*) minor change o) optimization +) feature added -) something removed !) fix !!) critical fix i) info e) experimantal -- 2012.07.26 - v0.42c +) added code for ATAPI BUSY wait auto-adjustment. We poll and measure BUSY-after-Interrupt delay. Next time we use this value to defere interrupt service instead of polling. -- 2012.07.26 - v0.42c1 !) undone newer FreeBSD changes in Intel UDMA setup. Caused significant performance degrade on ATAPI. -- 2012.07.27 - v0.42d !) fixed bug with setting ATAPI flag in AHCI IO flags for non-packet commands on ATAPI devices. -- 2012.07.27 - v0.42d1 !) fixed drive size reporting bug for CHS-only hardware. Was reported as 0, since UserAddresableBlocks is not valid. -- 2012.07.27 - v0.42d2 !!) fixed bug with concurrent use of InternalSrb by ATAPI internal requests and AHCI internal requests. Added AhciInternalSrb. -- 2012.07.28 - v0.42d3 !) added code for handling ATAPI on AHCI those generate interrupt, but do not clear BUSY and do not affect any bit in channel IS register. We should treat such case as in Native/Legacy ATA mode - just wait for BUSY release. *) optimized Interrupt disable/enable case in AHCI branch of AtapiDmaReinit() and AtaSetTransferMode() We shall not touch IE registers if no operaion is required. -- 2012.07.29 - v0.42d4 *) added logs to AHCI ATAPI pre-ready state checks. !) fixed code for ATAPI BUSY wait auto-adjustment. Delay time was not stored in LunExt properly +) added internal glabal option MaxIsrWait, which determines ATAPI BUSY wait threshold. If expected delay is above MaxIsrWait, interrupt service shall be immediately posted to DPC. It is defaulted to 40us for real hardware and to 200us for VBOX. Previously shreshold was hardcoded. -- 2012.07.29 - v0.42d5 +) added chglog.txt +) added ATA_REQ.ahci field to store error status observed in UniataAhciEndTransaction() *) added logs to UniataAhciEndTransaction() to study ATAPI completion in more details *) removed extra logs from AtapiCheckInterrupt__(), added in v0.42d4 -- 2012.07.29 - v0.42d5 +) added registry key for MaxIsrWait -- 2012.07.30 - v0.42d6 +) added FBS AHCI register definition (not used now) +) added IDENTIFY and COMMAND definitions from ATA/ATAPI Command Set 2 +) added SCSI ERASE10 command definition !) direct access to ATA Command register is replaced with call to AtaCommand() for AHCI compatibility *) added check for .Removable flag before sending SCSIOP_MEDIUM_REMOVAL +) added ByteCount extraction from RFIS for ATAPI devices. -- 2012.07.30 - v0.42e +) added MODE_SENSE6, Page CACHING support for HDDs +) added more logs and fixed ByteCount extraction from RFIS for ATAPI devices. -- 2012.07.30 - v0.42e1 *) added setting of DRIVE_SELECT to AHCI ATAPI branch in H2D setup -- 2012.07.31 - v0.42e2 +) added definition of H2D FIS structure +) added definition of FIS types (codes) *) clarified definition (size) of ACMD block in IDE_AHCI_CMD and PFIS, RFIS, SDBFIS in IDE_AHCI_RCV_FIS *) defined constants for FIS filed offsets IDX_AHCI_i_Xxxx *) added setting of IDE_DC_A_4BIT to AHCI H2D FIS setup !) fixed bug with incorrect setting of high-order LBA bits in DriveSelect register for LBA48 commands (should be 0 since 24-31 bits are stored in BlockNumberExp) *) used UniataAhciWaitCommandReady() in ATAPI 'Special case' completion to wait for CI release. *) changed AHCI ready wait granularity to 200 us (was 1 ms) in UniataAhciWaitCommandReady() -- 2012.08.01 - v0.42e3 *) added ACMD dump in UniataAhciBeginTransaction() for ATAPI debug *) removed (for DEBUG) AHCI prefetch on ATAPI +) added SACT AHCI register update before CI *) removed unnecessary CMD register update in UniataAhciBeginTransaction(). ACT+CI is enough i) analyzed SeaBIOS code v1.7 (http://code.coreboot.org/p/seabios/source/tree/bf079e16f8b0d016eb09787c2aebda359c41391a/src/ahci.c) *) added ATA_AHCI_P_IX_INF flag to list of allowed interrupts *) removed unnecessary RtlZeroMemory before UniataAhciSetupFIS_H2D in AHCI ATAPI branch -- 2012.08.01 - v0.42e4 +) (test) DMA-only mode for AHCI ATAPI -- 2012.08.01 - v0.42f *) (test) DMA-only mode for AHCI ATAPI SCSIOP_REQUEST_SENSE !) added handling for error-related IS bits. We should restart AHCI engine on all error conditions. otherwise device comes into indetermined state. *) adapter max transfer rate is now displayed in atactl -a output -- 2012.08.02 - v0.42f1 *) undone experimental changes with force DMA on AHCI ATAPI (will check later) *) AHCI abort sequence is moved to UniataAhciAbortOperation() +) implemented error handling in AHCI StopXxx() calls *) added use of AHCI ACT register to WaitReady and internal SendCommand -- 2012.08.02 - v0.42f2 +) added support for v12 and v16 READ/WRITE/VERIFY SCSI command emulation. It extends addressable space to 64bits (v16) and gives supports for HDDs above 2Tb, they use more than 32bits for LBA. *) undone all operations with AHCI ACT register -- 2012.08.02 - v0.42g !) fixed bug with incorrect ATA_AHCI_CMD_WRITE setting in AHCI ATAPI branch. +) added DMADIR bit support to ATAPI -- 2012.08.02 - v0.42g1 *) (test again) DMA-only mode for AHCI and SATA ATAPI !) fixed command version specific LBA decoding in IdeReadWrite -- 2012.08.03 - v0.42g2 *) fixes in LBA logging -- 2012.08.03 - v0.42g3 *) reduced amount of logs in AtapiCheckRegValue() !) fixed bug with 0 number of luns (in log only) in AtapiSetupLunPtrs() *) added logs to find why on some ICHxxx debug version with attached debugger crashes on controller reset -- 2012.08.04 - v0.42g4 *) added log to catch damaged Chan pointer in DeviceExtension (observed on Reset in Native/Legacy mode on ICHxxx) !) added handler for 'unexpected' ATAPI READY interrupt. *) added small delay after ATAPI PIO read, finished with BUSY status when all requested bytes are transferred. If BUSY is released, command is assumed to be