[SOLVED] SV FW 010 -> Crash

[johannes]

Hi there,

yesterday I updated my SV using the SVUPD010.APP to bring it to FW rev 010 and get the great Svethlana Speed improvement. Unfortunately after everything seemed fine (SVUPD010.APP reported current FW 006 correctly, confirmed correct update of both chips) I powercycled the Falcon and screen stayed dark.

So I hooked a VGA-Monitor to the falcon Videl output and I saw the "Exception Processing 2: Access Fault" message when loading the SV-Driver (SV_XBIOS.PRG) Address fault is $8001110B. See attached screenshot.

My Environment:
Falcon TOS 4.04, CT63 latest ABE/SDR and Boot 1.03c, SV was FW 006. Patched in 060 mode (under MiNT).

What I've tried so far:

1. Do not load driver: Boot to desktop possible (but only without svethlana driver which freezed the system, too).
2. Start SVUPD010.APP from desktop: Same Exception with slightly different addresses
3. Start SVUPD006.APP from desktop: "Supervidel not found".

today, i continued:

4. Program SV with Xilinx USB JTAG Cable and Impact software using the FW010. Programming successfull (according to IMpact software), but same outcome
5. Program SV with Xilinx USB JTAG Cable and Impact software uding the FW005. Same as above.

Somebody any ideas? would be greatly appreciated!

Thanks and cheers
Johannes

IMG_5039.JPG

[johannes]

Update: I was tinkering with different FW versions (not that much hope) and suddenly I could resurrect the SV with FW 009 (updated via JTAG cable). Once done that I tried to upgrade to 010 once more using the JTAG cable, but without luck.

Also, programming FW 010 with SVUPD010.APP was not successful. Same result: bricked SV. But I am glad, that I was able to revert to 009!

What can I change to give FW010 a new chance?

Thanks !
Johannes

[instream]

Did you replace your old Svethlana driver (v0.1) with the new one (v0.9)? The old driver does not work with the v010 FW, and the new driver will not work with the old FW v009 and older.

[johannes]

Yes, i replace the driver, but The ecxeption occurs before mint is loaded. The SV XBIOS is my first AUTO Folder Program.

[instream]

Is your SV_XBIOS.PRG 101.467 bytes?

[johannes]

Yes, 101.467 bytes. And the driver is not the only problem - when flashing to 010, not booting the driver and then calling SVUPD010.APP again, its the same error. Somehow I feel, my SV does not like FW010?!

Additional Info: with FW 009 (as well as with all previous) I got simultanious output on SV and normal Videl from powering on the falcon until SV_XBIOS.PRG is loaded. With FW010 SV output stays dark.

[Dal]

Out of interest, do you have the cable for the motherboard clocks installed?

[johannes]

No, currently I only use desktop Apps, so I didnt install it.

[instream]

ABE_V7F_SDR_V7D.zip[/attachment]I didn't recognize the address in the screen shot of your access fault until now when I looked through the VHDL code. I makes me suspect that you maybe have too old ABE/SDR versions on your CT60? I think I use the attached ones. Try them.

[instream]

Check first that they work with FW 009, then move on to FW 010.

[johannes]

Hi Instream,

thanks for your helping hand!!

I was on the latest ABE/SDR but of course, I tried your attached config files with my CT63.

The result is exactly the same:

- Programmed SV to 009
- Programmed CT63 to ABE 7F, SDR 7D
--> WORKS

- Programmed SV to 010
--> DOES NOT WORK, Same error message. Addresses are:

Status Register: $2300, bits to 1: SI1 I0
Program Counter: $010FC682
Basepage: $010F64A0
Cache register: $00000000
Address Fault: $8001110B
Fault Status Word: 00A50018, bits to 1: W SZ0 TM2 TM0 WE TTR

omitting stack, please drop a note if that might be usefull, too.

Additional Info: With V009, BOTH outputs (Falcon Videl and SuperVidel) show bootscreen, With V010 only Falcon Videl show bootscreen, and then SV_XBIOS.PRG hangs with above error.



Cheers
Johannes

[instream]

That is extremely odd. Since you with FW 010 don't get any indication that your SV can even snoop the Videl accesses to be able to display the boot screen, I would say that something has changed in the businterface of the VHDL code from FW 009 to FW 010, and that change didn't cause problems on my SV, but on yours. So I will have to compare 9 and 10 to to see if the bus timing has changed in any way.

[johannes]

I wonder if I am the only one with this kind of problem... Others here seem not to have any issues with v010. Thanks for digging into the code!

[Dal]

I've not upgraded mine yet. In light of this, I'm trying to schedule some time to give it a go though. In my case I have the clocks soldered so not sure what - if any - difference this makes.

[johannes]

I tried other approaches as well:

- Tos 1.03c -> Tos 2.01
- Set Clockspeed of CT63 to 66MHz

Same outcome...

Perhaps, instream can shed some light on the changes from 009 to 010.

[shoggoth]

I tried other approaches as well:

- Tos 1.03c -> Tos 2.01
- Set Clockspeed of CT63 to 66MHz

Same outcome...

Perhaps, instream can shed some light on the changes from 009 to 010.

I don't recommend using a CT TOS higher than 1.03c. It "should" "work", but there are just too many nasty hacks in those TOS versions - some of which was a real problem when developing the drivers.

[johannes]

Yeah, it was only a try to clutch at a straw. I will revert it to 1.03c

[Dal]

I updated mine today. No problem to report.

[johannes]

Good to hear. I am beginning to think about my skills

[instream]

I promised to compare the 009 code with 010. But I have been away and got the mandatory summer cold, so haben't (pun intended) looked into it yet.

[johannes]

No worries, I stand by for any new insights and possible workarounds. Thanks !!

[instream]

I have just uploaded some new FW files for you to try, at nature.atari.org. There are both FW010 files and FW011 (bigger superblitter FIFO and a small debug counter, so nothing major). Each FW version has 3 variants (different timing score) with different routing inside the FPGA. This can help you avoid the problems you got with the older FW010. Try them each in turn until you hopefully find a version that works for you. All variants work on my system. Don't mix mcs_0 files from one score set with a mcs_1 files from another score set.

[johannes]

Thanks!! I am currently in vacation, but will check as soon as I got back (mid September)!

[johannes]

Yeah! That did the trick for me! FW11, score 123 did work (90 did not work, I did not test 139). Thanks a lot! Speed test of Svethlana is impressive. Got 1.42 MB/s

Grateful cheers
Johannes

[instream]

Great!
Then the theory that some FPGAs have problems in different regions of the chip is correct. This is because they are quite heavily overclocked, from 166MHz to 200MHz. From know on we'll release every new version in three different variants, so you can try another one if the first choice fails.

/Torbjörn