[PATCH] soundwire: bus: add enumerated slave to device list
Currently slave devices are only added either from device tree or acpi entries. However lets say, there is wrong or no entry of a slave device in DT that is enumerated, then there is no way for user to know all the enumerated devices on the bus.
To fix this add slave device by default if there is no matching dt or acpi entry, so that we can see this in sysfs entry.
In my case I had a wrong address entry in DT, However I had no way to know what devices are actually enumerated on the bus!
Signed-off-by: Srinivas Kandagatla srinivas.kandagatla@linaro.org --- drivers/soundwire/bus.c | 1 + drivers/soundwire/bus.h | 2 ++ drivers/soundwire/bus_type.c | 6 ++++++ drivers/soundwire/slave.c | 4 ++-- 4 files changed, 11 insertions(+), 2 deletions(-)
diff --git a/drivers/soundwire/bus.c b/drivers/soundwire/bus.c index e6e0fb9a81b4..55d9c22c4ec5 100644 --- a/drivers/soundwire/bus.c +++ b/drivers/soundwire/bus.c @@ -699,6 +699,7 @@ static int sdw_program_device_num(struct sdw_bus *bus)
if (!found) { /* TODO: Park this device in Group 13 */ + sdw_slave_add(bus, &id, NULL); dev_err(bus->dev, "Slave Entry not found\n"); }
diff --git a/drivers/soundwire/bus.h b/drivers/soundwire/bus.h index 82484f741168..1517d6789dff 100644 --- a/drivers/soundwire/bus.h +++ b/drivers/soundwire/bus.h @@ -19,6 +19,8 @@ static inline int sdw_acpi_find_slaves(struct sdw_bus *bus) int sdw_of_find_slaves(struct sdw_bus *bus); void sdw_extract_slave_id(struct sdw_bus *bus, u64 addr, struct sdw_slave_id *id); +int sdw_slave_add(struct sdw_bus *bus, struct sdw_slave_id *id, + struct fwnode_handle *fwnode); int sdw_master_device_add(struct sdw_bus *bus, struct device *parent, struct fwnode_handle *fwnode); int sdw_master_device_del(struct sdw_bus *bus); diff --git a/drivers/soundwire/bus_type.c b/drivers/soundwire/bus_type.c index 6fba55898cf0..ac036223046f 100644 --- a/drivers/soundwire/bus_type.c +++ b/drivers/soundwire/bus_type.c @@ -84,6 +84,12 @@ static int sdw_drv_probe(struct device *dev) const struct sdw_device_id *id; int ret;
+ /* + * fw description is mandatory to bind + */ + if (!dev->fwnode || !dev->of_node) + return -ENODEV; + id = sdw_get_device_id(slave, drv); if (!id) return -ENODEV; diff --git a/drivers/soundwire/slave.c b/drivers/soundwire/slave.c index 0839445ee07b..24a16ebf9ae2 100644 --- a/drivers/soundwire/slave.c +++ b/drivers/soundwire/slave.c @@ -20,8 +20,8 @@ struct device_type sdw_slave_type = { .uevent = sdw_slave_uevent, };
-static int sdw_slave_add(struct sdw_bus *bus, - struct sdw_slave_id *id, struct fwnode_handle *fwnode) +int sdw_slave_add(struct sdw_bus *bus, + struct sdw_slave_id *id, struct fwnode_handle *fwnode) { struct sdw_slave *slave; int ret;
On 9/9/20 3:27 AM, Srinivas Kandagatla wrote:
Currently slave devices are only added either from device tree or acpi entries. However lets say, there is wrong or no entry of a slave device in DT that is enumerated, then there is no way for user to know all the enumerated devices on the bus.
Sorry Srinivas, I don't understand your point.
The sysfs entries will include all devices that are described in platform firmware (be it DT or ACPI).
If you add to sysfs entries unknown devices which happen to be present on the bus, then what? How would you identify them from the devices that are described in firmware?
Also the sysfs entries describe properties, but if you haven't bound a driver then how would this work?
I really feel this deserves more explanations on the problem statement and what you are hoping to achieve in this case.
Thanks!
To fix this add slave device by default if there is no matching dt or acpi entry, so that we can see this in sysfs entry.
