version 185 trap 0008

[Barry Landy]

I have now had 2 trap 8's in my T61 using version 185. Both times a memory stick was insrted in the USB 2 port.

The first trap screen showed corruption like an earlier one (under 184) that I posted; the second was not corruption and I will attach a screen shot.

I also noticed a possibly USB related entry in popuplog which I will also attach. I have put priority minor simply because it is not happening frequently.

[Barry Landy]

screenshot of trap 0008

[Lars Erdmann]

USBEHCD: readd mutex in AccNonIsoReq

[Lars Erdmann]

Did a change to USBEHCD (and other changes to USBUHCD.SYS,USBOHCD.SYS,USBEHCD.SYS).
Please try brandy.zip.
Since this is a race condition / timing issue you will need to test for quite some time.

[Barry Landy]

will do.

[Barry Landy]

One more trap 0008 with corruption. I had changed to the brandy set but not rebooted.

I was using unzip from a memoy stick (USB2) to update files on HDD.

(now rebooted of course and will keep monitoring)

[Barry Landy]

Still no trap 0008 to report

However, suspend resume no longer works.

Suspend (from the ACPI setup icon) works; but resume does not complete.

(I suspended AFTER ejecting a memory stick)

[Lars Erdmann]

Suspend/resume is known to cause problems with ACPI.PSD. There is not much I can do about it until the problems are sorted out on ACPI.PSD side.
I can not do a suspend/resume on my machine either.

[Barry Landy]

But suspend resume works ok on both earlier versions (184 AND 185)... ! so I assume that your fix to my trap 0008 did something to break resume?

[Lars Erdmann]

I haven't touched the code that is called on suspend/resume. That's all I can say.

[Barry Landy]

just a suggestion. Your last change was (I believe) to add a mutex. Could it be left locked on suspend/resume?

[Barry Landy]

Well well.

Booted OS2; did not use USB; (used CD instead); closed the lid expecting nothing to happen but it suspended. I tried resume and it worked.

Will try the opposite experiment tomorrow but that does make it look as though using a USB stick changes the suspend resume behaviour.

[Barry Landy]

later: after inserting a memory stick and writing to it, then ejecting it. lid close no longer suspended. However resume still works fine.

[Barry Landy]

One further trap 0008 using brandy.zip

[Barry Landy]

and another of the corruption type

[Lars Erdmann]

USBEHCD: readd mutex in AccNonIsoReq, misc. changes

[Lars Erdmann]

Please try brandy2.zip.
Also please add output of "pci.exe -D" so that I know your system components.
Is this a multi-core system ?

[Lars Erdmann]

Can you run 10.183 for some time ? I am not yet convinced that the trap 8 stems from the USB drivers.
Please run 10.183 for a couple of days and try to provoke the trap. It it occurs with 10.183 we know that the problem must be somewhere else.
As a second test, use the 10.185 package and only replace USBD.SYS with the one from the 10.183 package.

[Barry Landy]

Was out of touch for a couple of days. I am not sure what you would like me to do first....

re the question: it is a coreduo system but I am running with maxcpu=1 as with both cores active winos2 doesnt work.

[Barry Landy]

output of pci -d attached

[Barry Landy]

brandy.2 also failed (I got a trap 0008 first time of trying)

version 183 worked (one try).

I have noticed that the CPU meter (either with one or two cores) shows 99% most of the time for no obvious reason (both 185/185a/185b (the brandy versions) and 183)

[Barry Landy]

1) version 183 seems to work repeatedly

2) the 100% CPU is probably Firefox (so not relevant here)

[Lars Erdmann]

You will need to test 10.183 for at least a week of constant use. One day is too short to say for sure.

Also, what version of ACPI.PSD are you using ?
You might want to grab the newest version which is 3.21.02 (as always you will find it at the Mensys website via your "Software Choice" download opition).
Version 3.21.01 had fixed a problem where PCI devices where generating System Errors (where they should not). That could very well also lead to trap 8.

[Barry Landy]

I am indeed using 3.21.02. Will continue to test 183. However when I switched 185a/185b was trapping every time of trying sooner or later.

[Lars Erdmann]

USBEHCD: reverting various TD init stuff etc.

[Lars Erdmann]

Try brandy.3.zip in between. I have reverted some stuff to 10.183 level.
If that works (after at least one week of permanent use) I would like to change one other thing.

[Lars Erdmann]

Barry, can you please add a valid email address to the CC field so that you can get email notifications ?

[Lars Erdmann]

(only) USBEHCD (for now): change in timeout timer handling

[Lars Erdmann]

Try brandy.4.zip.

Please let me know if you get a trap 3. In that case, please post the trap screen.

[Barry Landy]

I have added the cc

Sorry for the delay : was travelling

I have now tried brandy.4 was seems good on the first trial (and no trap 3)

[Lars Erdmann]

You have to add a valid email address as CC and NOT your user name (unless you tell me that you successfully received this notification) !

Anyway, if brandy.4 is working ok in the next couple of days I want to add back in some things. Please let me know in a couple of days.

[Lars Erdmann]

By the way: do you have Theseus installed ? If not get it from here:
​http://www.os2site.com/sw/util/memory/thes4001.exe

Once you have installed it (you will just need to unzip the files contents) let me know.
I will then give you some instructions of what you need to do to examine the trap8.

One question in advance: are you using the IBM mouse driver or AMOUSE ? I guess you are using the IBM mouse driver, correct ?

[Lars Erdmann]

By the way: had you freshly updated ACPI.PSD when the trap8 started to show up ?

