diff options
Diffstat (limited to 'Documentation/ide')
-rw-r--r-- | Documentation/ide/ChangeLog.ide-tape.1995-2002 | 257 | ||||
-rw-r--r-- | Documentation/ide/ide-tape.txt | 146 |
2 files changed, 403 insertions, 0 deletions
diff --git a/Documentation/ide/ChangeLog.ide-tape.1995-2002 b/Documentation/ide/ChangeLog.ide-tape.1995-2002 new file mode 100644 index 000000000000..877fac8770b3 --- /dev/null +++ b/Documentation/ide/ChangeLog.ide-tape.1995-2002 @@ -0,0 +1,257 @@ +/* + * Ver 0.1 Nov 1 95 Pre-working code :-) + * Ver 0.2 Nov 23 95 A short backup (few megabytes) and restore procedure + * was successful ! (Using tar cvf ... on the block + * device interface). + * A longer backup resulted in major swapping, bad + * overall Linux performance and eventually failed as + * we received non serial read-ahead requests from the + * buffer cache. + * Ver 0.3 Nov 28 95 Long backups are now possible, thanks to the + * character device interface. Linux's responsiveness + * and performance doesn't seem to be much affected + * from the background backup procedure. + * Some general mtio.h magnetic tape operations are + * now supported by our character device. As a result, + * popular tape utilities are starting to work with + * ide tapes :-) + * The following configurations were tested: + * 1. An IDE ATAPI TAPE shares the same interface + * and irq with an IDE ATAPI CDROM. + * 2. An IDE ATAPI TAPE shares the same interface + * and irq with a normal IDE disk. + * Both configurations seemed to work just fine ! + * However, to be on the safe side, it is meanwhile + * recommended to give the IDE TAPE its own interface + * and irq. + * The one thing which needs to be done here is to + * add a "request postpone" feature to ide.c, + * so that we won't have to wait for the tape to finish + * performing a long media access (DSC) request (such + * as a rewind) before we can access the other device + * on the same interface. This effect doesn't disturb + * normal operation most of the time because read/write + * requests are relatively fast, and once we are + * performing one tape r/w request, a lot of requests + * from the other device can be queued and ide.c will + * service all of them after this single tape request. + * Ver 1.0 Dec 11 95 Integrated into Linux 1.3.46 development tree. + * On each read / write request, we now ask the drive + * if we can transfer a constant number of bytes + * (a parameter of the drive) only to its buffers, + * without causing actual media access. If we can't, + * we just wait until we can by polling the DSC bit. + * This ensures that while we are not transferring + * more bytes than the constant referred to above, the + * interrupt latency will not become too high and + * we won't cause an interrupt timeout, as happened + * occasionally in the previous version. + * While polling for DSC, the current request is + * postponed and ide.c is free to handle requests from + * the other device. This is handled transparently to + * ide.c. The hwgroup locking method which was used + * in the previous version was removed. + * Use of new general features which are provided by + * ide.c for use with atapi devices. + * (Programming done by Mark Lord) + * Few potential bug fixes (Again, suggested by Mark) + * Single character device data transfers are now + * not limited in size, as they were before. + * We are asking the tape about its recommended + * transfer unit and send a larger data transfer + * as several transfers of the above size. + * For best results, use an integral number of this + * basic unit (which is shown during driver + * initialization). I will soon add an ioctl to get + * this important parameter. + * Our data transfer buffer is allocated on startup, + * rather than before each data transfer. This should + * ensure that we will indeed have a data buffer. + * Ver 1.1 Dec 14 95 Fixed random problems which occurred when the tape + * shared an interface with another device. + * (poll_for_dsc was a complete mess). + * Removed some old (non-active) code which had + * to do with supporting buffer cache originated + * requests. + * The block device interface can now be opened, so + * that general ide driver features like the unmask + * interrupts flag can be selected with an ioctl. + * This is the only use of the block device interface. + * New fast pipelined operation mode (currently only on + * writes). When using the pipelined mode, the + * throughput can potentially reach the maximum + * tape supported throughput, regardless of the + * user backup program. On my tape drive, it sometimes + * boosted performance by a factor of 2. Pipelined + * mode is enabled by default, but since it has a few + * downfalls as well, you may want to disable it. + * A short explanation of the pipelined operation mode + * is available below. + * Ver 1.2 Jan 1 96 Eliminated pipelined mode race condition. + * Added pipeline read mode. As a result, restores + * are now as fast as