Bug#498293: linux-image-2.6.26-1-686: mos7840 USB serial driver fails to work and causes kernel OOPS with 2.6.26-4 - Debian

This is a discussion on Bug#498293: linux-image-2.6.26-1-686: mos7840 USB serial driver fails to work and causes kernel OOPS with 2.6.26-4 - Debian ; Package: linux-image-2.6.26-1-686 Version: 2.6.26-4 Severity: serious Justification: lenny/rc-policy I have a USB "Express" twin port serial adaptor, which identifies itself as USB ID 9710:7840 MosChip Semiconductor. This uses the mos7840 Moschip 7840/7820 USB Serial Driver. On linux-image-2.6.26-1-686 version 2.6.26-3 this ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: Bug#498293: linux-image-2.6.26-1-686: mos7840 USB serial driver fails to work and causes kernel OOPS with 2.6.26-4

  1. Bug#498293: linux-image-2.6.26-1-686: mos7840 USB serial driver fails to work and causes kernel OOPS with 2.6.26-4

    Package: linux-image-2.6.26-1-686
    Version: 2.6.26-4
    Severity: serious
    Justification: lenny/rc-policy

    I have a USB "Express" twin port serial adaptor, which identifies itself
    as USB ID 9710:7840 MosChip Semiconductor. This uses the mos7840 Moschip
    7840/7820 USB Serial Driver.

    On linux-image-2.6.26-1-686 version 2.6.26-3 this adaptor worked fine
    and neither of the following problems occurred.

    On linux-image-2.6.26-1-686 version 2.6.26-4 there are serious problems.

    First, the usb-serial ports do not work correctly. Whatever character
    I send to the tty, it emits the letter 'O' from the serial port.
    See the two strace attachments, which are from kernels 2.6.26-3 and
    2.6.26-4 respectively. They were generated using minicom talking to a
    serial device which just echoes the input and sends the prompt "Ready"
    when it receives a carriage return.

    Unfortunately I didn't type the same characters in each test, but
    if you diff the straces you can see that the tiocmget behaviour is
    substantially changed in what it returns, in addition to the incorrect
    character being sent.

    Secondly, on unplugging the adaptor, a kernel OOPS is caused due to a
    null pointer dereference. See dmesg extract also attached.

    Following this OOPS, lsusb just hangs and is not interruptible with
    control-C, and on attempting to shut down the machine the shutdown process
    hangs when closing down Bluetooth (the built-in Bluetooth adaptor also
    presents a USB interface). Magic sysrq key still works fortunately !

    A different usb-serial device using the ftdi-sio driver still works
    without error.

    I see the following in the package changelog and wonder if this change
    is related to the destabilisation ?

    linux-2.6 (2.6.26-4) unstable; urgency=low
    * Add stable release 2.6.26.3:
    - usb-serial: don't release unregistered minors

    Assigning this as "serious" in relation to Lenny rc policy, due to the
    fact that this kernel null pointer dereference seems to break the entire
    USB subsystem. Please reassign if this is inappropriate, thanks.

    Nick


  2. Bug#498293: linux-image-2.6.26-1-686: mos7840 USB serial driver fails to work and causes kernel OOPS with 2.6.26-4

    On Mon, Sep 08, 2008 at 08:38:20PM +0100, Nick Leverton wrote:
    > Package: linux-image-2.6.26-1-686
    > Version: 2.6.26-4
    > Severity: serious
    > Justification: lenny/rc-policy
    >
    > I have a USB "Express" twin port serial adaptor, which identifies itself
    > as USB ID 9710:7840 MosChip Semiconductor. This uses the mos7840 Moschip
    > 7840/7820 USB Serial Driver.
    >
    > On linux-image-2.6.26-1-686 version 2.6.26-3 this adaptor worked fine
    > and neither of the following problems occurred.
    >
    > On linux-image-2.6.26-1-686 version 2.6.26-4 there are serious problems.
    >
    > First, the usb-serial ports do not work correctly. Whatever character
    > I send to the tty, it emits the letter 'O' from the serial port.
    > See the two strace attachments, which are from kernels 2.6.26-3 and
    > 2.6.26-4 respectively. They were generated using minicom talking to a
    > serial device which just echoes the input and sends the prompt "Ready"
    > when it receives a carriage return.
    >
    > Unfortunately I didn't type the same characters in each test, but
    > if you diff the straces you can see that the tiocmget behaviour is
    > substantially changed in what it returns, in addition to the incorrect
    > character being sent.
    >
    > Secondly, on unplugging the adaptor, a kernel OOPS is caused due to a
    > null pointer dereference. See dmesg extract also attached.
    >
    > Following this OOPS, lsusb just hangs and is not interruptible with
    > control-C, and on attempting to shut down the machine the shutdown process
    > hangs when closing down Bluetooth (the built-in Bluetooth adaptor also
    > presents a USB interface). Magic sysrq key still works fortunately !
    >
    > A different usb-serial device using the ftdi-sio driver still works
    > without error.
    >
    > I see the following in the package changelog and wonder if this change
    > is related to the destabilisation ?
    >
    > linux-2.6 (2.6.26-4) unstable; urgency=low
    > * Add stable release 2.6.26.3:
    > - usb-serial: don't release unregistered minors


    The commit in question is
    http://git.kernel.org/?p=linux/kerne...18975d1562ab83

    Please try to a kernel build with that fix reverted. If it fixes
    the problem for you, please pass your debug data to Greg Kroah-Hartman
    , who's both one of the maintainers of the 2.6.26.x stable
    releases and the Linux USB maintainer.

    Cheers,
    Moritz



    --
    To UNSUBSCRIBE, email to debian-bugs-dist-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

+ Reply to Thread