device property: Introduce device_is_compatible()
Subsystems (5)
Andy Shevchenko
reviewerv2 cover device property: Introduce device_is_compatible()
Introduce a new helper to tell if device (node) is compatible to the given string value. This will help some drivers to get rid of unneeded OF APIs/etc and in may help others to be agnostic to OF/ACPI.
While doing it, I have noticed that ACPI_DEVICE_CLASS() macro seems defined in unsuitable location. Move it to the better one.
Last patch is an example of what the first two are doing.
The entire series can go, I believe, via ACPI (linux-pm) tree in case the last patch gets tag from the respective maintainer.
In v2:
- updated commit message and added kernel doc for a new API (Greg)
- also replaced acpi_device_get_match_data() with the agnostic API
- tried to keep header inclusions ordered (to some extent)
Andy Shevchenko (3):
ACPI: Move ACPI_DEVICE_CLASS() to mod_devicetable.h device property: Implement device_is_compatible()
ata: ahci_platform: Make code agnostic to OF/ACPI
drivers/ata/ahci_platform.c | 8 ++++---- include/linux/acpi.h | 14 -------------- include/linux/mod_devicetable.h | 13 +++++++++++++ include/linux/property.h | 12 ++++++++++++ 4 files changed, 29 insertions(+), 18 deletions(-)
Signature · 1 line
2.40.0.1.gaa8946217a0b
Andy Shevchenko
reviewerPATCH v2 1/3 ACPI: Move ACPI_DEVICE_CLASS() to mod_devicetable.h
The data type of struct acpi_device_id is defined in the mod_devicetable.h. It's suboptimal to require user with the almost agnostic code to include acpi.h solely for the macro that affects the data type defined elsewhere.
Taking into account the above and for the sake of consistency move ACPI_DEVICE_CLASS() to mod_devicetable.h.
Note, that with CONFIG_ACPI=n the ID table will be filed with data but it does not really matter because either it won't be used, or won't be compiled in some cases (when guarded by respective ifdeffery).
--- include/linux/acpi.h | 14 -------------- include/linux/mod_devicetable.h | 13 +++++++++++++ 2 files changed, 13 insertions(+), 14 deletions(-)
include/linux/acpi.h +0 -14
@@ -70,19 +70,6 @@ static inline void acpi_free_fwnode_static(struct fwnode_handle *fwnode)kfree (fwnode); } -/** -* ACPI_DEVICE_CLASS - macro used to describe an ACPI device with -* the PCI-defined class-code information -* -* @_cls : the class, subclass, prog-if triple for this device -* @_msk : the class mask for this device -* -* This macro is used to create a struct acpi_device_id that matches a -* specific PCI class. The .id and .driver_data fields will be left -* initialized with the default value. -*/ -#define ACPI_DEVICE_CLASS (_cls, _msk) .cls = (_cls), .cls_msk = (_msk), -static inline bool has_acpi_companion (struct device *dev) {return is_acpi_device_node (dev->fwnode);
@@ -782,7 +769,6 @@ const char *acpi_get_subsystem_id(acpi_handle handle);#define ACPI_COMPANION_SET (dev, adev) do { } while (0)#define ACPI_HANDLE (dev) (NULL)#define ACPI_HANDLE_FWNODE (fwnode) (NULL) -#define ACPI_DEVICE_CLASS (_cls, _msk) .cls = (0), .cls_msk = (0),#include <acpi/acpi_numa.h>
include/linux/mod_devicetable.h +13 -0
@@ -221,6 +221,19 @@ struct acpi_device_id {__u32 cls_msk; }; +/** +* ACPI_DEVICE_CLASS - macro used to describe an ACPI device with +* the PCI-defined class-code information +* +* @_cls : the class, subclass, prog-if triple for this device +* @_msk : the class mask for this device +* +* This macro is used to create a struct acpi_device_id that matches a +* specific PCI class. The .id and .driver_data fields will be left +* initialized with the default value. +*/ +#define ACPI_DEVICE_CLASS (_cls, _msk) .cls = (_cls), .cls_msk = (_msk), +#define PNP_ID_LEN 8#define PNP_MAX_DEVICES 8
Signature · 1 line
2.40.0.1.gaa8946217a0b
Re: [PATCH v2 1/3] ACPI: Move ACPI_DEVICE_CLASS() to mod_devicetable.h
Quoted 6 lines · Andy Shevchenko
The data type of struct
acpi_device_idis defined in themod_devicetable.h. It's suboptimal to require user with the almost agnostic code to includeacpi.hsolely for the macro that affects the data type defined elsewhere.
