From: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
To: Mark Brown <broonie@kernel.org>
Cc: rohkumar@qti.qualcomm.com, alsa-devel@alsa-project.org,
bgoswami@codeaurora.org, vinod.koul@linaro.org,
linux-kernel@vger.kernel.org, plai@codeaurora.org,
tiwai@suse.com, lgirdwood@gmail.com,
Ajit Pandey <ajitp@codeaurora.org>,
Liam Girdwood <liam.r.girdwood@linux.intel.com>,
Rohit kumar <rohitkr@codeaurora.org>,
asishb@codeaurora.org, srinivas.kandagatla@linaro.org
Subject: Re: [alsa-devel] [PATCH] ASoC: soc-core: Fix null pointer dereference in soc_find_component
Date: Mon, 14 Jan 2019 21:08:15 -0600 [thread overview]
Message-ID: <163e0e01-fcf9-6f9f-4317-e71bd9cb47b1@linux.intel.com> (raw)
In-Reply-To: <20190115000610.GM11073@sirena.org.uk>
On 1/14/19 6:06 PM, Mark Brown wrote:
> On Fri, Jan 11, 2019 at 03:49:08PM -0600, Pierre-Louis Bossart wrote:
>
>> Adding some traces I can see that the the platform name we use doesn't seem
>> compatible with your logic. All the Intel boards used a constant platform
>> name matching the PCI ID, see e.g. [1], which IIRC is used to bind
>> components. Liam, do you recall in more details if this is really required?
> That's telling me that either snd_soc_find_components() isn't finding
> components in the same way that we do when we bind things which isn't
> good or we're binding links without having fully matched everything on
> the link which also isn't good.
>
> Without a system that shows the issue I can't 100% confirm but I think
> it's the former - I'm fairly sure that those machines are relying on the
> component name being initialized to fmt_single_name() in
> snd_soc_component_initialize(). That is supposed to wind up using
> dev_name() (which would be the PCI address for a PCI device) as the
> basis of the name. What I can't currently see is how exactly that gets
> bound (or how any of the other links avoid trouble for that matter). We
> could revert and push this into cards but I would rather be confident
> that we understand what's going on, I'm not comfortable that we aren't
> just pushing the breakage around rather than fixing it. Can someone
> with an x86 system take a look and confirm exactly what's going on with
> binding these cards please?
I am actually not sure at all why we need the platform_name to be set in
Intel machine drivers.
I ran a 5mn experiment with SOF and we completely ignore what is set by
machine drivers, it's overridden by the component name:
dev_info(card->dev, "info: override FE DAI link %s\n",
card->dai_link[i].name);
/* override platform component */
if (snd_soc_init_platform(card, dai_link) < 0) {
dev_err(card->dev, "init platform error");
continue;
}
pr_err("plb: platform_name was %s, now %s\n",
dai_link->platform->name,
component->name);
dai_link->platform->name = component->name;
existing machine driver (info is incorrect btw, should be BE DAI link)
[ 36.628466] bxt-pcm512x bxt-pcm512x: info: override FE DAI link
SSP5-Codec
[ 36.628469] plb: platform_name was sof-audio, now sof-audio
[ 36.628772] bxt-pcm512x bxt-pcm512x: info: override FE DAI link iDisp1
[ 36.628773] plb: platform_name was 0000:00:0e.0, now sof-audio
[ 36.629111] bxt-pcm512x bxt-pcm512x: info: override FE DAI link iDisp2
[ 36.629112] plb: platform_name was 0000:00:0e.0, now sof-audio
[ 36.629422] bxt-pcm512x bxt-pcm512x: info: override FE DAI link iDisp3
[ 36.629423] plb: platform_name was 0000:00:0e.0, now sof-audio
machine driver with all platform_name commented out, no regression at all.
[ 15.839652] sof-audio sof-audio: created machine bxt-pcm512x
[ 15.846990] bxt-pcm512x bxt-pcm512x: info: override FE DAI link
SSP5-Codec
[ 15.846995] plb: platform_name was snd-soc-dummy, now sof-audio
[ 15.847325] bxt-pcm512x bxt-pcm512x: info: override FE DAI link iDisp1
[ 15.847326] plb: platform_name was snd-soc-dummy, now sof-audio
[ 15.847657] bxt-pcm512x bxt-pcm512x: info: override FE DAI link iDisp2
[ 15.847658] plb: platform_name was snd-soc-dummy, now sof-audio
[ 15.847974] bxt-pcm512x bxt-pcm512x: info: override FE DAI link iDisp3
[ 15.847974] plb: platform_name was snd-soc-dummy, now sof-audio
I checked for a bit longer with the Skylake driver, I also couldn't see
a difference with or without the platform_name set.
