Re: [PATCH 2/4] pinctrl: Fix pinmux_hog_maps when ctrl_dev_name isnot set

[Tony Lindgren]

* Tony Lindgren <tony@xxxxxxxxxxx> [120120 09:06]:
> * Stephen Warren <swarren@xxxxxxxxxx> [120120 08:24]:
> > Tony Lindgren wrote at Friday, January 20, 2012 9:17 AM:
> > > The ctrl_dev_name is optional for struct pinmux_map assuming
> > > that ctrl_dev is set. Without this patch we can get:
> > > 
> > > Unable to handle kernel NULL pointer dereference at virtual address 00000000
> > 
> > > diff --git a/drivers/pinctrl/pinmux.c b/drivers/pinctrl/pinmux.c
> > 
> > 
> > > @@ -992,9 +992,12 @@ int pinmux_hog_maps(struct pinctrl_dev *pctldev)
> > > 
> > >  	for (i = 0; i < pinmux_maps_num; i++) {
> > >  		struct pinmux_map const *map = &pinmux_maps[i];
> > > +		int match_found = 0;
> > > 
> > > -		if (((map->ctrl_dev == dev) ||
> > > -		     !strcmp(map->ctrl_dev_name, devname)) &&
> > > +		if (map->ctrl_dev_name && !strcmp(map->ctrl_dev_name, devname))
> > > +				match_found = 1;
> > > +
> > > +		if (((map->ctrl_dev == dev) || match_found) &&
> > >  		    map->hog_on_boot) {
> > >  			/* OK time to hog! */
> > >  			ret = pinmux_hog_map(pctldev, map);
> > 
> > Wouldn't it be better if match_found was true for all matching cases,
> > in other words for the new code to be:
> > 
> > 		int match_found = 0;
> > 
> > 		if (map->ctrl_dev_name && !strcmp(map->ctrl_dev_name, devname))
> > 				match_found = 1;
> > 		if (map->ctrl_dev == dev)
> > 				match_found = 1;
> > 		if (match_found && map->hog_on_boot) {
> 
> OK will take a look, that would make the code easier to read.

Here's this one updated, looks like we can keep it readable
even without match_found if we continue the search early based
on !map->hog_on_boot.

Regards,

Tony


From: Tony Lindgren <tony@xxxxxxxxxxx>
Date: Fri, 20 Jan 2012 07:43:53 -0800
Subject: [PATCH] pinctrl: Fix pinmux_hog_maps when ctrl_dev_name is not set

The ctrl_dev_name is optional for struct pinmux_map assuming
that ctrl_dev is set. Without this patch we can get:

Unable to handle kernel NULL pointer dereference at virtual address 00000000
...
(pinmux_hog_maps+0xa4/0x20c)
(pinctrl_register+0x2a4/0x378)
...

Fix this by adding adding a test for map->ctrl_dev.
Additionally move the test for map->ctrl_dev earlier
to optimize out the loop a bit.

Signed-off-by: Tony Lindgren <tony@xxxxxxxxxxx>

--- a/drivers/pinctrl/pinmux.c
+++ b/drivers/pinctrl/pinmux.c
@@ -993,9 +993,12 @@ int pinmux_hog_maps(struct pinctrl_dev *pctldev)
 	for (i = 0; i < pinmux_maps_num; i++) {
 		struct pinmux_map const *map = &pinmux_maps[i];
 
-		if (((map->ctrl_dev == dev) ||
-		     !strcmp(map->ctrl_dev_name, devname)) &&
-		    map->hog_on_boot) {
+		if (!map->hog_on_boot)
+			continue;
+
+		if ((map->ctrl_dev == dev) ||
+			(map->ctrl_dev_name &&
+				!strcmp(map->ctrl_dev_name, devname))) {
 			/* OK time to hog! */
 			ret = pinmux_hog_map(pctldev, map);
 			if (ret)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/