Quoted 8 lines
Taking into account the above and for the sake of consistency move
ACPI_DEVICE_CLASS()tomod_devicetable.h.Note, that with
CONFIG_ACPI=nthe ID table will be filed with data but it does not really matter because either it won't be used, or won't be compiled in some cases (when guarded by respective ifdeffery).Signed-off-by Andy Shevchenko <[email protected]>
or please let me know if you want me to apply this.
Quoted 64 lines
--- include/linux/acpi.h | 14 -------------- include/linux/mod_devicetable.h | 13 +++++++++++++ 2 files changed, 13 insertions(+), 14 deletions(-)include/linux/acpi.h +0
-14@@ -70,19 +70,6 @@ static inline void acpi_free_fwnode_static(struct fwnode_handle *fwnode)kfree (fwnode); } -/** -* ACPI_DEVICE_CLASS - macro used to describe an ACPI device with -* the PCI-defined class-code information -* -* @_cls : the class, subclass, prog-if triple for this device -* @_msk : the class mask for this device -* -* This macro is used to create a struct acpi_device_id that matches a -* specific PCI class. The .id and .driver_data fields will be left -* initialized with the default value. -*/ -#define ACPI_DEVICE_CLASS (_cls, _msk) .cls = (_cls), .cls_msk = (_msk), -static inline bool has_acpi_companion (struct device *dev) {return is_acpi_device_node (dev->fwnode);@@ -782,7 +769,6 @@ const char *acpi_get_subsystem_id(acpi_handle handle);#define ACPI_COMPANION_SET (dev, adev) do { } while (0)#define ACPI_HANDLE (dev) (NULL)#define ACPI_HANDLE_FWNODE (fwnode) (NULL) -#define ACPI_DEVICE_CLASS (_cls, _msk) .cls = (0), .cls_msk = (0),#include <acpi/acpi_numa.h> include/linux/mod_devicetable.h +13
-1@@ -221,6 +221,19 @@ struct acpi_device_id {__u32 cls_msk; }; +/** +* ACPI_DEVICE_CLASS - macro used to describe an ACPI device with +* the PCI-defined class-code information +* +* @_cls : the class, subclass, prog-if triple for this device +* @_msk : the class mask for this device +* +* This macro is used to create a struct acpi_device_id that matches a +* specific PCI class. The .id and .driver_data fields will be left +* initialized with the default value. +*/ +#define ACPI_DEVICE_CLASS (_cls, _msk) .cls = (_cls), .cls_msk = (_msk), +#define PNP_ID_LEN 8#define PNP_MAX_DEVICES 8 --2.40.0.1.gaa8946217a0b
Andy Shevchenko
reviewerRe: [PATCH v2 1/3] ACPI: Move ACPI_DEVICE_CLASS() to mod_devicetable.h
Quoted 15 lines · Rafael J. Wysocki
Quoted 13 lines · Andy Shevchenko
The data type of struct
acpi_device_idis defined in themod_devicetable.h. It's suboptimal to require user with the almost agnostic code to includeacpi.hsolely for the macro that affects the data type defined elsewhere.Taking into account the above and for the sake of consistency move
ACPI_DEVICE_CLASS()tomod_devicetable.h.Note, that with
CONFIG_ACPI=nthe ID table will be filed with data but it does not really matter because either it won't be used, or won't be compiled in some cases (when guarded by respective ifdeffery).
Quoted 5 lines
Quoted 1 line
Signed-off-by Andy Shevchenko <[email protected]>or please let me know if you want me to apply this.