backups. + * Optimized shared interface behavior. The new behavior + * typically results in better IDE bus efficiency and + * higher tape throughput. + * Pre-calculation of the expected read/write request + * service time, based on the tape's parameters. In + * the pipelined operation mode, this allows us to + * adjust our polling frequency to a much lower value, + * and thus to dramatically reduce our load on Linux, + * without any decrease in performance. + * Implemented additional mtio.h operations. + * The recommended user block size is returned by + * the MTIOCGET ioctl. + * Additional minor changes. + * Ver 1.3 Feb 9 96 Fixed pipelined read mode bug which prevented the + * use of some block sizes during a restore procedure. + * The character device interface will now present a + * continuous view of the media - any mix of block sizes + * during a backup/restore procedure is supported. The + * driver will buffer the requests internally and + * convert them to the tape's recommended transfer + * unit, making performance almost independent of the + * chosen user block size. + * Some improvements in error recovery. + * By cooperating with ide-dma.c, bus mastering DMA can + * now sometimes be used with IDE tape drives as well. + * Bus mastering DMA has the potential to dramatically + * reduce the CPU's overhead when accessing the device, + * and can be enabled by using hdparm -d1 on the tape's + * block device interface. For more info, read the + * comments in ide-dma.c. + * Ver 1.4 Mar 13 96 Fixed serialize support. + * Ver 1.5 Apr 12 96 Fixed shared interface operation, broken in 1.3.85. + * Fixed pipelined read mode inefficiency. + * Fixed nasty null dereferencing bug. + * Ver 1.6 Aug 16 96 Fixed FPU usage in the driver. + * Fixed end of media bug. + * Ver 1.7 Sep 10 96 Minor changes for the CONNER CTT8000-A model. + * Ver 1.8 Sep 26 96 Attempt to find a better balance between good + * interactive response and high system throughput. + * Ver 1.9 Nov 5 96 Automatically cross encountered filemarks rather + * than requiring an explicit FSF command. + * Abort pending requests at end of media. + * MTTELL was sometimes returning incorrect results. + * Return the real block size in the MTIOCGET ioctl. + * Some error recovery bug fixes. + * Ver 1.10 Nov 5 96 Major reorganization. + * Reduced CPU overhead a bit by eliminating internal + * bounce buffers. + * Added module support. + * Added multiple tape drives support. + * Added partition support. + * Rewrote DSC handling. + * Some portability fixes. + * Removed ide-tape.h. + * Additional minor changes. + * Ver 1.11 Dec 2 96 Bug fix in previous DSC timeout handling. + * Use ide_stall_queue() for DSC overlap. + * Use the maximum speed rather than the current speed + * to compute the request service time. + * Ver 1.12 Dec 7 97 Fix random memory overwriting and/or last block data + * corruption, which could occur if the total number + * of bytes written to the tape was not an integral + * number of tape blocks. + * Add support for INTERRUPT DRQ devices. + * Ver 1.13 Jan 2 98 Add "speed == 0" work-around for HP COLORADO 5GB + * Ver 1.14 Dec 30 98 Partial fixes for the Sony/AIWA tape drives. + * Replace cli()/sti() with hwgroup spinlocks. + * Ver 1.15 Mar 25 99 Fix SMP race condition by replacing hwgroup + * spinlock with private per-tape spinlock. + * Ver 1.16 Sep 1 99 Add OnStream tape support. + * Abort read pipeline on EOD. + * Wait for the tape to become ready in case it returns + * "in the process of becoming ready" on open(). + * Fix zero padding of the last written block in + * case the tape block size is larger than PAGE_SIZE. + * Decrease the default disconnection time to tn. + * Ver 1.16e Oct 3 99 Minor fixes. + * Ver 1.16e1 Oct 13 99 Patches by Arnold Niessen, + * niessen@iae.nl / arnold.niessen@philips.com + * GO-1) Undefined code in idetape_read_position + * according to Gadi's email + * AJN-1) Minor fix asc == 11 should be asc == 0x11 + * in idetape_issue_packet_command (did effect + * debugging output only) + * AJN-2) Added more debugging output, and + * added ide-tape: where missing. I would also + * like to add tape->name where possible + * AJN-3) Added different debug_level's + * via /proc/ide/hdc/settings + * "debug_level" determines amount of debugging output; + * can be changed using /proc/ide/hdx/settings + * 0 : almost no debugging output + * 1 : 0+output errors only + * 2 : 1+output all sensekey/asc + * 3 : 2+follow all chrdev related procedures + * 4 : 3+follow all procedures + * 5 : 4+include pc_stack rq_stack info + * 6 : 5+USE_COUNT updates + * AJN-4) Fixed timeout for retension in idetape_queue_pc_tail + * from 5 to 10 minutes + * AJN-5) Changed maximum number of blocks to skip when + * reading tapes with multiple consecutive write + * errors from 100 to 1000 in idetape_get_logical_blk + * Proposed changes to code: + * 1) output "logical_blk_num" via /proc + * 2) output "current_operation" via /proc + * 3) Either