Daniela's EIDE/SATA Driver

Daniela's S506 ADD - Release 1.8.15 ----------------------------------- NAME DaniS506.ADD - replacement for IBM1S506.ADD SYNOPSIS EIDE driver with support for Intel, VIA, ALi, SiS, CMD, Promise, Cyrix, HighPoint, AMD, Artop, SMSC, ServerWorks, Opti, Nvidia, National Semiconduc- tor, SiliconImage, ITE, Adaptec, ATI, Broadcom, JMicron, Marvell chipsets, and generic busmaster support for all SFF-8038i compliant chips. This covers virtually *any* EIDE controller built/sold in the past few years and even some just about to be released in the near future. Don't be confused by the generic chip names below! If you are uncertain about the actual controller to be supported, run the DumpIDE utility - it will tell you. See ChipList.txt for a full list of supported chipsets. ### note: ### This driver is *not* compatible to any of the software RAID setups offered ### by some so called IDE RAID or SATA RAID controllers (f.e. HPT, Promise ### FastTrack, AMI, VIA, Intel, SiS, you name it). None of these "fake RAIDs" ### have any hardware RAID features built in! ### Of course, you may run these chips in non-RAID BIOS setups (i.e. JBOD). *** Changes to the previous release are denoted by "***" tags. DESCRIPTION Since the original IBM1S506.ADD doesn't support the IDE controller found in the VIA 82C586(/A/B) south bridge beyond the basic features, I wasn't very happy with the performance of my 10Gig DTTA-351010 drive. The driver from VIA itself was no solution because it doesn't support all the advanced features found in todays hardware and locked up my system completely. So I sat down.... This release of my S506 driver is no longer based on the sources provided by IBM on their DDK site. It has the latest feature set as found in the current fixpacks and fixes later than these. My enhancements are: - (supports the Intel PIIX, just like IBM1S506 does) plus some PIIX4 variants unsupported by IBM1S506 - supports the Intel ICH/ICH0/ICH2/ICH3/ICH4/C-ICH/ICH5/ICH6/ICH7/ICH8 PATA controllers - supports the Intel ICH5/ICH6/ICH7/ICH8/ICH9/ICH10 SATA controllers - supports the Intel PCH SATA controllers - supports the Intel Management Engine IDE redirection controllers - supports the Intel SCH PATA controller ("Atom") (experimental) - supports the VIA PATA controllers - supports the VIA PATA/SATA RAID controllers (see note ###) - supports the ALi PATA controllers - supports the ALi 5281/5287/5289 PATA/SATA RAID controllers (see note ###) - supports the SiS 5513/5517/5518/180 PATA controllers - supports the SiS 180/181/182/1180/1182/1183/1184 (182 and above exp.) SATA RAID controllers (see note ###) - supports the CMD 643/646/648/649 PATA controllers (this includes OEMs like AMI) - supports the Promise Ultra33/66/100/TX2/133 PATA controllers - supports the Cyrix 5530 PATA controller (Cyrix MediaGX) - supports the HighPoint HPT302/366/368/370/371/372/374 PATA controllers - supports the AMD756/766/768/8111/GeodeLX PATA controllers - supports the ACARD ATP850/860/865/867 PATA controllers (ATP860 experimental) - supports the SMSC SLC90E66 PATA controller (PIIX4 look-alike) (exp.) - supports the ServerWorks/Broadcom PATA controllers - supports the Opti PATA controllers - supports the Nvidia nForce/2/3/4/500/600/700 PATA controllers important fix for broken BIOSes which cause system hangs - supports the Nvidia nForce2/3/4/500/600/700/M73/M79/M89 SATA controllers - supports the National Semiconductor Geode SCx200 PATA controller - supports the Silicon Image 680 PATA controller - supports the Silicon Image 3112/3114/3512 SATA controller (this includes OEMs like Adaptec) - supports the ITE 8211/8212/8213 PATA controllers (see note ###) - supports the ATI IXP200/300/400/600/700 PATA controllers - supports the ATI IXP300/400/600/700/800 SATA controllers - supports the NetCell/Revolution SyncRAID PATA/SATA Hardware-RAID Controller - supports the JMicron JMB360/361/363/365/366/368 PATA/SATA controllers - supports the Marvell 61xx PATA controllers - generic busmaster support for SFF-8038i compliant controllers - supports SATA <-> PATA adapters - supports SATA/SATAPI hotplugging - supports PCCard (formerly: PCMCIA) ATA/ATAPI controllers - supports ATA-Flash, Compact-Flash, CF+, SmartMedia, etc. in PCCard slots - supports internal DriveBays - supports the external DriveBay and other PCCard attached units - no need for chipset initialization by the BIOS (Intel, VIA, ALi, SiS, CMD, Promise, and HighPoint) - support for Ultra33/66/100/133 DMA modes according to the chip feature matrix (ChipList.txt) - support for 1.5GBit/s and 3.0 GBit/s SATA modes according to the chip feature matrix (ChipList.txt) - faster PIO due to 32bit IO - adjustable disk power management (device APM) - adjustable SATA link power management - sets device inactivity timers (POWERSV.SYS is no longer needed) - adjustable SMS length - adjustable drive noise level - optional write protection of the first track of a disk drive - unlocks full capacity of disks which are programmed to report a lower capacity to be compatible with older BIOSes - improved support for removable media drives (in particular Castlewood ORB EIDE and the Fujitsu M2541B MO drive) - supports devices which don't spin up automatically - optional limit data transfer rates for more reliability - adapts (feature removed: auto-adapts) chip timings to PCI clock - supposed to work with drive capacities up to 2 TiB (OS/2 ADD limit), tested with drives up to 2 TiB now - improved compatibility with current BIOSes (IBM1S506.ADD fails on an increasing number of recent BIOS implementations) - performance and feature tuning options - up to 16 PATA/SATA channels - enhanced S.M.A.R.T support - improved stability in case of SATA PHY communication errors - some fixes for bugs introduced with earlier releases - some fixes for bugs and omissions in the IBM code ;-) - work around some BIOS geometry issues - fixes possible data corruption with some VIA chips. - fixes a hang during initialization of some units which must not be touched for some time after completion of a couple of commands - support for high interrupts (IRQ16 and above) in APIC interrupt mode this requires ACPI.PSD v3 or later! - supports AHCI controllers in SATA legacy mode, when available - coexists with drivers that acquire controllers known to Danis506 including the soon to be released AHCI controller driver from Mensys. - fixes some SMP compatibility issues (thanks to Slavik Gnatenko) - rework SMP compatibility updates to avoid spinlocks (thanks to Slavik Gnatenko) *** - resolve ticket #8 (8086:2411 ICH1 controller #1 w/ATAPI device detect) *** - resolve ticket #8 (8086:244B ICH2 controller #1 w/ATAPI device detect) *** - fix Paulsbro SCH setup (8086:811A) to require word aligned DMA *** - correct ProbeChannel svn ticket #8 regression, fixes svn #16 bugtracker #3445, #3419 etc. ### Caution: users of DaniBoot v1.0 need to upgrade to version 1.0.2! ### Caution: users of DaniATAPI v0.3.x need to upgrade to version 0.4.0! ### Caution: other ATAPI related filter drivers no longer work due to necessary inter-driver interface changes! COMMAND LINE OPTIONS This driver supports most switches and options that IBM1S506 supports. For an explanation of these look into the OS/2 online help (by picking 'OS/2 Commands by name'/ 'BASEDEV...'/'IDE...' or by "help ibm1s506" from the command line). COMMAND LINE OVERVIEW Gerald Eischer did us the favour of creating a syntax diagram and a short overview of the command line options implemented in the driver Syntax Diagram -------------- +------------------+ +-- /I --+ +---------------+ V | V | V | BASEDEV=DaniS506.ADD --+----------------++-- /A:nn -+---+-------------++-> +-+- /!V -+------+ ^ +- /P:hhh ----+ | +- /!W -+ | | +- /IRQ:n ----+ +--- /!AA -------+ +--------+ +- /!BM ------+ +--- /PCLK:x ----+ | +- /FORCEGBM -+ +-+- /GBM -----+-+ | +- /PF:xxxx --+ | +- /FORCEGBM + | | +- /80WIRE ---+ +--- /SHUTDOWN --+ | +- /TO:nn ----+ +--- /DEBUG:n ---+ | +- /BAY ------+ +--- /!BIOS -----+ | +- /IT:x -----+ +--- /PCS:n -----+ | +- /LOC:spec -+ +--- /!BM -------+ | +--- /APM:n --+ +--- /PF:xxxx ---+ | +--- /LPM:n --+ +--- /IT:x ------+ | +--- /APM:n -----+ | +--- /LPM:n -----+ | | >--------+----------------------------------------+------------------| | ^ | +----------------------------------+ | | +--/I---+ +----------------------+ | | | V | V | | | +-+- /U:n +--+-------------------+-+-+-+ +- /T:n ------------+ +- /!SMS -----------+ +- /!DM ------------+ +- /IT:x -----------+ +- /MR:udp ---------+ +- /PF:xxxx --------+ +- /NL:n -----------+ +- /!SETMAX --------+ +- /WP:x -----------+ +- /!BM ------------+ +--- /APM:n --------+ +--- /LPM:n --------+ In most cases, options prefaced with ! can be used with the ! modifier to explicitly disable the option. Options are valid both with and without the ! modifer. The values in parentheses are the default values if the option is not present. Options may be prefixed by "-" instead of "/" as well. General options --------------- /!V /VL /VLL (quiet) (not) Verbose mode (less/lesser verbose) /!W /WL /WLL (quiet) same as /!V, but pauses screen /!AA (beep) No Audible Alerts /PCLK:x (33MHz) PCI Clock (DECIMAL) /GBM (off) Generic Busmaster Mode /FORCEGBM (off) Force Generic Busmaster mode /!SHUTDOWN (not) (not) Shutdown each attached device /DEBUG:n (0) Set debug level (DECIMAL) /!BIOS (call BIOS Disable all BIOS Int13 calls if required) /PCS (0) Support all found PCCard Slots /PCS:n (0) Support at most n PCCard Slots /!BM (ATA : enable) (not) Enable BusMaster DMA on all adapters (ATAPI: enable) /PF:xxxx (0) Register access timings /IT:x (BIOS) Inactivity timer (DECIMAL) /APM:x (BIOS) Advanced Power Management Level (DECIMAL) 1 (max. power saving) ... 254 (max. performance) /LPM:x (BIOS) SATA Link Power Management Level (DECIMAL) 0 (max. power saving) ... 3 (disabled, max. performance) Adapter options --------------- /A:nn (n/a) Adapter number (0, 1, .. 15) /LOC:bus:device.function#channel (n/a) assign PCI device and channel to this adapter /P:hhh (PCI) Base port address (HEX) /PORT:hhh same as /P /IRQ:n (PCI) IRQ level (DECIMAL) /I (off) Ignore adapter /!BM (ATA : enable) (not) Enable BusMaster DMA (ATAPI: enable) /PF:xxxx (0) Register access timings /80WIRE (autodetect) 80 wire cable present /FORCEGBM (off) Force Generic Busmaster mode /TO:n (DECIMAL) (5s/dynamic) Interrupt timeout /BAY (off) Adapter is connected to a drive bay /IT:x (BIOS) Inactivity timer (DECIMAL) /APM:x (BIOS) Advanced Power Management Level (DECIMAL) 1 (max. power saving) ... 254 (max. performance) /LPM:x (BIOS) SATA Link Power Management Level (DECIMAL) 0 (max. power saving) ... 3 (disabled, max. performance) Unit options ------------ /U:n (n/a) Unit number (0, 1) /UNIT:n same as /U /I (off) Ignore unit /T:n (5s) Error retry time in seconds (DECIMAL) /!SMS (autodetect) Disable block mode /SMS:n (autodetect) Set SMS block length to n /!DM (off) Prevents OS2DASD.DMD from managing the device /IT:x (BIOS) Inactivity timer (DECIMAL) /MR:udp (maximum) Limit data rate of a device /PF:xxxx (0) Register access timings /NL:n (current device noise level, n = 0 (faint) ... 126 (noisy) setting) /!SETMAX (enable unlock) disable unlocking of full drive capacity /WP, /WP:0 (off) enable write protection of complete disk or track 0 only /!BM (ATA : enable) (not) Enable BusMaster DMA (ATAPI: enable) /APM:x (BIOS) Advanced Power Management Level (DECIMAL) 1 (max. power saving) ... 254 (max. performance) /LPM:x (BIOS) SATA Link Power Management Level (DECIMAL) 0 (max. power saving) ... 3 (disabled, max. performance) Deprecated options ------------------ These options are no longer supported. If used, Danis506 will silently ignore the option. /!RMV Obsolete options ---------------- These options are no longer supported. Danis506 will report an error if these are used. /GEO /LBA /!CYLA /DC:n /!DSG /DT:x /NEC /PCLK:0 /MGAFIX /!R /FORCE /ATAPI /!ATAPI /PCMCIA Undocumented options -------------------- /LAT /FIXES /DEBUG COMMAND LINE DETAILS Beginning with release 1.0.8, this includes the undocumented switch /!SHUTDOWN found in the latest IBM1S506.ADD. At system shutdown, this IBM1S506 issues shutdown commands to each attached device. This behaviour is different from earlier releases. I've seen many reports about problems when the drives are shutdown. So I decided to no longer be compatible with IBM1S506 here. Danis506 *tolerates* the option /!SHUTDOWN on the command line, but it does nothing. If you want this new shutdown behaviour (on notebooks perhaps), then add the new option /SHUTDOWN. Rule of thumb: if your system doesn't suffer from any (possibly unknown) flaws, not a single command line option is required. The driver adapts itself to the environment and runs at maximum performance. In some situations, you may need some of the following options to achieve optimum performance: - ATAPI devices are initialized to PIO mode only if they appear to support only very early ATAPI specifications; the same is true if the IDE chip is known not to support PCI busmastering with ATAPI devices. In the former case you may try to run particular ATAPI devices using busmaster DMA by applying the proper /A:x /U:y /BM options. - a few chips need the /80WIRE option to enable ATA66 operation (see below). - if the DumpIDE utility detects an unsupported busmaster capable chip and thus recommends to set /GBM (see below). - you intend to hot-swap devices or use PCCard devices ------------------------------------------------------------------------- For all of us who get the impression today's computers are way too fast to keep up with, IBM did us the favour of adding another switch, /W (available for more than one year now). This switch replaces the /V switch, and just like /V, /W shows some very useful info about the EIDE system at boot time. But instead of throwing all of this stuff onto the screen and scrolling it off, /W writes it to the display in reasonable chunks and then *waits* for the user to hit a key after he/she has learned it by heart ;-) In order for /W to work you must have pausing on errors enabled in CONFIG.SYS, either by "PAUSEONERROR=yes" or by no PAUSEONERROR setting at all. To reduce screen clutter and avoid possible SYS2068 messages, some lines of the boot message may be suppressed by /VL (/WL) instead of /V (/W). If you prefer even less info at boot time, /VLL (/WLL) will print out the most basic messages only. Here is an example: With /V or /W it looks like this: Daniela's Bus Master IDE Driver for OS/2 Version 1.3.8 Controller:0 Port:01F0 