regression(?) in cxusb

regression(?) in cxusb

[Vincent McIntyre]

[-- Attachment #1: Type: text/plain, Size: 3737 bytes --]

Hi

I am seeing strange behaviour with the dvb-usb-cxusb.ko module.
Things have been working fine until just recently.

Target system is Ubuntu 16.04 LTS on amd64.
I am using their 'hwe' kernel series.

Symptoms:
 * linux-image-4.15.0-58-generic + media-build from a few weeks ago works
fine
 * linux-image-4.15.0-64-generic without media-build works fine
 * linux-image-4.15.0-64-generic + media build fails

The failure:
The media-build modules compile and install just fine,
but when I try to boot the kernel with those modules,
one of the devices in the machine is not fully detected.
It seems that the dvb-usb-cxusb.ko module is not loading
and so the ir receiver is not detected.

When I try to manually modprobe the module I get a really weird result:

$ sudo modprobe -v dvb-usb-cxusb
insmod /lib/modules/4.15.0-64-generic/kernel/drivers/media/mc/mc.ko
insmod
/lib/modules/4.15.0-64-generic/kernel/drivers/media/v4l2-core/videodev.ko
insmod
/lib/modules/4.15.0-64-generic/kernel/drivers/media/common/videobuf2/videobuf2-common.ko
insmod
/lib/modules/4.15.0-64-generic/kernel/drivers/media/common/videobuf2/videobuf2-v4l2.ko
insmod
/lib/modules/4.15.0-64-generic/kernel/drivers/media/common/videobuf2/videobuf2-memops.ko
insmod
/lib/modules/4.15.0-64-generic/kernel/drivers/media/common/videobuf2/videobuf2-vmalloc.ko
insmod
/lib/modules/4.15.0-64-generic/kernel/drivers/media/v4l2-core/v4l2-common.ko
modprobe: ERROR: could not insert 'dvb_usb_cxusb': Exec format error

$ file
/lib/modules/4.15.0-64-generic/kernel/drivers/media/usb/dvb-usb/dvb-usb-cxusb.ko
/lib/modules/4.15.0-64-generic/kernel/drivers/media/usb/dvb-usb/dvb-usb-cxusb.ko:
ELF 64-bit LSB relocatable, x86-64, version 1 (SYSV),
BuildID[sha1]=9234462bd2bef10fa72a648d31af572236589e9b, not stripped


When I built the modules for 4.15.0-58 media-build HEAD was

 commit 81bdb4532e08545386d2aa3a54a806afcb5050e2
 Author: Hans Verkuil <hverkuil-cisco@xs4all.nl>
 Date:   Thu Aug 22 08:25:00 2019 +0200

    Update backports/v4.6_i2c_mux.patch

When I built the modules for 4.15.0-64, the media-build HEAD was

 commit d75b29db1297d2475227cc8bada843542271e40d
 Author: Hans Verkuil <hverkuil-cisco@xs4all.nl>
 Date:   Tue Aug 27 09:42:03 2019 +0200

    Update backports/v4.0_fwnode.patch


The media-build warning in syslog reads:
(4.15.0-58)
  Latest git patches (needed if you report a bug to
linux-media@vger.kernel.org):
    577bbf23b758848f0c4a50d346460b690c753024 media: sunxi: Add A10 CSI
driver
    3f0e6df62213ede83e3b223d36594cbcc953cb08 media: sunxi: Refactor
the Makefile and Kconfig
    c5e8f4ccd7750487f4ce05d0e801637124f14609 media: dt-bindings:
media: Add Allwinner A10 CSI binding

(4.15.0-64)
  Latest git patches (needed if you report a bug to
linux-media@vger.kernel.org):
    6f51fdfd8229d5358c2d6e272cf73478866e8ddc media: videobuf-core.c:
poll_wait needs a non-NULL buf pointer
    20a438d53fd9d12a894161bc56cbeab7a9993c39 media: sun4i: Make
sun4i_csi_formats static
    968bce2f59ce462d296af28610628fe7e03b120a media: imx: remove unused
including <linux/version.h>

I've attached the outputs from 'journalctl -b -0' for two of the three
cases above (*),
which give more detail of the boot.

I can provide a complete log of the build against 4.15.0-64 if that
would be helpful.

Kind regards
Vince

(*) Only sending two to avoid the message getting blocked for being overlong,
there seems to be a 100,000 character limit.

$ cat /etc/os-release
NAME="Ubuntu"
VERSION="16.04.6 LTS (Xenial Xerus)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu 16.04.6 LTS"
VERSION_ID="16.04"
HOME_URL="http://www.ubuntu.com/"
SUPPORT_URL="http://help.ubuntu.com/"
BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"
VERSION_CODENAME=xenial
UBUNTU_CODENAME=xenial

[-- Attachment #2: journalctl.4.15.0-64+media.txt.gz --]
[-- Type: application/x-gzip, Size: 33571 bytes --]

Re: regression(?) in cxusb

[Vincent McIntyre]

[-- Attachment #1: Type: text/plain, Size: 39 bytes --]

Sorry forgot one of the attachments...

