On Tue, Nov 12, 2013 at 01:33:11PM +0100, David Henningsson wrote:

Hi,

This is a real oldie, but better late than never.

A while back, I discovered that this patch would help against a kernel panic when "cat /proc/asound/card0/codec0" was executed on this machine, which was running an fglrx driver.

I asked for a stable backport, but Takashi wanted testing to make sure this was actually fixed by the patch [1]. And now, several months later, I have this confirmation [2].

The test was conducted on an 3.5 Ubuntu kernel, based on the 3.5 extended stable tree, but would apply equally well to 3.4.

I'm attaching the backported patch - it applied cleanly, I've just added an extra section with an explanation.

Thank you David, I'll queue it for the 3.5 kernel.

Cheers, -- Luis

-- David Henningsson, Canonical Ltd. https://launchpad.net/~diwic

[1] http://mailman.alsa-project.org/pipermail/alsa-devel/2013-March/060132.html

[2] https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1212160/comments/4

From 76aabf83ca9c8cff2dd9e3bb5300e3778e0baf7d Mon Sep 17 00:00:00 2001 From: Takashi Iwai tiwai@suse.de Date: Wed, 20 Jun 2012 16:32:22 +0200 Subject: [PATCH] ALSA: hda - Move one-time init codes from generic_hdmi_init()

The codes to initialize work struct or create a proc interface should be called only once and never although it's called many times through the init callback. Move that stuff into patch_generic_hdmi() so that it's called only once.

Signed-off-by: Takashi Iwai tiwai@suse.de

(cherry picked from commit 8b8d654b55648561287bd8baca0f75f964a17038)

This is being backported to 3.4 and 3.5 due to prevention of kernel panic. The kernel panic was observed by executing 'cat /proc/asound/card0/codec0' on a particular machine, which was

• running in battery mode
• having an external HDMI monitor connected
• running the fglrx driver.

In short, the one-time init code initializes a delayed_work struct. Then, I believe the following happened:

1. the delayed_work was activated
2. the delayed_work struct was reinitialized by the init code
3. when the delayed_work timed out (ready to run the callback function), the kernel panic happened because the delayed_work struct had been corrupted in step 2.

BugLink: https://bugs.launchpad.net/bugs/1212160 Signed-off-by: David Henningsson david.henningsson@canonical.com

---

sound/pci/hda/patch_hdmi.c | 22 +++++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-)

diff --git a/sound/pci/hda/patch_hdmi.c b/sound/pci/hda/patch_hdmi.c index 02a6e3f..ef917bf 100644 --- a/sound/pci/hda/patch_hdmi.c +++ b/sound/pci/hda/patch_hdmi.c @@ -1282,23 +1282,34 @@ static int generic_hdmi_build_controls(struct hda_codec *codec) return 0; }

-static int generic_hdmi_init(struct hda_codec *codec) +static int generic_hdmi_init_per_pins(struct hda_codec *codec) { struct hdmi_spec *spec = codec->spec; int pin_idx;

for (pin_idx = 0; pin_idx < spec->num_pins; pin_idx++) { struct hdmi_spec_per_pin *per_pin = &spec->pins[pin_idx];

• hda_nid_t pin_nid = per_pin->pin_nid;
  

  struct hdmi_eld *eld = &per_pin->sink_eld;

• hdmi_init_pin(codec, pin_nid);
  

• snd_hda_jack_detect_enable(codec, pin_nid, pin_nid);
  

• per_pin->codec = codec; INIT_DELAYED_WORK(&per_pin->work, hdmi_repoll_eld); snd_hda_eld_proc_new(codec, eld, pin_idx); }

• return 0;

+}

+static int generic_hdmi_init(struct hda_codec *codec) +{

• struct hdmi_spec *spec = codec->spec;
• int pin_idx;
• for (pin_idx = 0; pin_idx < spec->num_pins; pin_idx++) {

• struct hdmi_spec_per_pin *per_pin = &spec->pins[pin_idx];
  

• hda_nid_t pin_nid = per_pin->pin_nid;
  

• hdmi_init_pin(codec, pin_nid);
  

• snd_hda_jack_detect_enable(codec, pin_nid, pin_nid);
  

• } snd_hda_jack_report_sync(codec); return 0;

} @@ -1343,6 +1354,7 @@ static int patch_generic_hdmi(struct hda_codec *codec) return -EINVAL; } codec->patch_ops = generic_hdmi_patch_ops;

• generic_hdmi_init_per_pins(codec);

  init_channel_allocations();

-- 1.7.9.5