Skip to contentSkip to conversation

PM / OPP: Add clock-latency-ns support

DISCUSSION Dan Carpenter 3 participants 6 messages Aug 10, 2015 – Aug 12, 2015

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

Hello Viresh Kumar,

The patch 3ca9bb33c627: "PM / OPP: Add clock-latency-ns support" from Jul 29, 2015, leads to the following static checker warning:

	drivers/base/power/opp.c:949 _opp_add_static_v2()
	warn: passing casted pointer '&new_opp->clock_latency_ns' to 'of_property_read_u32()' 64 vs 32.

drivers/base/power/opp.c

   946          new_opp->np = np;
   947          new_opp->dynamic = false;
   948          new_opp->available = true;
   949          of_property_read_u32(np, "clock-latency-ns",
   950                               (u32 *)&new_opp->clock_latency_ns);
                                            ^^^^^^^^^^^^^^^^^^^^^^^^^^

This code will break on 64 bit, big endian machines. I doin't know if that is an issue for this driver. I saw this was in the power/ directory and I spent a minute googling to see if PowerPC machines are big endian... :P

   951  
   952          ret = opp_get_microvolt(new_opp, dev);
   953          if (ret)
   954                  goto free_opp;
   955  

Also:

	drivers/base/power/opp.c:956 _opp_add_static_v2()
	warn: passing casted pointer '&new_opp->u_amp' to 'of_property_read_u32()' 64 vs 32.

regards,
dan carpenter

#1 lore raw
Quoted 27 lines · Dan Carpenter

Hello Viresh Kumar,

The patch 3ca9bb33c627: "PM / OPP: Add clock-latency-ns support" from Jul 29, 2015, leads to the following static checker warning:

	drivers/base/power/opp.c:949 _opp_add_static_v2()
	warn: passing casted pointer '&new_opp->clock_latency_ns' to 'of_property_read_u32()' 64 vs 32.

drivers/base/power/opp.c

   946          new_opp->np = np;
   947          new_opp->dynamic = false;
   948          new_opp->available = true;
   949          of_property_read_u32(np, "clock-latency-ns",
   950                               (u32 *)&new_opp->clock_latency_ns);
                                            ^^^^^^^^^^^^^^^^^^^^^^^^^^

This code will break on 64 bit, big endian machines. I doin't know if that is an issue for this driver. I saw this was in the power/ directory and I spent a minute googling to see if PowerPC machines are big endian... :P

   951  
   952          ret = opp_get_microvolt(new_opp, dev);
   953          if (ret)
   954                  goto free_opp;
   955  
Quoted 4 lines

Also:

	drivers/base/power/opp.c:956 _opp_add_static_v2()
	warn: passing casted pointer '&new_opp->u_amp' to 'of_property_read_u32()' 64 vs 32.

Hi Dan,

The problem is that the value here is of type 'unsigned long' which is 32/64 bit on 32/64 bit machines.

So, I looked at how other places in code has done it, and that's what I found.

Signature · 1 line
viresh
#2 lore raw
Quoted 30 lines · Viresh Kumar
Quoted 29 lines · Dan Carpenter

Hello Viresh Kumar,

The patch 3ca9bb33c627: "PM / OPP: Add clock-latency-ns support" from Jul 29, 2015, leads to the following static checker warning:

	drivers/base/power/opp.c:949 _opp_add_static_v2()
	warn: passing casted pointer '&new_opp->clock_latency_ns' to 'of_property_read_u32()' 64 vs 32.

drivers/base/power/opp.c

   946          new_opp->np = np;
   947          new_opp->dynamic = false;
   948          new_opp->available = true;
   949          of_property_read_u32(np, "clock-latency-ns",
   950                               (u32 *)&new_opp->clock_latency_ns);
                                            ^^^^^^^^^^^^^^^^^^^^^^^^^^

This code will break on 64 bit, big endian machines. I doin't know if that is an issue for this driver. I saw this was in the power/ directory and I spent a minute googling to see if PowerPC machines are big endian... :P

   951  
   952          ret = opp_get_microvolt(new_opp, dev);
   953          if (ret)
   954                  goto free_opp;
   955  

Also:

Quoted 7 lines
Quoted 2 lines
	drivers/base/power/opp.c:956 _opp_add_static_v2()
	warn: passing casted pointer '&new_opp->u_amp' to 'of_property_read_u32()' 64 vs 32.

Hi Dan,

The problem is that the value here is of type 'unsigned long' which is 32/64 bit on 32/64 bit machines.

Yep. It won't work on 64 bit big endian machines as described earlier.