In my case I had a wrong address entry in DT, However I had no way to know what devices are actually enumerated on the bus!
Signed-off-by: Srinivas Kandagatla srinivas.kandagatla@linaro.org
drivers/soundwire/bus.c | 1 + drivers/soundwire/bus.h | 2 ++ drivers/soundwire/bus_type.c | 6 ++++++ drivers/soundwire/slave.c | 4 ++-- 4 files changed, 11 insertions(+), 2 deletions(-)
diff --git a/drivers/soundwire/bus.c b/drivers/soundwire/bus.c index e6e0fb9a81b4..55d9c22c4ec5 100644 --- a/drivers/soundwire/bus.c +++ b/drivers/soundwire/bus.c @@ -699,6 +699,7 @@ static int sdw_program_device_num(struct sdw_bus *bus)
if (!found) { /* TODO: Park this device in Group 13 */
sdw_slave_add(bus, &id, NULL); dev_err(bus->dev, "Slave Entry not found\n");diff --git a/drivers/soundwire/bus.h b/drivers/soundwire/bus.h index 82484f741168..1517d6789dff 100644 --- a/drivers/soundwire/bus.h +++ b/drivers/soundwire/bus.h @@ -19,6 +19,8 @@ static inline int sdw_acpi_find_slaves(struct sdw_bus *bus) int sdw_of_find_slaves(struct sdw_bus *bus); void sdw_extract_slave_id(struct sdw_bus *bus, u64 addr, struct sdw_slave_id *id); +int sdw_slave_add(struct sdw_bus *bus, struct sdw_slave_id *id,
struct fwnode_handle *fwnode);diff --git a/drivers/soundwire/bus_type.c b/drivers/soundwire/bus_type.c index 6fba55898cf0..ac036223046f 100644 --- a/drivers/soundwire/bus_type.c +++ b/drivers/soundwire/bus_type.c @@ -84,6 +84,12 @@ static int sdw_drv_probe(struct device *dev) const struct sdw_device_id *id; int ret;
- /*
* fw description is mandatory to bind*/- if (!dev->fwnode || !dev->of_node)
return -ENODEV;- id = sdw_get_device_id(slave, drv); if (!id) return -ENODEV;
diff --git a/drivers/soundwire/slave.c b/drivers/soundwire/slave.c index 0839445ee07b..24a16ebf9ae2 100644 --- a/drivers/soundwire/slave.c +++ b/drivers/soundwire/slave.c @@ -20,8 +20,8 @@ struct device_type sdw_slave_type = { .uevent = sdw_slave_uevent, };
-static int sdw_slave_add(struct sdw_bus *bus,
struct sdw_slave_id *id, struct fwnode_handle *fwnode)+int sdw_slave_add(struct sdw_bus *bus,
struct sdw_slave_id *id, struct fwnode_handle *fwnode)
On 09/09/2020 14:37, Pierre-Louis Bossart wrote:
On 9/9/20 3:27 AM, Srinivas Kandagatla wrote:
Currently slave devices are only added either from device tree or acpi entries. However lets say, there is wrong or no entry of a slave device in DT that is enumerated, then there is no way for user to know all the enumerated devices on the bus.
Sorry Srinivas, I don't understand your point.
The sysfs entries will include all devices that are described in platform firmware (be it DT or ACPI).
yes that is true, but it will not include all the enumerated devices on the bus!
In my case on a new board I was trying to figure out what devices are on the bus even before even adding any device tree entries!
In second case I had a typo in the device tree entry and sysfs displayed devices with that typo rather than actual enumerated device id.
If you add to sysfs entries unknown devices which happen to be present on the bus, then what? How would you identify them from the devices that are described in firmware?
Both of them should be displayed in sysfs, core should be able to differentiate this based on the presence of fw_node or of_node and not bind!
Also the sysfs entries describe properties, but if you haven't bound a driver then how would this work?
This is would be informative, atleast in cases like me!
All I want to know is the list of enumerated devices on the bus, If doing this way is not the right thing, then am happy to try any suggestion!
For now I have managed to figure out enumerated device ids on the bus with this patch, I was hoping that other people would also hit such issue, so I sent this patch!
thanks, srini
I really feel this deserves more explanations on the problem statement and what you are hoping to achieve in this case.
