WinUAE 3.1.0 beta series

[mritter0]

Properly loaded config file and dump files.

[Toni Wilen]

Originally Posted by mritter0

Properly loaded config file and dump files.

Still wrong config and no winuaelog.txt. Load your config, click debugging checkbox in Paths panel, boot to WB. Save logs. There is no need to duplicate the crash. (It would only truncate log file)

Does http://www.winuae.net/files/b/winuae.zip still crash? (if it does, do not attach dump files, it will be useless)

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3100b13.zip

Beta 13:

- Mainboard SCSI controllers didn't initialize correctly (A3000/A4000T/CDTV) (b12)
- "Do not force full display reset.." update made it worse, should be better now (b12)
- I finally bothered to carefully re-examine AR3 freezer behavior, emulation updated, should now match real hardware, previously unsupported features like autofire work.
- If copper DMA is switched off mid-instruction, allow instruction to finish. Fixes AR2 weird copper list detection code.
- Added RDB mode HD geometry GUI support. (RDB mode geometry option = what IDE Identify Device and SCSI MODE SENSE Format Parameters/Rigid Disk Geometry returns). Very old controllers may require custom geometry.
- Added HD/CF IDE harddrive GUI option. Currently only difference is in IDE Identify Device word 0 identifier. (Cause for infamous hdtoolbox "Unit is not a disk (Type 7)" error)
- Added GUI IDE and SCSI version selection to hardfile panel. IDE has ATA-1, ATA-2+ (old default, returns ATA-6 to allow LBA48 support and more) and ATA-2+ Strict which drops ATA-1 feature that become optional in ATA-2, triggering A600/A1200/A4000 IDE driver bug that causes infamous "max transfer bug". (No, it isn't drive bug) SCSI has SCSI-1 and SCSI-2 options. Currently only real difference is INQUIRY version number, some old software (like original A590/A2091 install disk) require SCSI-1 version.
- Added Archos ADD-500 emulation.
- Added AdSCSI Advantage 2000 emulation.
- Internal enforcer emulation hit can break to debugger, enable/disable with 'fen' debugger command.
- SCSI Read Defect Data command: return empty defect header instead of Defect List Not Found error. (ADD-500 driver gets confused if Read Defect Data command does not return any data)
- WD33C93 SCSI devices didn't support multiple units (b12)
- Added hack that fixes max overscan scrolling in demo Seven Seas / Andromeda. (Demo has workaround for chipset bug that causes right edge to have a vertical gap if BPLCON1 is larger than zero. Hack is currently needed because glitch happens when hpos=0 and emulation is not designed to handle bitplane changes that cross virtual scanlines). Glitch was not emulated until few versions ago, demo worked only accidentally previously.
- Added Windows 10 TP bug workaround, for some reason dialog window is visible before WM_INITDIALOG message (documentation specifically says it won't be visible until after WM_INITDIALOG) and also dialog is set as not visible. Fixes quickly appearing and disappering empty WinUAE GUI panel each time GUI is opened.

AdSCSI Advantage 2000:
- 5380 based fake DMA.
- First 5380 based driver that uses chip's arbitrate bit instead of assuming AdSCSI is the only initiator.
- 1.6 ROM added. (Anyone have later? At least 1.9 exists)
- RDB parser is not fully compatible, does not load all filesystems (for example PFS3).

Archos ADD-500:
- Yet another 5380 based fake DMA SCSI controller.
- v1.21 ROM added.

[supaduper]

Your on a `roll` Toni , where would we all be without our brill Winuae ?, I know `up the Swanny without...................

Cheers Dude

[mark_k]

Originally Posted by Toni Wilen

- Added AdSCSI Advantage 2000 emulation.

For some reason, on booting this HDF with the AdSCSI 2000 config attached, the WinUAE debugger window opens.

hxxp://www.4!s!h!a!r!e!d.com/file/QcfdV0t6ce/AdSCSI_21MBhdftar.html

(The HDF is not necessarily supposed to work, I just took my working AdIDE A-Max HDF and removed the first sector data.)

[Toni Wilen]

Originally Posted by mark_k

For some reason, on booting this HDF with the AdSCSI 2000 config attached, the WinUAE debugger window opens.

Forgot to remove last minute debugging code..