IRQ:0E Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 VIA 571 PCI to IDE IC Rev:06 VID:1106 DID:0571 @33MHz Unit:0 Status:OK ATAPI Device Bus Master DMA DMA1/PIO3 Model:IOMEGA ZIP 250 ATAPI 51.G Controller:1 Port:0170 IRQ:0F Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 VIA 571 PCI to IDE IC Rev:06 VID:1106 DID:0571 @33MHz Unit:0 Status:OK ATAPI Device Bus Master DMA Ultra DMA2/PIO4 Model:Pioneer DVD-ROM ATAPIModel DVD-115 0108 E1.08 Unit:1 Status:OK SMS:16 LBA Bus Master DMA DMA2/PIO4 Identify Model:CASTLEWOOD ORB2-E USR-4.08 OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 534 4273 0 534 4273 Avail 4307184 H 128 16 0 0 128 16 OS2 4306176 S 63 63 0 0 63 63 % Used 99.98 Controller:2 Port:A000 IRQ:09 Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 CMD 649 PCI to IDE IC Rev:02 VID:1095 DID:0649 @33MHz Unit:0 Status:OK SMS:16 LBA Bus Master DMA Ultra DMA4/PIO4 BPB Model:IBM-DJNA-352030 J58OA30K OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 2482 39560 1023 39560 16383 Avail 39876480 H 255 16 255 255 16 16 OS2 39873330 S 63 63 63 63 63 63 % Used 99.99 Controller:3 Port:A800 IRQ:09 Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 CMD 649 PCI to IDE IC Rev:02 VID:1095 DID:0649 @33MHz Unit:0 Status:OK SMS:16 LBA NL:64 Bus Master DMA Ultra DMA5/PIO4 BIOS Model:IBM-DTLA-307030 TX4OA50C OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 14890 59560 620 59560 16383 Avail 60036480 H 64 16 64 0 16 16 OS2 60036480 S 63 63 63 0 63 63 % Used 100.00 Controller:4 Port:B400 IRQ:09 Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 HPT 370 PCI to IDE IC Rev:03 VID:1103 DID:0004 @33MHz Unit:0 Status:OK SMS:16 LBA Bus Master DMA DMA2/PIO4 BPB Model:QUANTUM FIREBALL_TM1280A A6B.2400 OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 621 2484 524 621 2484 Avail 2503872 H 64 16 255 64 64 16 OS2 2503872 S 63 63 63 63 63 63 % Used 100.00 Controller:5 Port:BC00 IRQ:09 Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 HPT 370 PCI to IDE IC Rev:03 VID:1103 DID:0004 @33MHz Unit:0 Status:OK SMS:16 LBA Bus Master DMA Ultra DMA2/PIO4 BPB Model:ST34310A 3.07 OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 524 8354 534 524 8354 Avail 8420832 H 255 16 128 255 255 16 OS2 8418060 S 63 63 63 63 63 63 % Used 99.97 But with /VL or /WL it's a little more comprehensive: Daniela's Bus Master IDE Driver for OS/2 Version 1.3.8 Controller:0 Port:01F0 IRQ:0E Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK ATAPI Device Bus Master DMA DMA1/PIO3 Model:IOMEGA ZIP 250 ATAPI 51.G Controller:1 Port:0170 IRQ:0F Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK ATAPI Device Bus Master DMA Ultra DMA2/PIO4 Model:Pioneer DVD-ROM ATAPIModel DVD-115 0108 E1.08 Unit:1 Status:OK SMS:16 LBA Bus Master DMA DMA2/PIO4 Identify Model:CASTLEWOOD ORB2-E USR-4.08 Controller:2 Port:A000 IRQ:09 Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK SMS:16 LBA Bus Master DMA Ultra DMA4/PIO4 BPB Model:IBM-DJNA-352030 J58OA30K Controller:3 Port:A800 IRQ:09 Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK SMS:16 LBA NL:64 Bus Master DMA Ultra DMA5/PIO4 BIOS Model:IBM-DTLA-307030 TX4OA50C Controller:4 Port:B400 IRQ:09 Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK SMS:16 LBA Bus Master DMA DMA2/PIO4 BPB Model:QUANTUM FIREBALL_TM1280A A6B.2400 Controller:5 Port:BC00 IRQ:09 Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK SMS:16 LBA Bus Master DMA Ultra DMA2/PIO4 BPB Model:ST34310A 3.07 And finally, with /VLL or /WLL it becomes very terse: Daniela's Bus Master IDE Driver for OS/2 Version 1.3.8 Model:IOMEGA ZIP 250 ATAPI 51.G Model:Pioneer DVD-ROM ATAPIModel DVD-115 0108 E1.08 Model:CASTLEWOOD ORB2-E USR-4.08 Model:IBM-DJNA-352030 J58OA30K Model:IBM-DTLA-307030 TX4OA50C Model:QUANTUM FIREBALL_TM1280A A6B.2400 Model:ST34310A 3.07 No matter of the verbosity level which may suppress some lines on screen, *all* messages are stored internally for later retrieval! Beginning with beta 7, the driver saves all messages possibly shown at boot-up internally. Users may redisplay these messages, or store them into a file later with commands like these: COPY IBMS506$ con: or COPY IBMS506$ anyfile Please, check the spelling! Some time ago, IBM decided to enable /SMS by default on most controller hardware. The /!SMS switch then disables this controller based default. But if the hard disk reports to be SMS capable, SMS is enabled regardless of the /!SMS switch. Since this might cause sporadic failures on some hardware as a few users reported, I changed the behaviour of the /!SMS switch to a full override which is honoured in any case. Sometimes it's not necessary to disable SMS mode, but just to limit the number of sectors transferred in one run to a lower amount. To implement this feature, the /SMS option is changed into /SMS:n, where n is the maximum number of sectors in one run. If n is 0 or 1, this is equivalent to /!SMS. Every hard disk has a maximum SMS run length defined by the manufacturer specific to each model. You cannot set a value above this individual maximum SMS run length. The driver sets up the actual SMS run length for a specific drive to the minimum of the user supplied value n and the maximum value reported by the drive. ------------------------------------------------------------------------- To take advantage of Ultra DMA mode 3 (44MB/s), Ultra DMA mode 4 (66MB/s), Ultra DMA mode 5 (100MB/s), or Ultra DMA mode 6 (133MB/s), *all* of these prerequisites must be true: - the EIDE controller must support the ATA-66/100 enhancements - the device must support the ATA-66/100 enhancements - the cable must be a 80 wire type (recommended in any case) - the motherboard must support the detection of the 80 wire cable (see #) - the EIDE controller must report a 80 wire cable present - the device must report a 80 wire cable present # some chipset manufacturers still don't assign dedicated pins to enable the detection of the cable type installed, and therefore don't define a standard procedure for driver writers to acquire this *vital* information. Any motherboard with a VIA596B, VIA686, or AMD756 south bridge is affected by that: the ATA-66 features aren't enabled by default in the driver because it has no way to detect the presence of a 80-wire cable reliably. In this case add /80WIRE to the *adapter* options if there is a 80-wire cable installed on that particular EIDE channel. This decision is completely up to *you*. Don't blame me for lost or corrupted data if you are wrong! As a workaround to solve this inconvenience with the chips mentioned, beginning with release 1.3.0, the driver checks the controller setup as done by the BIOS. If it finds a particular channel initialized to ATA-66 operation it assumes the BIOS positively checked for the presence of a 80 wire cable. ------------------------------------------------------------------------- The following options are features added by me: ------------------------------------------------------------------------- Support for drive bays and PCCard slots Many notebooks/laptops have so called drive bays built in to swap different devices (floppy, disk, CD/DVD, etc.) while powered on state. The option /BAY prepares the related IDE adapter for such a hot-swap. To make this safely work on OS/2, the support for APM (by means of DEVICE=APM.SYS) has to be installed. During APM suspend the device in the drive bay may be swapped The floppy driver and the different ATAPI filters made by IBM check the presence of appropriate devices after APM resume and operate accordingly. Beginning with release 1.1.9, DaniS506.ADD improves this hot-swap support. After APM resume the driver inspects these "special" adapters and reconfigures both the device and the controller exactly as it does at boot time. It then reports its findings to interested filter drivers controlling the units so that the filter driver can reconfigure itself appropriately. At present, only DaniATAPI.FLT supports this advanced reconfiguration protocol. Example: A notebook has got a drive bay with a swappable floppy drive, disk, ZIP, and DVD-CDR/W. The following lines in CONFIG.SYS make all these devices swappable and operational without powering down: BASEDEV=IBM1FLPY.ADD /A:0 /FORCE:1 /U:0 /F:1.44MB BASEDEV=DaniS506.ADD [other options] /A:1 /BAY (see note #) BASEDEV=DaniATAPI.FLT /A:1 /U:0 /TYPE:ZW note #: this assumes the secondary channel to be connected to the drive bay The ZIP drive and the hard disk show up as removable media unit in the drives folder after a "refresh removable media" command. Important: before swapping the units there *must* be an eject of the media (either by the "EJECT x:" command from the command line or fro the context menu of the drive icon)! Forgetting to do that may cause data loss because the file system caches may not yet be written back to the media. With some luck the command "ChkDsk x: /F" will repairs such damage without data loss. Beginning with release 1.3.0, the driver supports PCCard slots as well. This support will be enabled by means of the options /PCS or /PCS:n. With /PCS the driver checks how many PCCard slots are available from the socket services driver and creates the same number of "virtual" IDE adapters. If less IDE adapters for PCCards shall be created, then the required amount n may be set by the option /PCS:n. At most the corresponding number of PCCards with IDE units attached may be inserted into the PCCard slots at the same time. A "virtual" PCCard IDE adapter is implemented by a PCCard (the actual IDE controller) which is inserted into any one of the PCCard slots. Each supported PCCard inserted *concurrently* counts as one IDE channel. The driver attaches to OS/2's PCMCIA subsystem and takes over all PCCards with a function id of 4 (disk) and function extension 1 (ATA interface); this includes PCCard ATAPI devices as well (like the PCMCIA ZIP-250 or the Click!40 drives). Beginning with release 1.3.10 both ATA and ATAPI devices are supported (if DaniATAPI.FLT rev 0.3.0 or later is installed). PCCards may be arbitrarily inserted to and removed from any PCCard socket. The only prerequisite is a functional socket services driver and PCMCIA.SYS. Beware: older revisions of IBM's PCMCIA subsystem have defects which may cause failures to configure cards on sockets other than the first one. After inserting a PCCard into the slot, you'll likely hear some noises from the PCCard or its attached unit. Shortly after that you should hear a high pitched beep. This beep indicates that the driver has accepted the PCCard, has successfully assigned resource to it and initialized it, and was successful communicating with the attached unit. If you don't hear this beep, try reinserting the PCCard. A failed initialization will result in a two times low pitched beep. If you intend to remove a PCCard, don't forget to close the file system on the attached device by issuing "EJECT x:" from the command line or the drive's context menu. Failing to do so will leave the file system in an inconsistent "dirty" state which most likely needs to be corrected by CHKDSK to make the medium accessible again before the next use. "Eject" doesn't necessarily unload the medium (f.e. in case of the Click! drive), so don't be confused. Some drive activity will indicate the flushing of still unwritten file system cache contents back to the medium. The end of the flush operation is indicated by a middle pitched beep. Afterwards it is safe to remove the PCCard from the socket or swap the medium. Beginning with version 1.6.0 the driver handles PCCards from which the system sit going to be booted from and which are already initialized by the BIOS (for example the Sony Vaio Crusoe subnotebook with a bootable external CD-ROM). To do that some technical prerequisites must apply (the PCCard slot needs to be ExCA register compliant) and there must be no socket services driver loaded; the latter one is needed only because of errors in the socket services drivers which make booting from a BIOS initialized PCCard impossible. /BAY on SATA channels enables SATA and SATAPI device hotplugging. This is particularly useful with hot-swap disk enclosures. ------------------------------------------------------------------------- Modern disks have multiple ways to save power. At times of inactivity, internal circuits may be temporarily suspended or motors switched off on demand. These functions are summarized by the term "Advanced Power Management" (APM). The balance between disk access speed and power-saving is controllable. For this purpose, the driver implements the option /APM:n which may apply to single, multiple or all disks implementing the APM function. The valid range of the value 'n' is 1 to 255; increasing values mean better performance, and the value 255 switches off all power-saving functions. The meaning of any given value is on the manufacturer's discretion. Users have to look-up disk datasheets for this information by themselves, I won't reply to any questions regarding this topic! Rule of thumb: just try the value 128: it allows all power-saving tricks except for switching-off the disk's spindle motor. Because of that, the performance impact on the very first access after some time of inactivity is minimal. If a disk implements the APM feature, it should be strongly preferred over the inactivity timer described later on! Issuing the command "DiskInfo I" will reveal if a disk implements the APM feature: APM feature implemented, but inactive: Advanced power management level: disabled Commands/features: Enabled Supported Advanced Power Management feature set APM level 128 active: Advanced power management level: 128 Commands/features: Enabled Supported * Advanced Power Management feature set ---------------------------------------------------------------------- The SATA specification describes an optional feature to control the power consumption of a SATA link. The more circuits in the SATA link hardware are switched-off to save power the longer it takes to reactivate them. There is no generally available means to control the power mode of a SATA link from the controller side, but modern SATA disks often implement the feature to enable link power-saving modes at times of inactivity. The option /LPM:n (Link Power Management) enables this function for single, multiple or all disks which implement this feature and are connected to a SATA port which also has this feature implemented and enabled. The parameter 'n' may have the following values: 0: power-saving allowed according to the controller's BIOS setting 1: all power-saving modes allowed (unconditionally) 2: all