Hi all,
I have an HP pavilion ze4145 laptop running fedora core 1 with arjanv's
2.6.0-1.109 kernel. I have recently purchased an SMC 2532W-B 802.11b
wireless card. I am not sure that this is supported under linux, but
the problem I am having is rooted deeper than that:
when the redhat pcmcia script is run, the laptop locks up solid. sysrq
is enabled but i have not tried it yet, i'd have to look up the keys and
what they do. regardless, i have traced the problem to cardmgr itself,
cardctl works alright though it can't provide much info on the card. I
have pcmcia modularized as is most of the redhat kernel, and the modules
pcmcia_core, yenta_socket and ds are loaded. when cardmgr runs, it
locks the box up when it is inside the adjust_resources function in
cardmgr.c - at least on this 2.6.0 kernel, though the pcmcia script also
locks the box up on the 2.4 fedora kernels (2.4.22-1.2135.nptl and
2.4.22-1.2115.nptl, both fedora core 1 stock kernels).
the version of the pcmcia kernel stuff installed with fedora is 3.1.31 -
i have also installed the 3.2.7 userspace utilities (not kernel side,
the configuration process from pcmcia-cs-3.2.7 detected that pcmcia was
already enabled in the kernel and only installed the userspace stuff.)
the way in which i determined that cardmgr was at fault was by running
it by hand, simply cardmgr -v. I then traced it down further by adding
fprintfs to stderr to cardmgr.c - the specific line from which my box
never returns is
ret = ioctl(fd, DS_ADJUST_RESOURCE_INFO, &al->adj);
in adjust_resources() in cardmgr.c
the resource being adjusted is an io-range resource, and it's the second
one in the linked list that crashes my box, the first one succeeds just
fine
i also tried booting with noapic and acpi=off just to see if that had
anything to do with it, but no luck. I have not yet tried a stock
kernel.org kernel.
debugging this inside the ioctl is a bit out of my league, which is why
i have written this mail - any insight anyone else has (even if it's
just 'what the hell is wrong with you, you've screwed everything up by
doing xxx') would be much appreciated. Not being a kernel-hacker but
being a compsci major in school probably makes me too dangerous for my
own good. ;-) so if i've totally made a mess of things just tell me so.
I've attached the diff between my modified copy of cardmgr.c and the one
from pcmcia-cs-3.2.7. First diff I've ever made, so it could be wrong -
it's just fprintfs to see where it got while it was running. I'll study
up on the format of adjust_list_t and see if i can figure out exactly
which io range the code is trying to adjust and failing at.
thanks for any insights, all.