Thanks!
To fix this add slave device by default if there is no matching dt or acpi entry, so that we can see this in sysfs entry.
In my case I had a wrong address entry in DT, However I had no way to know what devices are actually enumerated on the bus!
Signed-off-by: Srinivas Kandagatla srinivas.kandagatla@linaro.org
drivers/soundwire/bus.c | 1 + drivers/soundwire/bus.h | 2 ++ drivers/soundwire/bus_type.c | 6 ++++++ drivers/soundwire/slave.c | 4 ++-- 4 files changed, 11 insertions(+), 2 deletions(-)
diff --git a/drivers/soundwire/bus.c b/drivers/soundwire/bus.c index e6e0fb9a81b4..55d9c22c4ec5 100644 --- a/drivers/soundwire/bus.c +++ b/drivers/soundwire/bus.c @@ -699,6 +699,7 @@ static int sdw_program_device_num(struct sdw_bus *bus) if (!found) { /* TODO: Park this device in Group 13 */ + sdw_slave_add(bus, &id, NULL); dev_err(bus->dev, "Slave Entry not found\n"); } diff --git a/drivers/soundwire/bus.h b/drivers/soundwire/bus.h index 82484f741168..1517d6789dff 100644 --- a/drivers/soundwire/bus.h +++ b/drivers/soundwire/bus.h @@ -19,6 +19,8 @@ static inline int sdw_acpi_find_slaves(struct sdw_bus *bus) int sdw_of_find_slaves(struct sdw_bus *bus); void sdw_extract_slave_id(struct sdw_bus *bus, u64 addr, struct sdw_slave_id *id); +int sdw_slave_add(struct sdw_bus *bus, struct sdw_slave_id *id, + struct fwnode_handle *fwnode); int sdw_master_device_add(struct sdw_bus *bus, struct device *parent, struct fwnode_handle *fwnode); int sdw_master_device_del(struct sdw_bus *bus); diff --git a/drivers/soundwire/bus_type.c b/drivers/soundwire/bus_type.c index 6fba55898cf0..ac036223046f 100644 --- a/drivers/soundwire/bus_type.c +++ b/drivers/soundwire/bus_type.c @@ -84,6 +84,12 @@ static int sdw_drv_probe(struct device *dev) const struct sdw_device_id *id; int ret; + /* + * fw description is mandatory to bind + */ + if (!dev->fwnode || !dev->of_node) + return -ENODEV;
id = sdw_get_device_id(slave, drv); if (!id) return -ENODEV; diff --git a/drivers/soundwire/slave.c b/drivers/soundwire/slave.c index 0839445ee07b..24a16ebf9ae2 100644 --- a/drivers/soundwire/slave.c +++ b/drivers/soundwire/slave.c @@ -20,8 +20,8 @@ struct device_type sdw_slave_type = { .uevent = sdw_slave_uevent, }; -static int sdw_slave_add(struct sdw_bus *bus, - struct sdw_slave_id *id, struct fwnode_handle *fwnode) +int sdw_slave_add(struct sdw_bus *bus, + struct sdw_slave_id *id, struct fwnode_handle *fwnode) { struct sdw_slave *slave; int ret;
Currently slave devices are only added either from device tree or acpi entries. However lets say, there is wrong or no entry of a slave device in DT that is enumerated, then there is no way for user to know all the enumerated devices on the bus.
Sorry Srinivas, I don't understand your point.
The sysfs entries will include all devices that are described in platform firmware (be it DT or ACPI).
yes that is true, but it will not include all the enumerated devices on the bus!
In my case on a new board I was trying to figure out what devices are on the bus even before even adding any device tree entries!
We've seen this before but dynamic debug provides all the information you need. see e.g. the logs from https://sof-ci.01.org/linuxpr/PR2425/build4447/devicetest/
jf-cml-rvp-sdw-1 kernel: [ 289.751974] soundwire sdw-master-0: Slave attached, programming device number jf-cml-rvp-sdw-1 kernel: [ 289.752121] soundwire sdw-master-0: SDW Slave Addr: 10025d070000 <<< HERE jf-cml-rvp-sdw-1 kernel: [ 289.752122] soundwire sdw-master-0: SDW Slave class_id 0, part_id 700, mfg_id 25d, unique_id 0, version 1
In second case I had a typo in the device tree entry and sysfs
displayed
devices with that typo rather than actual enumerated device id.
