Add sysfs interface files to hv_gpci device to expose system information
v4 cover Add sysfs interface files to hv_gpci device to expose system information
The hcall H_GET_PERF_COUNTER_INFO can be used to get data related to chips, dimms and system topology, by passing different counter request values.
Patchset adds sysfs files to "/sys/devices/hv_gpci/interface/" of hv_gpci pmu driver, which will expose system topology information using H_GET_PERF_COUNTER_INFO hcall. The added sysfs files are available for power10 and above platforms and needs root access to read the data.
Patches 1,3,5,7,9 adds sysfs interface files to the hv_gpci pmu driver, to get system topology information.
List of added sysfs files:
-> processor_bus_topology (Counter request value : 0xD0)
-> processor_config (Counter request value : 0x90)
-> affinity_domain_via_virtual_processor (Counter request value : 0xA0) -> affinity_domain_via_domain (Counter request value : 0xB0) -> affinity_domain_via_partition (Counter request value : 0xB1)
Patches 2,4,6,8,10 adds details of the newly added hv_gpci interface files listed above in the ABI documentation.
Patches 2,4,6,8,10 adds details of the newly added hv_gpci interface files listed above in the ABI documentation.
Changelog:
v3 -> v4
-> Add Reviewed-by tag from Athira Rajeev.
-> Correct typo by changing processor_bug_topology to processor_bus_topology
v2 -> v3
-> Make nit changes in documentation patches as suggested by Randy Dunlap.
v1 -> v2
-> Incase the HCALL fails with errors that can be resolve during runtime,
then only add sysinfo interface attributes to the interface_attrs attribute array. Even if one of the counter request value HCALL fails, don't add any sysinfo attribute to the interface_attrs attribute array. Add the code changes to make sure sysinfo interface added only when all the requirements met as suggested by Michael Ellerman.
-> Make changes in documentation, adds detail of errors type
which can be resolved at runtime as suggested by Michael Ellerman.
-> Add new enum and sysinfo_counter_request array to get required
counter request value in hv-gpci.c file.
-> Move the macros for interface attribute array index to hv-gpci.c, as
these macros currently only used in hv-gpci.c file.
Kajol Jain (10):
powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show
processor bus topology information
docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document
processor_bus_topology sysfs interface file
powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show
processor config information
docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document
processor_config sysfs interface file
powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity
domain via virtual processor information
docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document
affinity_domain_via_virtual_processor sysfs interface file
powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity
domain via domain information
docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document
affinity_domain_via_domain sysfs interface file
powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity
domain via partition information
docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document
affinity_domain_via_partition sysfs interface file
.../sysfs-bus-event_source-devices-hv_gpci | 160 +++++ arch/powerpc/perf/hv-gpci.c | 640 +++++++++++++++++- 2 files changed, 798 insertions(+), 2 deletions(-)
Signature · 1 line
2.39.3
PATCH v4 1/10 powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show processor bus topology information
The hcall H_GET_PERF_COUNTER_INFO with counter request value as PROCESSOR_BUS_TOPOLOGY(0XD0), can be used to get the system topology information. To expose the system topology information, patch adds sysfs file called "processor_bus_topology" to the "/sys/devices/hv_gpci/interface/" of hv_gpci pmu driver.
Add macro for PROCESSOR_BUS_TOPOLOGY counter request value in hv-gpci.c file. Also add a new function called "systeminfo_gpci_request", to make the H_GET_PERF_COUNTER_INFO hcall with added macro and populates the output buffer.
The processor_bus_topology sysfs file is only available for power10 and above platforms. Add a new function called "add_sysinfo_interface_files", which will add processor_bus_topology attribute in the interface_attrs array, only for power10 and above platforms.
Also add macro INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR in hv-gpci.c file, which points to the index of NULL placefolder, for processor_bus_topology attribute.
--- arch/powerpc/perf/hv-gpci.c | 184 +++++++++++++++++++++++++++++++++++- 1 file changed, 182 insertions(+), 2 deletions(-)
arch/powerpc/perf/hv-gpci.c +182 -2
@@ -102,6 +102,141 @@ static ssize_t cpumask_show(struct device *dev,return cpumap_print_to_pagebuf (true, buf, &hv_gpci_cpumask); } +/* Counter request value to retrieve system information */ +#define PROCESSOR_BUS_TOPOLOGY 0XD0 + +/* Interface attribute array index to store system information */ +#define INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR 6 + +static DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t)); + +static unsigned long systeminfo_gpci_request (u32 req,u32 starting_index, +u16 secondary_index,char *buf, +size_t *n,struct hv_gpci_request_buffer *arg) +{ +unsigned long ret; +size_t i, j; + + arg->params.counter_request =cpu_to_be32 (req); + arg->params.starting_index =cpu_to_be32 (starting_index); + arg->params.secondary_index =cpu_to_be16 (secondary_index); + + ret =plpar_hcall_norets (H_GET_PERF_COUNTER_INFO , +virt_to_phys (arg),HGPCI_REQ_BUFFER_SIZE ); + +/* +* ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', +* which means that the current buffer size cannot accommodate +* all the information and a partial buffer returned. +* hcall fails incase of ret value other than H_SUCCESS or H_PARAMETER. +* +* ret value as H_AUTHORITY implies that partition is not permitted to retrieve +* performance information, and required to set +* "Enable Performance Information Collection" option. +*/ +if (ret ==H_AUTHORITY ) +return -EPERM ; + +/* +* hcall can fail with other possible ret value like H_PRIVILEGE/H_HARDWARE +* because of invalid buffer-length/address or due to some hardware +* error. +*/ +if (ret && (ret !