i2c: mux: pca954x: allow management of device idle state via sysfs
Subsystems (2)
v4 cover i2c: mux: pca954x: allow management of device idle state via sysfs
From: Robert Shearman <[email protected]>
The behaviour, by default, to not deselect after each transfer is unsafe when there is a device with an address that conflicts with another device on another pca954x mux on the same parent bus, and it may not be convenient to use devicetree to set the deselect mux, e.g. when running on x86_64 when ACPI is used to discover most of the device hierarchy.
Therefore, provide the ability to set the idle state behaviour using a new sysfs file, idle_state as a complement to the method of instantiating the device via sysfs.
Changes in v4:
- also remove variables that are redundant as a result of platform data removal
- improve ABI file description field
Changes in v3:
- remove use of platform data to simplify the idle state by not having per-channel idle states.
- correct some comments
- fix a style issue regarding use of READ_ONCE
Changes in v2:
- change from exposing deselect mask to idle state with more options and less implementation specific
- fix style issues
- don't fail the create of the mux device if the device attribute file in sysfs failed to create
Robert Shearman (3):
i2c: mux: pca954x: remove support for unused platform data
i2c: mux: pca9541: remove support for unused platform data
i2c: mux: pca954x: allow management of device idle state via sysfs
.../ABI/testing/sysfs-bus-i2c-devices-pca954x | 20 ++++ drivers/i2c/muxes/i2c-mux-pca9541.c | 8 +- drivers/i2c/muxes/i2c-mux-pca954x.c | 106 +++++++++++++----- include/linux/platform_data/pca954x.h | 48 -------- 4 files changed, 100 insertions(+), 82 deletions(-) create mode 100644 Documentation/ABI/testing/sysfs-bus-i2c-devices-pca954x delete mode 100644 include/linux/platform_data/pca954x.h
Signature · 1 line
2.20.1
PATCH v4 1/3 i2c: mux: pca954x: remove support for unused platform data
From: Robert Shearman <[email protected]>
There are no in-tree users of the pca954x platform data and the per-channel deselect configuration complicates efforts to export the configuration to user-space in a way that could be applied to other muxes. Therefore, remove support for the pca954x platform data.
--- drivers/i2c/muxes/i2c-mux-pca954x.c | 23 +++-------------------- 1 file changed, 3 insertions(+), 20 deletions(-)
drivers/i2c/muxes/i2c-mux-pca954x.c +3 -20
@@ -46,7 +46,6 @@#include <linux/of.h> #include <linux/of_device.h> #include <linux/of_irq.h> -#include <linux/platform_data/pca954x.h> #include <linux/pm.h> #include <linux/slab.h> #include <linux/spinlock.h>
@@ -348,14 +347,13 @@ static int pca954x_probe(struct i2c_client *client, const struct i2c_device_id *id) {struct i2c_adapter *adap = client->adapter; -struct pca954x_platform_data *pdata =dev_get_platdata (&client->dev);struct device *dev = &client->dev;struct device_node *np = dev->of_node;bool idle_disconnect_dt;struct gpio_desc *gpio; -int num, force, class;struct i2c_mux_core *muxc;struct pca954x *data; +int num;int ret;if (!i2c_check_functionality (adap,I2C_FUNC_SMBUS_BYTE ))
@@ -422,24 +420,9 @@ static int pca954x_probe(struct i2c_client *client,/* Now create an adapter for each channel */ for (num =0 ; num < data->chip->nchans; num++) { -bool idle_disconnect_pd = false; - - force =0 ;/* dynamic adap number */ - class =0 ;/* no class by default */ -if (pdata) { -if (num < pdata->num_modes) { -/* force static number */ - force = pdata->modes[num].adap_id; - class = pdata->modes[num].class; - }else -/* discard unconfigured channels */ -break ; - idle_disconnect_pd = pdata->modes[num].deselect_on_exit; - } - data->deselect |= (idle_disconnect_pd || - idle_disconnect_dt) << num; + data->deselect |= idle_disconnect_dt << num; - ret =i2c_mux_add_adapter (muxc, force, num, class); + ret =i2c_mux_add_adapter (muxc,0 , num,0 );if (ret) goto fail_cleanup; }
Signature · 1 line
2.20.1
PATCH v4 2/3 i2c: mux: pca9541: remove support for unused platform data
In reply to Robert ShearmanFrom: Robert Shearman <[email protected]>
There are no in-tree users of the platform data, so remove it to simplify the code slightly.
