diff options
| author | Borislav Petkov <bbpetkov@yahoo.de> | 2008-02-02 13:56:48 -0500 |
|---|---|---|
| committer | Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> | 2008-02-02 13:56:48 -0500 |
| commit | 5ce78af4beeca6c80d8a4834f74297a8fa2e2253 (patch) | |
| tree | ca3af7db59981109475ba2dec4d0ea7e8250c742 /drivers/ide/ide-tape.c | |
| parent | dcb425f557eb01e73614f2292f0a92ce02e40e2a (diff) | |
ide-tape: move historical changelog to Documentation/ide/ChangeLog.ide-tape.1995-2002
Also, cleanup whitespace and update comments.
Bart:
- remove reference to drivers/block/ide.c
- move driver documentation to Documentation/ide/ide-tape.txt
Signed-off-by: Borislav Petkov <bbpetkov@yahoo.de>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Diffstat (limited to 'drivers/ide/ide-tape.c')
| -rw-r--r-- | drivers/ide/ide-tape.c | 414 |
1 files changed, 4 insertions, 410 deletions
diff --git a/drivers/ide/ide-tape.c b/drivers/ide/ide-tape.c index eff118ba217c..045bd2ae0c0f 100644 --- a/drivers/ide/ide-tape.c +++ b/drivers/ide/ide-tape.c | |||
| @@ -1,424 +1,18 @@ | |||
| 1 | /* | 1 | /* |
| 2 | * IDE ATAPI streaming tape driver. | ||
| 3 | * | ||
| 2 | * Copyright (C) 1995-1999 Gadi Oxman <gadio@netvision.net.il> | 4 | * Copyright (C) 1995-1999 Gadi Oxman <gadio@netvision.net.il> |
| 3 | * Copyright (C) 2003-2005 Bartlomiej Zolnierkiewicz | 5 | * Copyright (C) 2003-2005 Bartlomiej Zolnierkiewicz |
| 4 | * | 6 | * |
| 5 | * $Header$ | ||
| 6 | * | ||
| 7 | * This driver was constructed as a student project in the software laboratory | 7 | * This driver was constructed as a student project in the software laboratory |
| 8 | * of the faculty of electrical engineering in the Technion - Israel's | 8 | * of the faculty of electrical engineering in the Technion - Israel's |
| 9 | * Institute Of Technology, with the guide of Avner Lottem and Dr. Ilana David. | 9 | * Institute Of Technology, with the guide of Avner Lottem and Dr. Ilana David. |
| 10 | * | 10 | * |
| 11 | * It is hereby placed under the terms of the GNU general public license. | 11 | * It is hereby placed under the terms of the GNU general public license. |
| 12 | * (See linux/COPYING). | 12 | * (See linux/COPYING). |
| 13 | */ | ||
| 14 | |||
| 15 | /* | ||
| 16 | * IDE ATAPI streaming tape driver. | ||
| 17 | * | ||
| 18 | * This driver is a part of the Linux ide driver and works in co-operation | ||
| 19 | * with linux/drivers/block/ide.c. | ||
| 20 | * | ||
| 21 | * The driver, in co-operation with ide.c, basically traverses the | ||
| 22 | * request-list for the block device interface. The character device | ||
| 23 | * interface, on the other hand, creates new requests, adds them | ||
| 24 | * to the request-list of the block device, and waits for their completion. | ||
| 25 | * | ||
| 26 | * Pipelined operation mode is now supported on both reads and writes. | ||
| 27 | * | ||
| 28 | * The block device major and minor numbers are determined from the | ||
| 29 | * tape's relative position in the ide interfaces, as explained in ide.c. | ||
| 30 | * | ||
| 31 | * The character device interface consists of the following devices: | ||
| 32 | * | ||
| 33 | * ht0 major 37, minor 0 first IDE tape, rewind on close. | ||
| 34 | * ht1 major 37, minor 1 second IDE tape, rewind on close. | ||
| 35 | * ... | ||
| 36 | * nht0 major 37, minor 128 first IDE tape, no rewind on close. | ||
| 37 | * nht1 major 37, minor 129 second IDE tape, no rewind on close. | ||
| 38 | * ... | ||
| 39 | * | ||
| 40 | * Run linux/scripts/MAKEDEV.ide to create the above entries. | ||
| 41 | * | ||
| 42 | * The general magnetic tape commands compatible interface, as defined by | ||
| 43 | * include/linux/mtio.h, is accessible through the character device. | ||
