Skip to contentSkip to conversation

net: lora: sx1301: Fix radio SPI write

PATCH Ben Whitten 3 participants 3 messages Dec 31, 2018 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

+ Mark +linux-kernel

Quoted 22 lines · Andreas Färber

When converting to regmap_bus we omitted the write flag, rendering all sx125x register writes no-op.

---
 drivers/net/lora/sx130x_radio.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/net/lora/sx130x_radio.c +1 -0
index 6c94d13cd4db..e7b2df808c1a 100644
@@ -9,6 +9,7 @@
  * Copyright (c) 2013 Semtech-Cycleo
  */
 
+#include <linux/bitops.h>
 #include <linux/of.h>
 #include <linux/of_device.h>
 #include <linux/regmap.h>

drivers/net/lora/sx130x_radio.c static int sx1301_regmap_bus_write(void *context, unsigned int reg, · line 44

Quoted 5 lines
drivers/net/lora/sx130x_radio.c +1 -1
@@ -40,7 +41,7 @@ static int sx1301_regmap_bus_write(void *context, unsigned int reg,
        ret = regmap_write(priv->regmap, cs, 0);
        if (ret)
                return ret;
-       ret = regmap_write(priv->regmap, addr, reg);
+       ret = regmap_write(priv->regmap, addr, BIT(7) | reg);

Curious.
Isn't the (read/write)_flag_mask, set in the regmap_config, a property of the device connected to the regmap_bus?
I would have expected this to have been applied prior to passing to a regmap_bus call, surely the bus should only be concerned with transport?

It will work in our case as our current hardware uses 1 in bit 7 but it will not if we ever see other radios with different masks.

If the regmap_bus provider does need to apply the mask, can we get the end device regmap_config and apply it from there?
Mark what are your thoughts on the route to take?

Regards,
Ben Whitten

#2 lore raw

Am 31.12.18 um 01:14 schrieb Ben Whitten:

Quoted 19 lines

+ Mark +linux-kernel

Quoted 17 lines · Andreas Färber

When converting to regmap_bus we omitted the write flag, rendering all sx125x register writes no-op.

---
 drivers/net/lora/sx130x_radio.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/net/lora/sx130x_radio.c +0 -0
index 6c94d13cd4db..e7b2df808c1a 100644
@@ -9,6 +9,7 @@
  * Copyright (c) 2013 Semtech-Cycleo
  */
 
Quoted 14 lines

drivers/net/lora/sx130x_radio.c static int sx1301_regmap_bus_write(void *context, unsigned int reg, · line 44

Quoted 10 lines
drivers/net/lora/sx130x_radio.c +2 -1
+#include <linux/bitops.h>
 #include <linux/of.h>
 #include <linux/of_device.h>
 #include <linux/regmap.h>
@@ -40,7 +41,7 @@ static int sx1301_regmap_bus_write(void *context, unsigned int reg,
        ret = regmap_write(priv->regmap, cs, 0);
        if (ret)
                return ret;
-       ret = regmap_write(priv->regmap, addr, reg);
+       ret = regmap_write(priv->regmap, addr, BIT(7) | reg);

Curious.
Isn't the (read/write)_flag_mask, set in the regmap_config, a property of the device connected to the regmap_bus?

Yes, I assume so.

Quoted 2 lines

I would have expected this to have been applied prior to passing to a regmap_bus call, surely the bus should only be concerned with transport?

The argument the regmap_bus ops get is a reg though. It's up to our code to convert it to an address.

Quoted 5 lines

It will work in our case as our current hardware uses 1 in bit 7 but it will not if we ever see other radios with different masks.

If the regmap_bus provider does need to apply the mask, can we get the end device regmap_config and apply it from there?

I did not find a way to retrieve the struct regmap_config or its .{read,write}_flag_mask values through any regmap helper. Otherwise that would've been my first choice as well, since we do set it correctly.

We could add a regmap_config* field to struct sx130x_radio_device and have our probe function populate it, but given that we are implementing a regmap_bus here, #include "../../base/regmap/internal.h" while ugly may be the sanest choice. It still requires to add a .regmap field to struct sx130x_radio_device. Testing on the good interface went positive.

Regards,
Andreas

Quoted 4 lines

Mark what are your thoughts on the route to take?

Regards,
Ben Whitten

Signature · 3 lines
SUSE Linux GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Felix Imendörffer, Jane Smithard, Graham Norton
HRB 21284 (AG Nürnberg)
#3 lore raw
In reply to Ben Whitten
Quoted 3 lines · Ben Whitten

If the regmap_bus provider does need to apply the mask, can we get the end device regmap_config and apply it from there?
Mark what are your thoughts on the route to take?

If you're getting a byte stream in then it'll have the mask applied. If you're providing per-register read/write ops no formatting at all will be done, obviously in that case there's no need to bounce the information through the config.