power-saving modes allowed except for SLUMBER (unconditionally) 3: all power-saving switched off, maximum link performance Depending on the disk-controller combination, some values may exhibit failures or may have no effect. In general, the value zero is the best option. It's up to users not to use this function in case of problems. The driver can't do anything about that if this part of the SATA specification isn't correctly implemented in the given hardware. Issuing the command "DiskInfo I" will reveal if a disk implements the prerequisites to enable link power management: Prerequisites satisfied, but function inactive: Commands/features: Enabled Supported Device-initiated interface power management Prerequisites satisfied, power-saving mode switching active: Commands/features: Enabled Supported * Device-initiated interface power management Issuing "DumpIDE" will show the current power-saving mode of a given SATA link: SATA link 0: Status : 00000123 - device pres., PHY comm; 3.0Gb/s; interface active bzw. Status : 00000123 - device pres., PHY comm; 3.0Gb/s; i/f power PARTIAL bzw. Status : 00000123 - device pres., no PHY comm; 3.0Gb/s; i/f power SLUMBER ------------------------------------------------------------------------- Inactive SATA link ports are switched-off unconditionally for power-saving purposes. ------------------------------------------------------------------------- To use the inactivity timer feature simply add /IT:x to the *unit* parameters. x is the timeout value in minutes. A 0 disables the spin down completely. If you leave out this option, some devices will still spin down after a certain default inactivity time. For proper driver operation of this function, all of the following should be true: - in BIOS, APM must be enabled - in BIOS, hard disk power down timers must be disabled - in CONFIG.SYS, at least the statement DEVICE=APM.SYS must be present (in other words: OS/2 APM support must be installed) If these prerequisites are true, the driver can keep track of disk power down even in case of intermittant system APM suspends. This ensures quick error recovery by short command timeouts, but sufficiently long timeouts for drive spin-ups. If APM is not available in OS/2, the driver will prolong command timeouts to 20s for each command if a drive reports active power down modes or the user adds an /IT option; all other drives will be sent a command to disable automatic spin-down. ------------------------------------------------------------------------- To limit the data rate a device communicates at, add /MR:udp to the *unit* parameters. Choose the values for u, d, and p from this table: u: Ultra DMA 0 - disable Ultra DMA 1 - limit to Ultra DMA mode 0 2 - limit to Ultra DMA mode 1 3 - limit to Ultra DMA mode 2 4 - limit to Ultra DMA mode 3 5 - limit to Ultra DMA mode 4 6 - limit to Ultra DMA mode 5 d: Multi Word DMA 0 - disable MW DMA 1 - limit to MW DMA mode 0 (in S506.ADD effectively the same as 0) 2 - limit to MW DMA mode 1 3 - limit to MW DMA mode 2 p: PIO 0 - limit to PIO mode 0 1 - limit to PIO mode 1 (in S506.ADD effectively PIO 0) 2 - limit to PIO mode 2 (in S506.ADD effectively PIO 0) 3 - limit to PIO mode 3 (with flow control) 4 - limit to PIO mode 4 (with flow control) If you set u and d to 0 at the same time, this effectively disables busmastering on this device since no DMA mode is left to use. Leading zeroes may be left out. Many EIDE controllers don't have the capability to select PIO modes and multiword DMA modes independently because theses mode share the same timing registers. This is for historic reasons and the similarity of cycle timings. This applies to all controller chips mentioned above except for Promise Ultra and Cyrix MediaGX. Note: this applies to mixing transfer modes for the *same* device! The devices on an EIDE channel are independent from each other regarding the transfer modes (except for Intel 82371FB). To deal with this situation the driver selects the slower cycle time from the proposed PIO and MW-DMA modes. The net result can be seen from the table below: | no DMA MW-DMA 1 MW-DMA 2 Ultra-DMA x ------+---------------------------------------------------------------- PIO 0 | PIO0 PIO0/DMA1 PIO0/DMA2 PIO0/UDMAx (Intel,Promise, | Cyrix) PIO 0 | PIO0 PIO0/no DMA PIO0/no DMA PIO0/UDMAx (others) ------+---------------------------------------------------------------- PIO 3 | PIO3 PIO3/DMA1 PIO3/DMA2 PIO3/UDMAx (Promise,Cyrix) PIO 3 | PIO3 PIO3/DMA1 PIO3/DMA1 PIO3/UDMAx (others) ------+---------------------------------------------------------------- PIO 4 | PIO4 PIO4/DMA1 PIO4/DMA2 PIO4/UDMAx (Promise,Cyrix) PIO 4 | PIO4 PIO3/DMA1 PIO4/DMA2 PIO4/UDMAx (others) ------------------------------------------------------------------------- Some users are forced (or feel themselves forced ;-) to run the PCI bus on their motherboards at speeds other than the optimum 33MHz. For example, many Cyrix processors need very uncommon front side bus speeds which cannot be easily divided into 33MHz. This leads to either too slow EIDE timings or too fast ones. The former case may be disappointing, but the latter overclocks the EIDE interfaces and my introduce data distortions. If you know the actual PCI clock of your motherboard, you may add an additional driver option /PCLK:x (x is the PCI clock in MHz). The user supplied PCI clock is then sorted into one of four speed classes (25MHz, 33MHz, 37MHz, and 41MHz) which is reported at boot time. For each of these speed classes, the driver *tries* to select the optimum chip timings, but due to chipset limitations this is not possible in all cases. As most motherboards run their PCI bus at the optimum 33MHz clock today, you should use this switch only if really necessary! The technical background of this option: The so called Front Side Bus (FSB) is the bus the CPU communicates with the north bridge (which contains the memory controller, the cache controller, the AGP bridge, the PCI arbiter, the PCI bridge, and optionally some more devices). Depending on the design of the north bridge, the PCI clock is derived from the FSB clock by dividing this clock (synchronous design) by either an integral number (2, 3), or a fractional number (2, 2.5, 3), or it is derived from an independent clock source (asynchronous design, very uncommon). The PCI 2.1 spec (without its extensions) allows for 33MHz max. PCI clock. The VIA, the ALi, and most likely the SiS chipsets are capable of fractional dividers. The Intel chipsets are not. If you know, your PCI bus runs at the typical 33MHz clock (true for most of the mobo/CPU combinations sold today), don't apply this switch. If you know the PCI clock in advance (or like to tune the EIDE timings for some reason), you apply the /PCLK:n switch where n is the known PCI clock. The driver then selects the speed class from this value in the same way as from an autodetected value. These speed classes are the result of the typical FSB clocks and dividers: FSB clock / divider -> PCI clock -> class 50MHz / 2 -> 25MHz -> 25MHz 60MHz / 2 -> 30MHz -> 33MHz 66MHz / 2 -> 33MHz -> 33MHz 75MHz / 2 -> 37.5MHz -> 37MHz 75MHz / 2.5 -> 30MHz -> 33MHz 75MHz / 3 -> 25MHz -> 25MHz 83MHz / 2 -> 41.5MHz -> 41MHz 83MHz / 2.5 -> 33MHz -> 33MHz 95MHz / 3 -> 31.7MHz -> 33MHz 100MHz / 3 -> 33MHz -> 33MHz 112MHz / 3 -> 37.3MHz -> 37MHz 124MHz / 3 -> 41.3MHz -> 41MHz For the chosen speed class, the driver tries to select EIDE controller register values which don't exceed the PIO/DMA/Ultra ratings. My former driver revisions and IBM1S506 only know the proper register values for the 33MHz class. This runs the EIDE interface slower than possible in case of a 25MHz PCI clock, but exceeds the allowed timings in the 37MHz/41MHz cases. The former one is uncritical, but the latter one may cause failure of operation (better) or data corruption (worse). To give you an example how to use command line switches, look here: BASEDEV=DaniS506.ADD /V /A:0 /U:1 /MR:223 /IT:5 /A:1 /I This shows some info at boot time, limits the data rates to communicate with the primary slave to UDMA1/MWDMA1/PIO3 and stops the spindle motor after 5 minutes of inactivity on this device (which happens to be the IBM-DTTA351010 drive). It tells the driver also not to touch the secondary channel (adapter 1). ------------------------------------------------------------------------- The /GBM option enables the scan for busmaster capable PCI EIDE controllers if none of the 'well-known' chipsets (currently INTEL, VIA, ALi, SiS, CMD, Promise, HighPoint, AMD, SMSC, Artop, ServerWorks, CMD640, and RZ1000, the latter two are not busmaster capable!) is found. Before a chip is accepted by the driver to be operated in generic busmaster mode, it has to pass some sanity checks. Some BIOSes simply fail to initialize the controller appro- priately (seen on a board with an UMC8886BF). A rule of thumb: if Linux doesn't run a chip in DMA mode, it is not generic busmaster compatible. At present, EIDE controllers not supported by any S506 ADD run attached devices in PIO mode only, and with timings setup by the BIOS (most probably PIO0 if the device is hidden from the BIOS). The generic busmaster option enables usage of the busmaster machinery and device DMA modes, thus reducing CPU load by a large amount. In this mode, the ADD doesn't touch any chip registers other than the standard PCI command register (to enable PCI busmastering) and the registers associated with the busmaster engine. It has *no* way to enable Ultra-DMA modes in hardware if not done by the BIOS. If you think this should be the default behaviour of my driver, then simply change the /GBM option into the /FORCEGBM option. Then, *every* EIDE chip is treated as if it's not a 'well-known' chip. All programming of device timings in the EIDE controller has to be done by the BIOS. /FORCEGBM comes in two flavours, as a global option, and as an adapter option. If /FORCEGBM precedes all /A:x options, it applies to all adapters; if it follows an /A:x option, it applies to adapter x only. In some cases, /FORCEGBM can help to overcome stability problems because of the spec-conforming timings that the driver programs into the controller registers. In the recent past, some batches of boards popped up which are "tuned-down" by their BIOS to work reliably. In these cases, the "correct" timings are too fast and the boot process fails if these timings are set up by DaniS506.ADD. Adding /FORCEGBM prevents the driver from adjusting the timings so that the "special" BIOS timings are used instead. Examples: BASEDEV=DaniS506.ADD /W /GBM /A:0 /U:0 /MR:023 BASEDEV=DaniS506.ADD /W /FORCEGBM /A:0 /U:0 /MR:023 /A:0 /U:1 /MR:034 /IT:5 ------------------------------------------------------------------------- There is a new option /PF:xxxx which may be applied to an adapter or individual units. This option is meant to fine-tune register access timings and enable chipset-performance features. The option value is a bit field, whose individual bits need to be put together into the final option value. Bits 0-7 are meant for all PCI chipsets, bits 8-15 are chipset specific. In this driver release, the following settings are available: +---+---+---+---+---+---+---+---+---+---+---+---+---+---+-------+ |15 14 13 12 11 10 9 8 | 7 6 5 4 3 2 1 0 | +---+---+---+---+---+---+---+---+---+---+---+---+---+---+-------+ | unused | unused | (1) | +---+---+---+---+---+---+---+---+---+---+---+---+---+---+-------+ 1) Register access delay: 0 : 1000 ns (default, fixed for all non-PCI systems) 1 : 500 ns 2 : 250 ns 3 : no delay (works on most systems, check out yourself) The actual access delay used for a particular adapter channel is the largest of all attached units! BASEDEV=DaniS506.ADD /A:2 /PF:2 /U:0 /PF:3 If there is no slave device, then no register access delays are incurred because the individual unit setting /PF:3 overrides the adapter setting /PF:2. But if there is an additional slave device, this unit inherits the adapter option /PF:2. The largest delay of both units is the resulting one and thus each register access is delayed by 250ns. ------------------------------------------------------------------------- Modern ATA spec 6 compatible drives may support the so called acoustic management feature set. Using this you can trade off performance vs. noise (mostly noticable during quick sequences of random seeks). The new unit option /NL:n allows you to control this feature: the value n may be set to 0 ... 126, where '0' selects the faintest noise level with the biggest performance impact, while '126' selects best performance at the expense of noisiest operation. Drives are not required to support any value in between, but to select the next supported one. Drives store the latest programmed value internally and preserve it over power outages and resets. The driver will tell the current value read from the drive if you don't use /NL option, or the user selected value n if /NL:n is present. ------------------------------------------------------------------------- One user reported a boot failure in a particular situation which involved SCSI and ATA/ATAPI units. The solution was to avoid BIOS INT13 calls. The driver asks the BIOS about the geometry properties of the ATA drives. In general, this information is not required because the most authoritative source of the geometry info is the drive itself (by inspecting the MBR and BPBs). This implies that no BIOS INT13 calls are required in almost any scenario. The only exception is if your disk is completely blank, and you want to partition this blank disk by OS/2 FDISK with a geometry compatible to the BIOS one. So, if you prefer to deny the driver calling the BIOS, add the option /!BIOS. When doing so, the "strange" VDM seen with some process management utilities is gone as well. ------------------------------------------------------------------------- Later versions of FDISK and LVM tend to overwrite bootmanagers other than OS/2 BootManager on every write operation to the partition/LVM tables. In such cases it might be appropriate to write protect track zero of the boot drive (which contains the bootmanager code). This is possible either at bootup using the unit option /A:x /U:y /PROT0 ("PROT zero"), or later by means of the DiskInfo utility: DiskInfo <unit> p 1 enable write protection on <unit> DiskInfo <unit> p 0 disable write protection on <unit> "DiskInfo ?" shows all options in detail. ------------------------------------------------------------------------- Todays large disk drives often don't work properly with older BIOSes. Either the BIOS detects the wrong disk capacity, or sometimes the PC doesn't boot at all. In such cases, the cure of the problem is either to set a jumper at the disk or reprogram the disk such that it reports a smaller, trouble free capacity to the BIOS. Beginning with version 1.3.8 the driver is able to use the full disk capacity even if the disk is programmed to report a smaller capacity. This is done automatically; if you don't like that then you have to prevent the unlocking of the full capacity by means of the unit option /A:x /U:y /!SETMAX. To program a disk to report a smaller capacity, I recommend this process: 1) remove the disk from the BIOS setting. The PC will no longer hang. 2) boot from an OS/2 boot disk (or the like). 