=H_PARAMETER )) +return -EIO ; + +/* +* hcall H_GET_PERF_COUNTER_INFO populates the 'returned_values' +* to show the total number of counter_value array elements +* returned via hcall. +* hcall also populates 'cv_element_size' corresponds to individual +* counter_value array element size. Below loop go through all +* counter_value array elements as per their size and add it to +* the output buffer. +*/ +for (i =0 ; i <be16_to_cpu (arg->params.returned_values); i++) { + j = i *be16_to_cpu (arg->params.cv_element_size); + +for (; j < (i +1 ) *be16_to_cpu (arg->params.cv_element_size); j++) + *n +=sprintf (buf + *n,"%02x" , (u8 )arg->bytes[j]); + *n +=sprintf (buf + *n,"\n" ); + } + +if (*n >=PAGE_SIZE ) { +pr_info ("System information exceeds PAGE_SIZE\n" ); +return -EFBIG ; + } + +return ret; +} + +static ssize_t processor_bus_topology_show (struct device *dev,struct device_attribute *attr, +char *buf) +{ +struct hv_gpci_request_buffer *arg; +unsigned long ret; +size_t n =0 ; + + arg = (void *)get_cpu_var (hv_gpci_reqb); +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + +/* +* Pass the counter request value 0xD0 corresponds to request +* type 'Processor_bus_topology', to retrieve +* the system topology information. +* starting_index value implies the starting hardware +* chip id. +*/ + ret =systeminfo_gpci_request (PROCESSOR_BUS_TOPOLOGY ,0 ,0 , buf, &n, arg); + +if (!ret) +return n; + +if (ret !=H_PARAMETER ) + goto out; + +/* +* ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', which +* implies that buffer can't accommodate all information, and a partial buffer +* returned. To handle that, we need to make subsequent requests +* with next starting index to retrieve additional (missing) data. +* Below loop do subsequent hcalls with next starting index and add it +* to buffer util we get all the information. +*/ +while (ret ==H_PARAMETER ) { +int returned_values =be16_to_cpu (arg->params.returned_values); +int elementsize =be16_to_cpu (arg->params.cv_element_size); +int last_element = (returned_values -1 ) * elementsize; + +/* +* Since the starting index value is part of counter_value +* buffer elements, use the starting index value in the last +* element and add 1 to make subsequent hcalls. +*/ +u32 starting_index = arg->bytes[last_element +3 ] + + (arg->bytes[last_element +2 ] <<8 ) + + (arg->bytes[last_element +1 ] <<16 ) + + (arg->bytes[last_element] <<24 ) +1 ; + +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + + ret =systeminfo_gpci_request (PROCESSOR_BUS_TOPOLOGY , starting_index, +0 , buf, &n, arg); + +if (!ret) +return n; + +if (ret !=H_PARAMETER ) + goto out; + } + +return n; + +out: +put_cpu_var (hv_gpci_reqb); +return ret; +} +static DEVICE_ATTR_RO (kernel_version);static DEVICE_ATTR_RO (cpumask);
@@ -118,6 +253,11 @@ static struct attribute *interface_attrs[] = { &hv_caps_attr_expanded.attr, &hv_caps_attr_lab.attr, &hv_caps_attr_collect_privileged.attr, +/* +* This NULL is a placeholder for the processor_bus_topology +* attribute, set in init function if applicable. +*/ +NULL , NULL, };
@@ -143,8 +283,6 @@ static const struct attribute_group *attr_groups[] = { NULL, }; -static DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t)); -static unsigned long single_gpci_request (u32 req,u32 starting_index,u16 secondary_index,u8 version_in,u32 offset,u8 length,u64 *value)
@@ -325,6 +463,44 @@ static int hv_gpci_cpu_hotplug_init(void) ppc_hv_gpci_cpu_offline); } +static void add_sysinfo_interface_files (void ) +{ +struct device_attribute *attr =NULL ; +unsigned long ret; +struct hv_gpci_request_buffer *arg; + +/* Check for counter request type PROCESSOR_BUS_TOPOLOGY support */ + arg = (void *)get_cpu_var (hv_gpci_reqb); +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + + arg->params.counter_request =cpu_to_be32 (PROCESSOR_BUS_TOPOLOGY ); + + ret =plpar_hcall_norets (H_GET_PERF_COUNTER_INFO , +virt_to_phys (arg),HGPCI_REQ_BUFFER_SIZE ); + +put_cpu_var (hv_gpci_reqb); + +/* +* Add processor_bus_topology attribute in the interface_attrs +* attribute array, only for valid return types. +*/ +if (!ret || ret ==H_AUTHORITY || ret ==H_PARAMETER ) { + attr =kzalloc (sizeof (*attr),GFP_KERNEL ); + +if (!attr) { +pr_info ("Memory allocation failed for sysinfo interface files\n" ); +return ; + } + +sysfs_attr_init (&attr->attr); + attr->attr.name ="processor_bus_topology" ; + attr->attr.mode =0444 ; + attr->show = processor_bus_topology_show; + interface_attrs[INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR ] = &attr->attr; + }else +pr_devel ("hcall failed, with error: 0x%lx\n" , ret); +} +static int hv_gpci_init (void ) {int r;
@@ -388,6 +564,10 @@ static int hv_gpci_init(void)if (r)return r; +/* sysinfo interface files are only available for power10 and above platforms */ +if (PVR_VER (mfspr (SPRN_PVR )) >=PVR_POWER10 ) +add_sysinfo_interface_files (); +return 0 ; }
Signature · 1 line
2.39.3
PATCH v4 2/10 docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document processor_bus_topology sysfs interface file
In reply to Kajol JainAdd details of the new hv-gpci interface file called "processor_bus_topology" in the ABI documentation.
--- .../sysfs-bus-event_source-devices-hv_gpci | 32 +++++++++++++++++++ 1 file changed, 32 insertions(+)
Documentation/ABI/testing/sysfs-bus-event_source-devices-hv_gpci +32 -0
@@ -80,3 +80,35 @@ Contact: Linux on PowerPC Developer List <[email protected]> Description: read only This sysfs file exposes the cpumask which is designated to make HCALLs to retrieve hv-gpci pmu event counter data. + +What: /sys/devices/hv_gpci/interface/processor_bus_topology +Date: July 2023 +Contact: Linux on PowerPC Developer List <[email protected]> +Description: admin read only + This sysfs file exposes the system topology information by making HCALL + H_GET_PERF_COUNTER_INFO. The HCALL is made with counter request value + PROCESSOR_BUS_TOPOLOGY(0xD0). + + * This sysfs file will be created only for power10 and above platforms. + + * User needs root privileges to read data from this sysfs file. + + * This sysfs file will be created, only when the HCALL returns "H_SUCCESS", + "H_AUTHORITY" or "H_PARAMETER" as the return type. + + HCALL with return error type "H_AUTHORITY" can be resolved during + runtime by setting "Enable Performance Information Collection" option. + + * The end user reading this sysfs file must decode the content as per + underlying platform/firmware. + + Possible error codes while reading this sysfs file: + + * "-EPERM" : Partition is not permitted to retrieve performance information, + required to set "Enable Performance Information Collection" option. + + * "-EIO" : Can't retrieve system information because of invalid buffer length/invalid address + or because of some hardware error. Refer to getPerfCountInfo documentation for + more information. + + * "-EFBIG" : System information exceeds PAGE_SIZE.
