linux-kernel - Re: [PATCH] thermal: qoriq: Only enable sites that actually exist

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <4680217.GXAFRqVoOG@pliszka>
Date:   Sat, 15 Jan 2022 22:44:43 +0100
From:   Sebastian Krzyszkowiak <sebastian.krzyszkowiak@...i.sm>
To:     linux-pm@...r.kernel.org,
        Daniel Lezcano <daniel.lezcano@...aro.org>
Cc:     linux-kernel@...r.kernel.org,
        Andrey Smirnov <andrew.smirnov@...il.com>,
        Zhang Rui <rui.zhang@...el.com>,
        Amit Kucheria <amitk@...nel.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>, kernel@...i.sm
Subject: Re: [PATCH] thermal: qoriq: Only enable sites that actually exist

Hi Daniel,

On wtorek, 30 listopada 2021 16:13:02 CET Daniel Lezcano wrote:
> Hi Sebastian,
> 
> thanks for the fix.
> 
> On 29/11/2021 12:02, Sebastian Krzyszkowiak wrote:
> > On i.MX8MQ, enabling monitoring sites that aren't connected to anything
> > can cause unwanted side effects on some units. This seems to happen
> > once some of these sites report out-of-range readings and results in
> > sensor misbehavior, such as thermal zone readings getting stuck or even
> > suddenly reporting an impossibly high value, triggering emergency
> > shutdowns.
> > 
> > The datasheet lists all non-existent sites as "reserved" and doesn't
> > make any guarantees about being able to enable them at all, so let's
> > not do that.
> 
> The description of what does the patch is missing here.

Does this appended to commit description sound good?

"Instead, iterate over sensor DT nodes and only enable monitoring sites that 
are specified there prior to registering their thermal zones."

> > Fixes: 45038e03d633 ("thermal: qoriq: Enable all sensors before
> > registering them")
> > 
> > Signed-off-by: Sebastian Krzyszkowiak <sebastian.krzyszkowiak@...i.sm>
> > ---
> > 
> >  drivers/thermal/qoriq_thermal.c | 63 ++++++++++++++++++++++-----------
> >  1 file changed, 43 insertions(+), 20 deletions(-)
> > 
> > diff --git a/drivers/thermal/qoriq_thermal.c
> > b/drivers/thermal/qoriq_thermal.c index 73049f9bea25..ef0848849ee2 100644
> > --- a/drivers/thermal/qoriq_thermal.c
> > +++ b/drivers/thermal/qoriq_thermal.c
> > @@ -32,7 +32,6 @@
> > 
> >  #define TMR_DISABLE	0x0
> >  #define TMR_ME		0x80000000
> >  #define TMR_ALPF	0x0c000000
> > 
> > -#define TMR_MSITE_ALL	GENMASK(15, 0)
> > 
> >  #define REGS_TMTMIR	0x008	/* Temperature measurement 
interval Register */
> >  #define TMTMIR_DEFAULT	0x0000000f
> > 
> > @@ -129,33 +128,51 @@ static const struct thermal_zone_of_device_ops
> > tmu_tz_ops = {> 
> >  static int qoriq_tmu_register_tmu_zone(struct device *dev,
> >  
> >  				       struct qoriq_tmu_data 
*qdata)
> >  
> >  {
> > 
> > -	int id;
> > +	int ret = 0;
> > +	struct device_node *np, *child, *sensor_np;
> > 
> > -	if (qdata->ver == TMU_VER1) {
> > -		regmap_write(qdata->regmap, REGS_TMR,
> > -			     TMR_MSITE_ALL | TMR_ME | TMR_ALPF);
> > -	} else {
> > -		regmap_write(qdata->regmap, REGS_V2_TMSR, 
TMR_MSITE_ALL);
> > -		regmap_write(qdata->regmap, REGS_TMR, TMR_ME | 
TMR_ALPF_V2);
> > -	}
> > +	np = of_find_node_by_name(NULL, "thermal-zones");
> > +	if (!np)
> > +		return -ENODEV;
> > +
> > +	sensor_np = of_node_get(dev->of_node);
> > 
> > -	for (id = 0; id < SITES_MAX; id++) {
> > +	for_each_available_child_of_node(np, child) {
> > 
> >  		struct thermal_zone_device *tzd;
> > 
> > -		struct qoriq_sensor *sensor = &qdata->sensor[id];
> > -		int ret;
> > +		struct qoriq_sensor *sensor;
> > +		int id, site;
> > +
> > +		ret = thermal_zone_of_get_sensor_id(child, sensor_np, 
&id);
> > +
> > +		if (ret < 0) {
> > +			dev_err(dev, "failed to get valid sensor id: 
%d\n", ret);
> > +			of_node_put(child);
> > +			break;
> > +		}
> > 
> > +		sensor = &qdata->sensor[id];
> > 
> >  		sensor->id = id;
> > 
> > +		/* Enable monitoring */
> > +		if (qdata->ver == TMU_VER1) {
> > +			site = 0x1 << (15 - id);
> > +			regmap_update_bits(qdata->regmap, REGS_TMR,
> > +					   site | TMR_ME | 
TMR_ALPF,
> > +					   site | TMR_ME | 
TMR_ALPF);
> > +		} else {
> > +			site = 0x1 << id;
> > +			regmap_update_bits(qdata->regmap, 
REGS_V2_TMSR, site, site);
> > +			regmap_write(qdata->regmap, REGS_TMR, TMR_ME 
| TMR_ALPF_V2);
> > +		}
> 
> Why not create the site mask in the loop and then call once the block
> above out this loop?

That's how it worked in the past. The entire point of this patch is to not do 
that - otherwise we could simply revert 45038e03d633 and bring back the issue 
it was attempting to fix (bogus data being reported prior to the first reading).

> 
> >  		tzd = devm_thermal_zone_of_sensor_register(dev, id,
> >  		
> >  							   
sensor,
> >  							   
&tmu_tz_ops);
> > 
> > -		ret = PTR_ERR_OR_ZERO(tzd);
> > -		if (ret) {
> > -			if (ret == -ENODEV)
> > -				continue;
> > -
> > -			regmap_write(qdata->regmap, REGS_TMR, 
TMR_DISABLE);
> > -			return ret;
> > +		if (IS_ERR(tzd)) {
> > +			ret = PTR_ERR(tzd);
> > +			dev_err(dev, "failed to register thermal 
zone: %d\n", ret);
> > +			of_node_put(child);
> > +			break;
> > 
> >  		}
> >  		
> >  		if (devm_thermal_add_hwmon_sysfs(tzd))
> > 
> > @@ -164,7 +181,13 @@ static int qoriq_tmu_register_tmu_zone(struct device
> > *dev,> 
> >  	}
> > 
> > -	return 0;
> > +	of_node_put(sensor_np);
> > +	of_node_put(np);
> > +
> > +	if (ret)
> > +		regmap_write(qdata->regmap, REGS_TMR, TMR_DISABLE);
> > +
> > +	return ret;
> > 
> >  }
> >  
> >  static int qoriq_tmu_calibration(struct device *dev,

Cheers,
Sebastian
Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)