NetBSD Problem Report #12894

Received: (qmail 12092 invoked from network); 10 May 2001 12:25:02 -0000
Message-Id: <20010510122602.AE90F1110F@www.netbsd.org>
Date: Thu, 10 May 2001 05:26:02 -0700 (PDT)
From: rrnesttor@augustmail.com
Reply-To: rnestor@netbsd.org
To: gnats-bugs@gnats.netbsd.org
Subject: Early G3 and iMac USB keyboards don't work since the USB cleanup
X-Send-Pr-Version: www-1.0

>Number:         12894
>Category:       port-macppc
>Synopsis:       Early G3 and iMac USB keyboards don't work since the USB cleanup
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    port-macppc-maintainer
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu May 10 12:26:00 +0000 2001
>Closed-Date:    Thu Apr 15 03:11:36 +0000 2004
>Last-Modified:  Thu Apr 15 03:11:36 +0000 2004
>Originator:     Bob Nestor
>Release:        1.4, 1.5.1, -current (May 2001)
>Organization:
>Environment:
>Description:
Since the USB cleanup right after 1.4.2 the USB keyboard doesn't work on early model B&W G3s and iMacs. It does work in 1.4.2 and 1.4.3.  While booting the OF code notes that neither ADB nor USB ihandles were found and the system in -current now defaults to USB.  The prompts for root devices and dump device now works in -current but when the system goes multi-user the USB keyboard is dead.
>How-To-Repeat:
Attempt booting any version of NetBSD/macppc past 1.4.3 on an early model B&W G3 or iMac.
>Fix:
Unknown
>Release-Note:
>Audit-Trail:

From: Bob Nestor <rnestor@augustmail.com>
To: <gnats-bugs@netbsd.org>,
 "port-macppc" <port-macppc@NetBSD.ORG>
Cc:  
Subject: Re: port-macppc/12894
Date: Sat, 26 May 2001 11:31:34 -0500

 The USB keyboard on early model G3s and iMacs stopped working after the 
 USB code was cleaned up sometime between 1.4.2 and 1.5.  There is an open 
 PR on this problem (12894) and I have discovered some additional behavour 
 that may help isolate the source of the bug.

 Testing with -current (as of 5/25/2001) on a G3 with both ADB and USB 
 capability I have observed:

 With both ADB and USB keyboards installed the ADB is dead during boot but 
 the USB is responsive, however both are dead after the system has booted 
 to the login prompt.

 With only the ADB keyboard installed the ADB works during boot but is 
 dead after the system gets to the login prompt.  The USB keyboard is 
 usable if it is plugged in _AFTER_ the system gets to the login prompt.

 With only the USB keyboard installed the USB works during boot but is 
 dead after the system gets to the login prompt.  The ADB is also unusable 
 if it is plugged in at this time.  Unplugging the USB and plugging it 
 back in produces the expected USB events which are logged on the console, 
 but the keyboard is still unusable.

 It appears there is some undesired interaction between the ADB and USB 
 keyboard designation in the system.  Possibly some flag variable is 
 reversed?

 Hope this helps someone with a better understanding of the code.  As 
 always I'm happy to test patches if anyone has an idea.

 -bob

From: Bob Nestor <rnestor@augustmail.com>
To: <gnats-bugs@netbsd.org>,
 "port-macppc" <port-macppc@NetBSD.ORG>
Cc:  
Subject: Re: port-macppc/12894
Date: Sat, 26 May 2001 18:59:14 -0500

 I've done a bit more digging into the USB problem on early model G3s and 
 iMacs.  It appears to be a problem caused by the order devices are probed 
 and how the wscons devices are glued to physical devices. That logic 
 changed quite a bit between 1.4.3 and 1.5x/-current.

 Under 1.4.3A I see the line:
   wscons: wskbd1 glued to wsdisplay0 (console)
 this occurs after the ofb0 probe near the very end of the boot sequence.