[-- Attachment #2: journalctl.4.15.0-58+media.txt.gz --]
[-- Type: application/x-gzip, Size: 31509 bytes --]

Re: regression(?) in cxusb

[Frantisek Rysanek]

On 25 Sep 2019 at 20:31, Vincent McIntyre wrote:
> Hi
> 
> I am seeing strange behaviour with the dvb-usb-cxusb.ko module.
> Things have been working fine until just recently.
> 
> Target system is Ubuntu 16.04 LTS on amd64.
> I am using their 'hwe' kernel series.
> 
> Symptoms:
>  * linux-image-4.15.0-58-generic + media-build from a few weeks ago works
> fine
>  * linux-image-4.15.0-64-generic without media-build works fine
>  * linux-image-4.15.0-64-generic + media build fails
...
> insmod
> /lib/modules/4.15.0-64-generic/kernel/drivers/media/v4l2-core/v4l2-common.ko
> modprobe: ERROR: could not insert 'dvb_usb_cxusb': Exec format error
> 
Dear Vincent, I'm not a maitainer, just a fellow user who happens to 
keep an eye on this list.
Excuse my somewhat off topic response:

Would you consider compiling a fresh kernel?
What hardware are you trying to use with the cxusb driver?

The linux-media subsystem is pretty vibrant and fast-paced,
fairly serious changes happen between vanilla kernel releases.
In that context, 4.15 sounds oldish...

In the recent past, I recall support for the Mygica "T230C v2" (aka 
T230C2) moving from cxusb.c to dvbsky.c, plus a debate about
one or two unrelated fixes to that code (maybe one of them was a 
revert of a previous patch). I recall that one of the motivations for 
T230C2 to move from cxusb.c to dvbsky.c was better support for its IR 
sensor.
I believe this stuff will be present in Linux 5.4 (just pulled into 
the vanilla master, within the merge window that's just now open), 
possibly hasn't made it into 5.3.
Meanwhile, the directory structure of the source code keeps receiving 
"janitoring" changes etc...
=> I suggest that you test some newer kernel, possibly betatest some
5.4-pre/rc to see if those make a difference (if you cannot wait a 
few weeks for 5.4.0 release). Or try some 5.2 / 5.3 if you believe 
that the recent changes to cxusb/dvbsky are irrelevant to your HW.

Frank Rysanek

Re: regression(?) in cxusb

[JP]

On 9/25/19 8:19 PM, Frantisek Rysanek wrote:
> On 25 Sep 2019 at 20:31, Vincent McIntyre wrote:
>> Hi
>>
>> I am seeing strange behaviour with the dvb-usb-cxusb.ko module.
>> Things have been working fine until just recently.
>>
>> Target system is Ubuntu 16.04 LTS on amd64.
>> I am using their 'hwe' kernel series.
>>
>> Symptoms:
>>   * linux-image-4.15.0-58-generic + media-build from a few weeks ago works
>> fine
>>   * linux-image-4.15.0-64-generic without media-build works fine
>>   * linux-image-4.15.0-64-generic + media build fails
> ...
>> insmod
>> /lib/modules/4.15.0-64-generic/kernel/drivers/media/v4l2-core/v4l2-common.ko
>> modprobe: ERROR: could not insert 'dvb_usb_cxusb': Exec format error
>>
> Dear Vincent, I'm not a maitainer, just a fellow user who happens to
> keep an eye on this list.
> Excuse my somewhat off topic response:
>
> Would you consider compiling a fresh kernel?
> What hardware are you trying to use with the cxusb driver?
>
> The linux-media subsystem is pretty vibrant and fast-paced,
> fairly serious changes happen between vanilla kernel releases.
> In that context, 4.15 sounds oldish...
>
> In the recent past, I recall support for the Mygica "T230C v2" (aka
> T230C2) moving from cxusb.c to dvbsky.c, plus a debate about
> one or two unrelated fixes to that code (maybe one of them was a
> revert of a previous patch). I recall that one of the motivations for
> T230C2 to move from cxusb.c to dvbsky.c was better support for its IR
> sensor.
> I believe this stuff will be present in Linux 5.4 (just pulled into
> the vanilla master, within the merge window that's just now open),
> possibly hasn't made it into 5.3.
> Meanwhile, the directory structure of the source code keeps receiving
> "janitoring" changes etc...
> => I suggest that you test some newer kernel, possibly betatest some
> 5.4-pre/rc to see if those make a difference (if you cannot wait a
> few weeks for 5.4.0 release). Or try some 5.2 / 5.3 if you believe
> that the recent changes to cxusb/dvbsky are irrelevant to your HW.
>
> Frank Rysanek
>
>
Hi Vincent, Frank,

  Could be Ubuntu specific or quite possibly some corruption on local 
system.

Jan Pieter.