[okl4-developer] okl4 in qemu emulated gumstix environment (fixed!! but...)

Ashish Bijlani ashish.bijlani at gmail.com
Thu Aug 28 19:28:14 EST 2008 

i found the problem. linux kernel sys_opens /dev/console before
calling init in file "init/main.c". but "/etc/init.d/rcS" in rootfs
deletes /dev/console and creates a new /dev/console in an attempt to
change permissions. this breaks the link to /dev/console's inode that
was installed previsouly in init's PCB/FCB because FS will assign a
new inode to this newly created "/dev/console"  hence i couldn't see
any messages on serial once init would start. i was using okl4-2.1. it
seems latest release of okl4 viz okl4-2.1.1-patch9 fixes this issue
and code to delete /dev/console has been removed from new init.d/rcS

however, the console/shell freezes after sometime. any ideas why?
could this be because of another instance of ig_serial asi'm running
two instances of wombat?  how can i debug this?

-ashish

On Wed, Aug 27, 2008 at 2:11 AM, Geoffrey Lee <glee at ok-labs.com> wrote:
> On Wed, Aug 27, 2008 at 01:50:45AM -0400, Ashish Bijlani wrote:
>> my keystrokes are reaching ig_serial :
>>
>> NET: Registered protocol family 1
>> NET: Registered protocol family 17
>> NET: Registered protocol family 15
>> 802.1Q VLAN Support v1.8 Big_serial [ig_serial_open] :end
>> en Greear <greearb at candelatech.com>
>> All bugs added by David S. Miller <davem at redhat.com>
>> VFS: Mounted root (ext2 filesystem) readonly.
>> ig_serial [ig_serial_open] :start
>> vserial: init done (handle: 0, owner: 18001, mask: 2)
>> ig_serial [ig_serial_open] :end
>> ig_serial [ig_serial_open] :start
>> ig_serial [ig_serial_open] :end
>> ig_serial [ig_serial_open] :start
>> ig_serial [ig_serial_open] :end
>> ig_serial [ig_serial_open] :start
>> ig_serial [ig_serial_open] :end
>> ig_serial [ig_serial_write] :start
>> ig_serial [ig_serial_write] :start
>> ig_serial [ig_serial_write] :start
>> --- KD# User: breakin ---
>> > showqueue
>> Key: (X) blocked, <X> on CPU, {X} halted, !X! aborted
>>
>> [255]: (roottask)
>> [240]: (vtimer) (vrtc)
>> [200]: (event)
>> [110]: <vserial>
>> [100]: (L_timer) (L_timer)
>> [ 99]: (L_syscal) (L_syscal)
>> [ 98]: (L_1) L_1 (L_52)
>> idle : idle_thread
>>
>> > go
>> ig_serial [ig_serial_interrupt] :start
>> ig_serial [ig_serial_interrupt] :start
>>
>> what could be wrong now? should i pass console=ttyS0,115200n8 as
>> kernel command line arguments? where would /dev/console point to
>> without these args?
>
> Maybe your character is somehow not being passed down to the Linux
> tty layer.  The console= argument is not meaningful in this case,
> the driver is in vserial and ig_serial is only a client
> of vserial.
>
>>
>> i tested qemu gumstix serial emulation with pflash and linux. it seems
>> to be working fine.
>
> This is not sufficient, as it could still indicate an issue
> with qemu emulation.  The pxa serial driver used by vserial
> is written from scratch and is different to the driver
> used in native Linux.
>
>
>        -gl
>
>
>>
>> -ashish
>>
>> On Wed, Aug 27, 2008 at 12:14 AM, Ashish Bijlani
>> <ashish.bijlani at gmail.com> wrote:
>> > serial works fine on skyeye but network doesn't work because skyeye is
>> > missing smc9x card emulation which is used by gumstix...i'll debug
>> > serial handling in qemu...thanks geoff.
>> >
>> > -ashish
>> >
>> > On Wed, Aug 27, 2008 at 12:07 AM, Geoffrey Lee <glee at ok-labs.com> wrote:
>> >> On Tue, Aug 26, 2008 at 11:52:10PM -0400, Ashish Bijlani wrote:
>> >>> no i'm using the in-built ramdisk....i've not made any changes
>> >>>
>> >>
>> >> We do testing on both skyeye and hardware, so ig_serial should
>> >> be known to work.  Is it possible for you to either do your
>> >> work on skyeye or hardware?  I am guessing there might
>> >> be a possible issue with qemu's handling of the serial.
>> >>
>> >>
>> >>        -gl
>> >>
>> >>> On Tue, Aug 26, 2008 at 11:45 PM, Geoffrey Lee <glee at ok-labs.com> wrote:
