Date:	Mon, 2 Feb 2015 13:58:56 +0000
From:	Luke Kenneth Casson Leighton <lkcl@...l.net>
To:	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: 3.16.0-4-amd64 linux kernel userspace breakage

hi folks, i thought with the focus on "not breaking userspace" a
general report of two occurrences of exactly that occurring may be
appreciated.  firstly: as a long-standing linux user (and minor
contributor) i fully appreciate that this is not the "latest version"
however this is a live-running stable system (quantity 1) on which
compiling then testing absolute latest-and-greatest requires both some
thought and some scheduling for when it is non-disruptive.  your
patience and understanding therefore appreciated, taking this into
account if more info is required.

hardware: a standard (modern, 2014) 13in macbook pro with its
proprietary OS entirely removed, replaced with debian from grub-efi
boot.

last known good kernel: 3.13.

/proc and other entries: attached [dmesg is not included as it is
completely overrun with EMI-induced ATA resets/spikes]

first issue: compared to 3.13, pulseaudio is no longer functional
(does not even start up) despite applications being configured to
start it.

second issue: compared to 3.13, detection of microsd cards has
disappeared to the extent that there isn't even an entry in dmesg
indicating that the insertion ever occurred.

now, if this was just one issue (audio) i would have tolerated this
and not reported it.  however *two* separate issues means that i am
going to have to return to the older version of the linux kernel [with
a disruptive reboot where i have to terminate and lose state
information from over 40 xterms, and terminate one browser session
with over 200 tabs, 5 manually-started 3D CAD applications and shut
down one VM running PCB CAD software].

regarding the audio, the setup that i have is quite complex as it includes::

* pulseaudio to jack virtual redirection
* jack to a virtual alsa audio device with automatic "mixing"
* alsa with a virtual device as the default

the reason for the complexity is to ensure that one single ALSA
application does not dominate and take complete utter control of
audio, rendering other audio applications useless without a restart.
example: if firefox with a flash plugin gains control of the ALSA
audio device it is from then on completely impossible to use VLC...
unless firefox (with over 200 tabs) is terminated and restarted.

also adding to the complexity is the inclusion of skype, which forced
the addition of pulseaudio.

now, what *does* work on sound is the built-in audio from jackd.
firefox, vlc - all are operational.  however, with skype only having
the option to use pulseaudio, and pulseaudio no longer starting, i
have to return to 3.13 very soon.

if anyone would like to help investigate, please do ask questions and
i will happily run commands to provide more information.  however
please bear in mind that this is sufficiently disruptive to my work
that i will need to return to 3.13 (and stay there indefinitely)
within a few days at most.

l.

Download attachment "lspci" of type "application/octet-stream" (5340 bytes)

Download attachment "lsmod" of type "application/octet-stream" (5709 bytes)

Download attachment "cpuinfo" of type "application/octet-stream" (4036 bytes)

Download attachment "meminfo" of type "application/octet-stream" (1226 bytes)

Powered by blists - more mailing lists