Signature · 1 line
2.39.3
PATCH v4 3/10 powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show processor config information
In reply to Kajol JainThe hcall H_GET_PERF_COUNTER_INFO with counter request value as PROCESSOR_CONFIG(0X90), can be used to get the system processor configuration information. To expose the system processor config information, patch adds sysfs file called "processor_config" to the "/sys/devices/hv_gpci/interface/" of hv_gpci pmu driver.
Add enum and sysinfo_counter_request array to get required counter request value in hv-gpci.c file.
Also add a new function called "sysinfo_device_attr_create", which will create and return required device attribute to the add_sysinfo_interface_files function.
The processor_config sysfs file is only available for power10 and above platforms. Add a new macro called INTERFACE_PROCESSOR_CONFIG_ATTR, which points to the index of NULL placefolder, for processor_config attribute in the interface_attrs array. Also add macro INTERFACE_NULL_ATTR which points to index of NULL attribute in interface_attrs array.
--- arch/powerpc/perf/hv-gpci.c | 168 ++++++++++++++++++++++++++++++++---- 1 file changed, 153 insertions(+), 15 deletions(-)
arch/powerpc/perf/hv-gpci.c +153 -15
@@ -102,11 +102,21 @@ static ssize_t cpumask_show(struct device *dev,return cpumap_print_to_pagebuf (true, buf, &hv_gpci_cpumask); } -/* Counter request value to retrieve system information */ -#define PROCESSOR_BUS_TOPOLOGY 0XD0 -/* Interface attribute array index to store system information */ #define INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR 6 +#define INTERFACE_PROCESSOR_CONFIG_ATTR 7 +#define INTERFACE_NULL_ATTR 8 + +/* Counter request value to retrieve system information */ +enum { +PROCESSOR_BUS_TOPOLOGY , +PROCESSOR_CONFIG +}; + +static int sysinfo_counter_request[] = { + [PROCESSOR_BUS_TOPOLOGY ] =0xD0 , + [PROCESSOR_CONFIG ] =0x90 , +}; static DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t));
@@ -187,7 +197,8 @@ static ssize_t processor_bus_topology_show(struct device *dev, struct device_att* starting_index value implies the starting hardware * chip id. */ - ret =systeminfo_gpci_request (PROCESSOR_BUS_TOPOLOGY ,0 ,0 , buf, &n, arg); + ret =systeminfo_gpci_request (sysinfo_counter_request[PROCESSOR_BUS_TOPOLOGY ], +0 ,0 , buf, &n, arg);if (!ret)return n;
@@ -220,8 +231,76 @@ static ssize_t processor_bus_topology_show(struct device *dev, struct device_attmemset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); - ret =systeminfo_gpci_request (PROCESSOR_BUS_TOPOLOGY , starting_index, -0 , buf, &n, arg); + ret =systeminfo_gpci_request (sysinfo_counter_request[PROCESSOR_BUS_TOPOLOGY ], + starting_index,0 , buf, &n, arg); + +if (!ret) +return n; + +if (ret !=H_PARAMETER ) + goto out; + } + +return n; + +out: +put_cpu_var (hv_gpci_reqb); +return ret; +} + +static ssize_t processor_config_show (struct device *dev,struct device_attribute *attr, +char *buf) +{ +struct hv_gpci_request_buffer *arg; +unsigned long ret; +size_t n =0 ; + + arg = (void *)get_cpu_var (hv_gpci_reqb); +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + +/* +* Pass the counter request value 0x90 corresponds to request +* type 'Processor_config', to retrieve +* the system processor information. +* starting_index value implies the starting hardware +* processor index. +*/ + ret =systeminfo_gpci_request (sysinfo_counter_request[PROCESSOR_CONFIG ], +0 ,0 , buf, &n, arg); + +if (!ret) +return n; + +if (ret !=H_PARAMETER ) + goto out; + +/* +* ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', which +* implies that buffer can't accommodate all information, and a partial buffer +* returned. To handle that, we need to take subsequent requests +* with next starting index to retrieve additional (missing) data. +* Below loop do subsequent hcalls with next starting index and add it +* to buffer util we get all the information. +*/ +while (ret ==H_PARAMETER ) { +int returned_values =be16_to_cpu (arg->params.returned_values); +int elementsize =be16_to_cpu (arg->params.cv_element_size); +int last_element = (returned_values -1 ) * elementsize; + +/* +* Since the starting index is part of counter_value +* buffer elements, use the starting index value in the last +* element and add 1 to subsequent hcalls. +*/ +u32 starting_index = arg->bytes[last_element +3 ] + + (arg->bytes[last_element +2 ] <<8 ) + + (arg->bytes[last_element +1 ] <<16 ) + + (arg->bytes[last_element] <<24 ) +1 ; + +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + + ret =systeminfo_gpci_request (sysinfo_counter_request[PROCESSOR_CONFIG ], + starting_index,0 , buf, &n, arg);if (!ret)return n;
@@ -258,6 +337,11 @@ static struct attribute *interface_attrs[] = {* attribute, set in init function if applicable. */ NULL , +/* +* This NULL is a placeholder for the processor_config +* attribute, set in init function if applicable. +*/ +NULL , NULL, };
@@ -463,17 +547,24 @@ static int hv_gpci_cpu_hotplug_init(void) ppc_hv_gpci_cpu_offline); } -static void add_sysinfo_interface_files (void ) +static struct device_attribute *sysinfo_device_attr_create (int + sysinfo_interface_group_index,u32 req) {struct device_attribute *attr =NULL ;unsigned long ret;struct hv_gpci_request_buffer *arg; -/* Check for counter request type PROCESSOR_BUS_TOPOLOGY support */ +if (sysinfo_interface_group_index <INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR || + sysinfo_interface_group_index >=INTERFACE_NULL_ATTR ) { +pr_info ("Wrong interface group index for system information\n" ); +return NULL ; + } + +/* Check for given counter request value support */ arg = (void *)get_cpu_var (hv_gpci_reqb);memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); - arg->params.counter_request =cpu_to_be32 (PROCESSOR_BUS_TOPOLOGY ); + arg->params.counter_request =cpu_to_be32 (req); ret =plpar_hcall_norets (H_GET_PERF_COUNTER_INFO ,virt_to_phys (arg),HGPCI_REQ_BUFFER_SIZE );