3) program a smaller disk capacity by means of the Diskinfo utility: DiskInfo <unit> m <new capacity>, f.e. DiskInfo 0m m 32G For details see "DiskInfo ?" 4) reboot 5) if you like you may add the disk to the BIOS settings again INSTALLATION For an upgrade from a previously installed IDE driver (most likely IBM1S506.ADD) just copy DaniS506.ADD to bootdrive:\OS2\BOOT and rename IBM1S506.ADD (or the name of the previous IDE driver) to DaniS506.ADD in your CONFIG.SYS. For a first time install of an IDE driver you'll have to add this line to CONFIG.SYS: BASEDEV=DaniS506.ADD In this case your boot device probably is SCSI. To avoid confusion make sure to move the mentioned line *below* the SCSI driver. If you prefer chasing icons, you can do the copy and the modification of CONFIG.SYS using DDINSTAL ("OS/2 System"/"System Setup"/"Install/Remove" /"Device Driver"). Due to limits in DDINSTAL you have to disable the previously installed S506 base driver in CONFIG.SYS manually, either by deleting that line or putting a 'rem' statement in front of it. FILES DaniS506.ADD The driver itself DaniS506.DDP The device driver profile for DDINSTAL DaniS506.DOC The file you are reading right now :-) DumpIDE.EXE A utility to dump the current IDE registers DiskInfo.EXE A utility to show information about attached devices PROBLEM.TXT Problem reporting ChipList.txt Full listing of supported chipsets README Copyright and license notice COPYING.LESSER Lesser GPL License Smart\ smartctl.exe S.M.A.R.T. control utility smartctl.txt smartctl.exe User Guide DiskInfo\ DiskInfo.c Driver and disk info dump utility source code Identify.c Disk info dump utility HDParm.h Header used by DiskInfo.c BUGS Hopefully none, but who knows ... To report a bug, see PROBLEM.TXT I have no means to identify traps occurring outside this driver (module name different from 'IBMS506$' or not present). If a trap occurs within this driver, I need the following information: CS:EIP, EAX-EDI, DS and ES. My test systems at home (VIA and ALi) and a couple of other machines at work (different INTEL chips) run flawlessly with this driver, so I have virtually no chance to reproduce errors. If any of your EIDE devices is not detected at boot-up, please contact me. The driver has special debug code built in; I will tell you how to enable and gather these traces. If you experience problems with your hard disks after OS2DASD.DMD has been loaded (to be seen by pressing ALT-F2 when the white box appears on the screen), read the following section *very carefully*, it applies to you. I cannot solve your hardware problems with software. If you think your problems arise from bugs in the EIDE controller chip, you may contact me after gathering detailed info about this bug and its software solution. At present, I have a recent chip bug list from ALi and another one for the VIA 586 chipset created by myself. SOME WORDS OF WISDOM Much EIDE hardware you can buy (or already have bought) has *very* little design margins and is *very* susceptible to even the slightest signal distortion on the cables. Unlike SCSI, the EIDE bus has no termination resistors on both ends of the cable to reduce ringing or other signal reflections. Noise and crosstalk are other sources for bit errors during data transmission; careful grounding (both the data and the power connection) is paramount if you intend more than simple PIO speed. Power supplies are not uncommon for inducing noise! A very patient user found another source of data distortions after some very long and tedious reliability tests: the cache subsystem of his motherboard. He experienced bit errors which vanished after some fiddling within the BIOS settings of the cache timings (pipelining was the culprit in this case). This isn't something magic! Consider, a busmaster - especially when running at high transfer rates - puts an additional heavy burden on the cache logic in the chipset and the cache RAMs. Many random failures of OS/2 systems - even without busmastering of EIDE or SCSI controllers - have their origin here. If you think your data is trashed due to bugs in this driver, please take this into account: I made a lot of changes to IBM's source code in the driver initialization portion to support ALi, SiS, VIA, CMD, and Promise chipsets, and program spindown timers in the devices, and to select data transfer rates other than the fastest ones. Some of these changes could lead to traps during bootup depending on the actual parameter values used. I hope to have them fixed. These traps (identifiable by the module name 'IBMS506$') cannot corrupt any of the data because they happen before any write operation is issued to any drive. I'm sorry if you happen to have such an inconvenience. BUT: the code to carry out read and write requests either by PIO or busmaster assisted DMA is *unchanged*. In fact, all data transfers from the device into host memory or vice-versa are done 100% by the drive electronics and the chipset hardware on your motherboard (or 99% plus *one* CPU instruction in case of PIO). Driver code only steps in when any part of this hardware reports an error or wasn't able to fulfill the request. Keep in mind: PIO and MWDMA transfer modes have no means of detecting bit errors during data transfers, Ultra-DMA does CRC checking to maintain data integrity. So, instead of rushing to the fastest data transfer modes possible, limit the negotiated data transfer speed to the slowest available in each category (/MR:120). Using PIO4 on other than rock solid hardware which passed reliability tests is *very* dangerous. Since large portions of the OS2*.INI files cannot be flushed to disk at shutdown with error-checked Ultra-DMA, the ADD has to revert to the error-prone PIO mode. I consider this an OS/2 design flaw. If you run SYSBench or DISKIO, and experience unstable or rather low performance figures in the maximum data transfer rate test, you are seeing the effects of retries and interface resets. This is an indication of bit errors! As a reference, in the DISKIO 'Drive cache/bus transfer rate' test you should see the following results: Ultra-DMA mode 5 : 66MB/s or more Ultra-DMA mode 4 : 44MB/s or more Ultra-DMA mode 3 : 34MB/s or more Ultra-DMA mode 2 : 24MB/s or more Ultra-DMA mode 1 : 17MB/s or more Ultra-DMA mode 0 : 13MB/s or more MW-DMA mode 2 : 11MB/s or more MW-DMA mode 1 : 9MB/s or more SYSBench is based on the DISKIO code with some modifications. DISKIO does the worst case test on the last cylinder (falsely called track in both programs) using forward reads, SYSBench uses backward reads. Remember, different drives have different cache strategies built in, and reading backwards is a pretty uncommon access strategy with file system caches, so some drives are fooled in this test and the result is overly pessimistic. If you like to tune the data transfer speed on the cables to your drive, then run DISKIO or SYSBench without DMA rate limitation and look at the results reported for cylinder (track) 0. This is the maximum data transfer rate your drive is able to read off the media surface. Then select a transfer mode which matches this speed. I recommend Ultra DMA modes over MW-DMA modes. If the drive is DMA capable, select PIO mode 0 as non-DMA mode (for the rare cases where busmaster DMA cannot be used). If you see false device detects or even lockups at boot time, don't expect reliable data transfers thereafter. In any case, if you think you are experiencing data corruption, look at the output generated by the /W or /V switch, and then look for the data transfer modes used with your device. The first remedy to try will be disabling all fast transfer modes (by /!BM and/or /MR:). If you hear short beeps, then the data transfer state-machine encountered an exception: a high beep indicates a UDMA CRC failure, a low beep an error reported by the hardware which leads to a interface reset eventually, and a middle beep indicates a command timeout (usually caused by a spun-down hard disk). So, if your hardware works 'flawlessly' you will hear *no* sounds from the driver (except for spin-down events of course). With 'flawlessly' I mean 'without errors detectable by any instance'. If you prefer to ignore hardware errors indicated by these beeps, add the /!AA (no audible alerts) option. To protect the current working desktop from corruption, consider creating test copies of the x:\Desktop directory tree and the OS2*.INI files and pointing to these in a modified copy of your CONFIG.SYS located in x:\OS2\BOOT. For example, rename this copy to CONFIG.T and select this test configuration within the recovery choices screen by hitting the 'T' key. This way, you always have a working default boot system and a independent test system to play with. QUESTIONS & ANSWERS 1) Both my EIDE controller and my hard disk is ATA/66 (or ATA/100) capable, and a 80wire cable is installed. No matter what I am trying to do (reading the docs, applying the /80WIRE override switch), the driver allows no more than UltraDMA mode 2. What's wrong ? Most likely, your hard disk is setup to hide its ATA/66 (or ATA/100) capabilities. Some models are delivered with such a factory setup to prevent confusion of older BIOSes or controllers. Run "DiskInfo i" and look at word 88/058 reported from the disk in question. If it it something like xx07, the disk reports ATA/33 capability; if it is xx1F, it can do ATA/66, and if it is xx3F, then ATA/100 is available. If the reported capability is less than advertised, contact the disk manufacturer for a utility to enable the missing capabilities (often to be found on their web sites). 2) I experience random failures after wakeup from suspend. What can I do ? Toggle the "Delayed Transaction" setting at the PCI page in the BIOS. Some hardware doesn't implement this feature correctly. 3) I have a large (> 33GB) disk but FDISK or LVM report only a fraction of the true disk capacity. Where does that come from ? Run "COPY IBMS506$ con:" and look at the geometry values (example): Model:WDC WD400BB-00AUA1 18.20D18 OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 4865 65535 16383 65535 16383 Avail 78165360 H 255 16 16 255 16 16 OS2 78156225 S 63 63 63 63 63 63 % Used 99.98 If the "OS2:log" column shows a C value of 65535 then there is most likely a 16 bit overflow due to a badly chosen translation mode. Presumably, the H value will be 16 instead of the much more appropriate 255 (or 240). If you see the same H value in the "BPB:log" column then this is the translation geometry stored on the disk itself, in the partition table and/or the BPB sectors. In this case you *must* zero out the partition table to get rid of this unwise translation mode, a mere change in the BIOS settings is not sufficient! If there are only zeroes in the "BPB:log" column, then the disks doesn't have a valid partition table yet and the translation mode is taken from the BIOS. Change the BIOS setting to "LBA translation" mode. 4) I have problems reading CDs now which was working fine with earlier driver versions. Why ? Later drivers enable DMA modes on ATAPI devices now if the seem to follow sufficiently recent ATAPI implementation standards. Some ATAPI units still fail running on DMA, so you have to disable DMA busmastering for these units (/A:x /U:x /!BM). CREDITS I'd like to express my thanks to some outstanding people who made this piece of code a better one: Helen Terbizan for her never-ending efforts to make ALi support possible and her patience and perseverance during the tests. Mark Dodel for just the same patience and perseverance. Jens Glathe for his helpful tips and the courage to test the SiS code. Piotr Oniszczuk for his persistent reliability tests. Fred Springfield for his debugging of this pile of words. Gerald Eischer for creating the command line options diagram and table Mensys, BV and Serenity Systems International for providing financial support for on-going development and testing of this driver. The eComStation development and test teams for putting their systems at risk when testing new builds. TO DO Still left is to persuade IBM to incorporate these features in their distribution. done! IBM has officially taken the v1.6.6 sources. COPYRIGHT DaniS506 - Copyright (c) 1999-2008 Daniela Engert. All rights reserved. Portions Copyright (c) 2008, 2010 Steven Levine and Associates, Inc. All rights reserved. Portions Copyright (c) 2010 Mensys B.V., Inc. All rights reserved. STATUS OF PROGRAM DaniS506 is freeware distributed under the terms of the GNU Lesser General Public License. DISCLAIMER YOU ARE USING THIS PROGRAM AT YOUR OWN RISK! I don't take any responsibility for damages, problems, custodies, marital dis- putes, etc. resulting from use, inability to use, misuse, pos- session or non-possession of this program directly or indirectly. I also don't give any warranty for bug-free operation, fitness for a particular purpose or the appropriate behaviour of the program concerning animals, programmers and little children. THE SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION. Or, in a few words: If its good, I am responsible. If its bad, its all your fault. ;) Permission is granted to redistribute this program free of charge, provided it is distributed in the full archive with unmodified contents and no profit beyond the price of the media on which it is distributed is made. Exception to the last rule: It may be included on freeware/shareware collections on CD-ROM, as well as on magazine cover CD-ROMs. All trademarks mentioned anywhere around here are property of their owners and the like ... ORIGINAL AUTHOR Daniela Engert Internet: "Daniela Engert" <dani at ngrt dot de> CURRENT DEVELOPERS Steven H. Levine Internet: "Steven H. Levine" <steve53 at earthlink dot net> SUPPORT To request support, please enter a support ticket at http://svn.netlabs.org/xata/newticket You need a Netlabs login id to submit tickets. See http://svn.netlabs.org/xata for instruction on how to obtain a login id. eComStation users with a Support Contract can submit tickets to the eComStation Ticketing System at http://www.ecomstation.com/support/tickets/ticket_create.php Click the Support radio button to select the category. EPILOGUE Suggestions and bug-reports are always welcome. Well ... bug-reports are perhaps not *that* welcome ... ;) $Id: DaniS506.DOC 120 2013-04-24 01:04:13Z stevenhl $