>> >>> > On Tue, Aug 26, 2008 at 11:20:36PM -0400, Ashish Bijlani wrote:
>> >>> >> ig_serial fails to register in qemu emulated gumstix environemt. thats
>> >>> >> why i don't get console over ttyS0. I get the following message:
>> >>> >>
>> >>> >> kobject_add failed for ttyS0 with -EEXIST
>> >>> >
>> >>> > I doubt that's the cause of your error, we see this message
>> >>> > printed but the serial is fine for us.
>> >>> >
>> >>> >>
>> >>> >> i'm running okl4-2.1.1 with "ttyS0::askfirst:-/bin/sh" in
>> >>> >> /etc/inittab. any ideas why this error.
>> >>> >
>> >>> > Are you using the supplied ramdisk or are you using your own?
>> >>> >
>> >>> >        -gl
>> >>> >
>> >>> >>
>> >>> >> thanks,
>> >>> >> ashish
>> >>> >>
>> >>> >> On Tue, Aug 26, 2008 at 8:17 PM, Geoffrey Lee <glee at ok-labs.com> wrote:
>> >>> >> > On Tue, Aug 26, 2008 at 07:51:20AM -0400, Ashish Bijlani wrote:
>> >>> >> >> thanks for the suggestion geoff. btw, how do you emulate net when
>> >>> >> >> running on skyeye?
>> >>> >> >
>> >>> >> > I have not personally used the emulated networking support on
>> >>> >> > skyeye before, so I am not sure what's the best way to go about it.
>> >>> >> > They seem to have some README in the source tarball which documents
>> >>> >> > this however.
>> >>> >> >
>> >>> >> >> -ashish
>> >>> >> >
>> >>> >> >        -gl
>> >>> >> >
>> >>> >> >>
>> >>> >> >> On Tue, Aug 26, 2008 at 7:43 AM, Geoffrey Lee <glee at ok-labs.com> wrote:
>> >>> >> >> > On Tue, Aug 26, 2008 at 06:05:28AM -0400, Ashish Bijlani wrote:
>> >>> >> >> >> The latest code doesn't have networking for gumstix. CONFIG_SMC is
>> >>> >> >> >> missing in "linux/kernel-2.6.23-v2/l4linux_config_gumstix". Simply
>> >>> >> >> >> adding CONFIG_SMC doesn't work in the latest code base.
>> >>> >> >> >
>> >>> >> >> >
>> >>> >> >> > Generally to avoid missing dependencies and other things it is
>> >>> >> >> > best to edit the configuration file with menuconfig or similar
>> >>> >> >> > to modify the Linux configuration.
>> >>> >> >> >
>> >>> >> >> > My feeling is you may find it easier to forward port the bits required
>> >>> >> >> > to make the SMC driver work than deal with the limitations
>> >>> >> >> > of using an older OKL4 release.
>> >>> >> >> >
>> >>> >> >> >        -gl
>> >>> >> >> >
>> >>> >> >> >>
>> >>> >> >> >> -Ashish
>> >>> >> >> >>
>> >>> >> >> >> On Tue, Aug 26, 2008 at 6:02 AM, Geoffrey Lee <glee at ok-labs.com> wrote:
>> >>> >> >> >> > On Tue, Aug 26, 2008 at 05:55:28AM -0400, Ashish Bijlani wrote:
>> >>> >> >> >> >> I'm using okl4-1.4.1.1 and it doesn't seem to be having ig_serial
>> >>> >> >> >> >> under "linux/kernel/arch/l4/drivers", however it has "l4e_console". Is
>> >>> >> >> >> >> this normal? How do I enter KDB before executing /etc/init.d/rc?
>> >>> >> >> >> >
>> >>> >> >> >> > The serial driver in 1.4.1.1 is quite different to what's in the
>> >>> >> >> >> > current release, hence it is normal not to find ig_serial.c.  Best
>> >>> >> >> >> > thing you can do is trace the execution path inside your driver.
>> >>> >> >> >> >
>> >>> >> >> >> > You can manually call L4_KDB_Enter() to enter the debugger.
>> >>> >> >> >> >
>> >>> >> >> >> > By the way is there any reason you cannot use a newer version of
>> >>> >> >> >> > OKL4?
>> >>> >> >> >> >
>> >>> >> >> >> >
>> >>> >> >> >> >        -gl
>> >>> >> >> >> >
>> >>> >> >> >> >>
>> >>> >> >> >> >> Thanks,
>> >>> >> >> >> >> Ashish
>> >>> >> >> >> >>
>> >>> >> >> >> >> On Tue, Aug 26, 2008 at 5:49 AM, Geoffrey Lee <glee at ok-labs.com> wrote:
>> >>> >> >> >> >> > On Tue, Aug 26, 2008 at 05:43:51AM -0400, Ashish Bijlani wrote:
>> >>> >> >> >> >> >> Thanks for your quick reply. However, I'm using okl4-1.4 and I didn't
>> >>> >> >> >> >> >> change anything in it. I get shell using skyeye but no network as
>> >>> >> >> >> >> >> skyeye does not emulate smc91x network card. I don't get shell in
>> >>> >> >> >> >> >> qemu. The face that it runs fine on skyeye, indicates that ig_serial
>> >>> >> >> >> >> >> has been configured properly, isn't it?
