1268
1269         <para>
1270         Once it's determined that ATA bus errors have possibly
1271         occurred, lowering ATA bus transmission speed is one of
1272         actions which may alleviate the problem.  See <xref
1273         linkend="exrecReconf"/> for more information.
1274         </para>
1275
1276      </sect2>
1277
1278      <sect2 id="excatPCIbusErr">
1279         <title>PCI bus error</title>
1280
1281         <para>
1282         Data corruption or other failures during transmission over PCI
1283         (or other system bus).  For standard BMDMA, this is indicated
1284         by Error bit in the BMDMA Status register.  This type of
1285         errors must be logged as it indicates something is very wrong
1286         with the system.  Resetting host controller is recommended.
1287         </para>
1288
1289      </sect2>
1290
1291      <sect2 id="excatLateCompletion">
1292         <title>Late completion</title>
1293
1294         <para>
1295         This occurs when timeout occurs and the timeout handler finds
1296         out that the timed out command has completed successfully or
1297         with error.  This is usually caused by lost interrupts.  This
1298         type of errors must be logged.  Resetting host controller is
1299         recommended.
1300         </para>
1301
1302      </sect2>
1303
1304      <sect2 id="excatUnknown">
1305         <title>Unknown error (timeout)</title>
1306
1307         <para>
1308         This is when timeout occurs and the command is still
1309         processing or the host and device are in unknown state.  When
1310         this occurs, HSM could be in any valid or invalid state.  To
1311         bring the device to known state and make it forget about the
1312         timed out command, resetting is necessary.  The timed out
1313         command may be retried.
1314         </para>
1315
1316         <para>
1317         Timeouts can also be caused by transmission errors.  Refer to
1318         <xref linkend="excatATAbusErr"/> for more details.
1319         </para>
1320
1321      </sect2>
1322
1323      <sect2 id="excatHoplugPM">
1324         <title>Hotplug and power management exceptions</title>
1325
1326         <para>
1327         &lt;&lt;TODO: fill here&gt;&gt;
1328         </para>
1329
1330      </sect2>
1331
1332   </sect1>
1333
1334   <sect1 id="exrec">
1335      <title>EH recovery actions</title>
1336
1337      <para>
1338      This section discusses several important recovery actions.
1339      </para>
1340
1341      <sect2 id="exrecClr">
1342         <title>Clearing error condition</title>
1343
1344         <para>
1345         Many controllers require its error registers to be cleared by
1346         error handler.  Different controllers may have different
1347         requirements.
1348         </para>
1349
1350         <para>
1351         For SATA, it's strongly recommended to clear at least SError
1352         register during error handling.
1353         </para>
1354      </sect2>
1355
1356      <sect2 id="exrecRst">
1357         <title>Reset</title>
1358
1359         <para>
1360         During EH, resetting is necessary in the following cases.
1361         </para>
1362
1363         <itemizedlist>
1364
1365         <listitem>
1366         <para>
1367         HSM is in unknown or invalid state
1368         </para>
1369         </listitem>
1370
1371         <listitem>
1372         <para>
1373         HBA is in unknown or invalid state
1374         </para>
1375         </listitem>
1376
1377         <listitem>
1378         <para>
1379         EH needs to make HBA/device forget about in-flight commands
1380         </para>
1381         </listitem>
1382
1383         <listitem>
1384         <para>
1385         HBA/device behaves weirdly
1386         </para>
1387         </listitem>
1388
1389         </itemizedlist>
1390
1391         <para>
1392         Resetting during EH might be a good idea regardless of error
1393         condition to improve EH robustness.  Whether to reset both or
1394         either one of HBA and device depends on situation but the
1395         following scheme is recommended.
1396         </para>
1397
1398         <itemizedlist>
1399
1400         <listitem>
1401         <para>
1402         When it's known that HBA is in ready state but ATA/ATAPI
1403         device is in unknown state, reset only device.
1404         </para>
1405         </listitem>
1406
1407         <listitem>
1408         <para>
1409         If HBA is in unknown state, reset both HBA and device.
1410         </para>
1411         </listitem>
1412
1413         </itemizedlist>
1414
1415         <para>
1416         HBA resetting is implementation specific.  For a controller
1417         complying to taskfile/BMDMA PCI IDE, stopping active DMA
1418         transaction may be sufficient iff BMDMA state is the only HBA
1419         context.  But even mostly taskfile/BMDMA PCI IDE complying
1420         controllers may have implementation specific requirements and
1421         mechanism to reset themselves.  This must be addressed by
1422         specific drivers.
1423         </para>
1424
1425         <para>
1426         OTOH, ATA/ATAPI standard describes in detail ways to reset
1427         ATA/ATAPI devices.
1428         </para>
1429
1430         <variablelist>
1431
1432            <varlistentry><term>PATA hardware reset</term>
1433            <listitem>
1434            <para>
1435            This is hardware initiated device reset signalled with
1436            asserted PATA RESET- signal.  There is no standard way to
1437            initiate hardware reset from software although some
1438            hardware provides registers that allow driver to directly
1439            tweak the RESET- signal.
1440            </para>
1441            </listitem>
1442            </varlistentry>
1443
1444            <varlistentry><term>Software reset</term>
1445            <listitem>
1446            <para>
1447            This is achieved by turning CONTROL SRST bit on for at
1448            least 5us.  Both PATA and SATA support it but, in case of
1449            SATA, this may require controller-specific support as the
1450            second Register FIS to clear SRST should be transmitted
1451            while BSY bit is still set.  Note that on PATA, this resets
1452            both master and slave devices on a channel.