completed, otherwise further processing is posted to DPC. +) added SCSIOP_SERVICE_ACTION16/SCSIOP_SA_READ_CAPACITY16 support -- 2012.08.04 - v0.42g5 *) added logs to catch damaged (zero-filled) CDB on AHCI ATAPI READ *) removed unnecessary FIS logs from UniataAhciEndTransaction() !) fixed bug with unconditional ERROR state on AHCI ATAPI with zero transfer. -- 2012.08.05 - v0.42g6 *) SrbExtensionSize = sizeof(ATA_REQ) regardless of controller type (test) -- 2012.08.05 - v0.42g7 *) ATA_AHCI_CMD_PREFETCH is used again (test) -- 2012.08.05 - v0.42h !) fixed bug with missed call to MapError for AHCI ATAPI !) fixed bug with reading error status from improper place in AHCI ATAPI branch -- 2012.08.06 - v0.42h1 *) several fixes for making compilers happy -- 2012.08.07 - v0.42h2 *) several fixes for making compilers happy !) clarified use of SRB_STATUS_DATA_OVERRUN in AHCI branch -- 2012.08.07 - v0.42i !) fixes for SATA TransferRate report/detect *) added auto-update of controller MaxTransferMode according to actual established SATA transfer rate. This is done for incorrectly claimed controllers, +) added AHCI controller max. transfer rate detect. -- 2012.08.07 - v0.42i1 *) several fixes for making compilers happy -- 2012.08.07 - v0.42i2 !) fix for uninitialized DMA PRD in SATA ATAPI branch *) added more logs to find why on some ICHxxx debug version with attached debugger crashes on controller reset (chan pointer gets damaged) -- 2012.08.08 - v0.42i3 !) fixes for SATA speed reporting and representation. Doesn't affect actual interface speed. !) added minimum buffer size check in UNIATA IOCTL. TODO: perform such check in each IOCTL branch. -- 2012.08.09 - v0.42i4 *) minor update in AtapiSoftReset timings, need to fix problem with possible late BUSY after RESET. -- 2012.08.09 - v0.42j *) artificially broken to cause BusReset and find where chan pointer gets damaged -- 2012.08.09 - v0.42i5 !) rewrittem code for SRB_STATUS_DATA_OVERRUN reporting. Now check is performed in single place on CompleteRequest. *) transfer mode reporting is updated -- 2012.08.09 - v0.42i6 +) added PhyMode to LunExt and GETMODE. This value reflects actual bus transfer rate. Is used by atactl. *) rewritten OrigTransferMode behavior. Now it exactly reports maximum supported by device transfer rate. *) Ident -> MaxTransferMode code is moved to atapi.h and commonly used in both uniata and atactl *) transfer rate definition constants are moved to atapi.h -- 2012.08.09 - v0.42i7 *) added zero-filling of output buffer in atactl before calling IOCTL -- 2012.08.11 - v0.43 +) added buffer size check to all UniATA IOCTLs *) added channel info option to AdapterInfo IOCTL !) fixed bug with Physical/Logical channel numbering for Legacy controllers in AtapiChipInit() *) clarified HwRes check code (80-pin detection) +) channel transfer mode reporting to atactl -lx for controllers those have physically different channels -- 2012.08.11 - v0.43a !) fixed bug with buffer overflow in AdapterInfo IOCTL !) removed via_cable80(), since it doesn't actually work !) fixed bug with wrong channel info offset in AdapterInfo IOCTL -- 2012.08.12 - v0.43a1 !) fixed bugs with AdapterInfo IOCTL handling +) experimental ATA_PASS_THROUGH handling for AHCI -- 2012.08.14 - v0.43b +) added AHCI enable code for chipsets, those have GHC_AE disabled by default +) added MSI reset workaround code. *) MaxTransferMode is now checked against channel-specific value instead of controller-specific one *) merged ICH7 related updates from FreerBSD, e.g. operations with 0x94 DWORD register and SATA BAR+0x0C DWORD register *) removed unnecessary I1CH flag, indicating single channel controllers -- 2012.08.16 - v0.43c +) implemented SCSIOP_START_STOP_UNIT Immediate, StartStop flag and PowerCondition handling. ATA IDLE, SLEEP (power down) and STANDBY (stop spindel) commands are used. Device is returned from Sleep stete via HardReset PowerState variable is added to LunExt to track expected power state of the device. *) merged atactl compiler warning fixes from ReactOS (bug 7243) *) changed intel 40/80-pin cable check -- 2012.08.16 - v0.43c1 +) added AdvancedPowerMode registry option. 0 - disable 1 - min perf with standby 0x80 - min perf. w/o standby 0xfe - max performance (default) +) added AcousticMgmt registry option 0 - disable 0x80 - min acoustic level 0xfe - max performance (default) +) added IDLE/STANDBY timer init Note: some devices have non-zero default timer value (used when timer is disabled). Value of 0xfd is recommended in such case (max. possible, between 8 and 12 hours, vendor spec.) -- 2012.08.17 - v0.43c2 !) fixed bug with possible crash when chan pointer becomes NULL in AtapiStartIo. Happened when smartctl (from smartmontools) send IDENTIFY to non-existent or communication channel. !) fixed bug with improper device type check in ATA_PASS_THROUGH branch (checked for SATA instead of AHCI). This caused crash with smartctl !) fixed bug with 40/80-pin cable check on drives those do not report valid HwRes *) updates (clarification) of DriveNumber meaning in SMART/IDENTIFY IOCTLs -- 2012.08.18 - v0.43c3 !) fixed bug with confused Active and Supported modes from IDENTIFY. Bug caused drive to be limited to BIOS-selected transfer rate instead of maxumum available. *) implemented adaptive algorithm of handling DriveNumber in SMART IOCTLs to make smartctl happy. When invalid PathId/TargetId comes, if adapter has only 1 channel is is treated as primary or secondary IDE. Otherwise valid PathId/TargetId required. Note: Due to design bug in DISK.SYS (it doesn't send PathId down for IOCTLs) management tools should open ScsiPort%d instead of PhysicalDrive%d when possible and send IOCTL to proper SCSI address. -- 2012.08.18 - v0.43c4 *) implemented IDENTIFY checksum adjustment to make smartctl happy +) added power