Daniela's S506 ADD - Release 1.8.5 ---------------------------------- NAME DaniS506.ADD - replacement for IBM1S506.ADD SYNOPSIS EIDE driver with support for Intel, VIA, ALi, SiS, CMD, Promise, Cyrix, HighPoint, AMD, Artop, SMSC, ServerWorks, Opti, Nvidia, National Semiconduc- tor, SiliconImage, ITE, Adaptec, ATI, Broadcom, JMicron, Marvell chipsets, and generic busmaster support for all SFF-8038i compliant chips. This covers virtually *any* EIDE controller built/sold in the past few years and even some just about to be released in the near future. Don't be confused by the generic chip names below! If you are uncertain about your actual controller to be supported, run the DumpIDE utility - it will tell you. ### note: ### This driver is *not* compatible to any of the software RAID setups offered ### by some so called IDE RAID or SATA RAID controllers (f.e. HPT, Promise ### FastTrack, AMI, VIA, Intel, SiS, you name it). None of these "fake RAIDs" ### have any hardware RAID features built in! ### Of course, you may run these chips in non-RAID BIOS setups (i.e. JBOD). *** Changes to the previous release are denoted by "***" tags. DESCRIPTION Since the original IBM1S506.ADD doesn't support the IDE controller found in the VIA 82C586(/A/B) south bridge beyond the basic features, I wasn't very happy with the performance of my 10Gig DTTA-351010 drive. The driver from VIA itself was no solution because it doesn't support all the advanced features found in todays hardware and locked up my system completely. So I sat down.... This release of my S506 driver is no longer based on the sources provided by IBM on their DDK site. It has the latest feature set as found in the current fixpacks and fixes later than these. My enhancements are: - (supports the Intel PIIX, just like IBM1S506 does) plus some PIIX4 variants unsupported by IBM1S506 - supports the Intel ICH/ICH0/ICH2/ICH3/ICH4/C-ICH/ICH5/ICH6/ICH7/ICH8 PATA controllers - supports the Intel ICH5/ICH6/ICH7/ICH8/ICH9/ICH10 SATA controllers - supports the Intel Management Engine IDE redirection controllers *** - supports the Intel SCH PATA controller ("Atom") (experimental) - supports the VIA PATA controllers - supports the VIA PATA/SATA RAID controllers (see note ###) - supports the ALi PATA controllers - supports the ALi 5281/5287/5289 PATA/SATA RAID controllers (see note ###) - supports the SiS 5513/5517/5518/180 PATA controllers - supports the SiS 180/181/182/1180/1182/1183/1184 (182 and above exp.) SATA RAID controllers (see note ###) - supports the CMD 643/646/648/649 PATA controllers (this includes OEMs like AMI) - supports the Promise Ultra33/66/100/TX2/133 PATA controllers - supports the Cyrix 5530 PATA controller (Cyrix MediaGX) - supports the HighPoint HPT302/366/368/370/371/372/374 PATA controllers - supports the AMD756/766/768/8111/GeodeLX PATA controllers - supports the ACARD ATP850/860/865/867 PATA controllers (ATP860 experimental) - supports the SMSC SLC90E66 PATA controller (PIIX4 look-alike) (exp.) - supports the ServerWorks/Broadcom PATA controllers - supports the Opti PATA controllers - supports the Nvidia nForce/2/3/4/500/600/700 PATA controllers important fix for broken BIOSes which cause system hangs - supports the Nvidia nForce2/3/4/500/600/700 SATA controllers - supports the National Semiconductor Geode SCx200 PATA controller - supports the Silicon Image 680 PATA controller - supports the Silicon Image 3112/3114/3512 SATA controller (this includes OEMs like Adaptec) - supports the ITE 8211/8212 PATA controller (see note ###) - supports the ATI IXP200/300/400/600/700 PATA controllers - supports the ATI IXP300/400/600/700/800 SATA controllers - supports the NetCell/Revolution SyncRAID PATA/SATA Hardware-RAID Controller - supports the JMicron JMB360/361/363/365/366/368 PATA/SATA controllers - supports the Marvell 61xx PATA controllers - generic busmaster support for SFF-8038i compliant controllers - supports SATA <-> PATA adapters - supports SATA/SATAPI hotplugging - supports PCCard (formerly: PCMCIA) ATA/ATAPI controllers - supports ATA-Flash, Compact-Flash, CF+, SmartMedia, etc. in PCCard slots - supports internal DriveBays - supports the external DriveBay and other PCCard attached units - no need for chipset initialization by the BIOS (Intel, VIA, ALi, SiS, CMD, Promise, and HighPoint) - support for Ultra33/66/100/133 DMA modes according to the chip feature matrix (ChipList.txt) - support for 1.5GBit/s and 3.0 GBit/s SATA modes according to the chip feature matrix (ChipList.txt) - faster PIO due to 32bit IO - adjustable disk power management (device APM) - adjustable SATA link power management - sets device inactivity timers (POWERSV.SYS is no longer needed) - adjustable SMS length - adjustable drive noise level - optional write protection of the first track of a disk drive - unlocks full capacity of disks which are programmed to report a lower capacity to be compatible with older BIOSes - improved support for removable media drives (in particular Castlewood ORB EIDE and the Fujitsu M2541B MO drive) - supports devices which don't spin up automatically - optional limit data transfer rates for more reliability - adapts (feature removed: auto-adapts) chip timings to PCI clock - supposed to work with drive capacities up to 2 TiB (OS/2 ADD limit), tested with drives up to 2 TiB now - improved compatibility with current BIOSes (IBM1S506.ADD fails on an increasing number of recent BIOS implementations) - performance and feature tuning options - up to 16 PATA/SATA channels - enhanced S.M.A.R.T support - improved stability in case of SATA PHY communication errors - some fixes for bugs introduced with earlier releases - some fixes for bugs and omissions in the IBM code ;-) - work around some BIOS geometry issues - fixes possible data corruption with some VIA chips. - fixes a hang during initialization of some units which must not be touched for some time after completion of a couple of commands - support for high interrupts (IRQ16 and above) in APIC interrupt mode this requires ACPI.PSD v3 or later! ### Caution: users of DaniBoot v1.0 need to upgrade to version 1.0.2! ### Caution: users of DaniATAPI v0.3.x need to upgrade to version 0.4.0! ### Caution: other ATAPI related filter drivers no longer work due to necessary inter-driver interface changes! COMMAND LINE OPTIONS This driver supports most switches and options that IBM1S506 supports. For an explanation of these look into the OS/2 online help (by picking 'OS/2 Commands by name'/ 'BASEDEV...'/'IDE...' or by "help ibm1s506" from the command line). COMMAND LINE OVERVIEW Gerald Eischer did us the favour of creating a syntax diagram and a short overview of the command line options implemented in the driver *** Syntax Diagram -------------- +------------------+ +-- /I --+ +---------------+ V | V | V | BASEDEV=DaniS506.ADD --+----------------++-- /A:nn -+---+-------------++-> +-+- /!V -+------+ ^ +- /P:hhh ----+ | +- /!W -+ | | +- /IRQ:n ----+ +--- /!AA -------+ +--------+ +- /!BM ------+ +--- /PCLK:x ----+ | +- /FORCEGBM -+ +-+- /GBM -----+-+ | +- /PF:xxxx --+ | +- /FORCEGBM + | | +- /80WIRE ---+ +--- /SHUTDOWN --+ | +- /TO:nn ----+ +--- /DEBUG:n ---+ | +- /BAY ------+ +--- /!BIOS -----+ | +- /IT:x -----+ +--- /PCS:n -----+ | +- /LOC:spec -+ +--- /!BM -------+ | +--- /APM:n --+ +--- /PF:xxxx ---+ | +--- /LPM:n --+ +--- /IT:x ------+ | +--- /APM:n -----+ | +--- /LPM:n -----+ | | >--------+----------------------------------------+------------------| | ^ | +----------------------------------+ | | +--/I---+ +----------------------+ | | | V | V | | | +-+- /U:n +--+-------------------+-+-+-+ +- /T:n ------------+ +- /!SMS -----------+ +- /!DM ------------+ +- /IT:x -----------+ +- /MR:udp ---------+ +- /PF:xxxx --------+ +- /NL:n -----------+ +- /!SETMAX --------+ +- /WP:x -----------+ +- /!BM ------------+ +--- /APM:n --------+ +--- /LPM:n --------+ In most cases, options prefaced with ! can be used with the ! modifier to explicitly disable the option. Options are valid both with and without the ! modifer. The values in parentheses are the default values if the option is not present. Options may be prefixed by "-" instead of "/" as well. General options --------------- /!V /VL /VLL (quiet) (not) Verbose mode (less/lesser verbose) /!W /WL /WLL (quiet) same as /!V, but pauses screen /!AA (beep) No Audible Alerts /PCLK:x (33MHz) PCI Clock (DECIMAL) /GBM (off) Generic Busmaster Mode /FORCEGBM (off) Force Generic Busmaster mode /!SHUTDOWN (not) (not) Shutdown each attached device /DEBUG:n (0) Set debug level (DECIMAL) /!BIOS (call BIOS Disable all BIOS Int13 calls if required) /PCS (0) Support all found PCCard Slots /PCS:n (0) Support at most n PCCard Slots /!BM (ATA : enable) (not) Enable BusMaster DMA on all adapters (ATAPI: enable) /PF:xxxx (0) Register access timings /IT:x (BIOS) Inactivity timer (DECIMAL) /APM:x (BIOS) Advanced Power Management Level (DECIMAL) 1 (max. power saving) ... 254 (max. performance) /LPM:x (BIOS) SATA Link Power Management Level (DECIMAL) 0 (max. power saving) ... 3 (disabled, max. performance) Adapter options --------------- /A:nn (n/a) Adapter number (0, 1, .. 15) /LOC:bus:device.function#channel (n/a) assign PCI device and channel to this adapter /P:hhh (PCI) Base port address (HEX) /PORT:hhh same as /P /IRQ:n (PCI) IRQ level (DECIMAL) /I (off) Ignore adapter /!BM (ATA : enable) (not) Enable BusMaster DMA (ATAPI: enable) /PF:xxxx (0) Register access timings /80WIRE (autodetect) 80 wire cable present /FORCEGBM (off) Force Generic Busmaster mode /TO:n (DECIMAL) (5s/dynamic) Interrupt timeout /BAY (off) Adapter is connected to a drive bay /IT:x (BIOS) Inactivity timer (DECIMAL) /APM:x (BIOS) Advanced Power Management Level (DECIMAL) 1 (max. power saving) ... 254 (max. performance) /LPM:x (BIOS) SATA Link Power Management Level (DECIMAL) 0 (max. power saving) ... 3 (disabled, max. performance) Unit options ------------ /U:n (n/a) Unit number (0, 1) /UNIT:n same as /U /I (off) Ignore unit /T:n (5s) Error retry time in seconds (DECIMAL) /!SMS (autodetect) Disable block mode /SMS:n (autodetect) Set SMS block length to n /!DM (off) Prevents OS2DASD.DMD from managing the device /IT:x (BIOS) Inactivity timer (DECIMAL) /MR:udp (maximum) Limit data rate of a device /PF:xxxx (0) Register access timings /NL:n (current device noise level, n = 0 (faint) ... 126 (noisy) setting) /!SETMAX (enable unlock) disable unlocking of full drive capacity /WP, /WP:0 (off) enable write protection of complete disk or track 0 only /!BM (ATA : enable) (not) Enable BusMaster DMA (ATAPI: enable) /APM:x (BIOS) Advanced Power Management Level (DECIMAL) 1 (max. power saving) ... 254 (max. performance) /LPM:x (BIOS) SATA Link Power Management Level (DECIMAL) 0 (max. power saving) ... 3 (disabled, max. performance) Deprecated options ------------------ *** /!RMV Obsolete options ---------------- /GEO /LBA /!CYLA /DC:n /!DSG /DT:x /NEC /PCLK:0 /MGAFIX /!R /FORCE /ATAPI /!ATAPI /PCMCIA Undocumented options -------------------- /LAT /FIXES /DEBUG COMMAND LINE DETAILS Beginning with release 1.0.8, this includes the undocumented switch /!SHUTDOWN found in the latest IBM1S506.ADD. At system shutdown, this IBM1S506 issues shutdown commands to each attached device. This behaviour is different from earlier releases. I've seen many reports about problems when the drives are shutdown. So I decided to no longer be compatible with IBM1S506 here. DANIS506 *tolerates* the option /!SHUTDOWN on the commandline, but it does nothing. If you want this new shutdown behaviour (on notebooks perhaps), then add the new option /SHUTDOWN. Rule of thumb: if your system doesn't suffer from any (possibly unknown) flaws, not a single command line option is required. The driver adapts itself to the environment and runs at maximum performance. In some situations, you may need some of the following options to achieve optimum performance: - ATAPI devices are initialized to PIO mode only if they appear to support only very early ATAPI specifications; the same is true if the IDE chip is known not to support PCI busmastering with ATAPI devices. In the former case you may try to run particular ATAPI devices using busmaster DMA by applying the proper /A:x /U:y /BM options. - a few chips need the /80WIRE option to enable ATA66 operation (see below). - if the DumpIDE utility detects an unsupported busmaster capable chip and thus recommends to set /GBM (see below). - you intend to hot-swap devices or use PCCard devices ------------------------------------------------------------------------- For all of us who get the impression today's computers are way too fast to keep up with, IBM did us the favour of adding another switch, /W (available for more than one year now). This switch replaces the /V switch, and just like /V, /W shows some very useful info about the EIDE system at boot time. But instead of throwing all of this stuff onto the screen and scrolling it off, /W writes it to the display in reasonable chunks and then *waits* for the user to hit a key after he/she has learned it by heart ;-) In order for /W to work you must have pausing on errors enabled in CONFIG.SYS, either by "PAUSEONERROR=yes" or by no PAUSEONERROR setting at all. To reduce screen clutter and avoid possible SYS2068 messages, some lines of the boot message may be suppressed by /VL (/WL) instead of /V (/W). If you prefer even less info at boot time, /VLL (/WLL) will print out the most basic messages only. Here is an example: With /V or /W it looks like this: Daniela's Bus Master IDE Driver for OS/2 Version 1.3.8 Controller:0 Port:01F0 IRQ:0E Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 VIA 571 PCI to IDE IC Rev:06 VID:1106 DID:0571 @33MHz Unit:0 Status:OK ATAPI Device Bus Master DMA DMA1/PIO3 Model:IOMEGA ZIP 250 ATAPI 51.G Controller:1 Port:0170 IRQ:0F Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 VIA 571 PCI to IDE IC Rev:06 VID:1106 DID:0571 @33MHz Unit:0 Status:OK ATAPI Device Bus Master DMA Ultra DMA2/PIO4 Model:Pioneer DVD-ROM ATAPIModel DVD-115 0108 E1.08 Unit:1 Status:OK SMS:16 LBA Bus Master DMA DMA2/PIO4 Identify