1453            </para>
1454            </listitem>
1455            </varlistentry>
1456
1457            <varlistentry><term>EXECUTE DEVICE DIAGNOSTIC command</term>
1458            <listitem>
1459            <para>
1460            Although ATA/ATAPI standard doesn't describe exactly, EDD
1461            implies some level of resetting, possibly similar level
1462            with software reset.  Host-side EDD protocol can be handled
1463            with normal command processing and most SATA controllers
1464            should be able to handle EDD's just like other commands.
1465            As in software reset, EDD affects both devices on a PATA
1466            bus.
1467            </para>
1468            <para>
1469            Although EDD does reset devices, this doesn't suit error
1470            handling as EDD cannot be issued while BSY is set and it's
1471            unclear how it will act when device is in unknown/weird
1472            state.
1473            </para>
1474            </listitem>
1475            </varlistentry>
1476
1477            <varlistentry><term>ATAPI DEVICE RESET command</term>
1478            <listitem>
1479            <para>
1480            This is very similar to software reset except that reset
1481            can be restricted to the selected device without affecting
1482            the other device sharing the cable.
1483            </para>
1484            </listitem>
1485            </varlistentry>
1486
1487            <varlistentry><term>SATA phy reset</term>
1488            <listitem>
1489            <para>
1490            This is the preferred way of resetting a SATA device.  In
1491            effect, it's identical to PATA hardware reset.  Note that
1492            this can be done with the standard SCR Control register.
1493            As such, it's usually easier to implement than software
1494            reset.
1495            </para>
1496            </listitem>
1497            </varlistentry>
1498
1499         </variablelist>
1500
1501         <para>
1502         One more thing to consider when resetting devices is that
1503         resetting clears certain configuration parameters and they
1504         need to be set to their previous or newly adjusted values
1505         after reset.
1506         </para>
1507
1508         <para>
1509         Parameters affected are.
1510         </para>
1511
1512         <itemizedlist>
1513
1514         <listitem>
1515         <para>
1516         CHS set up with INITIALIZE DEVICE PARAMETERS (seldomly used)
1517         </para>
1518         </listitem>
1519
1520         <listitem>
1521         <para>
1522         Parameters set with SET FEATURES including transfer mode setting
1523         </para>
1524         </listitem>
1525
1526         <listitem>
1527         <para>
1528         Block count set with SET MULTIPLE MODE
1529         </para>
1530         </listitem>
1531
1532         <listitem>
1533         <para>
1534         Other parameters (SET MAX, MEDIA LOCK...)
1535         </para>
1536         </listitem>
1537
1538         </itemizedlist>
1539
1540         <para>
1541         ATA/ATAPI standard specifies that some parameters must be
1542         maintained across hardware or software reset, but doesn't
1543         strictly specify all of them.  Always reconfiguring needed
1544         parameters after reset is required for robustness.  Note that
1545         this also applies when resuming from deep sleep (power-off).
1546         </para>
1547
1548         <para>
1549         Also, ATA/ATAPI standard requires that IDENTIFY DEVICE /
1550         IDENTIFY PACKET DEVICE is issued after any configuration
1551         parameter is updated or a hardware reset and the result used
1552         for further operation.  OS driver is required to implement
1553         revalidation mechanism to support this.
1554         </para>
1555
1556      </sect2>
1557
1558      <sect2 id="exrecReconf">
1559         <title>Reconfigure transport</title>
1560
1561         <para>
1562         For both PATA and SATA, a lot of corners are cut for cheap
1563         connectors, cables or controllers and it's quite common to see
1564         high transmission error rate.  This can be mitigated by
1565         lowering transmission speed.
1566         </para>
1567
1568         <para>
1569         The following is a possible scheme Jeff Garzik suggested.
1570         </para>
1571
1572         <blockquote>
1573         <para>
1574         If more than $N (3?) transmission errors happen in 15 minutes,
1575         </para>
1576         <itemizedlist>
1577         <listitem>
1578         <para>
1579         if SATA, decrease SATA PHY speed.  if speed cannot be decreased,
1580         </para>
1581         </listitem>
1582         <listitem>
1583         <para>
1584         decrease UDMA xfer speed.  if at UDMA0, switch to PIO4,
1585         </para>
1586         </listitem>
1587         <listitem>
1588         <para>
1589         decrease PIO xfer speed.  if at PIO3, complain, but continue
1590         </para>
1591         </listitem>
1592         </itemizedlist>
1593         </blockquote>
1594
1595      </sect2>
1596
1597   </sect1>
1598
1599   </chapter>
1600
1601   <chapter id="PiixInt">
1602      <title>ata_piix Internals</title>
1603 !Idrivers/ata/ata_piix.c
1604   </chapter>
1605
1606   <chapter id="SILInt">
1607      <title>sata_sil Internals</title>
1608 !Idrivers/ata/sata_sil.c
1609   </chapter>
1610
1611   <chapter id="libataThanks">
1612      <title>Thanks</title>
1613   <para>
1614   The bulk of the ATA knowledge comes thanks to long conversations with
1615   Andre Hedrick (www.linux-ide.org), and long hours pondering the ATA
1616   and SCSI specifications.
1617   </para>
1618   <para>
1619   Thanks to Alan Cox for pointing out similarities
1620   between SATA and SCSI, and in general for motivation to hack on
1621   libata.
1622   </para>
1623   <para>
1624   libata's device detection
1625   method, ata_pio_devchk, and in general all the early probing was
1626   based on extensive study of Hale Landis's probe/reset code in his
1627   ATADRVR driver (www.ata-atapi.com).
1628   </para>
1629   </chapter>
1630
1631 </book>