management options to atactl o) completly rewritten and unified SCSI address extraction code in SMART IOCTLs -- 2012.08.18 - v0.43c5 !) fixed bug with reference to uninitialized chan pointer in SCSI address extraction code in SMART IOCTLs +) added new SMART IOCTLs to make smartctl happy (used for SCT): IOCTL_SCSI_MINIPORT_ENABLE_DISABLE_AUTO_OFFLINE IOCTL_SCSI_MINIPORT_READ_SMART_LOG IOCTL_SCSI_MINIPORT_WRITE_SMART_LOG +) buffer size check to SMART IOCTLs -- 2012.08.19 - v0.43c6 *) AdvancedPowerMode and AcousticMgmt are defaulted to max power saving without standby !) added validation code for Intel IDE I/O Config Register 0x54. Older controllers do not use it to reflect cable type -- 2012.08.19 - v0.43d !) changed Intel DMA setup code in order to handle standard modes correctly (according to specs). +) added special case for pseudo-UDMA3 setup on Intel PIIX4e -- 2012.08.20 - v0.43d1 !) fixed bug in AHCI ATA registers dump (invalid bit shift for CylHigh) o) unified ATA registers dump code (used in SMART and ATA PASSTHROUGH). +) added buffer size check to ATA PASSTHROUGH dump registers branch +) added check for CDB16 support (via IDENTIFY) and for requested CdbLength o) extra (unexpected) data wait in AtapiSendCommand in replaced with call to AtapiSuckDataPort2() +) added MODE_SENSE MODE_PAGE_POWER_CONDITION -- 2012.08.23 - v0.43d2 !) fixed bug with uninitialized chan->MaxTransferMode for VIA VT6421 PATA channel (was set to SA150 according to controller default) *) added logging of reported interrupt info in UniataFindBusMasterController() *) allowed 64-bit physical addresses for controllers with AHCI 64-bit support (experimental) -- 2012.08.25 - v0.43d3 !) fixed bug with device addressing for SMART/IDENTIFY on AHCI -- 2012.08.25 - v0.43d4 !) fixed uninitialized MemIo in UniataAhciDetect() *) pre-initialized 'status' in IdeSendCommand() to make compiler happy, other similar minor changes e) AltInit is not issued if HwInitialize() is called in context of ScsiPortInitialize() and we are sure about detect status under w2k+ -- 2012.08.26 - v0.43e *) InterruptMode is set explicitly to LevelSensetive for PCI controllers (since LevelSensetive == 0, was used by default) e) BM range for MasterDev (Legacy) controllres in not reported to ScsiPort via ConfigInfo *) adjusted setting of ConfigInfo options according to actual needs -- 2012.08.26 - v0.43e1 e) BM range for MasterDev (Legacy) controllres in not reported to ScsiPort via ConfigInfo (fixed, didn't work) *) logged PCI devices using 14 and 15 IRQ (conflicting with ISA ATA) -- 2012.08.27 - v0.43e2 *) fixed PIIX3/PIIX4 device names *) in XP .INFs %12% is replaced with %10%\System32\drivers, this fixes problem with 'Invalid INF section' +) added uata_comm.inf for UniATA virtual communication port -- 2012.08.28 - v0.43e3 e) Legacy IDE IO ports are now queried from Isa bus. PCIBus was used before. *) atactl can now display device info even if ATA/ATAPI identify failed, SCSI INQUIRY is used instead. -- 2012.08.29 - v0.43e4 +) added SCSIOP_REPORT_LUNS support for newer OSes e) uncommented code for fake controller to PCI part of Legacy IDE This is done to prevent system and/or other drivers from deinitializing PCI part of Legacy IDE (hangs) -- 2012.08.29 - v0.43e5 e) fake BM controller is inited with 1 bus and 1 lun (was all zeros) *) PCI IO enable code is moved to separate function since it is used in multiple places -- 2012.08.29 - v0.43e6 !) fix in device identification for fake PCI BM -- 2012.08.29 - v0.43e7 e) temporary disabled call to UniataEnableIoPCI() on Fake claim (fails) -- 2012.08.29 - v0.43e8 e) disabled Fake claim again -- 2012.08.29 - v0.43e9 e) re-enabled Fake claim e) used NULL HwInterrupt (fails) -- 2012.08.29 - v0.43f *) removed Fake claim +) added UniataEnableIoPCI() to AtapiChipInit() in global branch for w2k+ -- 2012.08.30 - v0.43f1 e) added UniataClaimLegacyPCIIDE() to claim Legacy PCI IDE with HalAssignSlotResources (fails) *) SCSIOP_REPORT_LUNS is emulated for HDD only -- 2012.08.30 - v0.43f2 !) added UniataClaimLegacyPCIIDE() is called BEFORE attempts to init Isa+PCI it works !) BM IO ranges are no longer reported with ISA parts. TODO: check if Fake claim works if called before ISA. -- 2012.08.30 - v0.43f3 e) check if Fake claim works if called before ISA. (doesn't work, Scsiport returns STATUS_DEVICE_DOES_NOT_EXIST) -- 2012.08.31 - v0.43f4 !!) reverted to UniataClaimLegacyPCIIDE(), this fixes resource conflict with other PCIIDE drivers -- 2012.08.31 - v0.43f5 !) fixed uata_comm.inf, now it works *) if PCI part cannot be acquired, controller is initialized in legacy ISA way without DMA -- 2012.09.18 - v0.44 !) fixed LogToDisplay by KtP *) changed VendorString definition to make newer compilers and cove checkers happy *) made an attempt to change .mak file and fix build problems after clean or build errors -- 2012.09.20 - v0.44a !) UniataEnableIoPci doesn't touch PCI Command reg. if it is not necessary (no update required), because on some systems (e.g. some ICH7) may produce interrupt storm. -- 2012.09.22 - v0.44b *) fixed bug with access to non-existent Lun in interrupt cleaning on init in PROBE_ON_INIT branch +) added interrupt cleaning code for !PROBE_ON_INIT branch +) added ICH7-specific SATA register access -- 2012.09.24 - v0.44b1 !) fixed .mak file (.res rebuild) *) experimental fix for SATA ATAPI, those do not interrupt after DMA packet command error (check for error immediately after sending packet) -- 2012.09.24 - v0.44b2 *) experimental fix for SATA ATAPI, check for DRQ (appeared, it is not asserted on error) -- 2012.09.25 - v0.44b3 *) removed experimental DMA-only mode for SATA ATAPI. Some devices hangs. -- 2012.10.06 - v0.44b4 *) fixed DMA->PIO condition in AtapiDmaInit() for non-AHCI SATA devices (was never switched) -- 2012.12.26 - v0.44c *) added more logs in AtapiSendCommand() to study error conditions after previous cmd execution failure. Now next command also treated as failed under VPC since ERR bit and Error Code is not cleared after PACKET command *) added call to BuildRequestSenseSrb() after immediately failed ATAPI command, I hope this will clear error status properly. -- 2012.12.28 - v0.44c1 !) fixed IDE interrupt reason processing bug. Appeared due to mistake during replacement of numeric values with constants. -- 2012.12.29 - v0.44c2 !) added INF patches by Gerhard Wiesinger, should fix problems with paths to binaries during setup. -- 2012.12.30 - v0.44c3 *) implemented smart BUSY up/down wait algorithm in AtapiSoftReset(). It should avoid unnecessary 1 second delay. -- 2013.03.20 - v0.44c4 !) fixed bug with incorrectly (-1) initialized deviceExtension->MaxTransferMode in UniataChipDetect() for SIS 961 !) fix candidate for unhandled Write PIO completion. Counters were not updated, this caused 'incomplete transfer' condition and unnecessary endless retry attempts. -- 2013.03.20 - v0.44d *) warning fixes -- 2013.04.05 - v0.44e *) warning fixes in atactl, PHAR -> PCCH +) added experimental BIOS handoff (BOH) support -- 2014.03.02 - v0.45 *) added DevIDs for newer ATI IPX700/800, Hudson-2 controllers, default them to SATA (non-AHCI) mode +) added (experimental) PATA/SATA and AHCI detection code for newer ATI IPX700/800, +) added nVidia AHCI controllers, default them to SATA (non-AHCI) mode *) added more PATA Marvell chips -- 2014.03.08 - v0.45a *) added "SIS 0183" controller !) fixed handler for SIS 018x SATA controllers without BAR4 -- 2014.03.08 - v0.45a1 *) added "SIS 1182" and "SIS 1183" controllers -- 2014.03.09 - v0.45a2 !) removed duplicates and wrong values from device list *) unified OS-specific INFs e) AHCI support is enabled in INFs by default *) SIS ATA controllers are listed in bm_list.h in order to be listed in INF *) listed VIA AHCI as supported !) fixed DevID for RZ 100x !) added handler for uninitialized BAR4 for all SATA chips +) added new Promise chips +) added channel number adjustment for Promise PATA e) partially added Promise SATA support *) added handler for 0 dma_count e) improved AHCI error handler. Do not retry after interface errors -- 2014.03.11 - v0.45a3 !) fixed bug with not updated uniata_ver.h when VER is specified *) added more logs in and around AtapiVirtToPhysAddr_() -- 2014.05.21 - v0.45a4 !) fixed bug with PCI device revision for UDMA-5 "SiS 630S" (improper detection) -- 2014.05.21 - v0.45a5 *) added "SiS 962" and "SiS 963" to list of known devices, programming specific flags are properly set for them. *) added more logs for flags and word-count in ISR -- 2014.08.05 - v0.45a6 e) ATAPI commands READ_CD and READ_CD_MFS are marked as DMA-capable. Managed by registry option AtapiDmaRawRead. -- 2014.08.11 - v0.45a7 *) Added workaround for "Marvell 9123" detection of AHCI channels. Chip reports 8 while it actually has 4. *) added more logs for SSD and AHCI channel detection. -- 2014.08.13 - v0.45a8 !) do not touch unimplemented AHCI channels in AtapiResetController() *) added logging to UniataAhciChanImplemented() -- 2014.09.06 - v0.45b !) fixed bug with missed AtaReq->WordsTransfered adjustment in IDE PIO Write branch in IdeReadWrite() !e) added handler for raised DRQ immediately after PIO read, seems we should try to read rest of data. see ReactOS BUG-8280 -- 2014.09.16 - v0.45b1 !e) changed Promise UDMA2+ programming sequence according to Linux sources. Should fix problem with high transfer rates. see ReactOS BUG-8280 -- 2014.10.17 - v0.45c !) Added check for valid BM status register on Intel chips in VBOX environment those report 'compatible' mode via 0x90 PCI reg. This is needed to workaround problem with AHCI-enabled controller without properly set 0x90. see ReactOS BUG-7020 !) Added LUN 1->2 reallocation in case of fall-back to 'compatible' mode on Intel chips according to 0x90 PCI reg. see ReactOS BUG-7020 -- 2014.10.20 - v0.45c1 !) fixed bug with ATA-style code without AHCI branch in AtapiEnableInterrupts() when we are in recursive disable state. see ReactOS BUG-7020 -- 2014.10.23 - v0.45d e) added experimental registry option Force80Pin to workaround controllers with broken 80-pin detection -- 2014.10.29 - v0.45d1 !) Fixed bug in option Force80Pin handling for Intel chips !) Fixed bug in DMA timing for ICH4 chips. Due to wrong programming we got assymetric transfer r/w rate in UDMA 5 - 80/16 Mb/s -- 2014.10.30 - v0.45e e) Added handling of "Exclude" option for separate channels for Compatible controllers e) Added handling of "Exclude" option for AHCI channels e) Partially added handling of "Exclude" option for ATA/SATA controllers (affects only total number of channels according to resulting port bitmap). e) Added "PortMask" option for AHCI channels and partially for SATA (like "Exclude"). This option can reduce available cannels. e) Added "NumberChannels" option for SATA, this option can only reduce available cannels. i) works on nVidia SATA MCP51 PATA/SATA -- 2015.02.08 - v0.45f, v0.45f1 *) Added logging of IR status for ATAPI -- 2015.02.08 - v0.45f2 *) Added logging of WordCount status and interrupt processing points for ATAPI e) Added AtapiSoftReset() on ATAPI errors if device reports wrong/unexpected state -- 2015.02.08 - v0.45f3 e) Added UniataSataClearErr() on ATAPI errors and controller reset e) Added extra wait for ready delay for SATA ATAPI reporting wrong device state after reset !) perform generic device reset for nVidia SATA controllers if device reports valid state. Previously, reset was never performed for SATA nVidia. -- 2015.02.09 - v0.45f4 e) Added UniataSataPhyEnable() with reset option for NVIDIA SATA controller reset !) fixed logical bug in nVidia device reset branch (still was never called) -- 2015.02.10 - v0.45f5 e) Added experimental Phy interrupt masking code for NVIDIA SATA during controller reset -- 2015.02.17 - v0.45f6 !) another experimental Phy interrupt masking code for NVIDIA SATA e) added experimental workaround for the case when the controller is fast enough to assert IRQ between sending PACKET command and GetBaseStatus(). We catch ATAPI interrupt with specific status (DRQ+IDLE, 0 WordCount, Write interrupt reason when Read is expected) and just continue execution. In general, we should mask interrupts around PACKET command and sending the packet itself, but on virtal machines it definitly cause loss of Data interrupt. Also, there is no problem on old uni- and dual-core systems. I don't know if this problem is related to number of CPU cores or to overal system performance. Interrupt masking behavior is managed by AtapiSendDisableIntr registry option and is defaulted to FALSE ROS BUG-9119 -- 2015.02.18 - v0.45f7 e) removed ATAPI state workaround from v0.45f6 (doesn't help, no interrupt appeared) e) try to force read data from ATAPI device on 'strange' condition -- 2015.02.18 - v0.45f8 e) try to wait for WordCount in DPC (ROS BUG-9119) -- 2015.02.18 - v0.45f9 e) force SATA ATAPI to send Packet in ISR (ROS BUG-9119) -- 2015.02.18 - v0.45f10 e) roll-back to v0.45f7 e) do not use AtaSetTransferMode() for SATA ATAPI devices according to Linux ata drivers (ROS BUG-9119) e) try blind reading data for ATAPI if no WordCount reported -- 2015.02.19 - v0.45f11 e) revert usage of AtaSetTransferMode() for SATA ATAPI devices. Problem with zero wordCount persists, but we get PIO timing problem. !) force read data from ATAPI device on 'strange' (zero wordCount, but raised DRQ) condition seems to work properly ROS BUG-9119 -- 2015.02.19 - v0.45g !) added check for ATAPI if IDE identify fails (some devices doesn't report ATAPI signature in wrong internal states) ROS BUG-9119 -- 2015.02.22 - v0.45g1 !) added check for Legacy ISA I/O ranges on non-MasterDev PCI controllers to avoid further conflict with ISA detection code (such condition was detected on QEmu) e) added check for valid value of DriveSelect register on SATA controllers w/o SATA regs. QEmu ignores DriveSelect and reports clone of Master on Slave. !) added delay between read cycles in AtapiSuckPort2() and AtapiSuckPortBuffer2() to avoid condition when DRQ is not raised again immediately after read. -- 2015.02.22 - v0.45g2 !) AtapiSuckPort2() is not called automatically if AtapiSuckDataPort2() get less data than expected. Rest of data may come later. ROS BUG-9119 -- 2015.02.22 - v0.45g3 !) added workaround for lost value of ConfigInfo->AtdiskSecondaryClaimed under ReactOS and not detected resource conflict in ScsiPortGetDeviceBase() QEmu expose disk controller as both PCI SATA and ISA IDE. We should not try ISA detect if PCI succeed with same I/O range !) fix for AtapiSuckPort2() in AtapiSuckPortBuffer(). Rest of data may come later. -- 2015.02.23 - v0.45g4 e) added setting of IDX_AHCI_P_CMD in UniataAhciBeginTransaction() along with CI, probably we need this for VBox only +) added read from AHCI registers after write to initiate flush (from Linux) -- 2015.02.23 - v0.45g5 *) undone setting of IDX_AHCI_P_CMD in UniataAhciBeginTransaction() along with CI. It doesn't help, interrupt is lost due to some other reason. e) used long wait for Drq in AtapiSuckPortBuffer(), short is not enough. We can't defer reading because no more interrupts will be generated and because BUSY is not set while some data is still expected. -- 2015.02.24 - v0.45h +) added Cyrix 5510/5520 support from Linux code -- 2015.02.24 - v0.45h1 +) added Cyrix 5530 and 5535 support from Linux code +) added handler for BusReset condition after SCSIOP_MECHANISM_STATUS Now driver sets DFLAGS_CHANGER_INITED in this case to prevent automatic use of unsupported command. -- 2015.04.06 - v0.45h2 !) setting of IDX_AHCI_P_CMD in UniataAhciBeginTransaction() along with CI is removed again. According to specs both these actions will start command. According to VBox logs command is executed twice. Interrupt loss is really caused by some other reason, but this but should be fixed anyway. -- 2015.04.06 - v0.45h3 !) added AtapiWaitForDrq() in ATAPI interrupt handler. Looks like it may be raised with delay (e.g. in QEmu) -- 2015.04.06 - v0.45h4 !) fixed bug with missing REQ_FLAG_DMA_OPERATION on succefull reinit in AtapiDmaSetup(). This prevented fallback to UDMA2 when higher modes doesn't work fine. -- 2015.04.07 - v0.45h5 *) minor log updates -- 2015.09.22 - v0.45h6 +) added BOCHS detection o) implemented quick device absence detection for BOCHS (0x00 status) !) unified IORES structure init via UniataInitIoRes() and UniataInitIoResEx() this affects problem with ReactOS boot under BOCHS (BUG 10176) e!) IORES->Proc check in IO routines replaced with ASSERT() also affets ReactOS BUG 10176 !) fixed bug with missing reinit dma after device reset and initial detection. !) added IO size alignment check before 32-bit PIO operations !) fixed bug with Intel timings (secondary channel) -- 2015.09.23 - v0.45h7 e) added optimization to avoid unnecessary IO with unsupported/missing PCI devices on 2nd and 3rd scan passes. -- 2015.09.23 - v0.45h8 +) sync with ReactOS repository eo) handle PRD and IO-buff double-buffering before sending CMD to device. Copying is moved from AtapiDmaStar() to AtapiDmaDBPreSync() e) #pragma pack(1) is used for BusMaster structures in 'bsmaster.h' -- 2015.09.23 - v0.45h9 !) #pragma pack() is used in all uniata headers to avoid cross-environemnt ambiguities (this was the root of the problem) -- 2015.09.23 - v0.45h10 *) merged IORES-related logging from parallel branch (39k forked from 39j) e!) fixed PCI bus scanning algorithm (it could stop after 1st pass even