@@ -481,24 +572,71 @@ static void add_sysinfo_interface_files(void)put_cpu_var (hv_gpci_reqb);/* -* Add processor_bus_topology attribute in the interface_attrs +* Add given counter request value attribute in the interface_attrs * attribute array, only for valid return types. */ if (!ret || ret ==H_AUTHORITY || ret ==H_PARAMETER ) { attr =kzalloc (sizeof (*attr),GFP_KERNEL );if (!attr) { -pr_info ("Memory allocation failed for sysinfo interface files\n" ); -return ; +pr_info ("Mem allocation failed, for sysinfo interface file" ); +return NULL ; }sysfs_attr_init (&attr->attr); - attr->attr.name ="processor_bus_topology" ; attr->attr.mode =0444 ; - attr->show = processor_bus_topology_show; - interface_attrs[INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR ] = &attr->attr; + +switch (sysinfo_interface_group_index) { +case INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR : + attr->attr.name ="processor_bus_topology" ; + attr->show = processor_bus_topology_show; +break ; +case INTERFACE_PROCESSOR_CONFIG_ATTR : + attr->attr.name ="processor_config" ; + attr->show = processor_config_show; +break ; + } }else pr_devel ("hcall failed, with error: 0x%lx\n" , ret); + +return attr; +} + +static void add_sysinfo_interface_files (void ) +{ +int sysfs_count; +struct device_attribute *attr[INTERFACE_NULL_ATTR -INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR ]; +int i; + + sysfs_count =INTERFACE_NULL_ATTR -INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR ; + +/* Get device attribute for a given counter request value */ +for (i =0 ; i < sysfs_count; i++) { + attr[i] =sysinfo_device_attr_create (i +INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR , + sysinfo_counter_request[i]); + +if (!attr[i]) + goto out; + } + +/* Add sysinfo interface attributes in the interface_attrs attribute array */ +for (i =0 ; i < sysfs_count; i++) + interface_attrs[i +INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR ] = &attr[i]->attr; + +return ; + +out: +/* +* The sysinfo interface attributes will be added, only if hcall passed for +* all the counter request values. Free the device attribute array incase +* of any hcall failure. +*/ +if (i >0 ) { +while (i >=0 ) { +kfree (attr[i]); + i--; + } + } }static int hv_gpci_init (void )
Signature · 1 line
2.39.3
PATCH v4 4/10 docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document processor_config sysfs interface file
In reply to Kajol JainAdd details of the new hv-gpci interface file called "processor_config" in the ABI documentation.
--- .../sysfs-bus-event_source-devices-hv_gpci | 32 +++++++++++++++++++ 1 file changed, 32 insertions(+)
Documentation/ABI/testing/sysfs-bus-event_source-devices-hv_gpci +32 -0
@@ -112,3 +112,35 @@ Description: admin read only more information. * "-EFBIG" : System information exceeds PAGE_SIZE. + +What: /sys/devices/hv_gpci/interface/processor_config +Date: July 2023 +Contact: Linux on PowerPC Developer List <[email protected]> +Description: admin read only + This sysfs file exposes the system topology information by making HCALL + H_GET_PERF_COUNTER_INFO. The HCALL is made with counter request value + PROCESSOR_CONFIG(0x90). + + * This sysfs file will be created only for power10 and above platforms. + + * User needs root privileges to read data from this sysfs file. + + * This sysfs file will be created, only when the HCALL returns "H_SUCCESS", + "H_AUTHORITY" or "H_PARAMETER" as the return type. + + HCALL with return error type "H_AUTHORITY" can be resolved during + runtime by setting "Enable Performance Information Collection" option. + + * The end user reading this sysfs file must decode the content as per + underlying platform/firmware. + + Possible error codes while reading this sysfs file: + + * "-EPERM" : Partition is not permitted to retrieve performance information, + required to set "Enable Performance Information Collection" option. + + * "-EIO" : Can't retrieve system information because of invalid buffer length/invalid address + or because of some hardware error. Refer to getPerfCountInfo documentation for + more information. + + * "-EFBIG" : System information exceeds PAGE_SIZE.
Signature · 1 line
2.39.3
PATCH v4 5/10 powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity domain via virtual processor information
In reply to Kajol JainThe hcall H_GET_PERF_COUNTER_INFO with counter request value as AFFINITY_DOMAIN_INFORMATION_BY_VIRTUAL_PROCESSOR(0XA0), can be used to get the system affinity domain via virtual processor information. To expose the system affinity domain via virtual processor information, patch adds sysfs file called "affinity_domain_via_virtual_processor" to the "/sys/devices/hv_gpci/interface/" of hv_gpci pmu driver.
The affinity_domain_via_virtual_processor sysfs file is only available for power10 and above platforms. Add a macro called INTERFACE_AFFINITY_DOMAIN_VIA_VP_ATTR, which points to the index of NULL placeholder, for affinity_domain_via_virtual_processor attribute in interface_attrs array. Also updated the value of INTERFACE_NULL_ATTR macro in hv-gpci.c file.