The first patch can be applied independently. But it may be better to apply all three via one tree (linux-pm was mentioned in the cover letter) as it shows the point.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Damien Le Moal
maintainerRe: [PATCH v2 1/3] ACPI: Move ACPI_DEVICE_CLASS() to mod_devicetable.h
In reply to Rafael J. WysockiQuoted 15 lines · Rafael J. Wysocki
Quoted 13 lines · Andy Shevchenko
The data type of struct
acpi_device_idis defined in themod_devicetable.h. It's suboptimal to require user with the almost agnostic code to includeacpi.hsolely for the macro that affects the data type defined elsewhere.Taking into account the above and for the sake of consistency move
ACPI_DEVICE_CLASS()tomod_devicetable.h.Note, that with
CONFIG_ACPI=nthe ID table will be filed with data but it does not really matter because either it won't be used, or won't be compiled in some cases (when guarded by respective ifdeffery).
Quoted 5 lines
Quoted 1 line
Signed-off-by Andy Shevchenko <[email protected]>or please let me know if you want me to apply this.
Probably better if you take the whole thing. But if needed, I can take this through the ata tree.
Signature · 2 lines
Damien Le Moal Western Digital Research
Andy Shevchenko
reviewerPATCH v2 2/3 device property: Implement device_is_compatible()
In reply to Andy ShevchenkoSome users want to use the struct device pointer to see if the device is compatible in terms of Open Firmware specifications, i.e. if it has a 'compatible' property and it matches to the given value. Provide inline helper for the users.
--- include/linux/property.h | 12 ++++++++++++ 1 file changed, 12 insertions(+)
include/linux/property.h +12 -0
@@ -85,6 +85,18 @@ bool fwnode_device_is_compatible(const struct fwnode_handle *fwnode, const charreturn fwnode_property_match_string (fwnode,"compatible" , compat) >=0 ; } +/** +* device_is_compatible - match 'compatible' property of the device with a given string +* @dev: Pointer to the struct device +* @compat: The string to match 'compatible' property with +* +* Returns: true if matches, otherwise false. +*/ +static inline bool device_is_compatible (const struct device *dev,const char *compat) +{ +return fwnode_device_is_compatible (dev_fwnode (dev), compat); +} + int fwnode_property_get_reference_args(const struct fwnode_handle *fwnode, const char *prop, const char *nargs_prop, unsigned int nargs, unsigned int index,
Signature · 1 line
2.40.0.1.gaa8946217a0b
Re: [PATCH v2 2/3] device property: Implement device_is_compatible()
Hi Andy,
Quoted 6 lines · Andy Shevchenko
Some users want to use the struct device pointer to see if the device is compatible in terms of Open Firmware specifications, i.e. if it has a 'compatible' property and it matches to the given value. Provide inline helper for the users.
Signed-off-by Andy Shevchenko <[email protected]>
Signature · 1 line
Sakari Ailus
Re: [PATCH v2 2/3] device property: Implement device_is_compatible()
In reply to Andy ShevchenkoQuoted 4 lines · Andy Shevchenko
Some users want to use the struct device pointer to see if the device is compatible in terms of Open Firmware specifications, i.e. if it has a 'compatible' property and it matches to the given value. Provide inline helper for the users.
IMO much useful wrapper. Thanks for the patch.
-Serge(y)
Quoted 33 lines
Signed-off-by Andy Shevchenko <[email protected]>--- include/linux/property.h | 12 ++++++++++++ 1 file changed, 12 insertions(+)include/linux/property.h +12
-0@@ -85,6 +85,18 @@ bool fwnode_device_is_compatible(const struct fwnode_handle *fwnode, const charreturn fwnode_property_match_string (fwnode,"compatible" , compat) >=0 ; } +/** +* device_is_compatible - match 'compatible' property of the device with a given string +* @dev: Pointer to the struct device +* @compat: The string to match 'compatible' property with +* +* Returns: true if matches, otherwise false. +*/ +static inline bool device_is_compatible (const struct device *dev,const char *compat) +{ +return fwnode_device_is_compatible (dev_fwnode (dev), compat); +} + int fwnode_property_get_reference_args(const struct fwnode_handle *fwnode, const char *prop, const char *nargs_prop, unsigned int nargs, unsigned int index,Signature · 1 line
2.40.0.1.gaa8946217a0b
Andy Shevchenko
reviewerRe: [PATCH v2 2/3] device property: Implement device_is_compatible()
Quoted 8 lines · Serge Semin
Quoted 4 lines · Andy Shevchenko
Some users want to use the struct device pointer to see if the device is compatible in terms of Open Firmware specifications, i.e. if it has a 'compatible' property and it matches to the given value. Provide inline helper for the users.