Remove the now unused pca954x.h platform data header.
--- drivers/i2c/muxes/i2c-mux-pca9541.c | 8 +---- include/linux/platform_data/pca954x.h | 48 --------------------------- 2 files changed, 1 insertion(+), 55 deletions(-) delete mode 100644 include/linux/platform_data/pca954x.h
drivers/i2c/muxes/i2c-mux-pca9541.c +1 -7
@@ -22,7 +22,6 @@#include <linux/i2c-mux.h> #include <linux/jiffies.h> #include <linux/module.h> -#include <linux/platform_data/pca954x.h> #include <linux/slab.h> /*
@@ -287,10 +286,8 @@ static int pca9541_probe(struct i2c_client *client, const struct i2c_device_id *id) {struct i2c_adapter *adap = client->adapter; -struct pca954x_platform_data *pdata =dev_get_platdata (&client->dev);struct i2c_mux_core *muxc;struct pca9541 *data; -int force;int ret;if (!i2c_check_functionality (adap,I2C_FUNC_SMBUS_BYTE_DATA ))
@@ -306,9 +303,6 @@ static int pca9541_probe(struct i2c_client *client,/* Create mux adapter */ - force =0 ; -if (pdata) - force = pdata->modes[0 ].adap_id; muxc =i2c_mux_alloc (adap, &client->dev,1 ,sizeof (*data),I2C_MUX_ARBITRATOR , pca9541_select_chan, pca9541_release_chan);
@@ -320,7 +314,7 @@ static int pca9541_probe(struct i2c_client *client,i2c_set_clientdata (client, muxc); - ret =i2c_mux_add_adapter (muxc, force,0 ,0 ); + ret =i2c_mux_add_adapter (muxc,0 ,0 ,0 );if (ret)return ret;
include/linux/platform_data/pca954x.h deleted +0 -48
48 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v4 3/3 i2c: mux: pca954x: allow management of device idle state via sysfs
In reply to Robert ShearmanFrom: Robert Shearman <[email protected]>
The behaviour, by default, to not deselect after each transfer is unsafe when there is a device with an address that conflicts with another device on another mux on the same parent bus, and it may not be convenient to use devicetree to set the deselect mux, e.g. when running on x86_64 when ACPI is used to discover most of the device hierarchy.
Therefore, provide the ability to set the idle state behaviour using a new sysfs file, idle_state as a complement to the method of instantiating the device via sysfs. The possible behaviours are disconnect, i.e. to deselect all channels from the mux, as-is (the default), i.e. leave the last channel selected, and set a predetermined channel.
The current behaviour of leaving the channel as-is after each transaction is preserved.
--- .../ABI/testing/sysfs-bus-i2c-devices-pca954x | 20 +++++ drivers/i2c/muxes/i2c-mux-pca954x.c | 85 +++++++++++++++++-- 2 files changed, 97 insertions(+), 8 deletions(-) create mode 100644 Documentation/ABI/testing/sysfs-bus-i2c-devices-pca954x
Documentation/ABI/testing/sysfs-bus-i2c-devices-pca954x new file +20 -0
@@ -0,0 +1,20 @@ +What: /sys/bus/i2c/.../idle_state +Date: January 2019 +KernelVersion: 5.2 +Contact: Robert Shearman <[email protected]> +Description: + Value that exists only for mux devices that can be + written to control the behaviour of the multiplexer on + idle. Possible values: + -2 - disconnect on idle, i.e. deselect the last used + channel, which is useful when there is a device + with an address that conflicts with another + device on another mux on the same parent bus. + -1 - leave the mux as-is, which is the most optimal + setting in terms of I2C operations and is the + default mode. + 0..<nchans> - set the mux to a predetermined channel, + which is useful if there is one channel that is + used almost always, and you want to reduce the + latency for normal operations after rare + transactions on other channels
drivers/i2c/muxes/i2c-mux-pca954x.c +77 -8
136 lines not on this page · raw message
Signature · 1 line
2.20.1
Hi Greg,
I was wondering if you think the below ABI addition looks sane to you? Or, should I ask someone else for a tag?