 Under 1.5 (-current) I don't see anything similar but didn't expect to 
 since the wscons_glue logic has been totally rewritten.  What appears to 
 be missing is any line that indicates the keyboard is tied to the 
 display.  If I boot with an ADB keyboard then switch to USB after the 
 system issues the login prompt I see the line:
    wskbd1 at ukbd0: console keyboard, using wsdisplay0
 and the USB keyboard is fully usable.  Looking at the order the devices 
 are probed it appears that ofb0/wsdiplay0 was probed and configured early 
 in the boot sequence, long before the eligible keyboard devices.  I'm not 
 sure, but this looks like it is where the problem might be.  On the iMac, 
 that doesn't have an ADB, a similar thing is probably happening.

 As a side note, the change in machdep.c that added the default call to 
 "ukbd_cnattach()" if the keyboard configuration couldn't be determined 
 from OpenFirmware isn't correct for systems with an ADB keyboard. I'm not 
 sure how to find out which keyboard is present, but on a system with old 
 firmware and no USB this will not configure a working console.

 -bob

From: Bob Nestor <rnestor@augustmail.com>
To: <gnats-bugs@netbsd.org>,
 "port-macppc" <port-macppc@NetBSD.ORG>
Cc:  
Subject: Re: port-macppc/12894
Date: Mon, 28 May 2001 12:49:04 -0500

 Discovered a bit more information on this problem.  The wscons glue code 
 logic was rewritten between 1.4.3 and 1.5 and this contributes to the 
 console keyboard and display not getting glued together.  In the old code 
 it was nessary to force a reattachment by calling wsdisplay_set_cons_kbd 
 in cninit in the machdep.c file.  This was done to delay wscons 
 associations until all the parent devices were configured.

 Under 1.5 the glue logic automatically defers associations if the parent 
 device(s) aren't yet configured. The hacked call to 
 wsdisplay_set_cons_kbd in machdep.c is stepping on the information used 
 in the deferred processing. By eliminating the hack the data isn't 
 getting trashed and the new deferred processing code in wscons works 
 correctly creating the proper associations between the console keyboard 
 and display.

 This still doesn't fix the problem of being able to use the USB keyboard 
 after the system displays the login prompt though.  It appears that the 
 USB keyboard interrupts are enabled during boot as evidenced by the 
 messages "ignored interrupt while polling" when the system is asking for 
 the boot device.  However once the system comes up and prompts for user 
 login the USB keyboard is dead.  The USB doesn't respond to 
 disconnect/connect interrupts either as it does if the system is brought 
 up with an ADB keyboard.  It appears this remaining problem is due to 
 interrupts being disabled on the USB, probably due to some timing error 
 during initialization.

 Can someone verify that the call in machdep.c to wsdisplay_set_cons_kbd 
 is no longer needed and remove it?  Does anyone have any ideas on the USB 
 interrupt problem?

 Thanks,
 -bob

From: Bob Nestor <rnestor@augustmail.com>
To: gnats-bugs@netbsd.org
Cc:  
Subject: Re: port-macppc/12894
Date: Sat, 31 May 2003 18:40:52 -0500

 Testing with 1.6 and 1.6I (9/28/02 sources) shows the problem has been 
 resolved on the B&W G3 (early model).

 Attempts to test with 1.6T (5/29/03 sources) shows a problem with 
 ofwboot on the B&W G3 (default catch) so it is impossible to say the 
 problem hasn't returned on this system.

 The 12"PB (867 G4) shows the problem described in this PR with 1.6 and 
 1.6I.  Unable to test the 1.6T snapshot on this system as ofwboot threw 
 a "decrementer exception" error.

 All these tests were run booting from CD using the command:
       "boot cd:,ofwboot.xcf netbsd.macppc"


From: Evgeni Dobrev <evgeni_dobrev@developer.bg>
To: gnats-bugs@NetBSD.org
Cc:  
Subject: Re: port-macppc/12894
Date: Sat, 13 Dec 2003 22:40:14 +0100

 The problem described in this PR occurs also on the iBook G4 with 
 1.6.1.  Seems like the USB devices also do not work since the LED on my 
 mouse would not light up.

