Sam440ep_setup breaks USB with Update 3

[ChrisH]

Thought I might as well keep this thread up-to-date, until I finish my tests: I found that USB keyboard & mouse can freeze even if I am NOT using a hub, so it just seems that a hub makes the problem much more likely.

Also, since the freezing can happen when using any app, but often seems to happen when I am typing in CygnusEd or when scrolling in MUI-OWB, I think this is due to USB having to handle "lots" of data at once. USB2 would send data even faster, so this might explain why Update 3 is so much less reliable. So I am currently testing with EHCI (aka USB2) disabled, to see if this helps.

Speculation: Since MUI-OWB (a heavy-duty app) makes the problem much more likely for me, I wonder if it is the CPU being used heavily & this is causing it to miss a USB interrupt? Or at least not respond in a timely manner to some USB messages, such that the USB stack gets confused (which might explain why un & re plugging the keyboard & mouse often does not fix the problem).

[sba]

Thought I might as well keep this thread up-to-date, until I finish my tests: I found that USB keyboard & mouse can freeze even if I am NOT using a hub, so it just seems that a hub makes the problem much more likely.

Also, since the freezing can happen when using any app, but often seems to happen when I am typing in CygnusEd or when scrolling in MUI-OWB, I think this is due to USB having to handle "lots" of data at once. USB2 would send data even faster, so this might explain why Update 3 is so much less reliable. So I am currently testing with EHCI (aka USB2) disabled, to see if this helps.

Generally, if you attach a full/low speed device (e.g., mouse) in to the ports directly (for SAM440), the ohci driver is involved. If you have a USB 2.0 hub in-between, the ehci driver is responsible for communicating with the device.

Speculation: Since MUI-OWB (a heavy-duty app) makes the problem much more likely for me, I wonder if it is the CPU being used heavily & this is causing it to miss a USB interrupt? Or at least not respond in a timely manner to some USB messages, such that the USB stack gets confused (which might explain why un & re plugging the keyboard & mouse often does not fix the problem).

To my knowledge, for devices operating within the USB specs, this shouldn't be an issue. Could you provide a new USB log in which you mark important events (mouse stop to work, reattached mouse, etc.)?

[ssolie]

Even better.. he can now open a proper bug report and attach it there...

[ChrisH]

@all
It looks like I am not the only one with this problem, as prowler reported almost identical issues (and solutions) here:
http://forum.hyperion-entertainment.biz ... 4893#p4783

@ssolie
I will certainly make a proper bug report... once I have a (hopefully) reproducible test case.

[ChrisH]

@all, ssolie, etc
It is beginning to look like ACube's "Sam440ep_setup" program is causing Update 3 to drop USB for me. Now, it would be easy to blame ACube, but as far as I am aware this is a valid speed-up. So I am more inclined to blame AmigaOS4, and will be making bug report once I am sure Sam440ep_setup is the cause.

[ChrisH]

@all
I am now fairly convinced that ACube's Sam440ep_setup is causing USB to be very unreliable with Update 3. A summary of my findings:

USB keyboard and/or mouse can stop working at any time. Un/re-plugging them usually does NOT fix the problem if they are connected via a USB hub (such as used by a KVM switch), while it occasionally does not fix the problem if no USB hub is used. This USB unreliability is made much more frequent (and so noticable) by the following:

1. The use of USB2.0 (disabling ehci.usbhcd greatly improves matters).
2. The use of MUI-OWB (but not RA-OWB!). Probably also certain other software (e.g. CygnusEd 5), but it is not clear what the common trigger is; although in most cases I am actively providing a lot of (keyboard or mouse) input to USB when it freezes.
3. The use of a USB hub may make things slightly worse.

If either of those first two is missing, then it can often take some hours before the problem occurs. But with all of those present, I can almost guarantee a problem within 30 minutes of MUI-OWB use (and sometimes much MUCH quicker - e.g. a few minutes).

EDIT: I also found that KingCON had occasionally crashing when I booted OS4. This seems to have stopped now that I have disabled Sam440ep_setup. I ought to check whether KingCON is another requirement for these problems to be more noticable...


A formal bug report will be forthcoming. My personal guess is that Sam440ep_setup is revealing some kind of race condition in USB, since Sam440ep_setup speeds-up some things. (I would therefore expect the freezing to happen very rarely without Sam440ep_setup, but it will take a lot of usage to confirm this.)

[DAX]

Could it be that USB2.0 activation (and a few other things) contributes to an overall heating of the antire system, which "eventually" leads to different kinds of malfunctionings?
As you know I had heating problems on my overclocked flex, but a small (and thus probably cluttered inside) computer case might give trouble to an EP model as well.
I'm, saying this because there are real malfuncionings which, however, do not take place on all systems, and my guess is that the lucky ones are well ventilated, OR, possibly using a Very good power supply coupled with more space (less clutter) inside.

If it is solely a kernel/library/driver problem, than it should happen to ALL, but as it happens just to some...

Just guessing.

P.S
Of course de-activating stuff would eliminate the heat as well.

[ChrisH]

@DAX
I am doubtful this is the cause in my case. I recently changed from a tiny case, to a slightly larger case with better ventillation & a completely different PSU. The problem remained unchanged.

Since I suspect some kind of race condition in USB is responsible, it would not be surprising if not everone saw it, even when using Sam440ep_setup. Although I do wonder whether the problem is somehow limited to ITX Sam440s (and not Flex Sam440s).

[xenic]

@DAX
I am doubtful this is the cause in my case. I recently changed from a tiny case, to a slightly larger case with better ventillation & a completely different PSU. The problem remained unchanged.

Since I suspect some kind of race condition in USB is responsible, it would not be surprising if not everone saw it, even when using Sam440ep_setup. Although I do wonder whether the problem is somehow limited to ITX Sam440s (and not Flex Sam440s).

Several people seemed to have problems with their SAM Flex 800 when updating to Update3. DAX solved the problem with additional cooling and I have stabalized my SAM by lowering the clock speed to 733 using HyperClock. I think DAX is wondering if the extra effeciency gained with Sam440ep_setup on your system with Update3 is pushing it to far and if CPU cooling might not solve the problem like it does on his system. If eliminating Sam440ep_setup turns out to only partially solve your problem (still get occasional problem) then CPU cooling might help. I might add that although my system is otherwise completely stable now, I still have issues using hubs and my keyboard/mouse are connected directly to the SAM USB ports now. One of my hubs doesn't work at all and the other causes my mouse not to be activated after a reboot or reset. My mouse doesn't even work through the non-functioning hub even if I plug it it after the system is up-and-running. I think part of your problem may be USB related and needs a USB fix.

[cha05e90]

Although I do wonder whether the problem is somehow limited to ITX Sam440s (and not Flex Sam440s).

If it really is a "ITX" effect, then not *all* boards are affected. My SAM440ep mini-ITX (667MHz) doesn't suffer from your described problem(s). BTW: I have still the Sam440ep_setup in my WBStartup.