Model:CASTLEWOOD ORB2-E USR-4.08 OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 534 4273 0 534 4273 Avail 4307184 H 128 16 0 0 128 16 OS2 4306176 S 63 63 0 0 63 63 % Used 99.98 Controller:2 Port:A000 IRQ:09 Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 CMD 649 PCI to IDE IC Rev:02 VID:1095 DID:0649 @33MHz Unit:0 Status:OK SMS:16 LBA Bus Master DMA Ultra DMA4/PIO4 BPB Model:IBM-DJNA-352030 J58OA30K OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 2482 39560 1023 39560 16383 Avail 39876480 H 255 16 255 255 16 16 OS2 39873330 S 63 63 63 63 63 63 % Used 99.99 Controller:3 Port:A800 IRQ:09 Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 CMD 649 PCI to IDE IC Rev:02 VID:1095 DID:0649 @33MHz Unit:0 Status:OK SMS:16 LBA NL:64 Bus Master DMA Ultra DMA5/PIO4 BIOS Model:IBM-DTLA-307030 TX4OA50C OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 14890 59560 620 59560 16383 Avail 60036480 H 64 16 64 0 16 16 OS2 60036480 S 63 63 63 0 63 63 % Used 100.00 Controller:4 Port:B400 IRQ:09 Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 HPT 370 PCI to IDE IC Rev:03 VID:1103 DID:0004 @33MHz Unit:0 Status:OK SMS:16 LBA Bus Master DMA DMA2/PIO4 BPB Model:QUANTUM FIREBALL_TM1280A A6B.2400 OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 621 2484 524 621 2484 Avail 2503872 H 64 16 255 64 64 16 OS2 2503872 S 63 63 63 63 63 63 % Used 100.00 Controller:5 Port:BC00 IRQ:09 Status:OK Bus Master DMA Scatter/Gather PCI Rev: 2.10 HPT 370 PCI to IDE IC Rev:03 VID:1103 DID:0004 @33MHz Unit:0 Status:OK SMS:16 LBA Bus Master DMA Ultra DMA2/PIO4 BPB Model:ST34310A 3.07 OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 524 8354 534 524 8354 Avail 8420832 H 255 16 128 255 255 16 OS2 8418060 S 63 63 63 63 63 63 % Used 99.97 But with /VL or /WL it's a little more comprehensive: Daniela's Bus Master IDE Driver for OS/2 Version 1.3.8 Controller:0 Port:01F0 IRQ:0E Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK ATAPI Device Bus Master DMA DMA1/PIO3 Model:IOMEGA ZIP 250 ATAPI 51.G Controller:1 Port:0170 IRQ:0F Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK ATAPI Device Bus Master DMA Ultra DMA2/PIO4 Model:Pioneer DVD-ROM ATAPIModel DVD-115 0108 E1.08 Unit:1 Status:OK SMS:16 LBA Bus Master DMA DMA2/PIO4 Identify Model:CASTLEWOOD ORB2-E USR-4.08 Controller:2 Port:A000 IRQ:09 Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK SMS:16 LBA Bus Master DMA Ultra DMA4/PIO4 BPB Model:IBM-DJNA-352030 J58OA30K Controller:3 Port:A800 IRQ:09 Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK SMS:16 LBA NL:64 Bus Master DMA Ultra DMA5/PIO4 BIOS Model:IBM-DTLA-307030 TX4OA50C Controller:4 Port:B400 IRQ:09 Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK SMS:16 LBA Bus Master DMA DMA2/PIO4 BPB Model:QUANTUM FIREBALL_TM1280A A6B.2400 Controller:5 Port:BC00 IRQ:09 Status:OK Bus Master DMA Scatter/Gather Unit:0 Status:OK SMS:16 LBA Bus Master DMA Ultra DMA2/PIO4 BPB Model:ST34310A 3.07 And finally, with /VLL or /WLL it becomes very terse: Daniela's Bus Master IDE Driver for OS/2 Version 1.3.8 Model:IOMEGA ZIP 250 ATAPI 51.G Model:Pioneer DVD-ROM ATAPIModel DVD-115 0108 E1.08 Model:CASTLEWOOD ORB2-E USR-4.08 Model:IBM-DJNA-352030 J58OA30K Model:IBM-DTLA-307030 TX4OA50C Model:QUANTUM FIREBALL_TM1280A A6B.2400 Model:ST34310A 3.07 No matter of the verbosity level which may suppress some lines on screen, *all* messages are stored internally for later retrieval! Beginning with beta 7, the driver saves all messages possibly shown at boot-up internally. Users may redisplay these messages, or store them into a file later with commands like these: COPY IBMS506$ con: or COPY IBMS506$ anyfile Please, check the spelling! Some time ago, IBM decided to enable /SMS by default on most controller hardware. The /!SMS switch then disables this controller based default. But if the harddisk reports to be SMS capable, SMS is enabled regardless of the /!SMS switch. Since this might cause sporadic failures on some hardware as a few users reported, I changed the behaviour of the /!SMS switch to a full override which is honoured in any case. Sometimes it's not necessary to disable SMS mode, but just to limit the number of sectors transferred in one run to a lower amount. To implement this feature, the /SMS option is changed into /SMS:n, where n is the maximum number of sectors in one run. If n is 0 or 1, this is equivalent to /!SMS. Every harddisk has a maximum SMS run length defined by the manufacturer specific to each model. You cannot set a value above this individual maximum SMS run length. The driver sets up the actual SMS run length for a specific drive to the minimum of the user supplied value n and the maximum value reported by the drive. ------------------------------------------------------------------------- To take advantage of Ultra DMA mode 3 (44MB/s), Ultra DMA mode 4 (66MB/s), Ultra DMA mode 5 (100MB/s), or Ultra DMA mode 6 (133MB/s), *all* of these prerequisites must be true: - the EIDE controller must support the ATA-66/100 enhancements - the device must support the ATA-66/100 enhancements - the cable must be a 80 wire type (recommended in any case) - the motherboard must support the detection of the 80 wire cable (see #) - the EIDE controller must report a 80 wire cable present - the device must report a 80 wire cable present # some chipset manufacturers still don't assign dedicated pins to enable the detection of the cable type installed, and therefore don't define a standard procedure for driver writers to acquire this *vital* information. Any motherboard with a VIA596B, VIA686, or AMD756 south bridge is affected by that: the ATA-66 features aren't enabled by default in the driver because it has no way to detect the presence of a 80-wire cable reliably. In this case add /80WIRE to the *adapter* options if there is a 80-wire cable installed on that particular EIDE channel. This decision is completely up to *you*. Don't blame me for lost or corrupted data if you are wrong! As a workaround to solve this inconvenience with the chips mentioned, beginning with release 1.3.0, the driver checks the controller setup as done by the BIOS. If it finds a particular channel initialized to ATA-66 operation it assumes the BIOS positively checked for the presence of a 80 wire cable. ------------------------------------------------------------------------- The following options are features added by me: ------------------------------------------------------------------------- Support for drive bays and PCCard slots Many notebooks/laptops have so called drive bays built in to swap different devices (floppy, disc, CD/DVD, etc.) while powered on state. The option /BAY prepares the related IDE adapter for such a hot-swap. To make this safely work on OS/2, the support for APM (by means of DEVICE=APM.SYS) has to be installed. During APM suspend the device in the drive bay may be swapped The floppy driver and the different ATAPI filters made by IBM check the presence of appropriate devices after APM resume and operate accordingly. Beginning with release 1.1.9, DaniS506.ADD improves this hot-swap support. After APM resume the driver inspects these "special" adapters and reconfigures both the device and the controller exactly as it does at boot time. It then reports its findings to interested filter drivers controlling the units so that the filter driver can reconfigure itself appropriately. At present, only DaniATAPI.FLT supports this advanced reconfiguration protocol. Example: A notebook has got a drive bay with a swappable floppy drive, disk, ZIP, and DVD-CDR/W. The following lines in CONFIG.SYS make all these devices swappable and operational without powering down: BASEDEV=IBM1FLPY.ADD /A:0 /FORCE:1 /U:0 /F:1.44MB BASEDEV=DaniS506.ADD [other options] /A:1 /BAY (see note #) BASEDEV=DaniATAPI.FLT /A:1 /U:0 /TYPE:ZW note #: this assumes the secondary channel to be connected to the drive bay The ZIP drive and the hard disk show up as removable media unit in the drives folder after a "refresh removable media" command. Important: before swapping the units there *must* be an eject of the media (either by the "EJECT x:" command from the command line or fro the context menu of the drive icon)! Forgetting to do that may cause data loss because the file system caches may not yet be written back to the media. With some luck the command "ChkDsk x: /F" will repairs such damage without data loss. Beginning with release 1.3.0, the driver supports PCCard slots as well. This support will be enabled by means of the options /PCS or /PCS:n. With /PCS the driver checks how many PCCard slots are available from the socket services driver and creates the same number of "virtual" IDE adapters. If less IDE adapters for PCCards shall be created, then the required amount n may be set by the option /PCS:n. At most the corresponding number of PCCards with IDE units attached may be inserted into the PCCard slots at the same time. A "virtual" PCCard IDE adapter is implemented by a PCCard (the actual IDE controller) which is inserted into any one of the PCCard slots. Each supported PCCard inserted *concurrently* counts as one IDE channel. The driver attaches to OS/2's PCMCIA subsystem and takes over all PCCards with a function id of 4 (disk) and function extension 1 (ATA interface); this includes PCCard ATAPI devices as well (like the PCMCIA ZIP-250 or the Click!40 drives). Beginning with release 1.3.10 both ATA and ATAPI devices are supported (if DaniATAPI.FLT rev 0.3.0 or later is installed). PCCards may be arbitrarily inserted to and removed from any PCCard socket. The only prerequisite is a functional socket services driver and PCMCIA.SYS. Beware: older revisions of IBM's PCMCIA subsystem have defects which may cause failures to configure cards on sockets other than the first one. After inserting a PCCard into the slot, you'll likely hear some noises from the PCCard or its attached unit. Shortly after that you should hear a high pitched beep. This beep indicates that the driver has accepted the PCCard, has successfully assigned resource to it and initialized it, and was successful communicating with the attached unit. If you don't hear this beep, try reinserting the PCCard. A failed initialization will result in a two times low pitched beep. If you intend to remove a PCCard, don't forget to close the file system on the attached device by issuing "EJECT x:" from the command line or the drive's context menu. Failing to do so will leave the file system in an inconsistent "dirty" state which most likely needs to be corrected by CHKDSK to make the medium accessable again before the next use. "Eject" doesn't necessarily unload the medium (f.e. in case of the Click! drive), so don't be confused. Some drive activity will indicate the flushing of still unwritten file system cache contents back to the medium. The end of the flush operation is indicated by a middle pitched beep. Afterwards it is safe to remove the PCCard from the socket or swap the medium. Beginning with version 1.6.0 the driver handles PCCards from which the system ist going to be booted from and which are already initialized by the BIOS (for example the Sony Vaio Crusoe subnotebook with a bootable external CD-Rom). To do that some technical prerequisites must apply (the PCCard slot needs to be ExCA register compliant) and there must be no socket services driver loaded; the latter one is needed only because of errors in the socket services drivers which make booting from a BIOS initialized PCCard impossible. /BAY on SATA channels enables SATA and SATAPI device hotplugging. This is particularly useful with hot-swap disk enclosures. ------------------------------------------------------------------------- Modern disks have multiple ways to save power. At times of inactivity, internal circuits may be temporarily suspended or motors switched off on demand. These functions are summarized by the term "Advanced Power Management" (APM). The balance between disk access speed and power-saving is controllable. For this purpose, the driver implements the option /APM:n which may apply to single, multiple or all disks implementing the APM function. The valid range of the value 'n' is 1 to 255; increasing values mean better performance, and the value 255 switches off all power-saving functions. The meaning of any given value is on the manufacturer's discretion. Users have to look-up disk datasheets for this information by themselves, I won't reply to any questions regarding this topic! Rule of thumb: just try the value 128: it allows all power-saving tricks except for switching-off the disk's spindle motor. Because of that, the performance impact on the very first access after some time of inactivity is minimal. If a disk implements the APM feature, it should be strongly preferred over the inactivity timer described later on! Issuing the command "DiskInfo I" will reveal if a disk implements the APM feature: APM feature implemented, but inactive: Advanced power management level: disabled Commands/features: Enabled Supported Advanced Power Management feature set APM level 128 active: Advanced power management level: 128 Commands/features: Enabled Supported * Advanced Power Management feature set ---------------------------------------------------------------------- The SATA specification describes an optional feature to control the power consumption of a SATA link. The more circuits in the SATA link hardware are switched-off to save power the longer it takes to reactivate them. There is no generally available means to control the power mode of a SATA link from the controller side, but modern SATA disks often implement the feature to enable link power-saving modes at times of inactivity. The option /LPM:n (Link Power Management) enables this function for single, multiple or all disks which implement this feature and are connected to a SATA port which also has this feature implemented and enabled. The parameter 'n' may have the following values: 0: power-saving allowed according to the controller's BIOS setting 1: all power-saving modes allowed (unconditionally) 2: all power-saving modes allowed except for SLUMBER (unconditionally) 3: all power-saving switched off, maximum link performance Depending on the disk-controller combination, some values may exhibit failures or may have no effect. In general, the value zero is the best option. It's up to users not to use this function in case of problems. The driver can't do anything about that if this part of the SATA specification isn't correctly implemented in the given hardware. Issuing the command "DiskInfo