--- arch/powerpc/perf/hv-gpci.c | 86 ++++++++++++++++++++++++++++++++++++- 1 file changed, 84 insertions(+), 2 deletions(-)
arch/powerpc/perf/hv-gpci.c +84 -2
@@ -105,17 +105,20 @@ static ssize_t cpumask_show(struct device *dev,/* Interface attribute array index to store system information */ #define INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR 6#define INTERFACE_PROCESSOR_CONFIG_ATTR 7 -#define INTERFACE_NULL_ATTR 8 +#define INTERFACE_AFFINITY_DOMAIN_VIA_VP_ATTR 8 +#define INTERFACE_NULL_ATTR 9/* Counter request value to retrieve system information */ enum {PROCESSOR_BUS_TOPOLOGY , -PROCESSOR_CONFIG +PROCESSOR_CONFIG , +AFFINITY_DOMAIN_VIA_VP ,/* affinity domain via virtual processor */ };static int sysinfo_counter_request[] = { [PROCESSOR_BUS_TOPOLOGY ] =0xD0 , [PROCESSOR_CONFIG ] =0x90 , + [AFFINITY_DOMAIN_VIA_VP ] =0xA0 , }; static DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t));
@@ -316,6 +319,76 @@ static ssize_t processor_config_show(struct device *dev, struct device_attributereturn ret; } +static ssize_t affinity_domain_via_virtual_processor_show (struct device *dev, +struct device_attribute *attr,char *buf) +{ +struct hv_gpci_request_buffer *arg; +unsigned long ret; +size_t n =0 ; + + arg = (void *)get_cpu_var (hv_gpci_reqb); +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + +/* +* Pass the counter request 0xA0 corresponds to request +* type 'Affinity_domain_information_by_virutal_processor', +* to retrieve the system affinity domain information. +* starting_index value refers to the starting hardware +* processor index. +*/ + ret =systeminfo_gpci_request (sysinfo_counter_request[AFFINITY_DOMAIN_VIA_VP ], +0 ,0 , buf, &n, arg); + +if (!ret) +return n; + +if (ret !=H_PARAMETER ) + goto out; + +/* +* ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', which +* implies that buffer can't accommodate all information, and a partial buffer +* returned. To handle that, we need to take subsequent requests +* with next secondary index to retrieve additional (missing) data. +* Below loop do subsequent hcalls with next secondary index and add it +* to buffer util we get all the information. +*/ +while (ret ==H_PARAMETER ) { +int returned_values =be16_to_cpu (arg->params.returned_values); +int elementsize =be16_to_cpu (arg->params.cv_element_size); +int last_element = (returned_values -1 ) * elementsize; + +/* +* Since the starting index and secondary index type is part of the +* counter_value buffer elements, use the starting index value in the +* last array element as subsequent starting index, and use secondary index +* value in the last array element plus 1 as subsequent secondary index. +* For counter request '0xA0', starting index points to partition id +* and secondary index points to corresponding virtual processor index. +*/ +u32 starting_index = arg->bytes[last_element +1 ] + (arg->bytes[last_element] <<8 ); +u16 secondary_index = arg->bytes[last_element +3 ] + + (arg->bytes[last_element +2 ] <<8 ) +1 ; + +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + + ret =systeminfo_gpci_request (sysinfo_counter_request[AFFINITY_DOMAIN_VIA_VP ], + starting_index, secondary_index, buf, &n, arg); + +if (!ret) +return n; + +if (ret !=H_PARAMETER ) + goto out; + } + +return n; + +out: +put_cpu_var (hv_gpci_reqb); +return ret; +} +static DEVICE_ATTR_RO (kernel_version);static DEVICE_ATTR_RO (cpumask);
@@ -342,6 +415,11 @@ static struct attribute *interface_attrs[] = {* attribute, set in init function if applicable. */ NULL , +/* +* This NULL is a placeholder for the affinity_domain_via_virtual_processor +* attribute, set in init function if applicable. +*/ +NULL , NULL, };
@@ -595,6 +673,10 @@ static struct device_attribute *sysinfo_device_attr_create(int attr->attr.name ="processor_config" ; attr->show = processor_config_show;break ; +case INTERFACE_AFFINITY_DOMAIN_VIA_VP_ATTR : + attr->attr.name ="affinity_domain_via_virtual_processor" ; + attr->show = affinity_domain_via_virtual_processor_show; +break ; } }else pr_devel ("hcall failed, with error: 0x%lx\n" , ret);
Signature · 1 line
2.39.3
PATCH v4 6/10 docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document affinity_domain_via_virtual_processor sysfs interface file
In reply to Kajol JainAdd details of the new hv-gpci interface file called "affinity_domain_via_virtual_processor" in the ABI documentation.
--- .../sysfs-bus-event_source-devices-hv_gpci | 32 +++++++++++++++++++ 1 file changed, 32 insertions(+)
Documentation/ABI/testing/sysfs-bus-event_source-devices-hv_gpci +32 -0
@@ -144,3 +144,35 @@ Description: admin read only more information. * "-EFBIG" : System information exceeds PAGE_SIZE. + +What: /sys/devices/hv_gpci/interface/affinity_domain_via_virtual_processor +Date: July 2023 +Contact: Linux on PowerPC Developer List <[email protected]> +Description: admin read only + This sysfs file exposes the system topology information by making HCALL + H_GET_PERF_COUNTER_INFO. The HCALL is made with counter request value + AFFINITY_DOMAIN_INFORMATION_BY_VIRTUAL_PROCESSOR(0xA0). + + * This sysfs file will be created only for power10 and above platforms. + + * User needs root privileges to read data from this sysfs file. + + * This sysfs file will be created, only when the HCALL returns "H_SUCCESS", + "H_AUTHORITY" or "H_PARAMETER" as the return type. + + HCALL with return error type "H_AUTHORITY" can be resolved during + runtime by setting "Enable Performance Information Collection" option. + + * The end user reading this sysfs file must decode the content as per + underlying platform/firmware. + + Possible error codes while reading this sysfs file: + + * "-EPERM" : Partition is not permitted to retrieve performance information, + required to set "Enable Performance Information Collection" option. + + * "-EIO" : Can't retrieve system information because of invalid buffer length/invalid address + or because of some hardware error. Refer to getPerfCountInfo documentation for + more information. + + * "-EFBIG" : System information exceeds PAGE_SIZE.
