jerky mouse still in 7.0-RELEASE

[Jim Pingle]

Kris Kennaway wrote:
> Teemu Korhonen wrote:
>> Did anyone find a solution to the "jerky mouse" -problem? It still 
>> exists in 7.0-RELEASE.
>>
>> I have pretty much exact same symptoms as in this post:
>> http://lists.freebsd.org/pipermail/freebsd-stable/2008-January/039599.html 
>>
> 
> Part of the problem here is that these "symptoms" are far too generic 
> for diagnosis and have a variety of known and unknown causes.
> 
> Some of the "known" causes include:
> 
> * Overloading the system transiently (e.g. if your KDE launches 30 
> processes at once, the system is going to be a bit sluggish for a few 
> seconds)
> 
> * Running powerd, which has poor interaction with interrupt delivery on 
> at least one user's system (might be an ACPI issue or hardware-specific).
> 
> * Performing lots of I/O to a non-mpsafe filesystem like msdosfs.
> 
> The other causes have so far resisted understanding.

I have a little more insight into the problem, at least in my case. Short 
version: it is a KVM/mouse detection issue.

I have one system that is experiencing a mouse problem, but as you said 
later in this thread it might be interrupt-related. The system in question 
is a 2xPIII-800 SMP system, running 7-STABLE from yesterday with the ULE 
scheduler. This happens to be my only 7-x box running X (so far), with a 
mouse that gets used.

The mouse hesitates at the console and in X. It will move for a half second 
or so, then stop for just as long. If I monitor processes with top or 
vmstat, it shows that moving the mouse causes the system to use ~50% CPU in 
interrupts. In X, the result is the same regardless of whether I use 
sysmouse or psm0 directly.

This system is hooked into a KVM. I found that if I booted with the system 
active on the KVM, the mouse worked fine. Also, while the mouse in a working 
state, the interrupts from the mouse never go above 3% CPU.

If I boot with another system active on the KVM instead, the mouse reverts 
to being jerky and in a state where it generates abnormally high levels of 
interrupts and hesitats.

The only difference in dmesg between the two boots is the model of the mouse 
as detected by the system:

# grep psm0 dmesg.iffymouse
psm0: <PS/2 Mouse> irq 12 on atkbdc0
psm0: [GIANT-LOCKED]
psm0: [ITHREAD]
psm0: model VersaPad, device ID 0

# grep psm0 dmesg.goodmouse
psm0: <PS/2 Mouse> irq 12 on atkbdc0
psm0: [GIANT-LOCKED]
psm0: [ITHREAD]
psm0: model NetMouse/NetScroll Optical, device ID 0

Note that in both cases, the model is incorrect compared to the actual mouse.

Jim