I" will reveal if a disk implements the prerequisites to enable link power management: Prerequisites satisfied, but function inactive: Commands/features: Enabled Supported Device-initiated interface power management Prerequisites satisfied, power-saving mode switching active: Commands/features: Enabled Supported * Device-initiated interface power management Issuing "DumpIDE" will show the current power-saving mode of a given SATA link: SATA link 0: Status : 00000123 - device pres., PHY comm; 3.0Gb/s; interface active bzw. Status : 00000123 - device pres., PHY comm; 3.0Gb/s; i/f power PARTIAL bzw. Status : 00000123 - device pres., no PHY comm; 3.0Gb/s; i/f power SLUMBER ------------------------------------------------------------------------- Inactive SATA link ports are switched-off unconditionally for power-saving purposes. ------------------------------------------------------------------------- To use the inactivity timer feature simply add /IT:x to the *unit* parameters. x is the timeout value in minutes. A 0 disables the spin down completely. If you leave out this option, some devices will still spin down after a certain default inactivity time. For proper driver operation of this function, all of the following should be true: - in BIOS, APM must be enabled - in BIOS, hard disk power down timers must be disabled - in CONFIG.SYS, at least the statement DEVICE=APM.SYS must be present (in other words: OS/2 APM support must be installed) If these prerequisites are true, the driver can keep track of disk power down even in case of intermittant system APM suspends. This ensures quick error recovery by short command timeouts, but sufficiently long timeouts for drive spin-ups. If APM is not available in OS/2, the driver will prolong command timeouts to 20s for each command if a drive reports active power down modes or the user adds an /IT option; all other drives will be sent a command to disable automatic spin-down. ------------------------------------------------------------------------- To limit the data rate a device communicates at, add /MR:udp to the *unit* parameters. Choose the values for u, d, and p from this table: u: Ultra DMA 0 - disable Ultra DMA 1 - limit to Ultra DMA mode 0 2 - limit to Ultra DMA mode 1 3 - limit to Ultra DMA mode 2 4 - limit to Ultra DMA mode 3 5 - limit to Ultra DMA mode 4 6 - limit to Ultra DMA mode 5 d: Multi Word DMA 0 - disable MW DMA 1 - limit to MW DMA mode 0 (in S506.ADD effectively the same as 0) 2 - limit to MW DMA mode 1 3 - limit to MW DMA mode 2 p: PIO 0 - limit to PIO mode 0 1 - limit to PIO mode 1 (in S506.ADD effectively PIO 0) 2 - limit to PIO mode 2 (in S506.ADD effectively PIO 0) 3 - limit to PIO mode 3 (with flow control) 4 - limit to PIO mode 4 (with flow control) If you set u and d to 0 at the same time, this effectively disables busmastering on this device since no DMA mode is left to use. Leading zeroes may be left out. Many EIDE controllers don't have the capability to select PIO modes and multiword DMA modes independantly because theses mode share the same timing registers. This is for historic reasons and the similarity of cycle timings. This applies to all controller chips mentioned above except for Promise Ultra and Cyrix MediaGX. Note: this applies to mixing transfer modes for the *same* device! The devices on an EIDE channel are independent from each other regarding the transfer modes (except for Intel 82371FB). To deal with this situation the driver selects the slower cycle time from the proposed PIO and MW-DMA modes. The net result can be seen from the table below: | no DMA MW-DMA 1 MW-DMA 2 Ultra-DMA x ------+---------------------------------------------------------------- PIO 0 | PIO0 PIO0/DMA1 PIO0/DMA2 PIO0/UDMAx (Intel,Promise, | Cyrix) PIO 0 | PIO0 PIO0/no DMA PIO0/no DMA PIO0/UDMAx (others) ------+---------------------------------------------------------------- PIO 3 | PIO3 PIO3/DMA1 PIO3/DMA2 PIO3/UDMAx (Promise,Cyrix) PIO 3 | PIO3 PIO3/DMA1 PIO3/DMA1 PIO3/UDMAx (others) ------+---------------------------------------------------------------- PIO 4 | PIO4 PIO4/DMA1 PIO4/DMA2 PIO4/UDMAx (Promise,Cyrix) PIO 4 | PIO4 PIO3/DMA1 PIO4/DMA2 PIO4/UDMAx (others) ------------------------------------------------------------------------- Some users are forced (or feel themselves forced ;-) to run the PCI bus on their motherboards at speeds other than the optimum 33MHz. For example, many Cyrix processors need very uncommon front side bus speeds which cannot be easily divided into 33MHz. This leads to either too slow EIDE timings or too fast ones. The former case may be disappointing, but the latter overclocks the EIDE interfaces and my introduce data distortions. If you know the actual PCI clock of your motherboard, you may add an additional driver option /PCLK:x (x is the PCI clock in MHz). The user supplied PCI clock is then sorted into one of four speed classes (25MHz, 33MHz, 37MHz, and 41MHz) which is reported at boot time. For each of these speed classes, the driver *tries* to select the optimum chip timings, but due to chipset limitations this is not possible in all cases. As most motherboards run their PCI bus at the optimum 33MHz clock today, you should use this switch only if really necessary! The technical background of this option: The so called Front Side Bus (FSB) is the bus the CPU communicates with the north bridge (which contains the memory controller, the cache controller, the AGP bridge, the PCI arbiter, the PCI bridge, and optionally some more devices). Depending on the design of the north bridge, the PCI clock is derived from the FSB clock by dividing this clock (synchronous design) by either an integral number (2, 3), or a fractional number (2, 2.5, 3), or it is derived from an independent clock source (asynchronous design, very uncommon). The PCI 2.1 spec (without its extensions) allows for 33MHz max. PCI clock. The VIA, the ALi, and most likely the SiS chipsets are capable of fractional dividers. The Intel chipsets are not. If you know, your PCI bus runs at the typical 33MHz clock (true for most of the mobo/CPU combinations sold today), don't apply this switch. If you know the PCI clock in advance (or like to tune the EIDE timings for some reason), you apply the /PCLK:n switch where n is the known PCI clock. The driver then selects the speed class from this value in the same way as from an autodetected value. These speed classes are the result of the typical FSB clocks and dividers: FSB clock / divider -> PCI clock -> class 50MHz / 2 -> 25MHz -> 25MHz 60MHz / 2 -> 30MHz -> 33MHz 66MHz / 2 -> 33MHz -> 33MHz 75MHz / 2 -> 37.5MHz -> 37MHz 75MHz / 2.5 -> 30MHz -> 33MHz 75MHz / 3 -> 25MHz -> 25MHz 83MHz / 2 -> 41.5MHz -> 41MHz 83MHz / 2.5 -> 33MHz -> 33MHz 95MHz / 3 -> 31.7MHz -> 33MHz 100MHz / 3 -> 33MHz -> 33MHz 112MHz / 3 -> 37.3MHz -> 37MHz 124MHz / 3 -> 41.3MHz -> 41MHz For the chosen speed class, the driver tries to select EIDE controller register values which don't exceed the PIO/DMA/Ultra ratings. My former driver revisions and IBM1S506 only know the proper register values for the 33MHz class. This runs the EIDE interface slower than possible in case of a 25MHz PCI clock, but exceeds the allowed timings in the 37MHz/41MHz cases. The former one is uncritical, but the latter one may cause failure of operation (better) or data corruption (worse). To give you an example how to use command line switches, look here: BASEDEV=DaniS506.ADD /V /A:0 /U:1 /MR:223 /IT:5 /A:1 /I This shows some info at boot time, limits the data rates to communicate with the primary slave to UDMA1/MWDMA1/PIO3 and stops the spindle motor after 5 minutes of inactivity on this device (which happens to be the IBM-DTTA351010 drive). It tells the driver also not to touch the secondary channel (adapter 1). ------------------------------------------------------------------------- The /GBM option enables the scan for busmaster capable PCI EIDE controllers if none of the 'well-known' chipsets (currently INTEL, VIA, ALi, SiS, CMD, Promise, HighPoint, AMD, SMSC, Artop, ServerWorks, CMD640, and RZ1000, the latter two are not busmaster capable!) is found. Before a chip is accepted by the driver to be operated in generic busmaster mode, it has to pass some sanity checks. Some BIOSes simply fail to initialize the controller appro- priately (seen on a board with an UMC8886BF). A rule of thumb: if Linux doesn't run a chip in DMA mode, it is not generic busmaster compatible. At present, EIDE controllers not supported by any S506 ADD run attached devices in PIO mode only, and with timings setup by the BIOS (most probably PIO0 if the device is hidden from the BIOS). The generic busmaster option enables usage of the busmaster machinery and device DMA modes, thus reducing CPU load by a large amount. In this mode, the ADD doesn't touch any chip registers other than the standard PCI command register (to enable PCI busmastering) and the registers associated with the busmaster engine. It has *no* way to enable Ultra-DMA modes in hardware if not done by the BIOS. If you think this should be the default behaviour of my driver, then simply change the /GBM option into the /FORCEGBM option. Then, *every* EIDE chip is treated as if it's not a 'well-known' chip. All programming of device timings in the EIDE controller has to be done by the BIOS. /FORCEGBM comes in two flavours, as a global option, and as an adapter option. If /FORCEGBM preeceds all /A:x options, it applies to all adapters; if it follows an /A:x option, it applies to adapter x only. In some cases, /FORCEGBM can help to overcome stability problems because of the spec-conforming timings that the driver programs into the controller registers. In the recent past, some batches of boards popped up which are "tuned-down" by their BIOS to work reliably. In these cases, the "correct" timings are too fast and the boot process fails if these timings are set up by DaniS506.ADD. Adding /FORCEGBM prevents the driver from adjusting the timings so that the "special" BIOS timings are used instead. Examples: BASEDEV=DaniS506.ADD /W /GBM /A:0 /U:0 /MR:023 BASEDEV=DaniS506.ADD /W /FORCEGBM /A:0 /U:0 /MR:023 /A:0 /U:1 /MR:034 /IT:5 ------------------------------------------------------------------------- There is a new option /PF:xxxx which may be applied to an adapter or individual units. This option is meant to fine-tune register access timings and enable chipset-performance features. The option value is a bit field, whose individual bits need to be put together into the final option value. Bits 0-7 are meant for all PCI chipsets, bits 8-15 are chipset specific. In this driver release, the following settings are available: +---+---+---+---+---+---+---+---+---+---+---+---+---+---+-------+ |15 14 13 12 11 10 9 8 | 7 6 5 4 3 2 1 0 | +---+---+---+---+---+---+---+---+---+---+---+---+---+---+-------+ | unused | unused | (1) | +---+---+---+---+---+---+---+---+---+---+---+---+---+---+-------+ 1) Register access delay: 0 : 1000 ns (default, fixed for all non-PCI systems) 1 : 500 ns 2 : 250 ns 3 : no delay (works on most systems, check out yourself) The actual access delay used for a particular adapter channel is the largest of all attached units! BASEDEV=DaniS506.ADD /A:2 /PF:2 /U:0 /PF:3 If there is no slave device, then no register access delays are incurred because the individual unit setting /PF:3 overrides the adapter setting /PF:2. But if there is an additional slave device, this unit inherits the adapter option /PF:2. The largest delay of both units is the resulting one and thus each register access is delayed by 250ns. ------------------------------------------------------------------------- Modern ATA spec 6 compatible drives may support the so called acoustic management feature set. Using this you can trade off performance vs. noise (mostly noticable during quick sequences of random seeks). The new unit option /NL:n allows you to control this feature: the value n may be set to 0 ... 126, where '0' selects the faintest noise level with the biggest performance impact, while '126' selects best performance at the expense of noisiest operation. Drives are not required to support any value in between, but to select the next supported one. Drives store the latest programmed value internally and preserve it over power outages and resets. The driver will tell the current value read from the drive if you don't use /NL option, or the user selected value n if /NL:n is present. ------------------------------------------------------------------------- One user reported a boot failure in a particular situation which involved SCSI and ATA/ATAPI units. The solution was to avoid BIOS INT13 calls. The driver asks the BIOS about the geometry properties of the ATA drives. In general, this information is not required because the most authorative source of the geometry info is the drive itself (by inspecting the MBR and BPBs). This implies that no BIOS INT13 calls are required in almost any scenario. The only exception is if your disk is completely blank, and you want to partition this blank disk by OS/2 FDISK with a geometry compatible to the BIOS one. So, if you prefer to deny the driver calling the BIOS, add the option /!BIOS. When doing so, the "strange" VDM seen with some process management utilities is gone as well. ------------------------------------------------------------------------- Later versions of FDISK and LVM tend to overwrite bootmanagers other than OS/2 BootManager on every write operation to the partition/LVM tables. In such cases it might be appropriate to write protect track zero of the boot drive (which contains the bootmanager code). This is possible either at bootup using the unit option /A:x /U:y /PROT0 ("PROT zero"), or later by means of the DiskInfo utility: DiskInfo <unit> p 1 enable write protection on <unit> DiskInfo <unit> p 0 disable write protection on <unit> "DiskInfo ?" shows all options in detail. ------------------------------------------------------------------------- Todays large disk drives often don't work properly with older BIOSes. Either the BIOS detects the wrong disk capacity, or sometimes the PC doesn't boot at all. In such cases, the cure of the problem is either to set a jumper at the disk or reprogram the disk such that it reports a smaller, troublefree capacity to the BIOS. Beginning with version 1.3.8 the driver is able to use the full disk capacity even if the disk is programmed to report a smaller capacity. This is done automatically; if you don't like that then you have to prevent the unlocking of the full capacity by means of the unit option /A:x /U:y /!SETMAX. To program a disk to report a smaller capacity, I recommend this process: 1) remove the disk from the BIOS setting. The PC will no langer hang. 2) boot from an OS/2 boot disk (or the like). 