[V2] mpc512x/clock: fix clk_get logic

Submitted by Wolfram Sang on Oct. 30, 2009, 5:53 p.m.

Details

Message ID 1256925231-21917-1-git-send-email-w.sang@pengutronix.de
State Changes Requested
Delegated to: Grant Likely
Headers show

Commit Message

Wolfram Sang Oct. 30, 2009, 5:53 p.m.
The current matching logic returns a clock even if only one out of two
arguments matches. This is wrong as devices may utilize more than one clock, so
only the first clock out of those is returned if the dev-match alone is
considered sufficent (noticed while working on the CAN driver). The proposed
new method will:

- return -EINVAL if both arguments are NULL
- skip the relevant check if one argument is NULL (first match wins)
- otherwise both arguments need to match

Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Cc: Mark Brown <broonie@sirena.org.uk>
Cc: Roel Kluin <roel.kluin@gmail.com>
Cc: Wolfgang Denk <wd@denx.de>
Cc: Grant Likely <grant.likely@secretlab.ca>
---

After Mark's valid comment, I'll try harder ;)

 arch/powerpc/platforms/512x/clock.c |   18 ++++++++++--------
 1 files changed, 10 insertions(+), 8 deletions(-)

Comments

Grant Likely Nov. 2, 2009, 5:48 p.m.
On Fri, Oct 30, 2009 at 10:53 AM, Wolfram Sang <w.sang@pengutronix.de> wrote:
> +       bool id_matched = !id;
> +       bool dev_matched = !dev;
[...]
> +                       dev_matched = true;
> +               if (id && strcmp(id, p->name) == 0)
> +                       id_matched = true;

Using bool/true/false doesn't seem to be a common pattern in the
kernel.  Anyone know what the winds of prevailing opinion are
regarding 'bool' in kernel code?

g.
Stephen Rothwell Nov. 2, 2009, 11:10 p.m.
Hi Grant,

On Mon, 2 Nov 2009 10:48:58 -0700 Grant Likely <grant.likely@secretlab.ca> wrote:
>
> Using bool/true/false doesn't seem to be a common pattern in the
> kernel.  Anyone know what the winds of prevailing opinion are
> regarding 'bool' in kernel code?

Its a good thing.
Grant Likely Nov. 2, 2009, 11:49 p.m.
On Mon, Nov 2, 2009 at 4:10 PM, Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> Hi Grant,
>
> On Mon, 2 Nov 2009 10:48:58 -0700 Grant Likely <grant.likely@secretlab.ca> wrote:
>>
>> Using bool/true/false doesn't seem to be a common pattern in the
>> kernel.  Anyone know what the winds of prevailing opinion are
>> regarding 'bool' in kernel code?
>
> Its a good thing.

Thanks,
g.

Patch hide | download patch | download mbox

diff --git a/arch/powerpc/platforms/512x/clock.c b/arch/powerpc/platforms/512x/clock.c
index 84544d0..4168457 100644
--- a/arch/powerpc/platforms/512x/clock.c
+++ b/arch/powerpc/platforms/512x/clock.c
@@ -53,19 +53,21 @@  static DEFINE_MUTEX(clocks_mutex);
 static struct clk *mpc5121_clk_get(struct device *dev, const char *id)
 {
 	struct clk *p, *clk = ERR_PTR(-ENOENT);
-	int dev_match = 0;
-	int id_match = 0;
+	/* If one argument is not given, skip its match */
+	bool id_matched = !id;
+	bool dev_matched = !dev;
 
-	if (dev == NULL || id == NULL)
-		return NULL;
+	/* We need at least one argument */
+	if (!dev && !id)
+		return ERR_PTR(-EINVAL);
 
 	mutex_lock(&clocks_mutex);
 	list_for_each_entry(p, &clocks, node) {
 		if (dev == p->dev)
-			dev_match++;
-		if (strcmp(id, p->name) == 0)
-			id_match++;
-		if ((dev_match || id_match) && try_module_get(p->owner)) {
+			dev_matched = true;
+		if (id && strcmp(id, p->name) == 0)
+			id_matched = true;
+		if (dev_matched && id_matched && try_module_get(p->owner)) {
 			clk = p;
 			break;
 		}