| 44 | * | ||
| 45 | * General ide driver configuration options, such as the interrupt-unmask | ||
| 46 | * flag, can be configured by issuing an ioctl to the block device interface, | ||
| 47 | * as any other ide device. | ||
| 48 | * | ||
| 49 | * Our own ide-tape ioctl's can be issued to either the block device or | ||
| 50 | * the character device interface. | ||
| 51 | * | ||
| 52 | * Maximal throughput with minimal bus load will usually be achieved in the | ||
| 53 | * following scenario: | ||
| 54 | * | ||
| 55 | * 1. ide-tape is operating in the pipelined operation mode. | ||
| 56 | * 2. No buffering is performed by the user backup program. | ||
| 57 | * | ||
| 58 | * Testing was done with a 2 GB CONNER CTMA 4000 IDE ATAPI Streaming Tape Drive. | ||
| 59 | * | ||
| 60 | * Ver 0.1 Nov 1 95 Pre-working code :-) | ||
| 61 | * Ver 0.2 Nov 23 95 A short backup (few megabytes) and restore procedure | ||
| 62 | * was successful ! (Using tar cvf ... on the block | ||
| 63 | * device interface). | ||
| 64 | * A longer backup resulted in major swapping, bad | ||
| 65 | * overall Linux performance and eventually failed as | ||
| 66 | * we received non serial read-ahead requests from the | ||
| 67 | * buffer cache. | ||
| 68 | * Ver 0.3 Nov 28 95 Long backups are now possible, thanks to the | ||
| 69 | * character device interface. Linux's responsiveness | ||
| 70 | * and performance doesn't seem to be much affected | ||
| 71 | * from the background backup procedure. | ||
| 72 | * Some general mtio.h magnetic tape operations are | ||
| 73 | * now supported by our character device. As a result, | ||
| 74 | * popular tape utilities are starting to work with | ||
| 75 | * ide tapes :-) | ||
| 76 | * The following configurations were tested: | ||
| 77 | * 1. An IDE ATAPI TAPE shares the same interface | ||
| 78 | * and irq with an IDE ATAPI CDROM. | ||
| 79 | * 2. An IDE ATAPI TAPE shares the same interface | ||
| 80 | * and irq with a normal IDE disk. | ||
| 81 | * Both configurations seemed to work just fine ! | ||
| 82 | * However, to be on the safe side, it is meanwhile | ||
| 83 | * recommended to give the IDE TAPE its own interface | ||
| 84 | * and irq. | ||
| 85 | * The one thing which needs to be done here is to | ||
| 86 | * add a "request postpone" feature to ide.c, | ||
| 87 | * so that we won't have to wait for the tape to finish | ||
| 88 | * performing a long media access (DSC) request (such | ||
| 89 | * as a rewind) before we can access the other device | ||
| 90 | * on the same interface. This effect doesn't disturb | ||
| 91 | * normal operation most of the time because read/write | ||
| 92 | * requests are relatively fast, and once we are | ||
| 93 | * performing one tape r/w request, a lot of requests | ||
| 94 | * from the other device can be queued and ide.c will | ||
| 95 | * service all of them after this single tape request. | ||
| 96 | * Ver 1.0 Dec 11 95 Integrated into Linux 1.3.46 development tree. | ||
| 97 | * On each read / write request, we now ask the drive | ||
| 98 | * if we can transfer a constant number of bytes | ||
| 99 | * (a parameter of the drive) only to its buffers, | ||
| 100 | * without causing actual media access. If we can't, | ||
| 101 | * we just wait until we can by polling the DSC bit. | ||
| 102 | * This ensures that while we are not transferring | ||
| 103 | * more bytes than the constant referred to above, the | ||
| 104 | * interrupt latency will not become too high and | ||
| 105 | * we won't cause an interrupt timeout, as happened | ||
| 106 | * occasionally in the previous version. | ||
| 107 | * While polling for DSC, the current request is | ||
| 108 | * postponed and ide.c is free to handle requests from | ||