solve or document the fact that `mt rewind' is + * required after reading from /dev/nhtx to be + * able to rmmod the idetape module; + * Also, sometimes an application finishes but the + * device remains `busy' for some time. Same cause ? + * Proposed changes to release-notes: + * 4) write a simple `quickstart' section in the + * release notes; I volunteer if you don't want to + * 5) include a pointer to video4linux in the doc + * to stimulate video applications + * 6) release notes lines 331 and 362: explain what happens + * if the application data rate is higher than 1100 KB/s; + * similar approach to lower-than-500 kB/s ? + * 7) 6.6 Comparison; wouldn't it be better to allow different + * strategies for read and write ? + * Wouldn't it be better to control the tape buffer + * contents instead of the bandwidth ? + * 8) line 536: replace will by would (if I understand + * this section correctly, a hypothetical and unwanted situation + * is being described) + * Ver 1.16f Dec 15 99 Change place of the secondary OnStream header frames. + * Ver 1.17 Nov 2000 / Jan 2001 Marcel Mol, marcel@mesa.nl + * - Add idetape_onstream_mode_sense_tape_parameter_page + * function to get tape capacity in frames: tape->capacity. + * - Add support for DI-50 drives( or any DI- drive). + * - 'workaround' for read error/blank block around block 3000. + * - Implement Early warning for end of media for Onstream. + * - Cosmetic code changes for readability. + * - Idetape_position_tape should not use SKIP bit during + * Onstream read recovery. + * - Add capacity, logical_blk_num and first/last_frame_position + * to /proc/ide/hd?/settings. + * - Module use count was gone in the Linux 2.4 driver. + * Ver 1.17a Apr 2001 Willem Riede osst@riede.org + * - Get drive's actual block size from mode sense block descriptor + * - Limit size of pipeline + * Ver 1.17b Oct 2002 Alan Stern <stern@rowland.harvard.edu> + * Changed IDETAPE_MIN_PIPELINE_STAGES to 1 and actually used + * it in the code! + * Actually removed aborted stages in idetape_abort_pipeline + * instead of just changing the command code. + * Made the transfer byte count for Request Sense equal to the + * actual length of the data transfer. + * Changed handling of partial data transfers: they do not + * cause DMA errors. + * Moved initiation of DMA transfers to the correct place. + * Removed reference to unallocated memory. + * Made __idetape_discard_read_pipeline return the number of + * sectors skipped, not the number of stages. + * Replaced errant kfree() calls with __idetape_kfree_stage(). + * Fixed off-by-one error in testing the pipeline length. + * Fixed handling of filemarks in the read pipeline. + * Small code optimization for MTBSF and MTBSFM ioctls. + * Don't try to unlock the door during device close if is + * already unlocked! + * Cosmetic fixes to miscellaneous debugging output messages. + * Set the minimum /proc/ide/hd?/settings values for "pipeline", + * "pipeline_min", and "pipeline_max" to 1. + */ diff --git a/Documentation/ide/ide-tape.txt b/Documentation/ide/ide-tape.txt new file mode 100644 index 000000000000..658f271a373f --- /dev/null +++ b/Documentation/ide/ide-tape.txt @@ -0,0 +1,146 @@ +/* + * IDE ATAPI streaming tape driver. + * + * This driver is a part of the Linux ide driver. + * + * The driver, in co-operation with ide.c, basically traverses the + * request-list for the block device interface. The character device + * interface, on the other hand, creates new requests, adds them + * to the request-list of the block device, and waits for their completion. + * + * Pipelined operation mode is now supported on both reads and writes. + * + * The block device major and minor numbers are determined from the + * tape's relative position in the ide interfaces, as explained in ide.c. + * + * The character device interface consists of the following devices: + * + * ht0 major 37, minor 0 first IDE tape, rewind on close. + * ht1 major 37, minor 1 second IDE tape, rewind on close. + * ... + * nht0 major 37, minor 128 first IDE tape, no rewind on close. + * nht1 major 37, minor 129 second IDE tape, no rewind on close. + * ... + * + * The general magnetic tape commands compatible interface, as defined by + * include/linux/mtio.h, is accessible through the character device. + * + * General ide driver configuration options, such as the interrupt-unmask + * flag, can be configured by issuing an ioctl to the block device interface, + * as any other ide device. + * + * Our own ide-tape ioctl's can be issued to either the block device or + * the character device interface. + * + * Maximal throughput with minimal bus load will usually be achieved in the + * following scenario: + * + * 1. ide-tape is operating in the pipelined operation mode. + * 2. No buffering is performed by the user backup program. + * + * Testing was done with a 2 GB CONNER CTMA 4000 IDE ATAPI Streaming Tape Drive. + * + * Here are some words from the first releases of