if some devices are detected) -- 2015.09.24 - v0.45i o) implemented quick device absence detection for all types of VM (0x00 status) o) stop detection process in IssueIdentify if after IDENTIFY command still reports 0x00 status. -- 2015.09.25 - v0.45i1 *) added more logs to IssueIdentify !) removed unhandled ATA_DMA attempt from Promice init code. e) temporary (for test only) removed HDD cache/acoustic/etc tuning (ROS BUG-8280) +) HDD tuning checks Capability flags -- 2015.09.27 - v0.45i2 *) added more logs to IssueIdentify (Security status) e) reverted HDD cache/acoustic/etc tuning (see 0.45i1) e) error/retry context is preserved after bus-reset and new attempts should be performed with lower transfer rate like after regular error condition. This affects IDE Read/Write operations only. -- 2015.09.29 - v0.45i3 *) added more logs to IssueIdentify capabilities (IORDY, SoftReset) e) used Linux code for Promise DMA init instead of FreeBSD !) fixed bug with unnecessary ExcludeFromSelect option in NT4 .INF -- 2015.10.06 - v0.45j !) Fall back to BM mode for Intel AHCI-capable chips w/o AccessRanges[5] -- 2015.10.06 - v0.45j1 !) Fall back to BM mode for all AHCI-capable chips w/o AccessRanges[5] -- 2016.01.18 - v0.45j2 e) detect wrong Bochs ATA DMA behavior. Versions prior to 2.6.8 expect ATA_DMA_READ *after* DMA start, however BM DMA specs declare opposite order. CTRFLAGS_DMA_BEFORE_R flag is introduced in order to manage this. (ROS BUG-6774) -- 2016.01.24 - v0.45j3 e) more logs for detect wrong Bochs ATA DMA behavior. (ROS BUG-6774) -- 2016.01.24 - v0.45j4 e) use CTRFLAGS_DMA_OPERATION instead of CTRFLAGS_DMA_ACTIVE in workaround condition (ROS BUG-6774) -- 2016.01.24 - v0.45j5 !) fixed bug with missing AtapiDmaStart() on HDD Read in IdeReadWrite() i) hangs on nVidia MCP51 (not sure that since this version) -- 2016.03.05 - v0.45j6 e) try use CTRFLAGS_DMA_BEFORE_R for ATAPI too !) move CTRFLAGS_DMA_BEFORE_R to global g_opt_BochDmaReadWorkaround e) do not use DMA on ATAPI under Bochs (ROS BUG-6774) !) fixed bug with completly cleared ChannelCtrlFlags on bus reset. Permanent bits are masked with CTRFLAGS_PERMANENT -- 2016.03.06 - v0.45j7 !) use g_opt_AtapiNoDma = TRUE for Bochs by default (ROS BUG-6774) e) try use g_opt_BochDmaReadWorkaround for both Read and Write operations (ROS BUG-6774) -- 2016.03.07 - v0.45j8 e) do not use g_opt_BochDmaReadWorkaround for Write operations since DMA Start before ATA WRITE cause Boch DMA command abort immediately after DMA Start w/o any error reports after ATA WRITE. !) Fixed bug in DMA Start / ATA WRITE sequence (missing DMA Start) for Bochs in workaround mode See ROS BUG-6774 -- 2016.03.07 - v0.45j9 eo) try use DWORD-IO for ATAPI -- 2016.03.07 - v0.45j10 *) added extra Bochs detection via ATA device S/N and FW (it is possible to meet no Boch VendorId on PCI bus) and absence of non-zero SubDevice/SubVendor on Intel chipsets +) Added old PIIX DMA support -- 2016.03.08 - v0.46 !) Fixed bug with PIO multiblock transfer (wordCount was not updated on each iteration), which may cause buffer overflow !) Fixed bug with uninitialized max_bcount on PIO transfers on LBA-enabled devices, which could cause unnecessary request fragmentation in PIO mode. e) do not try automatically initialize CD-changers, since it may cause ATAPI CD-ROM freeze. Let CD-ROM driver do it itself for its own risk (and knowledge). We shall jusy parse output on success. Original changer behavior can be reverted by UNIATA_INIT_CHANGERS define. (ROS BUG-9119) -) removed unnecessary calls of AtapiHwInitializeChanger() +) track last used Lun on the channel chan->last_cdev -- 2016.03.09 - v0.46a *) fixed KeNumberProcessors ROS/NT DDK incompatibility issue *) fixed g_opt_BochDmaReadWorkaround variable name (use 'Bochs') *) fixed DBG vs _DEBUG ROS compilation issue !) check for DRQ is taken back. However, it is not always asserted on error (v0.44b2), problem reported by RayeR -- 2016.03.09 - v0.46a1 e) force PIO on SATA ATAPI (experiment for Pioneer, ROS BUG-9119) -- 2016.03.12 - v0.46a2 e!) check ATAPI status (iReason) for CoD bit set before sending packet command and wait for it (experiment for Pioneer, ROS BUG-9119) *) added logging of DMADirRequired bit from ATAPI Identify -- 2016.03.12 - v0.46a3 !) IDE_STATUS_INDEX is excluded from Status vs AltStatus verification on ISA, extra log added -- 2016.03.13 - v0.46a4 !) IDE_STATUS_INDEX is excluded from StatusByte DRQ/Ready condition checks o) rewritten DRQ+CoD wait for ATAPI after ATA_PACKET command (ROS BUG-9119) -- 2016.03.14 - v0.46a5 e) change ATAPI regs setting order e) add DriveSelect + IDE_USE_LBA for ATAPI e) set ByteCount regs to 0 for ATAPI DMA -- 2016.03.14 - v0.46a6 e) allow DMA for short/unaligned data blocks for SATA ATAPI (ROS BUG-9119) -- 2016.03.14 - v0.46a7 *) reverted to normal values of ByteCount regs on ATAPI DMA (ROS BUG-9119) *) updated IDENTIFY structure according to ATA8-d2015r2-ATAATAPI o) unified ATAPI regs setting code *) added Error reg logging if DRQ+CoD never comes (ROS BUG-9119) -- 2016.03.14 - v0.46a8 e) added 16-byte alignment check for ATAPI DMA transfers e) added zero-filling of 'unused' ATAPI regs e) DRQ+CoD timeout is set to 1 sec e) defer ATAPI command packet sending to ISR for SATA ATAPI e) log port claiming operations on ISA init (we have problem with floppy.sys on ISA, reported by KtP) +) added IoRange cleanup on ISA init -- 2016.03.17 - v0.46b -) removed experimental changes for ROS BUG-9119: long DRQ+CoD delay, omited PIO/DMA setup for SATA ATAPI, forced DMA mode i) all actual improvements and bugfixes from v0.46ax are preserved -) removed extra AtaSetTransferMode() call in AMD/NVidia/VIA PIO setup !) VIA-specific PIO setup is now conditional in AMD/NVidia/VIA code !) fixed IO-range cleanup on unsuccessfull ISA detection attempt !) fixed out-or-range cleanup attempt on ISA detect (buffer overflow with possible crash later) -- 2016.03.19 - v0.46b1 +) wait 1us for BUSY raise after write command registers on "old" ISA systems. Original ATAPI.SYS does this. +) use long timeout 2s on hard reset after unsuccessfull soft reset on detect phase on "old" ISA systems. Original ATAPI.SYS does this. *) do not enumerate PCI bus if both IgnoreIsaCompatiblePci and IgnoreNativePci options are set. *) fixed several 'prefast' warnings -- 2016.03.19 - v0.46b2 !!) use ATA_IOSIZE and ATA_ALTIOSIZE for I/O range length in ISA detect part. Note, that historically used value of 2 for AltStatus range is definitly wrong. There must be 1. The ambiguity comes from rudimentary definition of IDE_REGISTERS_2 with unused DriveAddress member e) do not report 0x3f6 register usage on old ISA hardware to avoid interference with Floppy. However, it should use only 0x3f0-3f5,3f7 +) track last selected device via chan->last_devsel to avoid unnecessary I/O and possible wait for ready delay. Move SelectDrive() from macro to function. Create AtapiHardReset() for channel hard reset and related operations -- 2016.03.20 - v0.46b3 -) removed extra delays, appeared in 0.46b1, 0.46b2 +) added tunable delay in SelectDrive() after write to register e) try report 0x3f6 register usage even on old ISA !) fix bug with blocked UniataEnumBusMasterController() call !) declare IDE_REGISTERS_2 as union and use DriveAddress (which is now aliased with AltStatus) as Control register i) UniataClaimLegacyPCIIDE() doesn't work on NT4 SMP (by KtP) CSB5 device has ISA ports exposed via PCI BARs 0-3 -- 2016.03.20 - v0.46b4 !) fixed bug with sending packet command in ISR. This didn't happen on SATA because of error in code unification in v0.46a8 !) clarified AtaSetTransferMode() behavior for SATA devices. When device is switched from PIO to SATA_xxx mode, use highest UDMA mode from Identify data. e) use PIO0_TIMING delay between comamnd packet bytes when deal to ATAPI (ROS BUG-9119) -- 2016.03.20 - v0.46b5 !-) removed unnecessary ExFreePool() from UniataClaimLegacyPCIIDE() because buffer is deallocated inside HalAssignSlotResources() !) do not report 0x3f6 register usage on old ISA, even with correct 1-byte range length e) do not call UniataClaimLegacyPCIIDE() if ISA ranges are exposed via PCI BARs 0-3 -- 2016.03.21 - v0.46b6 !) according to newer FBSD/Linux code SiS chipset programming is not the same as nVidia e) try to disable NVQ option use (code seems to be much more different for NCQ mode) and clear NVQ flag in HwFlags in oder to operate in old-style mode. This option affects interrupt management registers layout. i) according to Linux sources, NVQ mode should be specially enabled. Otherwise use old-style mechanism. -- 2016.03.22 - v0.46b7 e) try to continue init if UniataClaimLegacyPCIIDE() fails on SMP HAL under NT3/4. i) Some NT3/4 SMP HALs cannot reallocate IO resources of BusMaster PCI controller Since nobody claimed PrimarySecondary yet, try init and claim them However it is not 100% safe way. !) fixed bug with DeviceID/VendorID in atactl output *) align DeviceID/VendorID length in atactl output !) fixed bug with ISA-on-PCI detect (used BMList[i] instead of newBMListPtr i) doesn't work (hangs) on MCP51 SATA/PATA -- 2016.03.22 - v0.46b8 -) removed old nVidia updates related to ROS BUG-9119 i) doesn't work (hangs) on MCP51 SATA/PATA -- 2016.03.22 - v0.46c i) roll-back to 45e +) commit general updates and optimizations from 0.46b8 expect nVidia-related, DriveSelect optimization, DWORD_IO, MECHANISM_STATUS, ATAPI iReason, error recovery context, ISA-related changes -- 2016.03.22 - v0.46c1 +) commit DriveSelect optimization +) commit fix for set (U)DMA mode on SATA i) MCP51 SATA/PATA ok (UDMA2 ATAPI + SATA150 HDD) -- 2016.03.23 - v0.46c2 +) commit DWORD_IO +) commit DMADir +) commit new VIA timings +) commit ISA-related changes (w/o PCI detect) -- 2016.03.23 - v0.46c3 +) commit MECHANISM_STATUS fix -- 2016.03.23 - v0.46c4 +) commit ISA-only fixes -- 2016.03.23 - v0.46c5 !) fixed check for conflicting IO resources (ISA vs PCI). Also, the check is unified and performed in AtapiCheckIOInterference() In some cases we could touch PCI-allocated ISA (BusMaster) IO resources on pure ISA detect. *) AtapiFindController() is renamed to AtapiFindIsaController() -- 2016.03.25 - v0.46c6-8 !) fixes in AtapiFindIsaController() scan algorithm -- 2016.03.25 - v0.46c9 +) commit SATA ATAPI-specific changes (ROS BUG-9119) -- 2016.03.25 - v0.46c10 +) commit new nVidia chipset setup i) we get back to 0.46b8 with Isa detect fixes and without very specific ATAPI continue-PIO-read code i) MCP51 SATA/PATA ok (UDMA2 ATAPI + SATA150 HDD), Bochs ok -- 2016.03.25 - v0.46d e) prefer AltInit if UniataClaimLegacyPCIIDE() fails for CSB5 (see above, 0.46b3) and similar cases *) fixed dev/vendor logs -- 2016.03.25 - v0.46d1 !) commited temporary commented resourceList allocation for HalAssignSlotResources(). e.g. ICH4 fails PCI init on UP system. i) in general, resource use claiming should be completly rewritten. -- 2016.03.26 - v0.46d2 !) always attempt to init Compatible controller if HalAssignSlotResources() fails and Primary/Secondary not claimed yet. This fixes problem with HALs other than default UP (hal.dll) -- 2016.03.27 - v0.46d3 *) Use compact data structure BUSMASTER_CONTROLLER_INFORMATION_BASE for known device list (just dev/vendor ID + flags). In user-mode and debug the structure is extnded with device name pointer. Original BUSMASTER_CONTROLLER_INFORMATION is now used for list of detected controllers only. This should reduce binary size and memory waste. +) added build_atacmd.exe utility for prebuilding AtaCommandFlags[] and AtaCommands48[] arrays. Now these tables are located in atacmd_map.h -) UniataInitAtaCommands() is