On Tue, 2007-07-17 at 09:16 +1000, Daniel M. Newman wrote:
> on 16/07/07 15:33 Soeren Sonnenburg said the following:
> > On Mon, 2007-07-16 at 10:32 +1000, Daniel M. Newman wrote:
> >
> > > Using kernel 2.6.22 and the mactel-linux revision 133 patch set from svn
> > > on a Macbook Pro 15in, there seem to be a couple of peculiarities in the
> > > trackpad behaviour.
> > >
> >
> > please try the attached kernel patch and make sure that you are using a
> > new X synaptics driver (e.g. 0.14.7~git20070517 )...
> >
> >
> I tried building the 0.14.7~git20070517 driver from source. It built
> fine, but it didn't play well with my Ubuntu distribution version
> 2:1.3.0.0.dfsg-6ubuntu2, failing with:
>
> (II) Loading /usr/lib/xorg/modules/input//synaptics_drv.so
> dlopen: /usr/lib/xorg/modules/input//synaptics_drv.so: undefined symbol:
> __stack_chk_fail_local
> (EE) Failed to load /usr/lib/xorg/modules/input//synaptics_drv.so
> so I reverted to the 0.14.6-0ubuntu7 driver, which was working ok.
too bad... this newer driver really resolved a couple of issues for me.
> > > 1. During boot-up, the appletouch module always loads normally.
> > > However, as X (using the fglrx driver 8.38.6) starts up I sometimes
> > > (about one time in five) get an error that the core pointer is
> > > unavailable. When this happens, the following shows up in syslog:
> > >
> > > kernel: BUG: scheduling while atomic: Xorg/0x00000001/2826
> > > kernel: [schedule+1267/2448] schedule+0x4f3/0x990
> > > kernel: [_spin_unlock+13/48] _spin_unlock+0xd/0x30
> > > kernel: [__wake_up_common+57/112] __wake_up_common+0x39/0x70
> > > kernel: [__wake_up+56/80] __wake_up+0x38/0x50
> > > kernel: [<f89d8f81>] irqmgr_wrap_shutdown+0xe1/0x150 [fglrx]
> > > kernel: [<f89c9f45>] firegl_takedown+0x625/0xc50 [fglrx]
> > > kernel: [proc_delete_inode+0/96] proc_delete_inode+0x0/0x60
> > > kernel: [<f89c900f>] firegl_release+0x12f/0x190 [fglrx]
> > > kernel: [<f89bd44f>] ip_firegl_release+0xf/0x20 [fglrx]
> > > kernel: [__fput+151/368] __fput+0x97/0x170
> > > kernel: [filp_close+73/128] filp_close+0x49/0x80
> > > kernel: [put_files_struct+156/192] put_files_struct+0x9c/0xc0
> > > kernel: [do_exit+314/2016] do_exit+0x13a/0x7e0
> > > kernel: [<f8a14be9>] IRQMGR_WorkerThreadRoutine+0x29/0x30 [fglrx]
> > > kernel: [<f89bab90>] kasThreadRoutineHelper+0x0/0x20 [fglrx]
> > > kernel: [<f89bab90>] kasThreadRoutineHelper+0x0/0x20 [fglrx]
> > > kernel: [<f89d8d2e>] IRQMGR_CallbackWrapper+0xe/0x20 [fglrx]
> > > kernel: [<f89bab90>] kasThreadRoutineHelper+0x0/0x20 [fglrx]
> > > kernel: [kernel_thread_helper+13/20] kernel_thread_helper+0xd/0x14
> > >
> > > Removing and re-inserting appletouch generally fixes this, although
> > > sometimes I have to do the remove and reinsert twice.
> > >
> >
> > this sounds more like a bug in fglrx... but I've yet to see this to
> > happen...
> >
> >
> No, I'm fairly sure it is to do with the appletouch, since it can be
> fixed by removing and re-inserting it. I think the fglrx component of
> the failure happens after the appletouch fails to provide a pointer
> device. Also, the bug seems to have some connection to the usb
> system, since it happens much more often if the machine is booted up
> with large numbers of usb devices connected, particularly if a usb hub
> is used. Maybe there is some sort of race in the allocation of usb
> devices ?
Could be, but as I've never seen it and I am also regularly connecting
my machine to a usb hub. So it would be great if you could isolate it a
bit more...
One last idea: is appletouch really loaded before you start X ? It could
be that X just does not find the touchpad device starts/stops X
triggering a bug in fglrx...
So you could try to explicitly load the appletouch module on startup and
also it might help to use this as the appletouch device line...
Option "Device"
"/dev/input/by-id/usb-Apple_Computer_Apple_Internal_Keyboard_._Trackpad-mouse"
Any appletouch errors in dmesg ?
> > > 2. I have my screen set to blank after a period of non-usage. When
> > > waking up after this happens, the trackpad cursor no longer follows
> > > finger movements reliably. Removing and re-inserting the appletouch
> > > module fixes this, but does not re-enable response to two- and three-
> > > finger taps. Restarting gdm fixes both problems.
> > >
> >
> > that should be fixed by the patch.
> >
> >
> Your patch fixed the problem, although sometimes it seems to slightly
> affect the sensitivity of the multi-finger taps after wakeup. It also
> prevents the svn patch
> "appletouch-shut-up-when-it-has-nothing-to-say.patch" from applying
> cleanly.
well the patch I sent you already includes the
appletouch-shut-up-when-it-has-nothing-to-say.patch + small fixes. and
it definitely does not change anything regarding multi-finger taps... so
that is a heisen-bug :-)
> > > 3. Trackpad button presses sometimes do not generate an immediate
> > > response. However, if the trackpad button is pressed and no response
> > > happens, a small movement of the cursor will cause the response to
> > > occur. It seems that the button press input is being queued, but the
> > > queue is not checked until trackpad motion is detected. I thought this
> > > might be due to my xorg.conf settings, but these seem to be in
> > > accordance with the various recommendations (see below).
> > >
> >
> > strange...
> >
> > Please report back whether it still happens with the attached patch and
> > a newer synaptics xorg driver...
> >
> This doesn't seem to happen now - maybe because the
> "appletouch-shut-up-when-it-has-nothing-to-say.patch" didn't install ?
well the shut-up patch could have triggered this, but as I said my patch
included that one + a fix when the touchpad 'wakes up' so it might have
been fixed by that patch too...
Soeren
-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Mactel-linux-devel mailing list
Mactel-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mactel-linux-devel