Cheers,
Peter
Quoted 200 lines · Robert Shearman
From: Robert Shearman <[email protected]>
The behaviour, by default, to not deselect after each transfer is unsafe when there is a device with an address that conflicts with another device on another mux on the same parent bus, and it may not be convenient to use devicetree to set the deselect mux, e.g. when running on
x86_64when ACPI is used to discover most of the device hierarchy.Therefore, provide the ability to set the idle state behaviour using a new sysfs file,
idle_stateas a complement to the method of instantiating the device via sysfs. The possible behaviours are disconnect, i.e. to deselect all channels from the mux, as-is (the default), i.e. leave the last channel selected, and set a predetermined channel.The current behaviour of leaving the channel as-is after each transaction is preserved.
Signed-off-by Robert Shearman <[email protected]>--- .../ABI/testing/sysfs-bus-i2c-devices-pca954x | 20 +++++ drivers/i2c/muxes/i2c-mux-pca954x.c | 85 +++++++++++++++++-- 2 files changed, 97 insertions(+), 8 deletions(-) create mode 100644 Documentation/ABI/testing/sysfs-bus-i2c-devices-pca954xDocumentation/ABI/testing/sysfs-bus-i2c-devices-pca954x new file +20
-0@@ -0,0 +1,20 @@ +What: /sys/bus/i2c/.../idle_state +Date: January 2019 +KernelVersion: 5.2 +Contact: Robert Shearman <[email protected]> +Description: + Value that exists only for mux devices that can be + written to control the behaviour of the multiplexer on + idle. Possible values: + -2 - disconnect on idle, i.e. deselect the last used + channel, which is useful when there is a device + with an address that conflicts with another + device on another mux on the same parent bus. + -1 - leave the mux as-is, which is the most optimal + setting in terms of I2C operations and is the + default mode. + 0..<nchans> - set the mux to a predetermined channel, + which is useful if there is one channel that is + used almost always, and you want to reduce the + latency for normal operations after rare + transactions on other channelsdrivers/i2c/muxes/i2c-mux-pca954x.c +77
-8@@ -49,6 +49,7 @@#include <linux/pm.h> #include <linux/slab.h> #include <linux/spinlock.h> +#include <dt-bindings/mux/mux.h> #define PCA954X_MAX_NCHANS 8@@ -84,7 +85,9 @@ struct pca954x {const struct chip_desc *chip;u8 last_chan;/* last register value */ -u8 deselect; +/* MUX_IDLE_AS_IS, MUX_IDLE_DISCONNECT or >= 0 for channel */ +s8 idle_state; +struct i2c_client *client;struct irq_domain *irq;@@ -253,15 +256,71 @@ static int pca954x_deselect_mux(struct i2c_mux_core *muxc, u32 chan) {struct pca954x *data =i2c_mux_priv (muxc);struct i2c_client *client = data->client; +s8 idle_state; + + idle_state =READ_ONCE (data->idle_state); +if (idle_state >=0 ) +/* Set the mux back to a predetermined channel */ +return pca954x_select_chan (muxc, idle_state); + +if (idle_state ==MUX_IDLE_DISCONNECT ) { +/* Deselect active channel */ + data->last_chan =0 ; +return pca954x_reg_write (muxc->parent, client, + data->last_chan); + } -if (!(data->deselect & (1 << chan))) -return 0 ; +/* otherwise leave as-is */ -/* Deselect active channel */ - data->last_chan =0 ; -return pca954x_reg_write (muxc->parent, client, data->last_chan); +return 0 ; +} + +static ssize_t idle_state_show (struct device *dev, +struct device_attribute *attr, +char *buf) +{ +struct i2c_client *client =to_i2c_client (dev); +struct i2c_mux_core *muxc =i2c_get_clientdata (client); +struct pca954x *data =i2c_mux_priv (muxc); + +return sprintf (buf,"%d\n" ,READ_ONCE (data->idle_state)); +} + +static ssize_t idle_state_store (struct device *dev, +struct device_attribute *attr, +const char *buf,size_t count) +{ +struct i2c_client *client =to_i2c_client (dev); +struct i2c_mux_core *muxc =i2c_get_clientdata (client); +struct pca954x *data =i2c_mux_priv (muxc); +int val; +int ret; + + ret =kstrtoint (buf,0 , &val); +if (ret <0 ) +return ret; + +if (val !