Signature · 1 line
2.39.3
PATCH v4 7/10 powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity domain via domain information
In reply to Kajol JainThe hcall H_GET_PERF_COUNTER_INFO with counter request value as AFFINITY_DOMAIN_INFORMATION_BY_DOMAIN(0XB0), can be used to get the system affinity domain via domain information. To expose the system affinity domain via domain information, patch adds sysfs file called "affinity_domain_via_domain" to the "/sys/devices/hv_gpci/interface/" of hv_gpci pmu driver.
Add new entry for AFFINITY_DOMAIN_VIA_DOM in sysinfo_counter_request array, which points to the counter request value "affinity_domain_via_domain" in hv-gpci.c file.
The affinity_domain_via_domain sysfs file is only available for power10 and above platforms. Add a macro called INTERFACE_AFFINITY_DOMAIN_VIA_DOM_ATTR, which points to the index of NULL placeholder, for affinity_domain_via_domain attribute in interface_attrs array. Also updated the value of INTERFACE_NULL_ATTR macro in hv-gpci.c file.
--- arch/powerpc/perf/hv-gpci.c | 80 ++++++++++++++++++++++++++++++++++++- 1 file changed, 79 insertions(+), 1 deletion(-)
arch/powerpc/perf/hv-gpci.c +79 -1
@@ -106,19 +106,22 @@ static ssize_t cpumask_show(struct device *dev,#define INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR 6#define INTERFACE_PROCESSOR_CONFIG_ATTR 7#define INTERFACE_AFFINITY_DOMAIN_VIA_VP_ATTR 8 -#define INTERFACE_NULL_ATTR 9 +#define INTERFACE_AFFINITY_DOMAIN_VIA_DOM_ATTR 9 +#define INTERFACE_NULL_ATTR 10/* Counter request value to retrieve system information */ enum {PROCESSOR_BUS_TOPOLOGY ,PROCESSOR_CONFIG ,AFFINITY_DOMAIN_VIA_VP ,/* affinity domain via virtual processor */ +AFFINITY_DOMAIN_VIA_DOM ,/* affinity domain via domain */ };static int sysinfo_counter_request[] = { [PROCESSOR_BUS_TOPOLOGY ] =0xD0 , [PROCESSOR_CONFIG ] =0x90 , [AFFINITY_DOMAIN_VIA_VP ] =0xA0 , + [AFFINITY_DOMAIN_VIA_DOM ] =0xB0 , }; static DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t));
@@ -389,6 +392,72 @@ static ssize_t affinity_domain_via_virtual_processor_show(struct device *dev,return ret; } +static ssize_t affinity_domain_via_domain_show (struct device *dev,struct device_attribute *attr, +char *buf) +{ +struct hv_gpci_request_buffer *arg; +unsigned long ret; +size_t n =0 ; + + arg = (void *)get_cpu_var (hv_gpci_reqb); +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + +/* +* Pass the counter request 0xB0 corresponds to request +* type 'Affinity_domain_information_by_domain', +* to retrieve the system affinity domain information. +* starting_index value refers to the starting hardware +* processor index. +*/ + ret =systeminfo_gpci_request (sysinfo_counter_request[AFFINITY_DOMAIN_VIA_DOM ], +0 ,0 , buf, &n, arg); + +if (!ret) +return n; + +if (ret !=H_PARAMETER ) + goto out; + +/* +* ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', which +* implies that buffer can't accommodate all information, and a partial buffer +* returned. To handle that, we need to take subsequent requests +* with next starting index to retrieve additional (missing) data. +* Below loop do subsequent hcalls with next starting index and add it +* to buffer util we get all the information. +*/ +while (ret ==H_PARAMETER ) { +int returned_values =be16_to_cpu (arg->params.returned_values); +int elementsize =be16_to_cpu (arg->params.cv_element_size); +int last_element = (returned_values -1 ) * elementsize; + +/* +* Since the starting index value is part of counter_value +* buffer elements, use the starting index value in the last +* element and add 1 to make subsequent hcalls. +*/ +u32 starting_index = arg->bytes[last_element +1 ] + + (arg->bytes[last_element] <<8 ) +1 ; + +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + + ret =systeminfo_gpci_request (sysinfo_counter_request[AFFINITY_DOMAIN_VIA_DOM ], + starting_index,0 , buf, &n, arg); + +if (!ret) +return n; + +if (ret !=H_PARAMETER ) + goto out; + } + +return n; + +out: +put_cpu_var (hv_gpci_reqb); +return ret; +} +static DEVICE_ATTR_RO (kernel_version);static DEVICE_ATTR_RO (cpumask);
@@ -420,6 +489,11 @@ static struct attribute *interface_attrs[] = {* attribute, set in init function if applicable. */ NULL , +/* +* This NULL is a placeholder for the affinity_domain_via_domain +* attribute, set in init function if applicable. +*/ +NULL , NULL, };
@@ -677,6 +751,10 @@ static struct device_attribute *sysinfo_device_attr_create(int attr->attr.name ="affinity_domain_via_virtual_processor" ; attr->show = affinity_domain_via_virtual_processor_show;break ; +case INTERFACE_AFFINITY_DOMAIN_VIA_DOM_ATTR : + attr->attr.name ="affinity_domain_via_domain" ; + attr->show = affinity_domain_via_domain_show; +break ; } }else pr_devel ("hcall failed, with error: 0x%lx\n" , ret);
Signature · 1 line
2.39.3
PATCH v4 8/10 docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document affinity_domain_via_domain sysfs interface file
In reply to Kajol JainAdd details of the new hv-gpci interface file called "affinity_domain_via_domain" in the ABI documentation.