hd.c, which are quoted + * in ide.c and apply here as well: + * + * | Special care is recommended. Have Fun! + * + * + * An overview of the pipelined operation mode. + * + * In the pipelined write mode, we will usually just add requests to our + * pipeline and return immediately, before we even start to service them. The + * user program will then have enough time to prepare the next request while + * we are still busy servicing previous requests. In the pipelined read mode, + * the situation is similar - we add read-ahead requests into the pipeline, + * before the user even requested them. + * + * The pipeline can be viewed as a "safety net" which will be activated when + * the system load is high and prevents the user backup program from keeping up + * with the current tape speed. At this point, the pipeline will get + * shorter and shorter but the tape will still be streaming at the same speed. + * Assuming we have enough pipeline stages, the system load will hopefully + * decrease before the pipeline is completely empty, and the backup program + * will be able to "catch up" and refill the pipeline again. + * + * When using the pipelined mode, it would be best to disable any type of + * buffering done by the user program, as ide-tape already provides all the + * benefits in the kernel, where it can be done in a more efficient way. + * As we will usually not block the user program on a request, the most + * efficient user code will then be a simple read-write-read-... cycle. + * Any additional logic will usually just slow down the backup process. + * + * Using the pipelined mode, I get a constant over 400 KBps throughput, + * which seems to be the maximum throughput supported by my tape. + * + * However, there are some downfalls: + * + * 1. We use memory (for data buffers) in proportional to the number + * of pipeline stages (each stage is about 26 KB with my tape). + * 2. In the pipelined write mode, we cheat and postpone error codes + * to the user task. In read mode, the actual tape position + * will be a bit further than the last requested block. + * + * Concerning (1): + * + * 1. We allocate stages dynamically only when we need them. When + * we don't need them, we don't consume additional memory. In + * case we can't allocate stages, we just manage without them + * (at the expense of decreased throughput) so when Linux is + * tight in memory, we will not pose additional difficulties. + * + * 2. The maximum number of stages (which is, in fact, the maximum + * amount of memory) which we allocate is limited by the compile + * time parameter IDETAPE_MAX_PIPELINE_STAGES. + * + * 3. The maximum number of stages is a controlled parameter - We + * don't start from the user defined maximum number of stages + * but from the lower IDETAPE_MIN_PIPELINE_STAGES (again, we + * will not even allocate this amount of stages if the user + * program can't handle the speed). We then implement a feedback + * loop which checks if the pipeline is empty, and if it is, we + * increase the maximum number of stages as necessary until we + * reach the optimum value which just manages to keep the tape + * busy with minimum allocated memory or until we reach + * IDETAPE_MAX_PIPELINE_STAGES. + * + * Concerning (2): + * + * In pipelined write mode, ide-tape can not return accurate error codes + * to the user program since we usually just add the request to the + * pipeline without waiting for it to be serviced. In case an error + * occurs, I will report it on the next user request. + * + * In the pipelined read mode, subsequent read requests or forward + * filemark spacing will perform correctly, as we preserve all blocks + * and filemarks which we encountered during our excess read-ahead. + * + * For accurate tape positioning and error reporting, disabling + * pipelined mode might be the best option. + * + * You can enable/disable/tune the pipelined operation mode by adjusting + * the compile time parameters below. + * + * + * Possible improvements. + * + * 1. Support for the ATAPI overlap protocol. + * + * In order to maximize bus throughput, we currently use the DSC + * overlap method which enables ide.c to service requests from the + * other device while the tape is busy executing a command. The + * DSC overlap method involves polling the tape's status register + * for the DSC bit, and servicing the other device while the tape + * isn't ready. + * + * In the current QIC development standard (December 1995), + * it is recommended that new tape drives will *in addition* + * implement the ATAPI overlap protocol, which is used for the + * same purpose - efficient use of the IDE bus, but is interrupt + * driven and thus has much less CPU overhead. + * + * ATAPI overlap is likely to be supported in most new ATAPI + * devices, including new ATAPI cdroms, and thus provides us + * a method by which we can achieve higher throughput when + * sharing a (fast) ATA-2 disk with any (slow) new ATAPI device. + */ |