=MUX_IDLE_AS_IS && val !=MUX_IDLE_DISCONNECT && + (val <0 || val >= data->chip->nchans)) +return -EINVAL ; + +i2c_lock_bus (muxc->parent,I2C_LOCK_SEGMENT ); + +WRITE_ONCE (data->idle_state, val); +/* +* Set the mux into a state consistent with the new +* idle_state. +*/ +if (data->last_chan || val !=MUX_IDLE_DISCONNECT ) + ret =pca954x_deselect_mux (muxc,0 ); + +i2c_unlock_bus (muxc->parent,I2C_LOCK_SEGMENT ); + +return ret <0 ? ret : count; } +static DEVICE_ATTR_RW (idle_state); +static irqreturn_t pca954x_irq_handler (int irq,void *dev_id) {struct pca954x *data = dev_id;@@ -328,8 +387,11 @@ static int pca954x_irq_setup(struct i2c_mux_core *muxc)static void pca954x_cleanup (struct i2c_mux_core *muxc) {struct pca954x *data =i2c_mux_priv (muxc); +struct i2c_client *client = data->client;int c, irq; +device_remove_file (&client->dev, &dev_attr_idle_state); +if (data->irq) {for (c =0 ; c < data->chip->nchans; c++) { irq =irq_find_mapping (data->irq, c);@@ -410,9 +472,12 @@ static int pca954x_probe(struct i2c_client *client, } data->last_chan =0 ;/* force the first selection */ + data->idle_state =MUX_IDLE_AS_IS ; idle_disconnect_dt = np &&of_property_read_bool (np,"i2c-mux-idle-disconnect" ); +if (idle_disconnect_dt) + data->idle_state =MUX_IDLE_DISCONNECT ; ret =pca954x_irq_setup (muxc);if (ret)@@ -420,8 +485,6 @@ static int pca954x_probe(struct i2c_client *client,/* Now create an adapter for each channel */ for (num =0 ; num < data->chip->nchans; num++) { - data->deselect |= idle_disconnect_dt << num; - ret =i2c_mux_add_adapter (muxc,0 , num,0 );if (ret) goto fail_cleanup;@@ -436,6 +499,12 @@ static int pca954x_probe(struct i2c_client *client, goto fail_cleanup; } +/* +* The attr probably isn't going to be needed in most cases, +* so don't fail completely on error. +*/ +device_create_file (dev, &dev_attr_idle_state); +dev_info (dev,"registered %d multiplexed busses for I2C %s %s\n" , num, data->chip->muxtype == pca954x_ismux ?"mux" :"switch" , client->name);
Quoted 3 lines · Peter Rosin
Hi Greg,
I was wondering if you think the below ABI addition looks sane to you?
I am not the i2c maintainer :)
Quoted 1 line
Or, should I ask someone else for a tag?
It's the middle of the merge window. None of us can do much other than focus on that until 5.1-rc1 is out.
thanks,
greg k-h
Hi Greg,
Quoted 2 lines
Quoted 1 line
I was wondering if you think the below ABI addition looks sane to you?
I am not the i2c maintainer :)
Peter is the i2c-mux maintainer, so I trust him very much on the I2C side of things. We just wondered if you'd notice a general flaw when exposing such an interface to userspace.
Quoted 4 lines
Quoted 1 line
Or, should I ask someone else for a tag?
It's the middle of the merge window. None of us can do much other than focus on that until 5.1-rc1 is out.
Sure, thanks anyway for the response.
Regards,
Wolfram
Quoted 8 lines · Wolfram Sang
Hi Greg,
Quoted 2 lines
Quoted 1 line
I was wondering if you think the below ABI addition looks sane to you?
I am not the i2c maintainer :)
Peter is the i2c-mux maintainer, so I trust him very much on the I2C side of things. We just wondered if you'd notice a general flaw when exposing such an interface to userspace.
Ah, I see. No, looks fine to me, it's a single value, well documented, should be just fine.
thanks,
greg k-h