| 109 | * the other device. This is handled transparently to | ||
| 110 | * ide.c. The hwgroup locking method which was used | ||
| 111 | * in the previous version was removed. | ||
| 112 | * Use of new general features which are provided by | ||
| 113 | * ide.c for use with atapi devices. | ||
| 114 | * (Programming done by Mark Lord) | ||
| 115 | * Few potential bug fixes (Again, suggested by Mark) | ||
| 116 | * Single character device data transfers are now | ||
| 117 | * not limited in size, as they were before. | ||
| 118 | * We are asking the tape about its recommended | ||
| 119 | * transfer unit and send a larger data transfer | ||
| 120 | * as several transfers of the above size. | ||
| 121 | * For best results, use an integral number of this | ||
| 122 | * basic unit (which is shown during driver | ||
| 123 | * initialization). I will soon add an ioctl to get | ||
| 124 | * this important parameter. | ||
| 125 | * Our data transfer buffer is allocated on startup, | ||
| 126 | * rather than before each data transfer. This should | ||
| 127 | * ensure that we will indeed have a data buffer. | ||
| 128 | * Ver 1.1 Dec 14 95 Fixed random problems which occurred when the tape | ||
| 129 | * shared an interface with another device. | ||
| 130 | * (poll_for_dsc was a complete mess). | ||
| 131 | * Removed some old (non-active) code which had | ||
| 132 | * to do with supporting buffer cache originated | ||
| 133 | * requests. | ||
| 134 | * The block device interface can now be opened, so | ||
| 135 | * that general ide driver features like the unmask | ||
| 136 | * interrupts flag can be selected with an ioctl. | ||
| 137 | * This is the only use of the block device interface. | ||
| 138 | * New fast pipelined operation mode (currently only on | ||
| 139 | * writes). When using the pipelined mode, the | ||
| 140 | * throughput can potentially reach the maximum | ||
| 141 | * tape supported throughput, regardless of the | ||
| 142 | * user backup program. On my tape drive, it sometimes | ||
| 143 | * boosted performance by a factor of 2. Pipelined | ||
| 144 | * mode is enabled by default, but since it has a few | ||
| 145 | * downfalls as well, you may want to disable it. | ||
| 146 | * A short explanation of the pipelined operation mode | ||
| 147 | * is available below. | ||
| 148 | * Ver 1.2 Jan 1 96 Eliminated pipelined mode race condition. | ||
| 149 | * Added pipeline read mode. As a result, restores | ||
| 150 | * are now as fast as backups. | ||
| 151 | * Optimized shared interface behavior. The new behavior | ||
| 152 | * typically results in better IDE bus efficiency and | ||
| 153 | * higher tape throughput. | ||
| 154 | * Pre-calculation of the expected read/write request | ||
| 155 | * service time, based on the tape's parameters. In | ||
| 156 | * the pipelined operation mode, this allows us to | ||
| 157 | * adjust our polling frequency to a much lower value, | ||
| 158 | * and thus to dramatically reduce our load on Linux, | ||
| 159 | * without any decrease in performance. | ||
| 160 | * Implemented additional mtio.h operations. | ||
| 161 | * The recommended user block size is returned by | ||
| 162 | * the MTIOCGET ioctl. | ||
| 163 | * Additional minor changes. | ||
| 164 | * Ver 1.3 Feb 9 96 Fixed pipelined read mode bug which prevented the | ||
| 165 | * use of some block sizes during a restore procedure. | ||
| 166 | * The character device interface will now present a | ||
| 167 | * continuous view of the media - any mix of block sizes | ||
| 168 | * during a backup/restore procedure is supported. The | ||
| 169 | * driver will buffer the requests internally and | ||
| 170 | * convert them to the tape's recommended transfer | ||