>> >>> >> >> >> >> >
>> >>> >> >> >> >> > Not necessarily, the simulator may sometimes behave differently
>> >>> >> >> >> >> > to how the hardware really behaves.  Please have a look around
>> >>> >> >> >> >> > the system to see if there are any clues of misbehavior, for
>> >>> >> >> >> >> > example, are you actually receiving any interrupts at all when
>> >>> >> >> >> >> > you hit a key on the serial?  Is the serial driver initialized
>> >>> >> >> >> >> > properly, or if vserial is present on the system at all?  Is there
>> >>> >> >> >> >> > any interaction between the vserial and OK Linux?  The KDB should be
>> >>> >> >> >> >> > quite helpful here.
>> >>> >> >> >> >> >
>> >>> >> >> >> >> >
>> >>> >> >> >> >> >        -gl
>> >>> >> >> >> >> >>
>> >>> >> >> >> >> >> -Ashish
>> >>> >> >> >> >> >>
>> >>> >> >> >> >> >> On Tue, Aug 26, 2008 at 5:39 AM, Geoffrey Lee <glee at ok-labs.com> wrote:
>> >>> >> >> >> >> >> > On Tue, Aug 26, 2008 at 05:04:57AM -0400, Ashish Bijlani wrote:
>> >>> >> >> >> >> >> >> Hi,
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> I'm able to boot OKL4 in QEMU emulated gumstix environment. However,
>> >>> >> >> >> >> >> >> I'm not able to get "Linux user startup!" and following messages on
>> >>> >> >> >> >> >> >> the qemu serial console. Any ideas why?
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> This is what I did:
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> cp build/image.boot /tftpboot/.
>> >>> >> >> >> >> >> >> ../qemu-gumstix/arm-softmmu/qemu-system-arm -M gumstix -hda
>> >>> >> >> >> >> >> >> ../connex-gumstix/flash -net nic -net tap,ifname=qtap0  -m 256
>> >>> >> >> >> >> >> >> -monitor null -serial stdio -m 289 -nographic
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> GUM>
>> >>> >> >> >> >> >> >> GUM> setenv ipaddr 10.0.0.231
>> >>> >> >> >> >> >> >> GUM> setenv serverip 10.0.0.29
>> >>> >> >> >> >> >> >> GUM> tftpboot a2000000 image.boot
>> >>> >> >> >> >> >> >> GUM> bootelf a2000000
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> The OKL4 kernel, wombat, iguana everything is loaded and booted
>> >>> >> >> >> >> >> >> properly. However, I don't see any messages after :
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> NET: Registered protocol family 15
>> >>> >> >> >> >> >> >> VFS: Mounted root (ext2 filesystem) readonly.
>> >>> >> >> >> >> >> >
>> >>> >> >> >> >> >> > This probably means ig_serial.c or the user-level serial driver
>> >>> >> >> >> >> >> > in drivers/ is misconfigured.  The kernel bootup messages get
>> >>> >> >> >> >> >> > shown via the KDB console, which gets routed to the driver
>> >>> >> >> >> >> >> > in the OKL4 kernel.  Once you get to Linux user startup
>> >>> >> >> >> >> >> > output gets routed through /dev/ttyS0, which is backed by
>> >>> >> >> >> >> >> > ig_serial.
>> >>> >> >> >> >> >> >
>> >>> >> >> >> >> >> >        -gl
>> >>> >> >> >> >> >> >
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> Am I missing something?
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> Thanks,
>> >>> >> >> >> >> >> >> Ashish
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >> _______________________________________________
>> >>> >> >> >> >> >> >> Developer mailing list
>> >>> >> >> >> >> >> >> Developer at okl4.org
>> >>> >> >> >> >> >> >> https://lists.okl4.org/mailman/listinfo/developer
>> >>> >> >> >> >> >> >>
>> >>> >> >> >> >> >> >
>> >>> >> >> >> >> >> > --
>> >>> >> >> >> >> >> >
>> >>> >> >> >> >> >> >
>> >>> >> >> >> >> >>
>> >>> >> >> >> >> >
>> >>> >> >> >> >> > --
>> >>> >> >> >> >> >
>> >>> >> >> >> >> >
>> >>> >> >> >> >>
>> >>> >> >> >> >
>> >>> >> >> >> > --
>> >>> >> >> >> >
>> >>> >> >> >> >
>> >>> >> >> >>
>> >>> >> >> >
>> >>> >> >> > --
>> >>> >> >> >
>> >>> >> >> >
>> >>> >> >>
>> >>> >> >
>> >>> >> > --
>> >>> >> >
>> >>> >> >
>> >>> >>
>> >>> >
>> >>> > --
>> >>> >
>> >>> >
>> >>>
>> >>
>> >> --
>> >>
>> >>
>> >
>>
>
> --
>
>

More information about the Developer mailing list