[mark_k]

Originally Posted by Toni Wilen

- Added RDB mode HD geometry GUI support. (RDB mode geometry option = what IDE Identify Device and SCSI MODE SENSE Format Parameters/Rigid Disk Geometry returns). Very old controllers may require custom geometry.

Would you be able to make the default manual geometry use the appropriate values if there's a matching entry in the default.geo file?

[mritter0]

Sorry. In a rush.

With the first beta you posted I have not had a crash yet. I have not tried with the second beta you posted.


EDIT: Just had a crash. It seems when just using programs it is fine. When I do some compiling with GCC is when it crashes.

EDIT2: After some use, it also does it after F12 and Restart; crashes immediately.

[mark_k]

With b13, use the attached config and the HDF from hxxp://www.4!s!h!a!r!e!d.com/file/flt-I6axce/testhdftar.html

[I don't know whether the specific HDF matters.]

Start emulation. Even with -log -scsiemulog there's no SCSI-related log output, which seems strange. The emulated system seems to hang at a black screen.

Press F12 and open the Hardfile Settings window for the HDF. Change controller type from GVP Series II to UAE, click OK then Reset. WinUAE crashes.

[Toni Wilen]

Originally Posted by mritter0

Sorry. In a rush.

With the first beta you posted I have not had a crash yet. I have not tried with the second beta you posted.


EDIT: Just had a crash. It seems when just using programs it is fine. When I do some compiling with GCC is when it crashes.

EDIT2: After some use, it also does it after F12 and Restart; crashes immediately.

Unfortunately I can't duplicate it and I can't see reason for the incorrect data that causes the crash. (and because it is OS4 config, it makes it even more difficult to debug)

Please attach new crash dump from b13.

Originally Posted by mark_k

[I don't know whether the specific HDF matters.]

Start emulation. Even with -log -scsiemulog there's no SCSI-related log output, which seems strange. The emulated system seems to hang at a black screen.

It is not HDF. Your config has GVP Series I and Series II enabled at the same time and they shared same data structures (I was going to update it in b12 update but forgot) which does not really work very well..

Fixed now.

Press F12 and open the Hardfile Settings window for the HDF. Change controller type from GVP Series II to UAE, click OK then Reset. WinUAE crashes.

"Don't do it" is the only possible reply now. (Enabling UAE boot rom on the fly causes issues)

[mark_k]

Originally Posted by Toni Wilen

It is not HDF. Your config has GVP Series I and Series II enabled at the same time and they shared same data structures (I was going to update it in b12 update but forgot) which does not really work very well..

Whoops, didn't realise I forgot to disable the Series I controller when I changed the config. It's very easy to forget that.

Would it be possible to make the names of enabled boards show in bold text in the SCSI/IDE/Boot ROM file drop-down gadget? Then the user can easily see which are enabled.

[mark_k]

Testing Supra A500 drive emulation I noticed a small issue when SupraHDTools issues a MODE SENSE command for page 3 (format device page):
raw_scsi: got command 1a (6 bytes)
1a.00.c3.00.ff.00.
SCSIEMU HD 0: 1A.00.C3.00.FF.00.00.00.00.00.00.00 CMDLEN=6 DATA=06806648
-> DATAOUT=36 ST=0 SENSELEN=0 REPLYLEN=36
REPLY: 23.00.00.08.00.00.A0.00.00.00.02.00.03.18.00.01.00.00.00.00.00.00.00.20.02.00.00.01.00.00.00.00.80.00.00.00.

The length field in the page data should be $16 not $18.

Edit: and should the tracks per zone field be 0000, not 0001?
And same length issue with page 4.

[mritter0]

It seems the culprit is Pete's FTP v1.2 program. Load to Workbench, F12, Restart, Start. Works fine.

Load to Workbench, start pFTP, connect to site, F12, Restart, Start, crash. Even if I quit pFTP properly it still crashes, but not as often.

Debugs show both with and without starting pFTP.

[Toni Wilen]

Originally Posted by mark_k

Testing Supra A500 drive emulation I noticed a small issue when SupraHDTools issues a MODE SENSE command for page 3 (format device page):
raw_scsi: got command 1a (6 bytes)
1a.00.c3.00.ff.00.
SCSIEMU HD 0: 1A.00.C3.00.FF.00.00.00.00.00.00.00 CMDLEN=6 DATA=06806648
-> DATAOUT=36 ST=0 SENSELEN=0 REPLYLEN=36
REPLY: 23.00.00.08.00.00.A0.00.00.00.02.00.03.18.00.01.00.00.00.00.00.00.00.20.02.00.00.01.00.00.00.00.80.00.00.00.

