PFS3AIO v3.1 Final

[Toni Wilen]

http://www.winuae.net/files/b/pfs3aio_v3.1_test3.zip

PFS3AIO 3.1 Test #3

- TD64 support check false negative fixed, it rejected many TD64 capable HD controllers. (Unsigned vs signed comparison..)
- Execute access mode test (TD64/NSD/DS) every time removable volume is inserted and also when formatting.
- Removed Direct SCSI scsi_Actual validation after read or write operation. Apparently not all SCSI controllers update it properly. Was added in 3.0.
- Fixed out of bounds buffer read when writing file to disk that does not fill last disk block completely. Generally harmless but can cause MuGuardianAngel hits and in worst case can cause accesses that cross end of RAM. (Original PFS3 bug/optimization)
- Always execute TD_MOTOR off after idle delay period. Previous pfs3aio versions only executed it if partition was located in removable media. Some HD controllers (At least Blizzard 2060) uses TD_MOTOR off to switch off activity led.

This is hopefully becomes final v3.1.

[Epo]

Toni, fantastic work as always. Your contribution to the Amiga community deserves a special medal.
I have a question regarding the versioning. For the whole archive you use 3.x but the file pfs3aio internally (also when added in the HDToolBox) shows 18.x or 19.x. Why the difference? Shouldn't be the whole package versioning same as the internal one like PFS3AIO v19.x Final? I think it would be less confusing but maybe I don't have the whole picture.

[Toni Wilen]

Originally Posted by Epo

Toni, fantastic work as always. Your contribution to the Amiga community deserves a special medal.
I have a question regarding the versioning. For the whole archive you use 3.x but the file pfs3aio internally (also when added in the HDToolBox) shows 18.x or 19.x. Why the difference? Shouldn't be the whole package versioning same as the internal one like PFS3AIO v19.x Final? I think it would be less confusing but maybe I don't have the whole picture.

It requires two versions simply because I want to keep v18 and v19 as internal structure versions which should only change when internal disk structure implements new feature or changes. v19 = >104G partition support and >4G file size support which are not v18 compatible if enabled.

Other version number changes each time new update is released, even if it is cosmetic change.

[Epo]

Other version number changes each time new update is released, even if it is cosmetic change.

So if there is a change (even cosmetic) the name of the package increments in version - I get that. But that change is due to what, the pfs3aio file change or something else? If to the pfs3aio then does it change internally too or keeps v19.0 all the way (if so why not 19.1, 19.2, etc.)? Help me understand this as it's a mistery for me for a while

[nexus]

im using this 3.1 test 3 on a fresh 3.9 with bb1-4 on a 300GB scsi on a warpengine with 2 4GB partitions
not sure what the deal is but once in a while i will get a error 2 write and a retry fixes and if i redo the write eg reinstalling something that issued the error it doesnt happen again
im using 0x50445303 with 300 set for buffers maxtransfer is 0x1FE00
i had to use thomas hddfix program to get the values for the hdd as the read configuration would only ever see 45gb
SEAGATE/HP-ST3300655LC/412751-016/BF3008B26C- 300GB 15K.5 U320 80-PIN HDD

besides that above this hdd is super fast and silent

[Toni Wilen]

Originally Posted by Epo

So if there is a change (even cosmetic) the name of the package increments in version - I get that. But that change is due to what, the pfs3aio file change or something else? If to the pfs3aio then does it change internally too or keeps v19.0 all the way (if so why not 19.1, 19.2, etc.)? Help me understand this as it's a mistery for me for a while

Forgot to reply.. Yes, I can bump revision number.

Originally Posted by nexus

im using this 3.1 test 3 on a fresh 3.9 with bb1-4 on a 300GB scsi on a warpengine with 2 4GB partitions
not sure what the deal is but once in a while i will get a error 2 write and a retry fixes and if i redo the write eg reinstalling something that issued the error it doesnt happen again
im using 0x50445303 with 300 set for buffers maxtransfer is 0x1FE00
i had to use thomas hddfix program to get the values for the hdd as the read configuration would only ever see 45gb
SEAGATE/HP-ST3300655LC/412751-016/BF3008B26C- 300GB 15K.5 U320 80-PIN HDD

besides that above this hdd is super fast and silent

Do you get same error (+full scsi sense data) if you use direct scsi mode? (dostype = PDS\3 = 50 44 53 03. Reformat is not needed.). It is most likely caused by random hardware glitch. (correct termination?)

[nexus]

i am using direct scsi identifier 0x50445303 and yes termination on warp and dvd drive with hdd in middle
its very random but any time it pops up and ask me to retry i do with no errors

the above is a new setup on A4K with a mediator which is also giving me issues everytime i start the voodoo icon or p96mode in prefs it crashes the amiga 99% when it doesnt everything is garbled on the desktop and only hitting redraw fixes

so not sure whats going on maybe i have some bad ram

[Toni Wilen]

Originally Posted by nexus

i am using direct scsi identifier 0x50445303 and yes termination on warp and dvd drive with hdd in middle
its very random but any time it pops up and ask me to retry i do with no errors

Don't you get error requester like this:

Write 2 Error X on block Y + Z
xx.xx.xx.xx.xx.xx.xx.xx.xx.xx.xx

? Or do you only get first line?

