OS3.2 boot delay in WinUAE

Condor-2-4-2 · 31 December 2021, 15:03

Hello,

I made two clean installations of AmigaOS3.2 - on one hdf file with FFS and another with PFS filesystem in WinUAE 4.9.0. I set the device number for both hdf files to 0. Each hdf has only one partition. Kickstart rom is kicka1200 v3.2.

After installation and reboot for both installations I noticed a delay about 12 seconds until the read access on the virtual drive begins and the OS boots. I can see the hdf access on the WinUAE on-screen display, which I activated. This delay happens for hard and soft reset.

I also used hdf files which I formated exactly the same way for AmigaOS3.1 installations with Kickstart rom version 3.1 and have no such delay. Here the disk access begins immidiatly after starting the WinUAE config. And I'm using the same WinUAE config settings as for the AmigaOS3.2 installations (CPU, Chipset, RAM, Hard drive settings).

Now my question is what can cause this long delay on booting AmigaOS3.2 and how can I avoid it?

I attached my WinUAE config for the AmigaOS3.2 installation and the startup-sequence, which is the original unedited one that comes with the installation.

Best regards and Happy 2022!

Condor-2-4-2

Toni Wilen · 31 December 2021, 18:11

It is OS 3.2 IDE scsi.device introducing annoying (probably technically correct but still annoying because it also ignores "ide killer") delay, slowing down booting even if there is some other autoconfig boot device.

Only workaround is to patch the ROM and disable the device driver.

Condor-2-4-2 · 31 December 2021, 21:43

Hello Toni,

you mentioned a workaround patch for the rom. Do you know if such a patch is available? Perhaps in the update 3.2.1?

And if not which alternate device would you suggest for my configuration? My intention is to use the virtual os to execute whdload and preinstalled games.

bubbob42 · 31 December 2021, 23:12

There are actually three delays. The second delay is caused by setpatch and was needed for some machines with ATAPI CDRs which would result in the IDE activity LED not working correctly. This has become optional for 3.2.1, so you should update.

The third delay is caused by mmu.library which will build the MMU tables (also triggered by setpatch), this can take quite some time on real machines with lots of RAM.

The first delay would be IDE detection - this has been increased after the first beta versions IIRC, because otherwise some drives with slow spin up times would require an additional manual reset or they would not be detected.

Toni Wilen · 31 December 2021, 23:27

Quote:

Originally Posted by bubbob42

The first delay would be IDE detection - this has been increased after the first beta versions IIRC, because otherwise some drives with slow spin up times would require an additional manual reset or they would not be detected.

Yes, but... KS 3.1 had 30s+ delay and it still skipped the delay immediately if "IDE killer" was in use. Apparently 3.2 delay is forced and ignores IDE killer. Making worst case situation the common case is not good. Common case is some kind of solid state drive that is ready before scsi.device initializes.

OS 3.9BB2 (possibly older ones too) had similar delay but it at least it was easy to disable because scsi.device update was setpatch loaded.

Toni Wilen · 01 January 2022, 10:01

Delay seems to go away if IDE hardware is disabled (it probably checks if Gayle ID is found). Advanced chipset -> untick compatible settings -> untick A600/A1200 IDE. It probably won't work with A4000 because it has no Gayle.

But missing Gayle might cause other side-effects..

Condor-2-4-2 · 01 January 2022, 20:25

In the meantime I updated my Amgia installation to version 3.2.1 with kickstart1200.47.102.rom. This didn't had an effect on the long boot delay. After that I deactivated A600/A1200 IDE in WinUAE chipset settings. This didn't bring any improvement for me too.

I attached my actual WinUAE config bellow.

thomas · 02 January 2022, 10:11

Just add a dummy HDF to the IDE bus.

RDB mode, install drive, delete all partitions.

Toni Wilen · 02 January 2022, 11:15

It does not help much. Even with drive installed, there is always about 5s delay before any IDE register accesses and then few more seconds before first RDB IDE read.

Latest 4.9.1 beta has a config file only hack that disables scsi.device completely: Almost instant boot from uaehf.device.