diff --git a/sound/soc/soc-core.c b/sound/soc/soc-core.c
index 0462b3ec977a..0fdf99ec17cd 100644
--- a/sound/soc/soc-core.c
+++ b/sound/soc/soc-core.c
@@ -918,7 +918,7 @@ static int soc_bind_dai_link(struct snd_soc_card *card,
if (!snd_soc_is_matching_component(dai_link->platform,
component))
continue;
-
+ pr_err("plb: binding with component_name %s \n",
component->name);
snd_soc_rtdcom_add(rtd, component);
}
@@ -1041,6 +1041,8 @@ static int snd_soc_init_platform(struct
snd_soc_card *card,
if (!platform)
return -ENOMEM;
+ pr_err("plb: %s: plaform->name set to %s\n", __func__,
+ dai_link->platform_name);
dai_link->platform = platform;
platform->name = dai_link->platform_name;
platform->of_node = dai_link->platform_of_node;
[ 1.345143] plb: snd_soc_init_platform: plaform->name set to 0000:00:1f.3
[ 1.345148] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: binding
CNL Audio
[ 1.345151] plb: binding with component_name 0000:00:1f.3
[ 1.345153] plb: snd_soc_init_platform: plaform->name set to 0000:00:1f.3
[ 1.345154] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: binding
CNL HDMI1
[ 1.345155] plb: binding with component_name 0000:00:1f.3
[ 1.345157] plb: snd_soc_init_platform: plaform->name set to 0000:00:1f.3
[ 1.345158] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: binding
CNL HDMI2
[ 1.345159] plb: binding with component_name 0000:00:1f.3
[ 1.345161] plb: snd_soc_init_platform: plaform->name set to 0000:00:1f.3
[ 1.345162] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: binding
CNL HDMI3
[ 1.345163] plb: binding with component_name 0000:00:1f.3
[ 1.349607] plb: snd_soc_init_platform: plaform->name set to (null)
[ 1.349613] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: binding
CNL Audio
[ 1.349617] plb: binding with component_name snd-soc-dummy
[ 1.349619] plb: binding with component_name snd-soc-dummy
[ 1.349621] plb: snd_soc_init_platform: plaform->name set to (null)
[ 1.349623] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: binding
CNL HDMI1
[ 1.349625] plb: binding with component_name snd-soc-dummy
[ 1.349626] plb: binding with component_name snd-soc-dummy
[ 1.349628] plb: snd_soc_init_platform: plaform->name set to (null)
[ 1.349631] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: binding
CNL HDMI2
[ 1.349632] plb: binding with component_name snd-soc-dummy
[ 1.349633] plb: binding with component_name snd-soc-dummy
[ 1.349635] plb: snd_soc_init_platform: plaform->name set to (null)
[ 1.349638] skl_hda_dsp_generic skl_hda_dsp_generic: ASoC: binding
CNL HDMI3
[ 1.349639] plb: binding with component_name snd-soc-dummy
[ 1.349641] plb: binding with component_name snd-soc-dummy
Audio playback works in both cases.
The funny thing is that the list of components is right in
/sys/kernel/debug/asoc.
My guess is that the required PCI component is already added when the
platform_name is used in soc_bind_dai_link() and snd_soc_rtdcom_add()
does nothing for the back-end, so the dailink platform_name has actually
zero influence on the outcome.
Another way to look at it is that we already provide the
dai_link->cpu_dai_name which is enough for soc_bind_dai_link() to find
the relevant component and as a result the dailink->platform_name seems
redundant/unnecessary. I am using the conditional here since this part
of the code (Intel driver included) seems to work by accident rather
than by design, and we'd need a set of additional tests before removing
these hard-coded initializations.
Does this help?
next prev parent reply other threads:[~2019-01-15 3:11 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-01-11 8:14 [PATCH] ASoC: soc-core: Fix null pointer dereference in soc_find_component Rohit kumar
2019-01-11 15:44 ` [alsa-devel] " Pierre-Louis Bossart
2019-01-11 21:49 ` Pierre-Louis Bossart
2019-01-12 6:07 ` Rohit kumar
2019-01-14 15:40 ` Liam Girdwood
2019-01-15 0:06 ` Mark Brown
2019-01-15 3:08 ` Pierre-Louis Bossart [this message]
2019-01-15 19:35 ` Pierre-Louis Bossart
2019-01-15 21:07 ` Mark Brown
2019-01-15 21:11 ` Matthias Reichl
2019-01-15 21:16 ` Pierre-Louis Bossart
2019-01-15 21:41 ` Mark Brown
2019-01-15 21:48 ` Matthias Reichl
2019-01-18 23:02 ` Pierre-Louis Bossart
[not found] ` <CAOReqxjhZAzOpr-bGcV6uxPsOEid--Ym2Y0YZMHkybgZSePvtQ@mail.gmail.com>
2019-01-19 1:15 ` Curtis Malainey
2019-01-21 18:30 ` Mark Brown
2019-01-22 20:11 ` Pierre-Louis Bossart
2019-01-23 1:36 ` Curtis Malainey
2019-01-23 2:01 ` Pierre-Louis Bossart
2019-01-24 18:44 ` Mark Brown
2019-01-24 19:07 ` Pierre-Louis Bossart
2019-01-24 19:26 ` Mark Brown
2019-01-25 1:32 ` Curtis Malainey
2019-01-21 19:17 ` Mark Brown
2019-01-14 23:26 ` Mark Brown
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=163e0e01-fcf9-6f9f-4317-e71bd9cb47b1@linux.intel.com \
--to=pierre-louis.bossart@linux.intel.com \
--cc=ajitp@codeaurora.org \
--cc=alsa-devel@alsa-project.org \
--cc=asishb@codeaurora.org \
--cc=bgoswami@codeaurora.org \
--cc=broonie@kernel.org \
--cc=lgirdwood@gmail.com \
--cc=liam.r.girdwood@linux.intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=plai@codeaurora.org \
--cc=rohitkr@codeaurora.org \
--cc=rohkumar@qti.qualcomm.com \
--cc=srinivas.kandagatla@linaro.org \
--cc=tiwai@suse.com \
--cc=vinod.koul@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).