Quoted 3 lines

So, I looked at how other places in code has done it, and that's what I found.

It's not portable. Sometimes we don't care about that because we know we don't care about 64 bit big endian systems. Hence, my email.

regards,
dan carpenter

#3 lore raw
Quoted 5 lines · Dan Carpenter
Quoted 3 lines · Viresh Kumar

The problem is that the value here is of type 'unsigned long' which is 32/64 bit on 32/64 bit machines.

Quoted 9 lines

Yep. It won't work on 64 bit big endian machines as described earlier.

Quoted 3 lines

So, I looked at how other places in code has done it, and that's what I found.

It's not portable. Sometimes we don't care about that because we know we don't care about 64 bit big endian systems. Hence, my email.

Making it portable should be simple enough by having a temporary variable of type u32 though.

---8<---

drivers/base/power/opp.c +5 -3
index 51b220e615d3..022715d1894c 100644
@@ -919,6 +919,7 @@ static int _opp_add_static_v2(struct device *dev, struct device_node *np)
 	struct dev_pm_opp *new_opp;
 	u64 rate;
 	int ret;
+	u32 val;
 
 	/* Hold our list modification lock here */
 	mutex_lock(&dev_opp_list_lock);
@@ -946,14 +947,15 @@ static int _opp_add_static_v2(struct device *dev, struct device_node *np)
 	new_opp->np = np;
 	new_opp->dynamic = false;
 	new_opp->available = true;
-	of_property_read_u32(np, "clock-latency-ns",
-			     (u32 *)&new_opp->clock_latency_ns);
+	of_property_read_u32(np, "clock-latency-ns", &val);
+	new_opp->clock_latency_ns = val;
 
 	ret = opp_get_microvolt(new_opp, dev);
 	if (ret)
 		goto free_opp;
 
-	of_property_read_u32(np, "opp-microamp", (u32 *)&new_opp->u_amp);
+	of_property_read_u32(np, "opp-microamp", &val);
+	new_opp->u_amp = val;
 
 	ret = _opp_add(dev, new_opp, dev_opp);
 	if (ret)
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#4 lore raw
Quoted 2 lines · Stephen Boyd

Making it portable should be simple enough by having a temporary variable of type u32 though.

Right.

@Dan: Does this look fine to you?

Message-Id: <40b3ad3c99c5fe1c50d997ba5418dd602c673f13.1439362565.git.viresh.kumar@linaro.org>
From: Viresh Kumar <[email protected]>
Date: Wed, 12 Aug 2015 12:20:49 +0530
Subject: [PATCH] PM / OPP: Fix static checker warning (broken 64bit big endian
systems)

Dan Carpenter reported (generated with static checker):

drivers/base/power/opp.c:949 _opp_add_static_v2()
warn: passing casted pointer '&new_opp->clock_latency_ns' to 'of_property_read_u32()' 64 vs 32.

This code will break on 64 bit, big endian machines.

Fix this by reading the value in a u32 type variable first and then assigning it to the unsigned long variable.

---
 drivers/base/power/opp.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/base/power/opp.c +5 -3
index 204c6c945168..a9e0af0dd9e5 100644
@@ -918,6 +918,7 @@ static int _opp_add_static_v2(struct device *dev, struct device_node *np)
 	struct device_opp *dev_opp;
 	struct dev_pm_opp *new_opp;
 	u64 rate;
+	u32 val;
 	int ret;
 
 	/* Hold our list modification lock here */
@@ -946,14 +947,15 @@ static int _opp_add_static_v2(struct device *dev, struct device_node *np)
 	new_opp->np = np;
 	new_opp->dynamic = false;
 	new_opp->available = true;
-	of_property_read_u32(np, "clock-latency-ns",
-			     (u32 *)&new_opp->clock_latency_ns);
+	of_property_read_u32(np, "clock-latency-ns", &val);
+	new_opp->clock_latency_ns = val;
 
 	ret = opp_get_microvolt(new_opp, dev);
 	if (ret)
 		goto free_opp;
 
-	of_property_read_u32(np, "opp-microamp", (u32 *)&new_opp->u_amp);
+	of_property_read_u32(np, "opp-microamp", &val);
+	new_opp->u_amp = val;
 
 	ret = _opp_add(dev, new_opp, dev_opp);
 	if (ret)
#5 lore raw
Quoted 7 lines · Viresh Kumar
Quoted 2 lines · Stephen Boyd

Making it portable should be simple enough by having a temporary variable of type u32 though.

Right.

@Dan: Does this look fine to you?

Looks good.

regards,
dan carpenter