| Borislav Petkov | 1056971 | 2008-04-27 15:38:28 +0200 | [diff] [blame] | 1 | IDE ATAPI streaming tape driver. | 
 | 2 |  | 
 | 3 | This driver is a part of the Linux ide driver. | 
 | 4 |  | 
 | 5 | The driver, in co-operation with ide.c, basically traverses the | 
 | 6 | request-list for the block device interface. The character device | 
 | 7 | interface, on the other hand, creates new requests, adds them | 
 | 8 | to the request-list of the block device, and waits for their completion. | 
 | 9 |  | 
 | 10 | The block device major and minor numbers are determined from the | 
 | 11 | tape's relative position in the ide interfaces, as explained in ide.c. | 
 | 12 |  | 
 | 13 | The character device interface consists of the following devices: | 
 | 14 |  | 
 | 15 | ht0		major 37, minor 0	first  IDE tape, rewind on close. | 
 | 16 | ht1		major 37, minor 1	second IDE tape, rewind on close. | 
 | 17 | ... | 
 | 18 | nht0		major 37, minor 128	first  IDE tape, no rewind on close. | 
 | 19 | nht1		major 37, minor 129	second IDE tape, no rewind on close. | 
 | 20 | ... | 
 | 21 |  | 
 | 22 | The general magnetic tape commands compatible interface, as defined by | 
 | 23 | include/linux/mtio.h, is accessible through the character device. | 
 | 24 |  | 
 | 25 | General ide driver configuration options, such as the interrupt-unmask | 
 | 26 | flag, can be configured by issuing an ioctl to the block device interface, | 
 | 27 | as any other ide device. | 
 | 28 |  | 
 | 29 | Our own ide-tape ioctl's can be issued to either the block device or | 
 | 30 | the character device interface. | 
 | 31 |  | 
 | 32 | Maximal throughput with minimal bus load will usually be achieved in the | 
 | 33 | following scenario: | 
 | 34 |  | 
 | 35 |      1.	ide-tape is operating in the pipelined operation mode. | 
 | 36 |      2.	No buffering is performed by the user backup program. | 
 | 37 |  | 
 | 38 | Testing was done with a 2 GB CONNER CTMA 4000 IDE ATAPI Streaming Tape Drive. | 
 | 39 |  | 
 | 40 | Here are some words from the first releases of hd.c, which are quoted | 
 | 41 | in ide.c and apply here as well: | 
 | 42 |  | 
 | 43 | | Special care is recommended.  Have Fun! | 
 | 44 |  | 
 | 45 | Possible improvements: | 
 | 46 |  | 
 | 47 | 1. Support for the ATAPI overlap protocol. | 
 | 48 |  | 
 | 49 | In order to maximize bus throughput, we currently use the DSC | 
 | 50 | overlap method which enables ide.c to service requests from the | 
 | 51 | other device while the tape is busy executing a command. The | 
 | 52 | DSC overlap method involves polling the tape's status register | 
 | 53 | for the DSC bit, and servicing the other device while the tape | 
 | 54 | isn't ready. | 
 | 55 |  | 
 | 56 | In the current QIC development standard (December 1995), | 
 | 57 | it is recommended that new tape drives will *in addition* | 
 | 58 | implement the ATAPI overlap protocol, which is used for the | 
 | 59 | same purpose - efficient use of the IDE bus, but is interrupt | 
 | 60 | driven and thus has much less CPU overhead. | 
 | 61 |  | 
 | 62 | ATAPI overlap is likely to be supported in most new ATAPI | 
 | 63 | devices, including new ATAPI cdroms, and thus provides us | 
 | 64 | a method by which we can achieve higher throughput when | 
 | 65 | sharing a (fast) ATA-2 disk with any (slow) new ATAPI device. |