The length field in the page data should be $16 not $18.

Edit: and should the tracks per zone field be 0000, not 0001?
And same length issue with page 4.

Lengths fixed. Not sure about field value, most of them are only used by ancient devices.. (new drivers don't care about those = obsolete stuff should match ancient device behavior)

Originally Posted by mritter0

It seems the culprit is Pete's FTP v1.2 program. Load to Workbench, F12, Restart, Start. Works fine.

Load to Workbench, start pFTP, connect to site, F12, Restart, Start, crash. Even if I quit pFTP properly it still crashes, but not as often.

Debugs show both with and without starting pFTP.

Does http://www.winuae.net/files/b/winuae.zip fix it?

[mritter0]

Yes, that fixed it. Thank you!

[mritter0]

Suggestion: When the zip debug archive is made, have it include the loaded config file, not the default one. I know I screwed it up a time or two not putting the correct config file in it thinking it was doing it that way.

[mark_k]

I've been having some problems with Supra controller emulation. I'm using the "Series II" Supra software with AMAB4 boot ROM.

Using supraboot.dms from www.l8r.net but with fixed/patched supradirect.device, and SupraTools as recently uploaded by clh333 (STII.DMS). [That SupraTools disk is a slightly earlier version than the l8r.net supraboot disk.]

I can partition the disk and install the DriveInit code to the RDB area by runing SupraTools:SupraAutoInstall. At that point the HD is partitioned and will (try to) auto-boot, but the partitions are not yet formatted.

On resetting and booting SupraBoot, it automatically mounts the (not yet formatted) partitions. It prompts for whether to copy Workbench files. Say n. Workbench loads. Open a CLI and type
Format DRIVE DH0: Name Supra0 NOICONS FFS QUICK
Format shows an error message:
Format failed: Error during format.

Running WinUAE with -log -scsiemulog, I noticed this in the log output:
raw_scsi_put_data but bus phase is 3!

If I do a non-quick format, the Formatting/Verifying text appears, and right at the end (presumably corresponding to whatever sectors Format QUICK wrote) there is the same Error during format message.

Configs for 500/XP and A2000 DMA controllers and HDF (partitioned and DriveInit code installed) attached.

[mark_k]

Originally Posted by Toni Wilen

- Added rare SupraDrive 2000 DMA emulation. Boot ROM are not available, some supra install disks have on-disk loadable driver which has bug in 2000 DMA model specific code that causes it to hang (uses wrong address register, overwrites one inportant byte in interrupt handler code..) Perfect.

Just a note for anyone testing Supra emulation with the Series II Supra software (supradirect.device)...

That bug applies to all Supra controllers, not just the 2000 DMA. It causes the interrupt handler code to be corrupted for A2000 WordSync, 500/XP ByteSync and 2000 DMA controllers. For the SupraDrive 4x4 and 500 it corrupts random memory, the address depending on whatever value FindConfigDev() left in A0.

Edit to add: After analysing the bug some more, it turns out that by pure luck (and the fact that FindConfigDev() returns with a pointer to the ConfigDev in A0, at least in both Kickstart 1.3 and 3.1), the bug is only fatal for the 2000 DMA.

[Toni Wilen]

Originally Posted by mritter0

Suggestion: When the zip debug archive is made, have it include the loaded config file, not the default one. I know I screwed it up a time or two not putting the correct config file in it thinking it was doing it that way.

Works for me (loaded some config, booted it, F12, Paths panel, save, config was correct).

Config file in zip is on the fly generated from currently active config (which is not necessarily exact same as you loaded if you changed any options after loading it)

[Toni Wilen]

Originally Posted by mark_k

Running WinUAE with -log -scsiemulog, I noticed this in the log output:
raw_scsi_put_data but bus phase is 3!

It is harmless. For some reason after all 512 bytes have been transferred using "fake dma" register, it writes dummy zero to 5380's data register. (which does not even do any real writes when using real 5380)

Formatting seems to work and all SCSI commands look fine. I don't know what the format command does, I probably sends some suprawhatever.device io command (that does not hit hardware) that returns something unexpected.