| 171 | * unit, making performance almost independent of the | ||
| 172 | * chosen user block size. | ||
| 173 | * Some improvements in error recovery. | ||
| 174 | * By cooperating with ide-dma.c, bus mastering DMA can | ||
| 175 | * now sometimes be used with IDE tape drives as well. | ||
| 176 | * Bus mastering DMA has the potential to dramatically | ||
| 177 | * reduce the CPU's overhead when accessing the device, | ||
| 178 | * and can be enabled by using hdparm -d1 on the tape's | ||
| 179 | * block device interface. For more info, read the | ||
| 180 | * comments in ide-dma.c. | ||
| 181 | * Ver 1.4 Mar 13 96 Fixed serialize support. | ||
| 182 | * Ver 1.5 Apr 12 96 Fixed shared interface operation, broken in 1.3.85. | ||
| 183 | * Fixed pipelined read mode inefficiency. | ||
| 184 | * Fixed nasty null dereferencing bug. | ||
| 185 | * Ver 1.6 Aug 16 96 Fixed FPU usage in the driver. | ||
| 186 | * Fixed end of media bug. | ||
| 187 | * Ver 1.7 Sep 10 96 Minor changes for the CONNER CTT8000-A model. | ||
| 188 | * Ver 1.8 Sep 26 96 Attempt to find a better balance between good | ||
| 189 | * interactive response and high system throughput. | ||
| 190 | * Ver 1.9 Nov 5 96 Automatically cross encountered filemarks rather | ||
| 191 | * than requiring an explicit FSF command. | ||
| 192 | * Abort pending requests at end of media. | ||
| 193 | * MTTELL was sometimes returning incorrect results. | ||
| 194 | * Return the real block size in the MTIOCGET ioctl. | ||
| 195 | * Some error recovery bug fixes. | ||
| 196 | * Ver 1.10 Nov 5 96 Major reorganization. | ||
| 197 | * Reduced CPU overhead a bit by eliminating internal | ||
| 198 | * bounce buffers. | ||
| 199 | * Added module support. | ||
| 200 | * Added multiple tape drives support. | ||
| 201 | * Added partition support. | ||
| 202 | * Rewrote DSC handling. | ||
| 203 | * Some portability fixes. | ||
| 204 | * Removed ide-tape.h. | ||
| 205 | * Additional minor changes. | ||
| 206 | * Ver 1.11 Dec 2 96 Bug fix in previous DSC timeout handling. | ||
| 207 | * Use ide_stall_queue() for DSC overlap. | ||
| 208 | * Use the maximum speed rather than the current speed | ||
| 209 | * to compute the request service time. | ||
| 210 | * Ver 1.12 Dec 7 97 Fix random memory overwriting and/or last block data | ||
| 211 | * corruption, which could occur if the total number | ||
| 212 | * of bytes written to the tape was not an integral | ||
| 213 | * number of tape blocks. | ||
| 214 | * Add support for INTERRUPT DRQ devices. | ||
| 215 | * Ver 1.13 Jan 2 98 Add "speed == 0" work-around for HP COLORADO 5GB | ||
| 216 | * Ver 1.14 Dec 30 98 Partial fixes for the Sony/AIWA tape drives. | ||
| 217 | * Replace cli()/sti() with hwgroup spinlocks. | ||
| 218 | * Ver 1.15 Mar 25 99 Fix SMP race condition by replacing hwgroup | ||
| 219 | * spinlock with private per-tape spinlock. | ||
| 220 | * Ver 1.16 Sep 1 99 Add OnStream tape support. | ||
| 221 | * Abort read pipeline on EOD. | ||
| 222 | * Wait for the tape to become ready in case it returns | ||
| 223 | * "in the process of becoming ready" on open(). | ||
| 224 | * Fix zero padding of the last written block in | ||
| 225 | * case the tape block size is larger than PAGE_SIZE. | ||
| 226 | * Decrease the default disconnection time to tn. | ||
| 227 | * Ver 1.16e Oct 3 99 Minor fixes. | ||
| 228 | * Ver 1.16e1 Oct 13 99 Patches by Arnold Niessen, | ||
| 229 | * niessen@iae.nl / arnold.niessen@philips.com | ||
| 230 | * GO-1) Undefined code in idetape_read_position | ||
| 231 | * according to Gadi's email | ||
| 232 | * AJN-1) Minor fix asc == 11 should be asc == 0x11 | ||
| 233 | * in idetape_issue_packet_command (did effect | ||
| 234 | * debugging output only) | ||