IMO much useful wrapper. Thanks for the patch.
Reviewed-by Serge Semin <[email protected]>
There is a v3. Do you mind to tag it as well?
Signature · 2 lines
With Best Regards, Andy Shevchenko
Andy Shevchenko
reviewerRe: [PATCH v2 2/3] device property: Implement device_is_compatible()
Quoted 11 lines · Andy Shevchenko
Quoted 8 lines · Serge Semin
Quoted 4 lines · Andy Shevchenko
Some users want to use the struct device pointer to see if the device is compatible in terms of Open Firmware specifications, i.e. if it has a 'compatible' property and it matches to the given value. Provide inline helper for the users.
IMO much useful wrapper. Thanks for the patch.
Reviewed-by Serge Semin <[email protected]>There is a v3. Do you mind to tag it as well?
Ah, and thank you for review! Appreciate it.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Re: [PATCH v2 2/3] device property: Implement device_is_compatible()
Quoted 8 lines · Andy Shevchenko
Quoted 7 lines · Andy Shevchenko
Quoted 6 lines · Serge Semin
Quoted 4 lines · Andy Shevchenko
Some users want to use the struct device pointer to see if the device is compatible in terms of Open Firmware specifications, i.e. if it has a 'compatible' property and it matches to the given value. Provide inline helper for the users.
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
IMO much useful wrapper. Thanks for the patch.
Reviewed-by Serge Semin <[email protected]>There is a v3. Do you mind to tag it as well?
Ah, and thank you for review! Appreciate it.
Always welcome.) I've resent the tags for the patches 2 and 3. Patch 1 is out of my competence, sorry.
-Serge(y)
Quoted 7 lines
Signature · 3 lines
-- With Best Regards, Andy Shevchenko
Andy Shevchenko
reviewerPATCH v2 3/3 ata: ahci_platform: Make code agnostic to OF/ACPI
In reply to Andy ShevchenkoWith the help of a new device_is_compatible() make the driver code agnostic to the OF/ACPI. This makes it neater. As a side effect the header inclusions is corrected (seems mod_devicetable.h was implicitly included).
--- drivers/ata/ahci_platform.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/ata/ahci_platform.c +4 -4
@@ -9,14 +9,14 @@*/ #include <linux/kernel.h> +#include <linux/mod_devicetable.h> #include <linux/module.h> #include <linux/pm.h> #include <linux/device.h> -#include <linux/of_device.h> #include <linux/platform_device.h> +#include <linux/property.h> #include <linux/libata.h> #include <linux/ahci_platform.h> -#include <linux/acpi.h> #include <linux/pci_ids.h> #include "ahci.h"
@@ -56,10 +56,10 @@ static int ahci_probe(struct platform_device *pdev)if (rc)return rc; -if (of_device_is_compatible (dev->of_node,"hisilicon,hisi-ahci" )) +if (device_is_compatible (dev,"hisilicon,hisi-ahci" )) hpriv->flags |=AHCI_HFLAG_NO_FBS |AHCI_HFLAG_NO_NCQ ; - port =acpi_device_get_match_data (dev); + port =device_get_match_data (dev);if (!port) port = &ahci_port_info;
Signature · 1 line
2.40.0.1.gaa8946217a0b
Re: [PATCH v2 3/3] ata: ahci_platform: Make code agnostic to OF/ACPI
Hi Andy,
Quoted 5 lines · Andy Shevchenko
With the help of a new
device_is_compatible()make the driver code agnostic to the OF/ACPI. This makes it neater. As a side effect the header inclusions is corrected (seemsmod_devicetable.hwas implicitly included).
You're wrapping the lines well before 75. Why?