-aubin
#
# Local PCMCIA Configuration File
#
#----------------------------------------------------------------------
# System resources available for PCMCIA devices
#include port 0x100-0x4ff, port 0xc00-0xcff
include port 0xfd00-0xfdff, port 0xfc00-0xfcff
#include memory 0xc0000-0xfffff
#include memory 0xa0000000-0xa0ffffff, memory 0x60000000-0x60ffffff
include memory 0x80000000-0x80000fff, memory 0xffeff000-0xffefffff
include memory 0xfbeff000-0xffefefff, memory 0x000d7000-0x000d7fff
# High port numbers do not always work...
# include port 0x1000-0x17ff
# Extra port range for IBM Token Ring
#include port 0xa00-0xaff
# Resources we should not use, even if they appear to be available
# First built-in serial port
exclude irq 3
exclude irq 4
#exclude irq 7
exclude irq 10
exclude irq 12
exclude irq 1
exclude irq 2
exclude irq 3
exclude irq 4
exclude irq 5
exclude irq 6
exclude irq 8
exclude irq 9
exclude irq 10
exclude irq 11
exclude irq 12
exclude irq 13
exclude irq 14
exclude irq 15
# Second built-in serial port
#exclude irq 3
# First built-in parallel port
#exclude irq 7
# PS/2 Mouse controller port, comment this out if you don't have a PS/2
# based mouse
#exclude irq 12
#
#----------------------------------------------------------------------
# Examples of options for loadable modules
# To fix sluggish network with IBM ethernet adapter...
#module "pcnet_cs" opts "mem_speed=600"
# Options for IBM Token Ring adapters
#module "ibmtr_cs" opts "mmiobase=0xd0000 srambase=0xd4000"
# Options for Raylink/WebGear driver: uncomment only one line...
# Generic ad-hoc network
module "ray_cs" opts "essid=ADHOC_ESSID hop_dwell=128 beacon_period=256 translate=1"
# Infrastructure network for older cards
#module "ray_cs" opts "net_type=1 essid=ESSID1"
# Infrastructure network for WebGear
#module "ray_cs" opts "net_type=1 essid=ESSID1 translate=1 hop_dwell=128 beacon_period=256"
# Options for WaveLAN/IEEE driver (AccessPoint mode)...
#module "wvlan_cs" opts "station_name=MY_PC"
# Options for WaveLAN/IEEE driver (ad-hoc mode)...
#module "wvlan_cs" opts "port_type=3 channel=1 station_name=MY_PC"
# Options for Xircom Netwave driver...
#module "netwave_cs" opts "domain=0x100 scramble_key=0x0"
On Sat, Jan 03, 2004 at 11:53:00PM -0500, Aubin LaBrosse wrote:
> when the redhat pcmcia script is run, the laptop locks up solid. sysrq
> is enabled but i have not tried it yet, i'd have to look up the keys and
> what they do.
Alt-Sysrq-Space will give you a summary, if the system is still alive
enough to accept sysrq.
> regardless, i have traced the problem to cardmgr itself,
> cardctl works alright though it can't provide much info on the card. I
> have pcmcia modularized as is most of the redhat kernel, and the modules
> pcmcia_core, yenta_socket and ds are loaded. when cardmgr runs, it
> locks the box up when it is inside the adjust_resources function in
> cardmgr.c
I suspect it may be another case where the kernel resource subsystem
doesn't actually know the full story of which ports are in use.
Chances are you're hitting some sort of system management port which
is then causing the system to lock up.
> the way in which i determined that cardmgr was at fault was by running
> it by hand, simply cardmgr -v. I then traced it down further by adding
> fprintfs to stderr to cardmgr.c - the specific line from which my box
> never returns is
>
> ret = ioctl(fd, DS_ADJUST_RESOURCE_INFO, &al->adj);
>
> in adjust_resources() in cardmgr.c
> the resource being adjusted is an io-range resource, and it's the second
> one in the linked list that crashes my box, the first one succeeds just
> fine
You could try finding out exactly which IO port(s) are causing the
problem by tweaking your PCMCIA config.opts file, by doing a binary
search in the affected IO region, and then reporting the result here.
The output of dmidecode and lspci -vxxx may also be useful.
Note that PCMCIA performs IO probes using a granularity of 8 bytes,
so there's no point going below that.
--
Russell King
Linux kernel 2.6 ARM Linux - http://www.arm.linux.org.uk/
maintainer of: 2.6 PCMCIA - http://pcmcia.arm.linux.org.uk/
2.6 Serial core
On Sun, 2004-01-04 at 15:53, Aubin LaBrosse wrote:
> Hi all,
>
> I have an HP pavilion ze4145 laptop running fedora core 1 with arjanv's
> 2.6.0-1.109 kernel. I have recently purchased an SMC 2532W-B 802.11b
> wireless card. I am not sure that this is supported under linux, but
> the problem I am having is rooted deeper than that:
>
> when the redhat pcmcia script is run, the laptop locks up solid. sysrq
> is enabled but i have not tried it yet, i'd have to look up the keys and
> what they do. regardless, i have traced the problem to cardmgr itself,
> cardctl works alright though it can't provide much info on the card. I
> have pcmcia modularized as is most of the redhat kernel, and the modules
> pcmcia_core, yenta_socket and ds are loaded. when cardmgr runs, it
> locks the box up when it is inside the adjust_resources function in
> cardmgr.c - at least on this 2.6.0 kernel, though the pcmcia script also
> locks the box up on the 2.4 fedora kernels (2.4.22-1.2135.nptl and
> 2.4.22-1.2115.nptl, both fedora core 1 stock kernels).