| 235 | * AJN-2) Added more debugging output, and | ||
| 236 | * added ide-tape: where missing. I would also | ||
| 237 | * like to add tape->name where possible | ||
| 238 | * AJN-3) Added different debug_level's | ||
| 239 | * via /proc/ide/hdc/settings | ||
| 240 | * "debug_level" determines amount of debugging output; | ||
| 241 | * can be changed using /proc/ide/hdx/settings | ||
| 242 | * 0 : almost no debugging output | ||
| 243 | * 1 : 0+output errors only | ||
| 244 | * 2 : 1+output all sensekey/asc | ||
| 245 | * 3 : 2+follow all chrdev related procedures | ||
| 246 | * 4 : 3+follow all procedures | ||
| 247 | * 5 : 4+include pc_stack rq_stack info | ||
| 248 | * 6 : 5+USE_COUNT updates | ||
| 249 | * AJN-4) Fixed timeout for retension in idetape_queue_pc_tail | ||
| 250 | * from 5 to 10 minutes | ||
| 251 | * AJN-5) Changed maximum number of blocks to skip when | ||
| 252 | * reading tapes with multiple consecutive write | ||
| 253 | * errors from 100 to 1000 in idetape_get_logical_blk | ||
| 254 | * Proposed changes to code: | ||
| 255 | * 1) output "logical_blk_num" via /proc | ||
| 256 | * 2) output "current_operation" via /proc | ||
| 257 | * 3) Either solve or document the fact that `mt rewind' is | ||
| 258 | * required after reading from /dev/nhtx to be | ||
| 259 | * able to rmmod the idetape module; | ||
| 260 | * Also, sometimes an application finishes but the | ||
| 261 | * device remains `busy' for some time. Same cause ? | ||
| 262 | * Proposed changes to release-notes: | ||
| 263 | * 4) write a simple `quickstart' section in the | ||
| 264 | * release notes; I volunteer if you don't want to | ||
| 265 | * 5) include a pointer to video4linux in the doc | ||
| 266 | * to stimulate video applications | ||
| 267 | * 6) release notes lines 331 and 362: explain what happens | ||
| 268 | * if the application data rate is higher than 1100 KB/s; | ||
| 269 | * similar approach to lower-than-500 kB/s ? | ||
| 270 | * 7) 6.6 Comparison; wouldn't it be better to allow different | ||
| 271 | * strategies for read and write ? | ||
| 272 | * Wouldn't it be better to control the tape buffer | ||
| 273 | * contents instead of the bandwidth ? | ||
| 274 | * 8) line 536: replace will by would (if I understand | ||
| 275 | * this section correctly, a hypothetical and unwanted situation | ||
| 276 | * is being described) | ||
| 277 | * Ver 1.16f Dec 15 99 Change place of the secondary OnStream header frames. | ||
| 278 | * Ver 1.17 Nov 2000 / Jan 2001 Marcel Mol, marcel@mesa.nl | ||
| 279 | * - Add idetape_onstream_mode_sense_tape_parameter_page | ||
| 280 | * function to get tape capacity in frames: tape->capacity. | ||
| 281 | * - Add support for DI-50 drives( or any DI- drive). | ||
| 282 | * - 'workaround' for read error/blank block around block 3000. | ||
| 283 | * - Implement Early warning for end of media for Onstream. | ||
| 284 | * - Cosmetic code changes for readability. | ||
| 285 | * - Idetape_position_tape should not use SKIP bit during | ||
| 286 | * Onstream read recovery. | ||
| 287 | * - Add capacity, logical_blk_num and first/last_frame_position | ||
| 288 | * to /proc/ide/hd?/settings. | ||
| 289 | * - Module use count was gone in the Linux 2.4 driver. | ||
| 290 | * Ver 1.17a Apr 2001 Willem Riede osst@riede.org | ||
| 291 | * - Get drive's actual block size from mode sense block descriptor | ||
| 292 | * - Limit size of pipeline | ||
| 293 | * Ver 1.17b Oct 2002 Alan Stern <stern@rowland.harvard.edu> | ||
| 294 | * Changed IDETAPE_MIN_PIPELINE_STAGES to 1 and actually used | ||
| 295 | * it in the code! | ||
| 296 | * Actually removed aborted stages in idetape_abort_pipeline | ||
| 297 | * instead of just changing the command code. | ||