That's a feature, not a bug? We use what address the platform firmware provides. If it's inaccurate then nothing can work.
If you add to sysfs entries unknown devices which happen to be present on the bus, then what? How would you identify them from the devices that are described in firmware?
Both of them should be displayed in sysfs, core should be able to differentiate this based on the presence of fw_node or of_node and not bind!
Core yes but user not so much. If the intent is to list the devices present on the bus, your patch still requires manual work.
Also the sysfs entries describe properties, but if you haven't bound a driver then how would this work?
This is would be informative, atleast in cases like me!
All I want to know is the list of enumerated devices on the bus, If doing this way is not the right thing, then am happy to try any suggestion!
For now I have managed to figure out enumerated device ids on the bus with this patch, I was hoping that other people would also hit such issue, so I sent this patch!
Now I get your point but a) you already have a dynamic debug trace to list all devices b) adding 'undeclared' devices would make things quite murky and is only half of the solution. We already struggle because we already have 'ghost' devices in sysfs that are not physically present, and no way to differentiate between the two. If we did add those entries, then we'd need two new sysfs attributes such as 'declared' and 'enumerated'.
On 09/09/2020 15:39, Pierre-Louis Bossart wrote:
Currently slave devices are only added either from device tree or acpi entries. However lets say, there is wrong or no entry of a slave device in DT that is enumerated, then there is no way for user to know all the enumerated devices on the bus.
Sorry Srinivas, I don't understand your point.
The sysfs entries will include all devices that are described in platform firmware (be it DT or ACPI).
yes that is true, but it will not include all the enumerated devices on the bus!
In my case on a new board I was trying to figure out what devices are on the bus even before even adding any device tree entries!
We've seen this before but dynamic debug provides all the information you need. see e.g. the logs from https://sof-ci.01.org/linuxpr/PR2425/build4447/devicetest/
jf-cml-rvp-sdw-1 kernel: [ 289.751974] soundwire sdw-master-0: Slave attached, programming device number jf-cml-rvp-sdw-1 kernel: [ 289.752121] soundwire sdw-master-0: SDW Slave Addr: 10025d070000 <<< HERE
Yes, I have noticed this too! This will be printed for every call to sdw_extract_slave_id()!
...
Now I get your point but a) you already have a dynamic debug trace to list all devices b) adding 'undeclared' devices would make things quite murky and is only half of the solution. We already struggle because we already have 'ghost' devices in sysfs that are not physically present, and no way to differentiate between the two. If we did add those entries, then we'd need two new sysfs attributes such as 'declared' and 'enumerated'.
I totally agree with you on dealing with the undeclared devices, which is unnecessary mess! May be we could make the enumerated devices discovery bit more verbose!
--srini
On 9/9/20 10:54 AM, Srinivas Kandagatla wrote:
On 09/09/2020 15:39, Pierre-Louis Bossart wrote:
Currently slave devices are only added either from device tree or acpi entries. However lets say, there is wrong or no entry of a slave device in DT that is enumerated, then there is no way for user to know all the enumerated devices on the bus.
Sorry Srinivas, I don't understand your point.
The sysfs entries will include all devices that are described in platform firmware (be it DT or ACPI).
yes that is true, but it will not include all the enumerated devices on the bus!
In my case on a new board I was trying to figure out what devices are on the bus even before even adding any device tree entries!
We've seen this before but dynamic debug provides all the information you need. see e.g. the logs from https://sof-ci.01.org/linuxpr/PR2425/build4447/devicetest/
jf-cml-rvp-sdw-1 kernel: [ 289.751974] soundwire sdw-master-0: Slave attached, programming device number jf-cml-rvp-sdw-1 kernel: [ 289.752121] soundwire sdw-master-0: SDW Slave Addr: 10025d070000 <<< HERE
Yes, I have noticed this too! This will be printed for every call to sdw_extract_slave_id()!
...