Quoted 2 lines
Signed-off-by Andy Shevchenko <[email protected]>
Signature · 1 line
Sakari Ailus
Andy Shevchenko
reviewerRe: [PATCH v2 3/3] ata: ahci_platform: Make code agnostic to OF/ACPI
Quoted 8 lines · Sakari Ailus
Quoted 5 lines · Andy Shevchenko
With the help of a new
device_is_compatible()make the driver code agnostic to the OF/ACPI. This makes it neater. As a side effect the header inclusions is corrected (seemsmod_devicetable.hwas implicitly included).You're wrapping the lines well before 75. Why?
Didn't pay attention to that much. Is it a problem? Should I send a new version because of that?
Quoted 3 lines
Quoted 1 line
Signed-off-by Andy Shevchenko <[email protected]>Reviewed-by Sakari Ailus <[email protected]>
Thank you!
Signature · 2 lines
With Best Regards, Andy Shevchenko
Sakari Ailus
reviewerRe: [PATCH v2 3/3] ata: ahci_platform: Make code agnostic to OF/ACPI
Quoted 8 lines · Andy Shevchenko
Quoted 7 lines · Sakari Ailus
Quoted 5 lines · Andy Shevchenko
With the help of a new
device_is_compatible()make the driver code agnostic to the OF/ACPI. This makes it neater. As a side effect the header inclusions is corrected (seemsmod_devicetable.hwas implicitly included).
Quoted 4 lines
Quoted 1 line
You're wrapping the lines well before 75. Why?
Didn't pay attention to that much. Is it a problem? Should I send a new version because of that?
I guess not. But it's a good practice to wrap at 75 instead.
Signature · 1 line
Sakari Ailus
Sakari Ailus
reviewerRe: [PATCH v2 3/3] ata: ahci_platform: Make code agnostic to OF/ACPI
In reply to Andy ShevchenkoHi Andy,
Quoted 33 lines · Andy Shevchenko
With the help of a new
device_is_compatible()make the driver code agnostic to the OF/ACPI. This makes it neater. As a side effect the header inclusions is corrected (seemsmod_devicetable.hwas implicitly included).Signed-off-by Andy Shevchenko <[email protected]>--- drivers/ata/ahci_platform.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)drivers/ata/ahci_platform.c +2
-2@@ -9,14 +9,14 @@*/ #include <linux/kernel.h> +#include <linux/mod_devicetable.h> #include <linux/module.h> #include <linux/pm.h> #include <linux/device.h> -#include <linux/of_device.h> #include <linux/platform_device.h> +#include <linux/property.h> #include <linux/libata.h> #include <linux/ahci_platform.h> -#include <linux/acpi.h> #include <linux/pci_ids.h> #include "ahci.h"
drivers/ata/ahci_platform.c static int ahci_probe(struct platform_device *pdev) · line 62
Quoted 9 lines
drivers/ata/ahci_platform.c +2
-2@@ -56,10 +56,10 @@ static int ahci_probe(struct platform_device *pdev)if (rc)return rc; -if (of_device_is_compatible (dev->of_node,"hisilicon,hisi-ahci" )) +if (device_is_compatible (dev,"hisilicon,hisi-ahci" )) hpriv->flags |=AHCI_HFLAG_NO_FBS |AHCI_HFLAG_NO_NCQ ; - port =acpi_device_get_match_data (dev); + port =device_get_match_data (dev);
There are just a handful of users for acpi_device_get_match_data() in the tree. The code could be moved to acpi_fwnode_device_get_match_data() after coverting these. May be out of scope of this set though.
Quoted 3 lines
drivers/ata/ahci_platform.c +0
-0if (!port) port = &ahci_port_info;
Signature · 3 lines
Kind regards, Sakari Ailus
Andy Shevchenko
reviewerRe: [PATCH v2 3/3] ata: ahci_platform: Make code agnostic to OF/ACPI
Quoted 1 line · Sakari Ailus
On Fri, Jun 09, 2023 at 06:49:00PM +0300, Andy Shevchenko wrote:
...
Quoted 10 lines
drivers/ata/ahci_platform.c static int ahci_probe(struct platform_device *pdev) · line 62
Quoted 6 lines
drivers/ata/ahci_platform.c +2
-2-if (of_device_is_compatible (dev->of_node,"hisilicon,hisi-ahci" )) +if (device_is_compatible (dev,"hisilicon,hisi-ahci" )) hpriv->flags |=AHCI_HFLAG_NO_FBS |AHCI_HFLAG_NO_NCQ ; - port =acpi_device_get_match_data (dev); + port =device_get_match_data (dev);There are just a handful of users for
acpi_device_get_match_data()in the tree. The code could be moved toacpi_fwnode_device_get_match_data()after coverting these. May be out of scope of this set though.