[Barry Landy]

I did not see an option to add an email address but wll look harder.

I do not have Theseus installed but will do so.

I believe it is the IBM mouse but I have gtried both in the past (in connection with getting the mouse to work after resume).

I dont THINK I had freshly updated ACPI when the trap 8 started to show up but it is quite a long time ago when I had the first one

[Barry Landy]

1) Please point to an option to add email as a cc 2) OS2site needs a login which I dont have so that's something else to investigate.

[Lars Erdmann]

The CC field is a text entry field. You can enter anything you want. If you give me an email address I will add it for you.

Did you get the email notifications or not ?

[Lars Erdmann]

USBEHCD, "SetBits","ResetBits": changed to a "conservative" implementation

[Lars Erdmann]

Forget brandy.4.zip. Go for brandy.5.zip.

I am starting to believe that you might have problems with the new ACPI.PSD version.

[Barry Landy]

OK will try brandy.5 (but brandy.4 seemed good). The OS2SITE says "ask around" to find out how to get a signon; can you tell me?

I cant see a text entry for cc only a check box : I did add my email to preferences just now will that help?

If not: bl10@…

[Lars Erdmann]

USBEHCD, "RemovePeriodicQH": backing out freeing QH

[Lars Erdmann]

Forget about brandy.5.zip. Go for brandy.6.zip

[Lars Erdmann]

You should now get notifications.

Yes, maybe it helps to add your email address to the preferences. I have not set up this TRAC system.

[Barry Landy]

Now getting notifications. Maybe it was the preferences.

optimistically trying to do more than one thing at once I had traps with both brandy.4 and .5 (using WIFI in parallel with the USB device).

Will try .6 (quite tedious because of all the reboots!! post traps)

[Barry Landy]

OK. One successful try of brandy.6 : will keep trying.

[Lars Erdmann]

I am beginning to believe that you have more than just a USB problem.
You should check which devices share IRQs with the various USB HCs.
Additionally you should consider backlevelling ACPI.PSD to some older version and observe if that gets you rid of the traps. If yes, you will then need to address this problem to the ACPI bug tracker.

[Barry Landy]

I have attached the output of rmview /irq. The sharing of 20 between e1000e (wifi) and UHCD seems suspicious. I will also try backlevelling ACPI to 3.20.03

[Barry Landy]

1) It is quite plain that the WIFI and the USB interfere with each other (eg, before the system died locked up I got a "USB device not ejected cleanly" (cant remember the exact words) when the device had not been touched bit data was going down the WIFI route. This I assume is an ACPI problem? the old version 3.20.03 was no different.

2) I think brandy.6 is OK so long as I dont use the WIFI.....

[Barry Landy]

David Azarewich suggested that the E or X centre USB widget might be responsible for the problems observed Friday. I remover the widget and that indeed turns out to be the case.

Using the latest brandy.6 remains OK

Would you like me to revert to the original 185?

[Lars Erdmann]

No, don't bother to revert back to 10.185. Just wait for 10.186. Effectively I have now backlevelled most of the things that I changed starting from 10.185.
At least you now know that you can stick to the most recent ACPI version.

Let me know if I can close this ticket.

[Lars Erdmann]

Removing the USB widget and instead using the daemon program usbmsdd.exe fixed the problem.
There was no problem with the USB HC drivers.

[Barry Landy]

Sorry could not reply earlier (Jewish New Year). Quite happy that you closed the ticket.

[Lars Erdmann]

What driver are you using for your WiFI device ? Is it fairly new ? If yes you should raise a bug against eComStation so that someone can investigate if that WiFI driver works ok or not.

[Barry Landy]

Its a Genmac driver GENM32W and not new. The e1000e is the wired ethernet port.

WIFI seems to work OK if I dont have the USB widget.

[Barry Landy]

Version 186 seems to work well for me (without the USB widget!). If I get problems I will open a new ticket.

[Lars Erdmann]

One last thing:

1) please install Theseus. See comment #27
2) start Theseus, select System->General System->Device Drivers
3) search for the line that reports "01F38" as its data segment limit (second column)
4) tell me which driver (name) it is (fifth column). If you should find more then one with that data segment limit, tell me all of them (you can search in the window with Ctrl-F / Ctrl-A)

Thanks !

[Barry Landy]

Installed Theseus 4 (thanks for sorting out the need to login to get it)

I found no drivers ending at that address.

[Barry Landy]

After spending a week trying to recreate the conditions that caused a trap (after adding a dump partition) I have succeeded by formatting the memory stick JFS (HPFS/FAT32 do not seem to provoke this problem).

Currently with version 185 of the USB software and with the USB Widget in place.

This is a non corruption dump and I can attach a mobile phone pic of the trap. CS:EIP is 07c0:000009b7

Who would like me to do what with the 2Gb dump I hope was taken?

[Barry Landy]

To catch up with stuff posted mainly to the newsgroup.

When trying with version 185 and the USB widget in place to provoke a dump I discovered another effect.

Writing a long series of writes (86 files 86Mb) from HDD to a JFS formatted USB stick I sometimes get a trap 8 (though I have not yet managed to provoke it when dumping was set up) and sometimes I get a hang of the write process (but NOT the system). In the hang state if I manually eject the stick the system gives a trap 8.

I have a dump of that and have passed it to Stephen Levine.

I will continue trying to provoke a trap 8 from the write process as it might be a different problem for a little and then return to version 196 and see if any traps occur. If I generate a trap with 186 I will start a new ticket.

[Barry Landy]

moments after posting that I succeeeded in dumping a trap 8 and will arrange to process it.