Now I get your point but a) you already have a dynamic debug trace to list all devices b) adding 'undeclared' devices would make things quite murky and is only half of the solution. We already struggle because we already have 'ghost' devices in sysfs that are not physically present, and no way to differentiate between the two. If we did add those entries, then we'd need two new sysfs attributes such as 'declared' and 'enumerated'.
I totally agree with you on dealing with the undeclared devices, which is unnecessary mess!
It's not necessarily that bad. - if the intent is to have a single platform firmware that can deal with different boards, it's a good thing. - but if it's just sloppy platform firmware that just does copy-paste from platform to platform then indeed it becomes a mess.
May be we could make the enumerated devices discovery bit more verbose!
Maybe adding a device number sysfs entry would help, e.g. reporting NotAttched or a value in [0,11] would tell you if the device is actually present.
On 09-09-20, 12:00, Pierre-Louis Bossart wrote:
On 9/9/20 10:54 AM, Srinivas Kandagatla wrote:
On 09/09/2020 15:39, Pierre-Louis Bossart wrote:
Currently slave devices are only added either from device tree or acpi entries. However lets say, there is wrong or no entry of a slave device in DT that is enumerated, then there is no way for user to know all the enumerated devices on the bus.
Sorry Srinivas, I don't understand your point.
The sysfs entries will include all devices that are described in platform firmware (be it DT or ACPI).
yes that is true, but it will not include all the enumerated devices on the bus!
In my case on a new board I was trying to figure out what devices are on the bus even before even adding any device tree entries!
We've seen this before but dynamic debug provides all the information you need. see e.g. the logs from https://sof-ci.01.org/linuxpr/PR2425/build4447/devicetest/
jf-cml-rvp-sdw-1 kernel: [ 289.751974] soundwire sdw-master-0: Slave attached, programming device number jf-cml-rvp-sdw-1 kernel: [ 289.752121] soundwire sdw-master-0: SDW Slave Addr: 10025d070000 <<< HERE
Yes, I have noticed this too! This will be printed for every call to sdw_extract_slave_id()!
...
Now I get your point but a) you already have a dynamic debug trace to list all devices b) adding 'undeclared' devices would make things quite murky and is only half of the solution. We already struggle because we already have 'ghost' devices in sysfs that are not physically present, and no way to differentiate between the two. If we did add those entries, then we'd need two new sysfs attributes such as 'declared' and 'enumerated'.
I totally agree with you on dealing with the undeclared devices, which is unnecessary mess!
It's not necessarily that bad.
- if the intent is to have a single platform firmware that can deal with
different boards, it's a good thing.
- but if it's just sloppy platform firmware that just does copy-paste from
platform to platform then indeed it becomes a mess.
May be we could make the enumerated devices discovery bit more verbose!
Maybe adding a device number sysfs entry would help, e.g. reporting NotAttched or a value in [0,11] would tell you if the device is actually present.
Agreed, I cooked this patch to report verbose device status, let me know if you are okay with this. I think this would be useful regardless of current discussion.
On Db845c I see:
root@linaro-alip:/sys/bus/soundwire/devices# cat sdw:0:217:2010:0:1/status Attached root@linaro-alip:/sys/bus/soundwire/devices# cat sdw:0:217:2010:0:2/status Attached
diff --git a/drivers/soundwire/sysfs_slave.c b/drivers/soundwire/sysfs_slave.c index f510071b0add..3b2765f10024 100644 --- a/drivers/soundwire/sysfs_slave.c +++ b/drivers/soundwire/sysfs_slave.c @@ -97,8 +97,27 @@ static ssize_t modalias_show(struct device *dev, } static DEVICE_ATTR_RO(modalias);
+#define SDW_SLAVE_MAX (SDW_SLAVE_RESERVED + 1) + +static const char *const slave_status[SDW_SLAVE_MAX] = { + [SDW_SLAVE_UNATTACHED] = "UNATTACHED", + [SDW_SLAVE_ATTACHED] = "Attached", + [SDW_SLAVE_ALERT] = "Alert", + [SDW_SLAVE_RESERVED] = "Reserved", +}; + +static ssize_t status_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct sdw_slave *slave = dev_to_sdw_dev(dev); + + return sprintf(buf, "%s\n", slave_status[slave->status]); +} +static DEVICE_ATTR_RO(status); + static struct attribute *slave_attrs[] = { &dev_attr_modalias.attr, + &dev_attr_status.attr, NULL, }; ATTRIBUTE_GROUPS(slave);
May be we could make the enumerated devices discovery bit more verbose!
