[PATCH] i.MX31: mxc-rnga: implement waiting for data in driver

[PATCH] i.MX31: mxc-rnga: implement waiting for data in driver

[Michael Thalmeier]

mxc_rnga_data_present dit not use the wait parameter which caused a build
warning about initialization from incompatible pointer type.

Apart from this warning it is cleaner to implement the interface as intendet
and wait in the driver until data is available or a timeout has occured.

Signed-off-by: Michael Thalmeier <michael.thalmeier@hale.at>
---
 drivers/char/hw_random/mxc-rnga.c |   16 +++++++++++-----
 1 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/char/hw_random/mxc-rnga.c b/drivers/char/hw_random/mxc-rnga.c
index 187c6be..39fb446 100644
--- a/drivers/char/hw_random/mxc-rnga.c
+++ b/drivers/char/hw_random/mxc-rnga.c
@@ -25,6 +25,7 @@
 #include <linux/platform_device.h>
 #include <linux/hw_random.h>
 #include <linux/io.h>
+#include <linux/delay.h>
 
 /* RNGA Registers */
 #define RNGA_CONTROL			0x00
@@ -60,14 +61,19 @@
 
 static struct platform_device *rng_dev;
 
-static int mxc_rnga_data_present(struct hwrng *rng)
+static int mxc_rnga_data_present(struct hwrng *rng, int wait)
 {
-	int level;
+	int level, i;
 	void __iomem *rng_base = (void __iomem *)rng->priv;
 
-	/* how many random numbers is in FIFO? [0-16] */
-	level = ((__raw_readl(rng_base + RNGA_STATUS) &
-			RNGA_STATUS_LEVEL_MASK) >> 8);
+	for (i = 0; i < 20; i++) {
+		/* how many random numbers is in FIFO? [0-16] */
+		level = ((__raw_readl(rng_base + RNGA_STATUS) &
+				RNGA_STATUS_LEVEL_MASK) >> 8);
+		if (level || !wait)
+			break;
+		udelay(10);
+	}
 
 	return level > 0 ? 1 : 0;
 }
-- 
1.7.7.6



--
Scanned by MailScanner.

[PATCH] i.MX31: mxc-rnga: implement waiting for data in driver

[Fabio Estevam]

On Tue, Feb 21, 2012 at 10:13 AM, Michael Thalmeier
<michael.thalmeier@hale.at> wrote:

> -static int mxc_rnga_data_present(struct hwrng *rng)
> +static int mxc_rnga_data_present(struct hwrng *rng, int wait)

This looks good, but ...

> ?{
> - ? ? ? int level;
> + ? ? ? int level, i;
> ? ? ? ?void __iomem *rng_base = (void __iomem *)rng->priv;
>
> - ? ? ? /* how many random numbers is in FIFO? [0-16] */
> - ? ? ? level = ((__raw_readl(rng_base + RNGA_STATUS) &
> - ? ? ? ? ? ? ? ? ? ? ? RNGA_STATUS_LEVEL_MASK) >> 8);
> + ? ? ? for (i = 0; i < 20; i++) {

Why the magic "20" here?

It would be better to add a proper timeout mechanism instead, such as
time_after(jiffies, timeout)

[PATCH] i.MX31: mxc-rnga: implement waiting for data in driver

[Michael Thalmeier]

On 2012-02-21 14:29, Fabio Estevam wrote:

> On Tue, Feb 21, 2012 at 10:13 AM, Michael Thalmeier
> <michael.thalmeier@hale.at> wrote:
> 
>> -static int mxc_rnga_data_present(struct hwrng *rng)
>> +static int mxc_rnga_data_present(struct hwrng *rng, int wait)
> 
> This looks good, but ...
> 
>>  {
>> -       int level;
>> +       int level, i;
>>        void __iomem *rng_base = (void __iomem *)rng->priv;
>>
>> -       /* how many random numbers is in FIFO? [0-16] */
>> -       level = ((__raw_readl(rng_base + RNGA_STATUS) &
>> -                       RNGA_STATUS_LEVEL_MASK) >> 8);
>> +       for (i = 0; i < 20; i++) {
> 
> Why the magic "20" here?
> 
> It would be better to add a proper timeout mechanism instead, such as
> time_after(jiffies, timeout)
> 


I am absolutely with you.
The point is only that this is the behaviour of nearly all hw_random
drivers, and I basically just copied it over into this driver.


--
Scanned by MailScanner.

[PATCH] i.MX31: mxc-rnga: implement waiting for data in driver

[Arnd Bergmann]

On Tuesday 21 February 2012, Michael Thalmeier wrote:
> On 2012-02-21 14:29, Fabio Estevam wrote:
> 
> > On Tue, Feb 21, 2012 at 10:13 AM, Michael Thalmeier
> > <michael.thalmeier@hale.at> wrote:
> > 
> >> -static int mxc_rnga_data_present(struct hwrng *rng)
> >> +static int mxc_rnga_data_present(struct hwrng *rng, int wait)
> > 
> > This looks good, but ...
> > 
> >>  {
> >> -       int level;
> >> +       int level, i;
> >>        void __iomem *rng_base = (void __iomem *)rng->priv;
> >>
> >> -       /* how many random numbers is in FIFO? [0-16] */
> >> -       level = ((__raw_readl(rng_base + RNGA_STATUS) &
> >> -                       RNGA_STATUS_LEVEL_MASK) >> 8);
> >> +       for (i = 0; i < 20; i++) {
> > 
> > Why the magic "20" here?
> > 
> > It would be better to add a proper timeout mechanism instead, such as
> > time_after(jiffies, timeout)
> > 
> I am absolutely with you.
> The point is only that this is the behaviour of nearly all hw_random
> drivers, and I basically just copied it over into this driver.

Hmm, I guess they are all wrong then ;-)

It would be nice to move the retry loop into common code where it
would be easier to change.

Note that comparing jiffies is not going to help here because the
maximum delay in the loop is less than a jiffy.

	Arnd