My guess is that this delay was most likely added in OS39 (BB2 which added broken LBA support?) and it is not OS 3.1.4+ new feature.

Michael · 02 January 2022, 16:41

The magic option is 4.9.1 and config file entry

scsidevice_disable=true

and it kind of works, minimalistic delays!

Not sure about any side effect

Condor-2-4-2 · 02 January 2022, 20:19

It also works for me. After changing to WinUAE 4.9.1 beta 3 and adding the line scsidevice_disable=true in the WinUAE config file the drive access now starts immidiatly after executing the config file. The booting of the AmigaOS3.2 itself is very fast.

But I also didn't know if this method can cause undesirable side effects as Michael wrotes before.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Amiga 3000 scsi boot initial delay	dschallock	support.Hardware	12	16 August 2021 08:45
winuae os3.9 main configuration won't boot anymore	honx	support.WinUAE	21	18 May 2020 20:11
Delay Icon Load of Removable Media during Boot	Pheonix	support.Apps	8	12 September 2016 18:55
Minimize delay in WinUAE?	Photon	support.WinUAE	7	06 December 2005 21:19
Boot delay option	Bloodwych	request.UAE Wishlist	9	25 January 2004 23:01

31 December 2021, 18:11	#2
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,506	It is OS 3.2 IDE scsi.device introducing annoying (probably technically correct but still annoying because it also ignores "ide killer") delay, slowing down booting even if there is some other autoconfig boot device. Only workaround is to patch the ROM and disable the device driver.

31 December 2021, 21:43	#3
Condor-2-4-2 Registered User Join Date: Mar 2017 Location: Germany Posts: 89	Hello Toni, you mentioned a workaround patch for the rom. Do you know if such a patch is available? Perhaps in the update 3.2.1? And if not which alternate device would you suggest for my configuration? My intention is to use the virtual os to execute whdload and preinstalled games.

31 December 2021, 23:12	#4
bubbob42 Registered User Join Date: Oct 2012 Location: Germany Posts: 585	There are actually three delays. The second delay is caused by setpatch and was needed for some machines with ATAPI CDRs which would result in the IDE activity LED not working correctly. This has become optional for 3.2.1, so you should update. The third delay is caused by mmu.library which will build the MMU tables (also triggered by setpatch), this can take quite some time on real machines with lots of RAM. The first delay would be IDE detection - this has been increased after the first beta versions IIRC, because otherwise some drives with slow spin up times would require an additional manual reset or they would not be detected.

01 January 2022, 10:01	#6
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,506	Delay seems to go away if IDE hardware is disabled (it probably checks if Gayle ID is found). Advanced chipset -> untick compatible settings -> untick A600/A1200 IDE. It probably won't work with A4000 because it has no Gayle. But missing Gayle might cause other side-effects..

02 January 2022, 10:11	#8
thomas Registered User Join Date: Jan 2002 Location: Germany Posts: 6,985	Just add a dummy HDF to the IDE bus. RDB mode, install drive, delete all partitions.

02 January 2022, 11:15	#9
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,506	It does not help much. Even with drive installed, there is always about 5s delay before any IDE register accesses and then few more seconds before first RDB IDE read. Latest 4.9.1 beta has a config file only hack that disables scsi.device completely: Almost instant boot from uaehf.device. My guess is that this delay was most likely added in OS39 (BB2 which added broken LBA support?) and it is not OS 3.1.4+ new feature.

02 January 2022, 16:41	#10
Michael A1260T/PPC/BV/SCSI/NET Join Date: Jan 2013 Location: Moscow / Russia Posts: 839	The magic option is 4.9.1 and config file entry scsidevice_disable=true and it kind of works, minimalistic delays! Not sure about any side effect

02 January 2022, 20:19	#11
Condor-2-4-2 Registered User Join Date: Mar 2017 Location: Germany Posts: 89	It also works for me. After changing to WinUAE 4.9.1 beta 3 and adding the line scsidevice_disable=true in the WinUAE config file the drive access now starts immidiatly after executing the config file. The booting of the AmigaOS3.2 itself is very fast. But I also didn't know if this method can cause undesirable side effects as Michael wrotes before.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)