| 298 | * Made the transfer byte count for Request Sense equal to the | ||
| 299 | * actual length of the data transfer. | ||
| 300 | * Changed handling of partial data transfers: they do not | ||
| 301 | * cause DMA errors. | ||
| 302 | * Moved initiation of DMA transfers to the correct place. | ||
| 303 | * Removed reference to unallocated memory. | ||
| 304 | * Made __idetape_discard_read_pipeline return the number of | ||
| 305 | * sectors skipped, not the number of stages. | ||
| 306 | * Replaced errant kfree() calls with __idetape_kfree_stage(). | ||
| 307 | * Fixed off-by-one error in testing the pipeline length. | ||
| 308 | * Fixed handling of filemarks in the read pipeline. | ||
| 309 | * Small code optimization for MTBSF and MTBSFM ioctls. | ||
| 310 | * Don't try to unlock the door during device close if is | ||
| 311 | * already unlocked! | ||
| 312 | * Cosmetic fixes to miscellaneous debugging output messages. | ||
| 313 | * Set the minimum /proc/ide/hd?/settings values for "pipeline", | ||
| 314 | * "pipeline_min", and "pipeline_max" to 1. | ||
| 315 | * | ||
| 316 | * Here are some words from the first releases of hd.c, which are quoted | ||
| 317 | * in ide.c and apply here as well: | ||
| 318 | * | ||
| 319 | * | Special care is recommended. Have Fun! | ||
| 320 | * | ||
| 321 | */ | ||
| 322 | |||
| 323 | /* | ||
| 324 | * An overview of the pipelined operation mode. | ||
| 325 | * | ||
| 326 | * In the pipelined write mode, we will usually just add requests to our | ||
| 327 | * pipeline and return immediately, before we even start to service them. The | ||
| 328 | * user program will then have enough time to prepare the next request while | ||
| 329 | * we are still busy servicing previous requests. In the pipelined read mode, | ||
| 330 | * the situation is similar - we add read-ahead requests into the pipeline, | ||
| 331 | * before the user even requested them. | ||
| 332 | * | ||
| 333 | * The pipeline can be viewed as a "safety net" which will be activated when | ||
| 334 | * the system load is high and prevents the user backup program from keeping up | ||
| 335 | * with the current tape speed. At this point, the pipeline will get | ||
| 336 | * shorter and shorter but the tape will still be streaming at the same speed. | ||
| 337 | * Assuming we have enough pipeline stages, the system load will hopefully | ||
| 338 | * decrease before the pipeline is completely empty, and the backup program | ||
| 339 | * will be able to "catch up" and refill the pipeline again. | ||
| 340 | * | ||
| 341 | * When using the pipelined mode, it would be best to disable any type of | ||
| 342 | * buffering done by the user program, as ide-tape already provides all the | ||
| 343 | * benefits in the kernel, where it can be done in a more efficient way. | ||
| 344 | * As we will usually not block the user program on a request, the most | ||
| 345 | * efficient user code will then be a simple read-write-read-... cycle. | ||
| 346 | * Any additional logic will usually just slow down the backup process. | ||
| 347 | * | ||
| 348 | * Using the pipelined mode, I get a constant over 400 KBps throughput, | ||
| 349 | * which seems to be the maximum throughput supported by my tape. | ||
| 350 | * | ||
| 351 | * However, there are some downfalls: | ||
| 352 | * | ||
| 353 | * 1. We use memory (for data buffers) in proportional to the number | ||
| 354 | * of pipeline stages (each stage is about 26 KB with my tape). | ||
| 355 | * 2. In the pipelined write mode, we cheat and postpone error codes | ||
| 356 | * to the user task. In read mode, the actual tape position | ||
| 357 | * will be a bit further than the last requested block. | ||
| 358 | * | ||
| 359 | * Concerning (1): | ||
| 360 | * | ||