Maybe adding a device number sysfs entry would help, e.g. reporting NotAttched or a value in [0,11] would tell you if the device is actually present.
Agreed, I cooked this patch to report verbose device status, let me know if you are okay with this. I think this would be useful regardless of current discussion.
On Db845c I see:
root@linaro-alip:/sys/bus/soundwire/devices# cat sdw:0:217:2010:0:1/status Attached root@linaro-alip:/sys/bus/soundwire/devices# cat sdw:0:217:2010:0:2/status Attached
looks like we are all aligned on the idea, I have a similar patch to at https://github.com/thesofproject/linux/pull/2426
The difference is that the sysfs status and device_number is added even without a driver probe and when there's no firmware description. sysfs is currently only added after the driver probe, which wouldn't work for the case Srinivas was trying to deal with.
but the way you dealt the status below is better than the switch case I used, so will merge this into my code.
Srinivas' patch needs a fix for ACPI platforms, won't probe otherwise since we don't have an of_node. If that's alright with everyone I can submit a patchset that gathers the 3 contributions.
diff --git a/drivers/soundwire/sysfs_slave.c b/drivers/soundwire/sysfs_slave.c index f510071b0add..3b2765f10024 100644 --- a/drivers/soundwire/sysfs_slave.c +++ b/drivers/soundwire/sysfs_slave.c @@ -97,8 +97,27 @@ static ssize_t modalias_show(struct device *dev, } static DEVICE_ATTR_RO(modalias);
+#define SDW_SLAVE_MAX (SDW_SLAVE_RESERVED + 1)
+static const char *const slave_status[SDW_SLAVE_MAX] = {
- [SDW_SLAVE_UNATTACHED] = "UNATTACHED",
- [SDW_SLAVE_ATTACHED] = "Attached",
- [SDW_SLAVE_ALERT] = "Alert",
- [SDW_SLAVE_RESERVED] = "Reserved",
+};
+static ssize_t status_show(struct device *dev,
struct device_attribute *attr, char *buf)+{
- struct sdw_slave *slave = dev_to_sdw_dev(dev);
- return sprintf(buf, "%s\n", slave_status[slave->status]);
+} +static DEVICE_ATTR_RO(status);
- static struct attribute *slave_attrs[] = { &dev_attr_modalias.attr,
- &dev_attr_status.attr, NULL, }; ATTRIBUTE_GROUPS(slave);
On 10-09-20, 09:02, Pierre-Louis Bossart wrote:
May be we could make the enumerated devices discovery bit more verbose!
Maybe adding a device number sysfs entry would help, e.g. reporting NotAttched or a value in [0,11] would tell you if the device is actually present.
Agreed, I cooked this patch to report verbose device status, let me know if you are okay with this. I think this would be useful regardless of current discussion.
On Db845c I see:
root@linaro-alip:/sys/bus/soundwire/devices# cat sdw:0:217:2010:0:1/status Attached root@linaro-alip:/sys/bus/soundwire/devices# cat sdw:0:217:2010:0:2/status Attached
looks like we are all aligned on the idea, I have a similar patch to at https://github.com/thesofproject/linux/pull/2426
The difference is that the sysfs status and device_number is added even without a driver probe and when there's no firmware description. sysfs is currently only added after the driver probe, which wouldn't work for the case Srinivas was trying to deal with.
Okay sound fine
but the way you dealt the status below is better than the switch case I used, so will merge this into my code.
Why merge? That patch can remain independent and you can add device_number patch on top and another one for moving sysfs creation without a driver probe, these three sound like three different patches to me.
Srinivas' patch needs a fix for ACPI platforms, won't probe otherwise since we don't have an of_node. If that's alright with everyone I can submit a patchset that gathers the 3 contributions.
Yes one series should be good, but lets keep one change in a patch please
participants (3)
-
Pierre-Louis Bossart -
Srinivas Kandagatla -
Vinod Koul