Received-SPF: pass (google.com: domain of linux-kernel+bounces-87765-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
From: songxiebing <soxiebing@163.com>
To: tiwai@suse.de
Cc: alsa-devel@alsa-project.org,
	linux-kernel@vger.kernel.org,
	songxiebing@kylinos.cn,
	soxiebing@163.com,
	tiwai@suse.com
Subject: [PATCH v2] ALSA: hda: optimize the probe codec process
Date: Fri,  1 Mar 2024 09:18:41 +0800
Message-Id: <20240301011841.7247-1-soxiebing@163.com>
In-Reply-To: <878r35huf0.wl-tiwai@suse.de>
References: <878r35huf0.wl-tiwai@suse.de>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

From: songxiebing <songxiebing@kylinos.cn>

In azx_probe_codecs function, when bus->codec_mask is becomes to 0(no codecs),
execute azx_init_chip, bus->codec_mask will be initialized to a value again,
this causes snd_hda_codec_new function to run, the process is as follows:
-->snd_hda_codec_new
-->snd_hda_codec_device_init
-->snd_hdac_device_init---snd_hdac_read_parm(...AC_PAR_VENDOR_ID) 2s
		       ---snd_hdac_read_parm(...AC_PAR_VENDOR_ID) 2s
		       ---snd_hdac_read_parm(...AC_PAR_SUBSYSTEM_ID) 2s
		       ---snd_hdac_read_parm(...AC_PAR_REV_ID) 2s
		       ---snd_hdac_read_parm(...AC_PAR_NODE_COUNT) 2s
when no codecs, read communication is error, each command will be polled for
2 second, a total of 10s, it is easy to some problem.
like this:
  2 [   14.833404][ 6] [  T164] hda 0006:00: Codec #0 probe error; disabling it...
  3 [   14.844178][ 6] [  T164] hda 0006:00: codec_mask = 0x1
  4 [   14.880532][ 6] [  T164] hda 0006:00: too slow response, last cmd=0x0f0000
  5 [   15.891988][ 6] [  T164] hda 0006:00: too slow response, last cmd=0x0f0000
  6 [   16.978090][ 6] [  T164] hda 0006:00: too slow response, last cmd=0x0f0001
  7 [   18.140895][ 6] [  T164] hda 0006:00: too slow response, last cmd=0x0f0002
  8 [   19.135516][ 6] [  T164] hda 0006:00: too slow response, last cmd=0x0f0004
 10 [   19.900086][ 6] [  T164] hda 0006:00: no codecs initialized
 11 [   45.573398][ 2] [    C2] watchdog: BUG: soft lockup - CPU#2 stuck for 22s! [kworker/2:0:25]

Here, when bus->codec_mask is 0, use a direct break to avoid execute snd_hda_codec_new function.

Signed-off-by: songxiebing <songxiebing@kylinos.cn>
---
Changes in v2: fix some non-ASCII comma letters.
---
 sound/pci/hda/hda_controller.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/sound/pci/hda/hda_controller.c b/sound/pci/hda/hda_controller.c
index e5c53ad..e1ece0a 100644
--- a/sound/pci/hda/hda_controller.c
+++ b/sound/pci/hda/hda_controller.c
@@ -1474,6 +1474,9 @@ int azx_probe_codecs(struct azx *chip, unsigned int max_slots)
 				dev_warn(chip->card->dev,
 					 "Codec #%d probe error; disabling it...\n", c);
 				bus->codec_mask &= ~(1 << c);
+				/* no codecs */
+				if (bus->codec_mask == 0)
+					break;
 				/* More badly, accessing to a non-existing
 				 * codec often screws up the controller chip,
 				 * and disturbs the further communications.
-- 
2.7.4