Re: [PATCHv9 1/3] Runtime Interpreted Power Sequences

On Sat, Nov 17, 2012 at 07:55:45PM +0900, Alexandre Courbot wrote:

Some device drivers (e.g. panel or backlights) need to follow precise
sequences for powering on and off, involving GPIOs, regulators, PWMs
and other power-related resources, with a defined powering order and
delays to respect between steps. These sequences are device-specific,
and do not belong to a particular driver - therefore they have been
performed by board-specific hook functions so far.

With the advent of the device tree and of ARM kernels that are not
board-tied, we cannot rely on these board-specific hooks anymore but
need a way to implement these sequences in a portable manner. This patch
introduces a simple interpreter that can execute such power sequences
encoded either as platform data or within the device tree. It also
introduces first support for regulator, GPIO and PWM resources.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Reviewed-by: Stephen Warren <redacted>
Reviewed-by: Mark Brown <redacted>
Tested-by: Thierry Reding <redacted>
Tested-by: Stephen Warren <redacted>
Acked-by: Thierry Reding <redacted>
---

This looks almost perfect!

Just a few final notes, again mostly about the build system bits.

[...]

This really needs a proper Kconfig description and a help text, shortly
describing the purpose of the subsystem.

We usually don't write file names in the comments.

+ *
+ * Author: Alexandre Courbot [off-list ref]
+ *

[...]

+	return 0;
+}
+EXPORT_SYMBOL_GPL(power_seq_run);
+
+/* defined in power_seq_*.c files */

Why not place the decls into the _priv.h file?.. I understand that this
might be a temporary stuff until we have something better for ops
registration, but still, I believe it belongs to the header file.

Things should not depend on _priv.h's includes. I.e. I see this file uses
of_ routines, so it should include <linux/of.h>. <linux/delay.h> for
udelay_range(), etc.

Also, usually we place "private" headers at the very end, not at the
beginning.

+
+#ifdef CONFIG_OF
+static int of_power_seq_parse_delay(struct device_node *node,
+				    struct power_seq *seq,
+				    unsigned int step_nbr,
+				    struct power_seq_resource *res)
+{
+	struct power_seq_step *step = &seq->steps[step_nbr];
+	int err;
+
+	err = of_property_read_u32(node, "delay",
+				   &step->delay.delay);
+	if (err < 0)
+		power_seq_err(seq, step_nbr, "error reading delay property\n");
+
+	return err;
+}
+#else
+#define of_power_seq_parse_delay NULL
+#endif

[...]

+#define power_seq_err(seq, step_nbr, format, ...)			\
+	dev_err(seq->set->dev, "%s[%d]: " format, seq->id, step_nbr,	\
+	##__VA_ARGS__);
+
+#ifdef CONFIG_OF
+int of_power_seq_parse_enable_properties(struct device_node *node,
+					 struct power_seq *seq,
+					 unsigned int step_nbr, bool *enable);

Um, I believe you don't need CONFIG_OF here. (If it's about 'struct
device_node', then as I see it in of.h, the header always declares it,
even for the !OF case.)

+#endif
+
+/**

[...]

Would be really great if you could get rid of the #ifdefs in the .c files.

To do so, in the makefile you could write something like this:

power_seq-$(CONFIG_REGULATOR) += regulator.o

And in the header file (as explained above), you'd have

#ifdef CONFIG_REGULATOR
#define ...REGULATOR_OPS &power_seq_regulator_ops
#else
#define ...REGULATOR_OPS NULL
#endif

Or something along these lines...

Thanks,
Anton.