removed -- 2016.03.29 - v0.46d4 !) fixed bug with improper mode value in setting (U)DMA mode on SATA e-) try do not touch chipset/SATA registers on nVidia MCP51 (10de:0266) ROS BUG-9119 e) try enable/disable interrupts in AtaSetTransferMode() for SMP systems. e) set g_opt_AtapiSendDisableIntr to TRUE on SMP systems. -- 2016.03.31 - v0.46d5 +) if we do not touch chipset/SATA registers on nVidia MCP51 (10de:0266) my notebook freezes on init (both UP/SMP modes) i) SMP still gets frozen in GUI -- 2016.04.02 - v0.46d6 +) read DWORD-aligned part of data using DwordIo and then unaligned part in regular way. i) When ATAPI 64k PIO read is requested we may have 0xfffe byte count reported for 0x10000 bytes in single interrupt. It is not allowed to read entire block with DwordIo intead of wait for the last word. -- 2016.04.03 - v0.46d7 !) device list BusMasterAdapters[] and command map AtaCommandFlags[], AtaCommands48[] are compiled in special stubs bm_devs.cpp and atacmd_map.cpp in order to avoid data duplicates in final binary. Also, this prevents ReactOS build errors. e+) added Windows 7 inf build -- 2016.04.05 - v0.46d8 !) fixed .inf build process (2k/2k3 order of copying) -) removed w7 .inf build, 2k3 if fixed instestead +) added build dependency for .inf i) in general, .inf and xxx_ver.h build process and dependencies must be rewritten !) fixed bug with not reinitialized portBase and Irq on ISA bus scan. Only first value was used for all probes +) added protection from invalid ISA registry setting. Both portBase and Irq must be spefified for ISA channels. Otherwise registry settings are treated as invalid and hard-coded defaults are used. i) SMP boot problem was caused by BIOS legacy USB keyboard emulation. This option must be disabled for NT4/NT3.51 -- 2016.06.28 - v0.46e i) We MUST register 2nd ISR for UP systems. This is needed for cases when multichannel controller generate interrupt while we are still in its ISR for other channle's interrupt. New interrupt must be detected and queued for further processing. If we do not do this, system will not route this interrupt to main ISR (since it is busy) and we shall get to infinite loop looking for interrupt handler. +) try to identify CrashDump case and tune init process according to this environment. !) fixed 'control' register uninitialized case for I82371FB, ROS-11157 !) fixed bug with buffer overflow when SCSIOP_READ_CAPACITY produce multiple interrupts and we have wordCount truncated to 0, but non-zero actual transfer length (previously read block) inside last ISR call. AtaReq->WordsTransfered must be used instead. See ROS-11296 -- 2016.08.28 - v0.46e1 !) fixed bug with wrong CrashDump mode condition check, which prevented ATAPI detecton on regular init. -- 2017.01.06 - v0.46e2 !) fixed bug with unitialized page table for valid memory addresses. We should touch user buffer in AtapiVirtToPhysAddr_(). See ROS-11894 -- 2017.04.07 - v0.46e3 !) fixed bug with not handled 0-value of CurrentSectorsPerTrack/NumberOfCurrentHeads under PCem VM *) added check for 0-value of SectorsPerTrack/NumberOfHeads -- 2017.12.11 - v0.46e4 !) fixed bug with not "nVidia nForce MCP61 S2" (use generic interrupt check) See ROS-11773 *) write zero to IDX_IO2_o_Control before sending ATAPI command. See ROS-11773 -- 2017.12.17 - v0.46e5 !) fixed bug with double memory allocation for under-4G buffer in AtapiDmaAllocate() in AHCI mode See ROS-13988 -- 2018.03.18 - v0.46e6 e!) treat AHCI non-fatal condition on READ operations without error bits set as buffer underrun (INFS, ATA_AHCI_P_IX_INF) e) check for LBA32 writes on 2TB+ HDDs (cause data corruption), just write to log now !) fix for Intel ICH9 (2921, 2926) compability flags (should be non-AHCI dual channel). See ROS-13310 +) added JMicron dual-port support + SATA/PATA detect e) 80-pin detection is always issued at the end of AtapiChipInit() regardless of controller SATA support flag since we may have mixed SATA/PATA ports on the same controller -- 2018.03.22 - v0.46e7 !) fix in JMicron dual-port support, do not try to update LUNs since they are not initialized yet o) use pre-cached values of 0x40 and 0x80 regs on JMicron before general vs channel-specific mode checks (thanks to Torsten K -- 2018.07.26 - v0.46e7x +) adapted https://github.com/reactos/reactos/pull/673/commits/4da824887556fa374a5b77cb3fdddc28d2e07f23 64-bit patch (use ULONGIO_PTR, ULONG_PTR) -- 2019.01.21 - v0.46e8x e) AHCI v.1.3, v.1.3.1 are experimentally listed as supported (success) -- 2019.03.10 - v0.46e8 +) merged v0.46e7, v0.46e7x and v0.46e8x -- 2019.03.19 - v0.47 !) fix for .mak files under Wine build enfironment if not exist "$(SomeDIR)/$(NULL)" must be changed to .... "$(SomeDIR)" +) added a lot of AMD, Intel and Marvell AHCI chipsets, e) added UNIATA_AHCI_ALT_SIG flag. Some AHCI returns signature in PSIG rather than FIS on soft reset !) call UniataExpectChannelInterrupt() in IdeVerify() +) added ASMedia, ThunderX and Annapurna AHCI chips e) added workaround for AHCI chips those do not setup port mask. So enable ports manually according to port count register !) fixed Win/Wine .mak path/dependence compatibility issues e) added UNIATA_RAID_CONTROLLER flag to RAID controllers. Note, UniATA doesn't attach to RAIDs since SkipRaids option is set to 1 by default +) added interface-type specific registry settings under PATA, SATA and AHCI sub-hives of Parameters. 2019.03.26 - v0.47a !) AtapiSetupLunPtrs() was not called after UniataAllocateLunExt() in Intel SATA/AHCI detect code, CORE-15886 *) CR/LF -> CR in driver sources !) fix for (UNIATA_AHCI | UNIATA_AHCI) instead of (UNIATA_SATA | UNIATA_AHCI) in AtapiRegCheckDevValue() *) remove 0x prefix when printing AHCI version *) pcidump\head\driver\pcidump.h dependency is conditional. Local copy of pcidump.h is used by default