ns558 mis-detects gameport
Hi All,
I am passing on this information at the request of Daniel Drake (Gentoo kernel ebuild maintainer).
My hardware is an ASRock K7S41GX motherboard with Athlon XP2200+ CPU running 2.6.12 on Gentoo GNU/Linux 2005.0. My gamepad is an Heroic HC 3100 2-axis, 4-button digital model with Turbo features.
The CVS version string of ns558.c is:
$Id: ns558.c,v 1.43 2002/01/24 19:23:21 vojtech Exp $
My motherboard features a generic PC/ISA gameport at BIOS-selectable addresses of 0x200 or 0x208. I have built my kernel (using Gentoo's genkernel) to include the Joystick Interface, Generic PC/ISA Gameport and Analog Joystick support as modules which are loaded at boot by coldplug/hotplug logic.
If I manually modprobe ns558 (which loads gameport), analog and joydev after boot my gameport is detected. If I let coldplug/hotplug load the modules at boot then ns558 fails to detect my gameport.
If I unload, and then reload, ns558 using coldplug/hotplug at boot then ns558 detects my gameport correctly. My module loading setup and dmesg output for a ns558 insert-remove-insert cycle are as follows:
options analog map=gamepad
above analog joydev
pre-install analog modprobe -r ns558; modprobe ns558
gameport: NS558 ISA Gameport is isa0200/gameport0, io 0x201, speed 806kHz
pnp: Device 00:0a disabled.
ns558: probe of 00:0a failed with error -16
gameport: kgameportd exiting
pnp: Device 00:0a activated.
gameport: NS558 PnP Gameport is pnp00:0a/gameport0, io 0x200, speed 806kHz
input: Analog 2-axis 4-button gamepad at pnp00:0a/gameport0 [TSC timer, 1786
MHz clock, 1299 ns res]
At https://www.redhat.com/archives/fedora-list/2005-January/msg04967.html the same problem is reported for 2.6.10 on Fedora.
Is a fix or workaround, other than what I'm doing already, available for this problem?
Regards,
Neil Darlow
Signature · 1 line
Anti-virus scanned by ClamAV-0.86.1 - http://www.clamav.net
Quoted 37 lines · Neil Darlow
Hi All,
I am passing on this information at the request of Daniel Drake (Gentoo kernel ebuild maintainer).
My hardware is an ASRock K7S41GX motherboard with Athlon XP2200+ CPU running 2.6.12 on Gentoo GNU/Linux 2005.0. My gamepad is an Heroic HC 3100 2-axis, 4-button digital model with Turbo features.
The CVS version string of
ns558.cis:
$Id: ns558.c,v 1.43 2002/01/24 19:23:21 vojtech Exp $My motherboard features a generic PC/ISA gameport at BIOS-selectable addresses of
0x200or0x208. I have built my kernel (using Gentoo's genkernel) to include the Joystick Interface, Generic PC/ISA Gameport and Analog Joystick support as modules which are loaded at boot by coldplug/hotplug logic.If I manually modprobe ns558 (which loads gameport), analog and joydev after boot my gameport is detected. If I let coldplug/hotplug load the modules at boot then ns558 fails to detect my gameport.
If I unload, and then reload, ns558 using coldplug/hotplug at boot then ns558 detects my gameport correctly. My module loading setup and dmesg output for a ns558 insert-remove-insert cycle are as follows:
options analog map=gamepad
above analog joydev
pre-install analog modprobe -r ns558; modprobe ns558gameport: NS558 ISA Gameport is isa0200/gameport0, io
0x201, speed 806kHz
pnp: Device 00:0a disabled.
ns558: probe of 00:0a failed with error -16
gameport: kgameportd exiting
pnp: Device 00:0a activated.
gameport: NS558 PnP Gameport is pnp00:0a/gameport0, io0x200, speed 806kHz
input: Analog 2-axis 4-button gamepad at pnp00:0a/gameport0 [TSC timer, 1786
Quoted 7 lines
MHz clock, 1299 ns res]At https://www.redhat.com/archives/fedora-list/2005-January/msg04967.html the same problem is reported for 2.6.10 on Fedora.
Is a fix or workaround, other than what I'm doing already, available for this problem?
ns558 first probes for legacy ISA gameports, and then for PnP gameports. The problem, as you can see above is that in the first case it finds your gameport as a legacy one, and then proceeds to probe for PnP devices. That fails and causes the legacy gameport to be disabled by the PnP subsystem.
On the second try, the gameport is disabled. The legacy probe doesn't find it, and the PnP probe succeeds, and enables it. It then works.
In the current input GIT tree there is a patch to reverse the order of probing (PnP first) for exactly this reason. I expect 2.6.13 should have the fix.
As a workaround, you can try disabling the gameport in BIOS. The legacy probe won't see it, and the PnP probe might enable it just fine.
