Changeset 589 in xtideuniversalbios for trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeCommand.asm

Timestamp:

May 22, 2016, 12:26:57 PM (8 years ago)

Author:

krille_n_

Message:

Changes:

• BIOS: Fixed a purely cosmetic bug from r542 where, in builds containing MODULE_EBIOS, the boot menu would display an incorrect drive size (0.4 kB with MODULE_STRINGS_COMPRESSED or 0.5 kB without) for old drives with no support for LBA.
• Fixed a bug from r392 where Vision_DetectAndReturnIDinAXandPortInDXifControllerPresent would return the ID in AL instead of AH (if DANGEROUS_DETECTION had been defined).
• Fixed a bug from r587 in AdvAtaInit.asm that would prevent detection of QDI Vision controllers.
• Also changed how the QDI Vision IDs are defined (removed the need for shifting) to avoid confusion. This fixed a potential bug from r587 in AdvAtaInit.asm where some IDs were not being shifted.
• Fixed a bug in PDC20x30.asm from r587 where GetPdcIDtoAX would not return with the IDE base port in DX so DisablePdcProgrammingMode would fail.
• Made some changes to ModuleDependency.inc and other files so that MODULE_ADVANCED_ATA now requires USE_386. Consequently it is no longer included in the regular AT-builds, only in the 386_8k-build.
• Moved the UNROLL_SECTORS_IN_CX_TO_xWORDS macros from IDE_8bit.inc to IdeIO.inc which means it's now possible to build a BIOS without MODULE_8BIT_IDE.
• XTIDECFG: Added a minimum DOS version check (since it needs DOS version 2+) to allow the program to quit gracefully in the unlikely scenario where someone tries to run it under DOS version 1.
• Made some changes to Drive.asm to improve drive enumeration. The old method using GET_DOS_DRIVE_PARAMETER_BLOCK_FOR_SPECIFIC_DRIVE worked well in Windows XP but not in Windows 98 SE (in Windows or in DOS mode). The two problems were; 1) The function call would access the drives which on single floppy drive systems would cause Windows to swap between A: and B: (throwing a blue screen asking the user to insert a disk etc). 2) Only floppy drives and FAT16 drives would be available in the list of drives, no FAT32/optical/network drives.
• Improved code in IdeControllerMenu.asm so that the default port addresses for all IDE interfaces are now restored when (re-)selecting the (same) type of IDE device.
• Also made it impossible to select a device type unless the required module is included in the loaded BIOS.
• The version check done when loading a BIOS now uses the FLASH_SIGNATURE definition from Version.inc. Any changes affecting RomVars now only requires updating that definition. This means that changes to RomVars must be implemented in both the BIOS and XTIDECFG before being committed to the repository.
• Added a compatibility fix for 3Com 3C503 cards to the ROM checksumming code in Buffers.asm (Buffers_GenerateChecksum).
• SerDrive: Made some minor changes to file names and paths to improve compatibility with case sensitive environments.
• BIOSDRVS: Made a minor size optimization which as a side effect also makes it compatible with all DOS versions including DOS version 1.
• Library: Renamed the WAIT_RETRACE_IF_NECESSARY_THEN macro to CALL_WAIT_FOR_RETRACE_IF_NECESSARY_THEN and made a tail-call-optimized version of it (JMP_WAIT_FOR_RETRACE_IF_NECESSARY_THEN).
• A speed optimization to the eRCL_IM macro for 386 and higher. This change breaks emulation in the sense that the macro will fail when given a memory operand as the first parameter.
• Other minor optimizations and fixes.

File:

• trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeCommand.asm (modified) (3 diffs)

trunk/XTIDE_Universal_BIOS/Src/Device/IDE/IdeCommand.asm

@@ -33 +33 @@
 IdeCommand_ResetMasterAndSlaveController:
     ; HSR0: Set_SRST
-    call    AccessDPT_GetDeviceControlByteToAL
-    or      al, FLG_DEVCONTROL_SRST | FLG_DEVCONTROL_nIEN   ; Set Reset bit
+; Used to be:
+;   call    AccessDPT_GetDeviceControlByteToAL
+;   or      al, FLG_DEVCONTROL_SRST | FLG_DEVCONTROL_nIEN   ; Set Reset bit
+; Is now:
+    mov     al, FLG_DEVCONTROL_SRST | FLG_DEVCONTROL_nIEN
+; ---
     OUTPUT_AL_TO_IDE_CONTROL_BLOCK_REGISTER     DEVICE_CONTROL_REGISTER_out
     mov     ax, HSR0_RESET_WAIT_US
@@ -40 +44 @@
 
     ; HSR1: Clear_wait
-    call    AccessDPT_GetDeviceControlByteToAL
-    or      al, FLG_DEVCONTROL_nIEN
-    and     al, ~FLG_DEVCONTROL_SRST                        ; Clear reset bit
+; Used to be:
+;   call    AccessDPT_GetDeviceControlByteToAL
+;   or      al, FLG_DEVCONTROL_nIEN
+;   and     al, ~FLG_DEVCONTROL_SRST                        ; Clear reset bit
+; Is now:
+    mov     al, FLG_DEVCONTROL_nIEN
+; ---
     OUTPUT_AL_TO_IDE_CONTROL_BLOCK_REGISTER     DEVICE_CONTROL_REGISTER_out
     mov     ax, HSR1_RESET_WAIT_US
@@ -51 +59 @@
     jmp     IdeWait_PollStatusFlagInBLwithTimeoutInBH
 
+; *FIXME* AccessDPT_GetDeviceControlByteToAL currently always returns with
+; AL cleared (0) or with only bit 1 set (FLG_DEVCONTROL_nIEN = 2).
+; The commented away instructions above sets FLG_DEVCONTROL_nIEN anyway
+; making the call to AccessDPT_GetDeviceControlByteToAL redundant.
+; I have left this code as is since I don't know if it's a mistake
+; (from all the way back to r150) or if it's coded this way in anticipation
+; of some future changes to AccessDPT_GetDeviceControlByteToAL.
 
 ;--------------------------------------------------------------------