Why do we need that one if we can use device_get_match_data() directly? It will be also flexible in case one of OF code will need something like this (custom info structure for the respective compatible string). That said, I don't think we need to change to acpi_*() whatever.
Signature · 2 lines
With Best Regards, Andy Shevchenko
Sakari Ailus
reviewerRe: [PATCH v2 3/3] ata: ahci_platform: Make code agnostic to OF/ACPI
Hi Andy,
Quoted 9 lines · Andy Shevchenko
Quoted 1 line · Sakari Ailus
On Fri, Jun 09, 2023 at 06:49:00PM +0300, Andy Shevchenko wrote:
...
Quoted 4 lines
Quoted 4 lines
drivers/ata/ahci_platform.c +1
-1-if (of_device_is_compatible (dev->of_node,"hisilicon,hisi-ahci" )) +if (device_is_compatible (dev,"hisilicon,hisi-ahci" )) hpriv->flags |=AHCI_HFLAG_NO_FBS |AHCI_HFLAG_NO_NCQ ;
Quoted 8 lines
Quoted 6 lines
drivers/ata/ahci_platform.c static int ahci_probe(struct platform_device *pdev) · line 62
Quoted 2 lines
drivers/ata/ahci_platform.c +1
-1- port =acpi_device_get_match_data (dev); + port =device_get_match_data (dev);There are just a handful of users for
acpi_device_get_match_data()in the tree. The code could be moved toacpi_fwnode_device_get_match_data()after coverting these. May be out of scope of this set though.Why do we need that one if we can use
device_get_match_data()directly?
That was what I wanted to point your attention to. ;-)
Quoted 3 lines
It will be also flexible in case one of OF code will need something like this (custom info structure for the respective compatible string). That said, I don't think we need to change to acpi_*() whatever.
I agree.
Signature · 3 lines
Kind regards, Sakari Ailus
Re: [PATCH v2 3/3] ata: ahci_platform: Make code agnostic to OF/ACPI
In reply to Andy ShevchenkoQuoted 5 lines · Andy Shevchenko
With the help of a new
device_is_compatible()make the driver code agnostic to the OF/ACPI. This makes it neater. As a side effect the header inclusions is corrected (seemsmod_devicetable.hwas implicitly included).
I don't think the driver will get to be fully agnostic after this patch because for instance the ahci_platform_get_resources() method directly uses the OF-available functions, walks over the OF subnodes, touches the OF-properties, etc. So AFAICS in order to be fully OF/ACPI agnostic the entire libahci_platform.o driver needs to be converted too, but it's not trivial at all.
Anyway as a start this patch looks good.
-Serge(y)
Quoted 44 lines
Signed-off-by Andy Shevchenko <[email protected]>--- drivers/ata/ahci_platform.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)drivers/ata/ahci_platform.c +4
-4@@ -9,14 +9,14 @@*/ #include <linux/kernel.h> +#include <linux/mod_devicetable.h> #include <linux/module.h> #include <linux/pm.h> #include <linux/device.h> -#include <linux/of_device.h> #include <linux/platform_device.h> +#include <linux/property.h> #include <linux/libata.h> #include <linux/ahci_platform.h> -#include <linux/acpi.h> #include <linux/pci_ids.h> #include "ahci.h" @@ -56,10 +56,10 @@ static int ahci_probe(struct platform_device *pdev)if (rc)return rc; -if (of_device_is_compatible (dev->of_node,"hisilicon,hisi-ahci" )) +if (device_is_compatible (dev,"hisilicon,hisi-ahci" )) hpriv->flags |=AHCI_HFLAG_NO_FBS |AHCI_HFLAG_NO_NCQ ; - port =acpi_device_get_match_data (dev); + port =device_get_match_data (dev);if (!port) port = &ahci_port_info;Signature · 1 line
2.40.0.1.gaa8946217a0b