Signature · 2 lines
Vojtech Pavlik SuSE Labs, SuSE CR
Hi Vojtech,
Quoted 3 lines · Vojtech Pavlik
In the current input GIT tree there is a patch to reverse the order of probing (PnP first) for exactly this reason. I expect 2.6.13 should have the fix.
Daniel, is it worth backporting this fix for gentoo-sources-2.6.12 so others aren't bitten or will we have to wait for 2.6.13?
Quoted 2 lines
As a workaround, you can try disabling the gameport in BIOS. The legacy probe won't see it, and the PnP probe might enable it just fine.
Disabling the gameport in my BIOS seems to take it off the bus and PnP doesn't see it at all.
Regards,
Neil Darlow
Signature · 1 line
Anti-virus scanned by ClamAV-0.86.1 - http://www.clamav.net
Quoted 11 lines · Neil Darlow
Hi Vojtech,
Quoted 3 lines · Vojtech Pavlik
In the current input GIT tree there is a patch to reverse the order of probing (PnP first) for exactly this reason. I expect 2.6.13 should have the fix.
Daniel, is it worth backporting this fix for gentoo-sources-2.6.12 so others aren't bitten or will we have to wait for 2.6.13?
Sure, but only after you have confirmed it fixes the problem for you. I've attached the patch. Please let me know how you get on.
Thanks,
Daniel
From: Vojtech Pavlik <[email protected]>
Date: Sun, 29 May 2005 07:25:01 +0000 (-0500)
Subject: Input: Probe PnP gameports first, ISA after that.
X-Git-Url: http://www.kernel.org/git/gitweb.cgi?p=linux/kernel/git/dtor/input.git;a=commitdiff;h=f6397cecadc52779902bdd8f8cd3ea5af3a19ad1
Input: Probe PnP gameports first, ISA after that.
---
drivers/input/gameport/ns558.c +6 -6
@@ -258,18 +258,18 @@ static int __init ns558_init(void) {int i =0 ; +if (pnp_register_driver (&ns558_pnp_driver) >=0 ) + pnp_registered =1 ; +/* -* Probe ISA ports first so that PnP gets to choose free port addresses -* not occupied by the ISA ports. +* Probe ISA ports after PnP, so that PnP ports that are already +* enabled get detected as PnP. This may be suboptimal in multi-device +* configurations, but saves hassle with simple setups. */ while (ns558_isa_portlist[i])ns558_isa_probe (ns558_isa_portlist[i++]); -if (pnp_register_driver (&ns558_pnp_driver) >=0 ) - pnp_registered =1 ; - -return (list_empty (&ns558_list) && !pnp_registered) ? -ENODEV :0 ; }
Quoted 38 lines · Neil Darlow
Hi All,
I am passing on this information at the request of Daniel Drake (Gentoo kernel ebuild maintainer).
My hardware is an ASRock K7S41GX motherboard with Athlon XP2200+ CPU running 2.6.12 on Gentoo GNU/Linux 2005.0. My gamepad is an Heroic HC 3100 2-axis, 4-button digital model with Turbo features.
The CVS version string of
ns558.cis:
$Id: ns558.c,v 1.43 2002/01/24 19:23:21 vojtech Exp $My motherboard features a generic PC/ISA gameport at BIOS-selectable addresses of
0x200or0x208. I have built my kernel (using Gentoo's genkernel) to include the Joystick Interface, Generic PC/ISA Gameport and Analog Joystick support as modules which are loaded at boot by coldplug/hotplug logic.If I manually modprobe ns558 (which loads gameport), analog and joydev after boot my gameport is detected. If I let coldplug/hotplug load the modules at boot then ns558 fails to detect my gameport.
If I unload, and then reload, ns558 using coldplug/hotplug at boot then ns558 detects my gameport correctly. My module loading setup and dmesg output for a ns558 insert-remove-insert cycle are as follows:
options analog map=gamepad
above analog joydev
pre-install analog modprobe -r ns558; modprobe ns558gameport: NS558 ISA Gameport is isa0200/gameport0, io
0x201, speed 806kHz
pnp: Device 00:0a disabled.
ns558: probe of 00:0a failed with error -16
gameport: kgameportd exiting
pnp: Device 00:0a activated.
gameport: NS558 PnP Gameport is pnp00:0a/gameport0, io0x200, speed 806kHz
input: Analog 2-axis 4-button gamepad at pnp00:0a/gameport0 [TSC timer, 1786
Quoted 8 lines
MHz clock, 1299 ns res]At https://www.redhat.com/archives/fedora-list/2005-January/msg04967.html the same problem is reported for 2.6.10 on Fedora.
Is a fix or workaround, other than what I'm doing already, available for this problem?
I assume this is some sort of ordering/dependency problem. I don't think we're going to get onto fixing it for 2.6.13, I'm afraid.
It would really help if you could raise a bug report at bugzilla.kernel.org so that it doesn't get forgotten. In that report, please identify the most recent kernel version whcih worked correctly, if any.
Thanks.