 I booted the CD with the command :
         "boot cd:,ofwboot.xcf netbsd.macppc"

State-Changed-From-To: open->feedback 
State-Changed-By: mrg 
State-Changed-When: Fri Apr 2 14:28:58 UTC 2004 
State-Changed-Why:  
does it seem that this problem is fixed in -current? 

thanks 

From: Bob Nestor <rnestor@augustmail.com>
To: gnats-bugs@gnats.netbsd.org
Cc: port-macppc-maintainer@netbsd.org, mrg@netbsd.org
Subject: Re: port-macppc/12894
Date: Fri, 2 Apr 2004 18:27:44 -0600

 On Apr 2, 2004, at 8:29 AM, mrg@netbsd.org wrote:

 > Synopsis: Early G3 and iMac USB keyboards don't work since the USB 
 > cleanup
 >
 > State-Changed-From-To: open->feedback
 > State-Changed-By: mrg
 > State-Changed-When: Fri Apr 2 14:28:58 UTC 2004
 > State-Changed-Why:
 > does it seem that this problem is fixed in -current?
 >
 > thanks

 I tried this with a 2.0A system built from sources fetched on March 28, 
 2004.  The USB keyboard still doesn't work.  Also the Openfirmware boot 
 code is reporting Default Catch errors and the IDE disk reports lost 
 interrupt errors.  The last version that had a working USB, Boot code 
 and IDE disk for this system (B&W G3) was 1.6K.

 I'd be happy to run tests and try things, but I'm not of much use in 
 tracking down Boot code, USB or IDE disk handler errors.  I'm more of 
 an applications-type coder.

 -bob

State-Changed-From-To: feedback->open 
State-Changed-By: mrg 
State-Changed-When: Sat Apr 3 08:16:14 UTC 2004 
State-Changed-Why:  
problem still exists. 

From: Bob Nestor <rnestor@augustmail.com>
To: mrg@netbsd.org
Cc: port-macppc-maintainer@netbsd.org, gnats-bugs@gnats.netbsd.org
Subject: Re: port-macppc/12894
Date: Wed, 14 Apr 2004 20:29:58 -0500

 On Apr 2, 2004, at 6:27 PM, Bob Nestor wrote:

 > On Apr 2, 2004, at 8:29 AM, mrg@netbsd.org wrote:
 >
 >> Synopsis: Early G3 and iMac USB keyboards don't work since the USB 
 >> cleanup
 >>
 >> State-Changed-From-To: open->feedback
 >> State-Changed-By: mrg
 >> State-Changed-When: Fri Apr 2 14:28:58 UTC 2004
 >> State-Changed-Why:
 >> does it seem that this problem is fixed in -current?
 >>
 >> thanks
 >
 > I tried this with a 2.0A system built from sources fetched on March 
 > 28, 2004.  The USB keyboard still doesn't work.  Also the Openfirmware 
 > boot code is reporting Default Catch errors and the IDE disk reports 
 > lost interrupt errors.  The last version that had a working USB, Boot 
 > code and IDE disk for this system (B&W G3) was 1.6K.
 >
 > I'd be happy to run tests and try things, but I'm not of much use in 
 > tracking down Boot code, USB or IDE disk handler errors.  I'm more of 
 > an applications-type coder.

 I have an updated status on this problem.  A 2.0C kernel built from 
 sources of April 13th has a working USB on my B&W G3.  I'm no longer 
 seeing the IDE lost interrupt errors and it appears I can actually 
 access the disk -- "disklabel wd0" doesn't cause an endless stream of 
 errors, I hear the disk seeking and I get printed results consistent 
 with the disk partitioning.

 Version 1.9 of ofwboot has a problem booting from CDROM, but it works 
 over the network.  I haven't tried from disk as the IDE disk was 
 non-functional under NetBSD (for me) until this build.

 All in all this is fantastic progress!  At this point I believe the PR 
 should be closed and the original problem is no longer present.

 -bob

State-Changed-From-To: open->closed 
State-Changed-By: mrg 
State-Changed-When: Thu Apr 15 03:11:27 UTC 2004 
State-Changed-Why:  
problem solved, thanks! 
>Unformatted:

Home	PR Database Search