| 361 | * 1. We allocate stages dynamically only when we need them. When | ||
| 362 | * we don't need them, we don't consume additional memory. In | ||
| 363 | * case we can't allocate stages, we just manage without them | ||
| 364 | * (at the expense of decreased throughput) so when Linux is | ||
| 365 | * tight in memory, we will not pose additional difficulties. | ||
| 366 | * | ||
| 367 | * 2. The maximum number of stages (which is, in fact, the maximum | ||
| 368 | * amount of memory) which we allocate is limited by the compile | ||
| 369 | * time parameter IDETAPE_MAX_PIPELINE_STAGES. | ||
| 370 | * | ||
| 371 | * 3. The maximum number of stages is a controlled parameter - We | ||
| 372 | * don't start from the user defined maximum number of stages | ||
| 373 | * but from the lower IDETAPE_MIN_PIPELINE_STAGES (again, we | ||
| 374 | * will not even allocate this amount of stages if the user | ||
| 375 | * program can't handle the speed). We then implement a feedback | ||
| 376 | * loop which checks if the pipeline is empty, and if it is, we | ||
| 377 | * increase the maximum number of stages as necessary until we | ||
| 378 | * reach the optimum value which just manages to keep the tape | ||
| 379 | * busy with minimum allocated memory or until we reach | ||
| 380 | * IDETAPE_MAX_PIPELINE_STAGES. | ||
| 381 | * | ||
| 382 | * Concerning (2): | ||
| 383 | * | ||
| 384 | * In pipelined write mode, ide-tape can not return accurate error codes | ||
| 385 | * to the user program since we usually just add the request to the | ||
| 386 | * pipeline without waiting for it to be serviced. In case an error | ||
| 387 | * occurs, I will report it on the next user request. | ||
| 388 | * | ||
| 389 | * In the pipelined read mode, subsequent read requests or forward | ||
| 390 | * filemark spacing will perform correctly, as we preserve all blocks | ||
| 391 | * and filemarks which we encountered during our excess read-ahead. | ||
| 392 | * | ||
| 393 | * For accurate tape positioning and error reporting, disabling | ||
| 394 | * pipelined mode might be the best option. | ||
| 395 | * | ||
| 396 | * You can enable/disable/tune the pipelined operation mode by adjusting | ||
| 397 | * the compile time parameters below. | ||
| 398 | */ | ||
| 399 | |||
| 400 | /* | ||
| 401 | * Possible improvements. | ||
| 402 | * | ||
| 403 | * 1. Support for the ATAPI overlap protocol. | ||
| 404 | * | ||
| 405 | * In order to maximize bus throughput, we currently use the DSC | ||
| 406 | * overlap method which enables ide.c to service requests from the | ||
| 407 | * other device while the tape is busy executing a command. The | ||
| 408 | * DSC overlap method involves polling the tape's status register | ||
| 409 | * for the DSC bit, and servicing the other device while the tape | ||
| 410 | * isn't ready. | ||
| 411 | * | ||
| 412 | * In the current QIC development standard (December 1995), | ||
| 413 | * it is recommended that new tape drives will *in addition* | ||
| 414 | * implement the ATAPI overlap protocol, which is used for the | ||
| 415 | * same purpose - efficient use of the IDE bus, but is interrupt | ||
| 416 | * driven and thus has much less CPU overhead. | ||
| 417 | * | 13 | * |
| 418 | * ATAPI overlap is likely to be supported in most new ATAPI | 14 | * For a historical changelog see |
| 419 | * devices, including new ATAPI cdroms, and thus provides us | 15 | * Documentation/ide/ChangeLog.ide-tape.1995-2002 |
| 420 | * a method by which we can achieve higher throughput when | ||
| 421 | * sharing a (fast) ATA-2 disk with any (slow) new ATAPI device. | ||
| 422 | */ | 16 | */ |
| 423 | 17 | ||
| 424 | #define IDETAPE_VERSION "1.19" | 18 | #define IDETAPE_VERSION "1.19" |