2 = Direct SCSI mode. Actual error code is Y.

[nexus]

yes like the 1st line next time it happens ill take a pic

[nexus]

but no matter what hdtoolbox never can read the drive configuration every time i hit that button its always set to 45GB
so i have to manually do the geometry im also using a 80 to 50pin adapter
though in the device selection screen does read drive firmware and size of 279GB
this was done through emergency disk and cdrom

the error never happened on os3.9 install it started after when installing boingbags from what i remember each bag i had extracted to my second 4GB partition Work: and during extract and running each boingbag i got a error once during extracting and installing
the 1st time it happened i redid what i had done eg extract the lha and install the boing bag the error didnt happen again till i did the next boingbag
so moving the lha from a CD to Work: was no error but from Work: to Workbench: partitions it did

[Toni Wilen]

which hdtoolbox version? (afaik only 3.9bb2 version supports very large drives without strange glitches)

About direct scsi error:

It is likely that "normal" access methods do few automatic retries if it detects SCSI bus related error before returning error to caller (hides the random error).

Direct SCSI passes the command directly to drive, without any pre/post processing or retries. I can add automatic retries but first I'd like to see exactly what kind of error gets reported by the drive.

[nexus]

i used hdtoolbox on emergency disk the 1st amigaos3.9 cd i also tried what ever hdtoolbox version is on the amiga911 emergency disk same results also check4GB says both are ok even though i can do 7.8GB
anyways too try and force this for you it happened on boot

write 2 error 173 on block 8360396 + 2
70.00.00.00.00.00.00.0A.00.00.00.00.00.00.00.00.00.00

[Toni Wilen]

Originally Posted by nexus

i used hdtoolbox on emergency disk the 1st amigaos3.9 cd
anyways too try and force this for you it happened on boot

write 2 error 173 on block 8360396 + 2
70.00.00.00.00.00.00.0A.00.00.00.00.00.00.00.00.00.00

173 = 173 - 128 (if this field is >= 128 it is HD_SCSICMD IO error code) = 45 = HFERR_BadStatus.

Sense data tells that nothing is wrong.. So in other words, it wasn't the drive that reported the error. It probably means HD controller's SCSI chip detected some unexpected SCSI bus related problem.

I guess I can add automatic retry (1-2 times max) if HD_SCSICMD returns error but SCSI sense data reports no errors.

[nexus]

not sure what if anything i might have done wrong but im a newb at doing this stuff on a amiga so it is possible
i have termination on warpengine and a new scsi cable with dvdrom on the end termination jumper on
even tried with term power on the hdd and warp
ofc im having other issues setting this up aswell soo????
like setting up this mediator with just a voodoo crash fest and i tried 2 mediators
wasnt sure if the hdd above issue was a issue with this setup
this is a new amiga i never setup before so not sure if there are other issues

[Toni Wilen]

Originally Posted by nexus

not sure what if anything i might have done wrong but im a newb at doing this stuff on a amiga so it is possible

I don't think you have done anything wrong (If termination is correct). Cause most likely is some hardware glitch, it may be even "normal" because most filesystem don't use direct scsi.

Try without CD drive.

I have resourced disassembly of warpdrive.device, I can check (later) if it does some extra checks and does retries in normal CMD_READ routine.

[nexus]

should i continue using this version or go back cause at this moment im about to redo everything cause of this RTG issue

[Toni Wilen]

Originally Posted by nexus

should i continue using this version or go back cause at this moment im about to redo everything cause of this RTG issue

It shouldn't change anything. Direct SCSI error check has always been there, only format of requester has changed in 3.x versions (previously it only told that some read or write error happened and that was all..).

[nexus]

here is another that happened while copying workbench to work partition

write 2 error 173 on block 86327 + 6
F0.00.04.00.83.B9.8F.0A.00.00.00.00.03.00.08.80.00.17

but i guess write 2 error 173 is always same
seems to happen when moving a bunch of files

[Toni Wilen]

Key value (byte 2) = 4 = "Hardware Error"
ASC and ASCQ values (bytes 12,13) = 03 00 = "Write fault"

(https://en.wikipedia.org/wiki/Key_Code_Qualifier)

This confirms there is some random SCSI related hardware problem or incompatibility.

[nexus]

here is a decoder http://scsi.ev-en.org/
after a reformat wipe of hdd redoing everything fresh install right after reboot into 1st 3.9boot i tried coping the workbench to another partition
4150373+26
F0 00 04 00 3F 54 65 0A 00 00 00 00 03 00 08 80 00 17
errors dont happen during the install
this time was 2 small 2GB partitions on same harddrive with another new 80 to 50 adapter just trying to rule possibilities out in my issues atm
the harddrives are also new though i messed up my 1st using warpengine scsitools now hdtoolbox sees it but cant handle the device type
so i guess that drive needs to be redone on a PC with seagates tools

though these drives have selectable even byte sector sizes from 512 to 528 bytes/sector set by firmware and im not sure if this firmware on the drive sets 512 or not
so not sure if that causes issues
thomas hddfixtool shows the correct size for 512 which is what i use