>
> the version of the pcmcia kernel stuff installed with fedora is 3.1.31 -
> i have also installed the 3.2.7 userspace utilities (not kernel side,
> the configuration process from pcmcia-cs-3.2.7 detected that pcmcia was
> already enabled in the kernel and only installed the userspace stuff.)
>
> the way in which i determined that cardmgr was at fault was by running
> it by hand, simply cardmgr -v. I then traced it down further by adding
> fprintfs to stderr to cardmgr.c - the specific line from which my box
> never returns is
>
> ret = ioctl(fd, DS_ADJUST_RESOURCE_INFO, &al->adj);
>
> in adjust_resources() in cardmgr.c
> the resource being adjusted is an io-range resource, and it's the second
> one in the linked list that crashes my box, the first one succeeds just
> fine
>
> i also tried booting with noapic and acpi=off just to see if that had
> anything to do with it, but no luck. I have not yet tried a stock
> kernel.org kernel.
>
> debugging this inside the ioctl is a bit out of my league, which is why
> i have written this mail - any insight anyone else has (even if it's
> just 'what the hell is wrong with you, you've screwed everything up by
> doing xxx') would be much appreciated. Not being a kernel-hacker but
> being a compsci major in school probably makes me too dangerous for my
> own good. ;-) so if i've totally made a mess of things just tell me so.
> I've attached the diff between my modified copy of cardmgr.c and the one
> from pcmcia-cs-3.2.7. First diff I've ever made, so it could be wrong -
> it's just fprintfs to see where it got while it was running. I'll study
> up on the format of adjust_list_t and see if i can figure out exactly
> which io range the code is trying to adjust and failing at.
>
> thanks for any insights, all.
>
> -aubin
I have an HP nx9005 laptop which also hung quite badly during pcmcia
startup. I found a website with a config.opts which seems to work around
this problem. It may or may not work for you, but it might be worth a
try. This is what I used:
http://www.consultmatt.co.uk/nx9005/config.php
- Peter.
On Sun, 2004-01-11 at 17:34, Peter Lieverdink wrote:
>
> http://www.consultmatt.co.uk/nx9005/config.php
>
> - Peter.
I'll keep that one in mind as well, thanks Peter. Mathieu's reply
(earlier in this thread) fixed it for me, i used the config he put in
his mail. thanks to all of you for your help and insight!
Russell, would it be useful to you if I tracked it down more precisely
anyway? the config mathieu posted worked for me so i didn't look too
deep, but if you'd like me to try and narrow it down so that we know
which io space causes problems on these machines i could do that.
-aubin
On Sun, Jan 11, 2004 at 06:00:11PM -0500, Aubin LaBrosse wrote:
> On Sun, 2004-01-11 at 17:34, Peter Lieverdink wrote:
> > http://www.consultmatt.co.uk/nx9005/config.php
> >
> > - Peter.
>
> I'll keep that one in mind as well, thanks Peter. Mathieu's reply
> (earlier in this thread) fixed it for me, i used the config he put in
> his mail. thanks to all of you for your help and insight!
>
> Russell, would it be useful to you if I tracked it down more precisely
> anyway? the config mathieu posted worked for me so i didn't look too
> deep, but if you'd like me to try and narrow it down so that we know
> which io space causes problems on these machines i could do that.
Personally, I don't care very much - it's useful to know the affected
port range vs the machine, so when other people have the same problem
we don't have to re-investigate.
I'd rather there was some way to automatically mark these pesky regions
in the resource manager so we didn't have to do this, but I guess that's
going to be too much.
--
Russell King
Linux kernel 2.6 ARM Linux - http://www.arm.linux.org.uk/
maintainer of: 2.6 PCMCIA - http://pcmcia.arm.linux.org.uk/
2.6 Serial core