3) program a smaller disk capacity by means of the Diskinfo utility: DiskInfo <unit> m <new capacity>, f.e. DiskInfo 0m m 32G For details see "DiskInfo ?" 4) reboot 5) if you like you may add the disk to the BIOS settings again INSTALLATION For an upgrade from a previously installed IDE driver (most likely IBM1S506.ADD) just copy DaniS506.ADD to bootdrive:\OS2\BOOT and rename IBM1S506.ADD (or the name of the previous IDE driver) to DaniS506.ADD in your CONFIG.SYS. For a first time install of an IDE driver you'll have to add this line to CONFIG.SYS: BASEDEV=DaniS506.ADD In this case your boot device probably is SCSI. To avoid confusion make sure to move the mentioned line *below* the SCSI driver. If you prefer chasing icons, you can do the copy and the modification of CONFIG.SYS using DDINSTAL ("OS/2 System"/"System Setup"/"Install/Remove" /"Device Driver"). Due to limits in DDINSTAL you have to disable the previously installed S506 base driver in CONFIG.SYS manually, either by deleting that line or putting a 'rem' statement in front of it. FILES DaniS506.ADD The driver itself. DaniS506.DDP The device driver profile for DDINSTAL DaniS506.DOC The file you are reading right now :-) DumpIDE.EXE A utility to dump the current IDE registers DiskInfo.EXE A utility to show some information about the attached devices \nodebug\DaniS506.ADD A smaller driver version without debug traces \nodebug\cmprssd\DaniS506.ADD The same as above, but LXlite compressed. This one is *not* guaranteed to work on every system. \source\DiskInfo.c Guess ;-) BUGS Hopefully none, but who knows ... !!! Bug reports without IDE configuration and transfer modes involved, and !!! without the output from the dump utility, are pretty useless. You should !!! create these outputs in a running *and* the faulty system. The easiest way !!! to do this is to redirect the screen output from the dump utility to a file like this: DumpIDE > Good.dmp or DumpIDE > Bad.dmp Many people refer to driver revisions by file dates. Don't do that for these reasons: - I don't keep track of file dates - I build several drivers almost each day - file dates can easly be changed - there are many different calendars in this world The only valid reference is the driver's version number. It is shown on boot up if the /V or /W switch is given. If you can boot to a command line !!! the output of "COPY IBMS506$ anyfile" will list the driver version in !!! the first line. If you cannot boot, then run "BLDLEVEL DaniS506.ADD" which will output something like this: Signature: @#DANI:1.2#@##1## 4.7.2000 10:45:10 .... Vendor: DANI Revision: 1.02 Date/Time: 4.7.2000 10:45:10 Build Machine: Nachtigall File Version: 1.2.5 <------------------------------+ Description: Adapter Driver for ST506/IDE DASD | | This is the full version number -----------------------+ I have no means to identify traps occuring outside this driver (module name different from 'IBMS506$' or not present). If a trap occurs within this driver, I need the following information: CS:EIP, EAX-EDI, DS and ES. My test systems at home (VIA and ALi) and a couple of other machines at work (different INTEL chips) run flawlessly with this driver, so I have virtually no chance to reproduce errors. If any of your EIDE devices is not detected at boot-up, please contact me. The driver has special debug code built in; I will tell you how to enable and gather these traces. If you experience problems with your hard disks after OS2DASD.DMD has been loaded (to be seen by pressing ALT-F2 when the white box appears on the screen), read the following section *very carefully*, it applies to you. I cannot solve your hardware problems with software. If you think your problems arise from bugs in the EIDE controller chip, you may contact me after gathering detailed info about this bug and its software solution. At present, I have a recent chip bug list from ALi and another one for the VIA 586 chipset created by myself. SOME WORDS OF WISDOM Much EIDE hardware you can buy (or already have bought) has *very* little design margins and is *very* susceptable to even the slightest signal distortion on the cables. Unlike SCSI, the EIDE bus has no termination resistors on both ends of the cable to reduce ringing or other signal reflections. Noise and crosstalk are other sources for bit errors during data transmission; careful grounding (both the data and the power connection) is paramount if you intend more than simple PIO speed. Power supplies are not uncommon for inducing noise! A very patient user found another source of data distortions after some very long and tedious reliability tests: the cache subsystem of his motherboard. He experienced bit errors which vanished after some fiddling within the BIOS settings of the cache timings (pipelining was the culprit in this case). This isn't something magic! Consider, a busmaster - especially when running at high transfer rates - puts an additional heavy burden on the cache logic in the chipset and the cache RAMs. Many random failures of OS/2 systems - even without busmastering of EIDE or SCSI controllers - have their origin here. If you think your data is trashed due to bugs in this driver, please take this into account: I made a lot of changes to IBM's source code in the driver initialization portion to support ALi, SiS, VIA, CMD, and Promise chipsets, and program spindown timers in the devices, and to select data transfer rates other than the fastest ones. Some of these changes could lead to traps during bootup depending on the actual parameter values used. I hope to have them fixed. These traps (identifiable by the module name 'IBMS506$') cannot corrupt any of the data because they happen before any write operation is issued to any drive. I'm sorry if you happen to have such an inconvenience. BUT: the code to carry out read and write requests either by PIO or busmaster assisted DMA is *unchanged*. In fact, all data transfers from the device into host memory or vice-versa are done 100% by the drive electronics and the chipset hardware on your motherboard (or 99% plus *one* CPU instruction in case of PIO). Driver code only steps in when any part of this hardware reports an error or wasn't able to fulfill the request. Keep in mind: PIO and MWDMA transfer modes have no means of detecting bit errors during data transfers, Ultra-DMA does CRC checking to maintain data integrity. So, instead of rushing to the fastest data transfer modes possible, limit the negotiated data transfer speed to the slowest available in each category (/MR:120). Using PIO4 on other than rock solid hardware which passed reliability tests is *very* dangerous. Since large portions of the OS2*.INI files cannot be flushed to disk at shutdown with error-checked Ultra-DMA, the ADD has to revert to the error-prone PIO mode. I consider this an OS/2 design flaw. If you run SYSBench or DISKIO, and experience unstable or rather low performance figures in the maximum data transfer rate test, you are seeing the effects of retries and interface resets. This is an indication of bit errors! As a reference, in the DISKIO 'Drive cache/bus transfer rate' test you should see the following results: Ultra-DMA mode 5 : 66MB/s or more Ultra-DMA mode 4 : 44MB/s or more Ultra-DMA mode 3 : 34MB/s or more Ultra-DMA mode 2 : 24MB/s or more Ultra-DMA mode 1 : 17MB/s or more Ultra-DMA mode 0 : 13MB/s or more MW-DMA mode 2 : 11MB/s or more MW-DMA mode 1 : 9MB/s or more SYSBench is based on the DISKIO code with some modifications. DISKIO does the worst case test on the last cylinder (falsely called track in both programs) using forward reads, SYSBench uses backward reads. Remember, different drives have different cache strategies built in, and reading backwards is a pretty uncommon access strategy with file system caches, so some drives are fooled in this test and the result is overly pessimistic. If you like to tune the data transfer speed on the cables to your drive, then run DISKIO or SYSBench without DMA rate limitation and look at the results reported for cylinder (track) 0. This is the maximum data transfer rate your drive is able to read off the media surface. Then select a transfer mode which matches this speed. I recommend Ultra DMA modes over MW-DMA modes. If the drive is DMA capable, select PIO mode 0 as non-DMA mode (for the rare cases where busmaster DMA cannot be used). If you see false device detects or even lockups at boot time, don't expect reliable data transfers thereafter. In any case, if you think you are experiencing data corruption, look at the output generated by the /W or /V switch, and then look for the data transfer modes used with your device. The first remedy to try will be disabling all fast transfer modes (by /!BM and/or /MR:). If you hear short beeps, then the data transfer state-machine encountered an exception: a high beep indicates a UDMA CRC failure, a low beep an error reported by the hardware which leads to a interface reset eventually, and a middle beep indicates a command timeout (usually caused by a spun-down harddisk). So, if your hardware works 'flawlessly' you will hear *no* sounds from the driver (except for spin-down events of course). With 'flawlessly' I mean 'without errors detectable by any instance'. If you prefer to ignore hardware errors indicated by these beeps, add the /!AA (no audible alerts) option. To protect the current working desktop from corruption, consider creating test copies of the x:\Desktop directory tree and the OS2*.INI files and pointing to these in a modified copy of your CONFIG.SYS located in x:\OS2\BOOT. For example, rename this copy to CONFIG.T and select this test configuration within the recovery choices screen by hitting the 'T' key. This way, you always have a working default boot system and a independent test system to play with. QUESTIONS & ANSWERS 1) Both my EIDE controller and my harddisc is ATA/66 (or ATA/100) capable, and a 80wire cable is installed. No matter what I am trying to do (reading the docs, applying the /80WIRE override switch), the driver allows no more than UltraDMA mode 2. What's wrong ? Most likely, your harddisk is setup to hide its ATA/66 (or ATA/100) capabilities. Some models are delivered with such a factory setup to prevent confusion of older BIOSes or controllers. Run "DiskInfo i" and look at word 88/058 reported from the disk in question. If it it something like xx07, the disk reports ATA/33 capability; if it is xx1F, it can do ATA/66, and if it is xx3F, then ATA/100 is available. If the reported capability is less than advertised, contact the disk manufacturer for a utility to enable the missing capabilities (often to be found on their web sites). 2) I experience random failures after wakeup from suspend. What can I do ? Toggle the "Delayed Transaction" setting at the PCI page in the BIOS. Some hardware doesn't implement this feature correctly. 3) I have a large (> 33GB) disk but FDISK or LVM report only a fraction of the true disk capacity. Where does that come from ? Run "COPY IBMS506$ con:" and look at the geometry values (example): Model:WDC WD400BB-00AUA1 18.20D18 OS2:log phys BIOS:log BPB:log IDE:log phys Total Sectors C 4865 65535 16383 65535 16383 Avail 78165360 H 255 16 16 255 16 16 OS2 78156225 S 63 63 63 63 63 63 % Used 99.98 If the "OS2:log" column shows a C value of 65535 then there is most likely a 16 bit overflow due to a badly chosen translation mode. Presumably, the H value will be 16 instead of the much more appropriate 255 (or 240). If you see the same H value in the "BPB:log" column then this is the translation geometry stored on the disk itself, in the partition table and/or the BPB sectors. In this case you *must* zero out the partition table to get rid of this unwise translation mode, a mere change in the BIOS settings is not sufficient! If there are only zeroes in the "BPB:log" column, then the disks doesn't have a valid partition table yet and the translation mode is taken from the BIOS. Change the BIOS setting to "LBA translation" mode. 4) I have problems reading CDs now which was working fine with earlier driver versions. Why ? Later drivers enable DMA modes on ATAPI devices now if the seem to follow sufficiently recent ATAPI implementation standards. Some ATAPI units still fail running on DMA, so you have to disable DMA busmastering for these units (/A:x /U:x /!BM). CREDITS I'd like to express my thanks to some outstanding people who made this piece of code a better one: Helen Terbizan for her neverending efforts to make ALi support possible and her patience and perseverance during the tests. Mark Dodel for just the same patience and perseverance. Jens Glathe for his helpful tips and the courage to test the SiS code. Piotr Oniszczuk for his persistent reliability tests. Fred Springfield for his debugging of this pile of words. Gerald Eischer for creating the command line options diagram and table TO DO Still left is to persuade IBM to incorporate these features in their distribution. done! IBM has officially taken the v1.6.6 sources. COPYRIGHT DaniS506 - Copyright (c) 1999-2008 Daniela Engert. All rights reserved. STATUS OF PROGRAM DaniS506 is freeware. DISCLAIMER YOU ARE USING THIS PROGRAM AT YOUR OWN RISK! I don't take any responsibillity for damages, problems, custodies, marital dis- putes, etc. resulting from use, inability to use, misuse, pos- session or non-possession of this program directly or indirectly. I also don't give any warranty for bug-free operation, fitness for a particular purpose or the appropriate behaviour of the program concerning animals, programers and little children. THE SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION. Or, in a few words: If its good, I am responsible. If its bad, its all your fault. ;) Permission is granted to redistribute this program free of charge, provided it is distributed in the full archive with unmodified contents and no profit beyond the price of the media on which it is distributed is made. Exception to the last rule: It may be included on freeware/shareware collections on CD-ROM, as well as on magazine cover CD-ROMs. All trademarks mentioned anywhere around here are property of their owners and the like ... AUTHOR Daniela Engert Internet: "Daniela Engert" <dani@ngrt.de> EPILOGUE Suggestions and bug-reports are always welcome. Well ... bug-reports are perhaps not *that* welcome ... ;)