--- .../sysfs-bus-event_source-devices-hv_gpci | 32 +++++++++++++++++++ 1 file changed, 32 insertions(+)
Documentation/ABI/testing/sysfs-bus-event_source-devices-hv_gpci +32 -0
@@ -176,3 +176,35 @@ Description: admin read only more information. * "-EFBIG" : System information exceeds PAGE_SIZE. + +What: /sys/devices/hv_gpci/interface/affinity_domain_via_domain +Date: July 2023 +Contact: Linux on PowerPC Developer List <[email protected]> +Description: admin read only + This sysfs file exposes the system topology information by making HCALL + H_GET_PERF_COUNTER_INFO. The HCALL is made with counter request value + AFFINITY_DOMAIN_INFORMATION_BY_DOMAIN(0xB0). + + * This sysfs file will be created only for power10 and above platforms. + + * User needs root privileges to read data from this sysfs file. + + * This sysfs file will be created, only when the HCALL returns "H_SUCCESS", + "H_AUTHORITY" or "H_PARAMETER" as the return type. + + HCALL with return error type "H_AUTHORITY" can be resolved during + runtime by setting "Enable Performance Information Collection" option. + + * The end user reading this sysfs file must decode the content as per + underlying platform/firmware. + + Possible error codes while reading this sysfs file: + + * "-EPERM" : Partition is not permitted to retrieve performance information, + required to set "Enable Performance Information Collection" option. + + * "-EIO" : Can't retrieve system information because of invalid buffer length/invalid address + or because of some hardware error. Refer to getPerfCountInfo documentation for + more information. + + * "-EFBIG" : System information exceeds PAGE_SIZE.
Signature · 1 line
2.39.3
PATCH v4 9/10 powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity domain via partition information
In reply to Kajol JainThe hcall H_GET_PERF_COUNTER_INFO with counter request value as AFFINITY_DOMAIN_INFORMATION_BY_PARTITION(0XB1), can be used to get the system affinity domain via partition information. To expose the system affinity domain via partition information, patch adds sysfs file called "affinity_domain_via_partition" to the "/sys/devices/hv_gpci/interface/" of hv_gpci pmu driver.
Add new entry for AFFINITY_DOMAIN_VIA_PAR in sysinfo_counter_request array, which points to the counter request value "affinity_domain_via_partition" in hv-gpci.c file. Also add a new function called "affinity_domain_via_partition_result_parse" to parse the hcall result and store it in output buffer.
The affinity_domain_via_partition sysfs file is only available for power10 and above platforms. Add a macro called INTERFACE_AFFINITY_DOMAIN_VIA_PAR_ATTR, which points to the index of NULL placeholder, for affinity_domain_via_partition attribute in interface_attrs array. Also updated the value of INTERFACE_NULL_ATTR macro in hv-gpci.c file.
--- arch/powerpc/perf/hv-gpci.c | 160 +++++++++++++++++++++++++++++++++++- 1 file changed, 159 insertions(+), 1 deletion(-)
arch/powerpc/perf/hv-gpci.c +159 -1
@@ -107,7 +107,8 @@ static ssize_t cpumask_show(struct device *dev,#define INTERFACE_PROCESSOR_CONFIG_ATTR 7#define INTERFACE_AFFINITY_DOMAIN_VIA_VP_ATTR 8#define INTERFACE_AFFINITY_DOMAIN_VIA_DOM_ATTR 9 -#define INTERFACE_NULL_ATTR 10 +#define INTERFACE_AFFINITY_DOMAIN_VIA_PAR_ATTR 10 +#define INTERFACE_NULL_ATTR 11/* Counter request value to retrieve system information */ enum {
@@ -115,6 +116,7 @@ enum {PROCESSOR_CONFIG ,AFFINITY_DOMAIN_VIA_VP ,/* affinity domain via virtual processor */ AFFINITY_DOMAIN_VIA_DOM ,/* affinity domain via domain */ + AFFINITY_DOMAIN_VIA_PAR, /* affinity domain via partition */ }; static int sysinfo_counter_request[] = {
@@ -122,6 +124,7 @@ static int sysinfo_counter_request[] = { [PROCESSOR_CONFIG] = 0x90, [AFFINITY_DOMAIN_VIA_VP] = 0xA0, [AFFINITY_DOMAIN_VIA_DOM] = 0xB0, + [AFFINITY_DOMAIN_VIA_PAR] = 0xB1, }; static DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t));
@@ -458,6 +461,152 @@ static ssize_t affinity_domain_via_domain_show(struct device *dev, struct devicereturn ret; } +static void affinity_domain_via_partition_result_parse (int returned_values, +int element_size,char *buf,size_t *last_element, +size_t *n,struct hv_gpci_request_buffer *arg) +{ +size_t i =0 , j =0 ; +size_t k, l, m; +uint16_t total_affinity_domain_ele, size_of_each_affinity_domain_ele; + +/* +* hcall H_GET_PERF_COUNTER_INFO populates the 'returned_values' +* to show the total number of counter_value array elements +* returned via hcall. +* Unlike other request types, the data structure returned by this +* request is variable-size. For this counter request type, +* hcall populates 'cv_element_size' corresponds to minimum size of +* the structure returned i.e; the size of the structure with no domain +* information. Below loop go through all counter_value array +* to determine the number and size of each domain array element and +* add it to the output buffer. +*/ +while (i < returned_values) { + k = j; +for (; k < j + element_size; k++) + *n +=sprintf (buf + *n,"%02x" , (u8 )arg->bytes[k]); + *n +=sprintf (buf + *n,"\n" ); + + total_affinity_domain_ele = (u8 )arg->bytes[k -2 ] <<8 | (u8 )arg->bytes[k -3 ]; + size_of_each_affinity_domain_ele = (u8 )arg->bytes[k] <<8 | (u8 )arg->bytes[k -1 ]; + +for (l =0 ; l < total_affinity_domain_ele; l++) { +for (m =0 ; m < size_of_each_affinity_domain_ele; m++) { + *n +=sprintf (buf + *n,"%02x" , (u8 )arg->bytes[k]); + k++; + } + *n +=sprintf (buf + *n,"\n" ); + } + + *n +=sprintf (buf + *n,"\n" ); + i++; + j = k; + } + + *last_element = k; +} + +static ssize_t affinity_domain_via_partition_show (struct device *dev,struct device_attribute *attr, +char *buf) +{ +struct hv_gpci_request_buffer *arg; +unsigned long ret; +size_t n =0 ; +size_t last_element =0 ; +u32 starting_index; + + arg = (void *)get_cpu_var (hv_gpci_reqb); +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + +/* +* Pass the counter request value 0xB1 corresponds to counter request +* type 'Affinity_domain_information_by_partition', +* to retrieve the system affinity domain by partition information. +* starting_index value refers to the starting hardware +* processor index. +*/ + arg->params.counter_request =cpu_to_be32 (sysinfo_counter_request[AFFINITY_DOMAIN_VIA_PAR ]); + arg->params.starting_index =cpu_to_be32 (0 ); + + ret =plpar_hcall_norets (H_GET_PERF_COUNTER_INFO , +virt_to_phys (arg),HGPCI_REQ_BUFFER_SIZE ); + +if (!ret) + goto parse_result; + +/* +* ret value as 'H_PARAMETER' implies that the current buffer size +* can't accommodate all the information, and a partial buffer +* returned. To handle that, we need to make subsequent requests +* with next starting index to retrieve additional (missing) data. +* Below loop do subsequent hcalls with next starting index and add it +* to buffer util we get all the information. +*/ +while (ret ==H_PARAMETER ) { +affinity_domain_via_partition_result_parse ( +be16_to_cpu (arg->params.returned_values) -1 , +be16_to_cpu (arg->params.cv_element_size), buf, + &last_element, &n, arg); + +if (n >=PAGE_SIZE ) { +put_cpu_var (hv_gpci_reqb); +pr_debug ("System information exceeds PAGE_SIZE\n" ); +return -EFBIG ; + } + +/* +* Since the starting index value is part of counter_value +* buffer elements, use the starting_index value in the last +* element and add 1 to make subsequent hcalls. +*/ + starting_index = (u8 )arg->bytes[last_element] <<8 | + (u8 )arg->bytes[last_element +1 ]; + +memset (arg,0 ,HGPCI_REQ_BUFFER_SIZE ); + arg->params.counter_request =cpu_to_be32 ( + sysinfo_counter_request[AFFINITY_DOMAIN_VIA_PAR ]); + arg->params.starting_index =cpu_to_be32 (starting_index); + + ret =plpar_hcall_norets (H_GET_PERF_COUNTER_INFO , +virt_to_phys (arg),HGPCI_REQ_BUFFER_SIZE ); + +if (ret && (ret !=H_PARAMETER )) + goto out; + } + +parse_result: +affinity_domain_via_partition_result_parse ( +be16_to_cpu (arg->params.returned_values), +be16_to_cpu (arg->params.cv_element_size), + buf, &last_element, &n, arg); + +put_cpu_var (hv_gpci_reqb); +return n; + +out: +put_cpu_var (hv_gpci_reqb); + +/* +* ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', +* which means that the current buffer size cannot accommodate +* all the information and a partial buffer returned. +* hcall fails incase of ret value other than H_SUCCESS or H_PARAMETER. +* +* ret value as H_AUTHORITY implies that partition is not permitted to retrieve +* performance information, and required to set +* "Enable Performance Information Collection" option. +*/ +if (ret ==H_AUTHORITY ) +return -EPERM ; + +/* +* hcall can fail with other possible ret value like H_PRIVILEGE/H_HARDWARE +* because of invalid buffer-length/address or due to some hardware +* error. +*/ +return -EIO ; +} +static DEVICE_ATTR_RO (kernel_version);static DEVICE_ATTR_RO (cpumask);
@@ -494,6 +643,11 @@ static struct attribute *interface_attrs[] = {* attribute, set in init function if applicable. */ NULL , +/* +* This NULL is a placeholder for the affinity_domain_via_partition +* attribute, set in init function if applicable. +*/ +NULL , NULL, };
@@ -755,6 +909,10 @@ static struct device_attribute *sysinfo_device_attr_create(int attr->attr.name ="affinity_domain_via_domain" ; attr->show = affinity_domain_via_domain_show;break ; +case INTERFACE_AFFINITY_DOMAIN_VIA_PAR_ATTR : + attr->attr.name ="affinity_domain_via_partition" ; + attr->show = affinity_domain_via_partition_show; +break ; } }else pr_devel ("hcall failed, with error: 0x%lx\n" , ret);
Signature · 1 line
2.39.3
PATCH v4 10/10 docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document affinity_domain_via_partition sysfs interface file
In reply to Kajol JainAdd details of the new hv-gpci interface file called "affinity_domain_via_partition" in the ABI documentation.
--- .../sysfs-bus-event_source-devices-hv_gpci | 32 +++++++++++++++++++ 1 file changed, 32 insertions(+)
Documentation/ABI/testing/sysfs-bus-event_source-devices-hv_gpci +32 -0
@@ -208,3 +208,35 @@ Description: admin read only more information. * "-EFBIG" : System information exceeds PAGE_SIZE. + +What: /sys/devices/hv_gpci/interface/affinity_domain_via_partition +Date: July 2023 +Contact: Linux on PowerPC Developer List <[email protected]> +Description: admin read only + This sysfs file exposes the system topology information by making HCALL + H_GET_PERF_COUNTER_INFO. The HCALL is made with counter request value + AFFINITY_DOMAIN_INFORMATION_BY_PARTITION(0xB1). + + * This sysfs file will be created only for power10 and above platforms. + + * User needs root privileges to read data from this sysfs file. + + * This sysfs file will be created, only when the HCALL returns "H_SUCCESS", + "H_AUTHORITY" or "H_PARAMETER" as the return type. + + HCALL with return error type "H_AUTHORITY" can be resolved during + runtime by setting "Enable Performance Information Collection" option. + + * The end user reading this sysfs file must decode the content as per + underlying platform/firmware. + + Possible error codes while reading this sysfs file: + + * "-EPERM" : Partition is not permitted to retrieve performance information, + required to set "Enable Performance Information Collection" option. + + * "-EIO" : Can't retrieve system information because of invalid buffer length/invalid address + or because of some hardware error. Refer to getPerfCountInfo documentation for + more information. + + * "-EFBIG" : System information exceeds PAGE_SIZE.
Signature · 1 line
2.39.3
Michael Ellerman
AppliedQuoted 10 lines · Kajol Jain
The hcall
H_GET_PERF_COUNTER_INFOcan be used to get data related to chips, dimms and system topology, by passing different counter request values.
Patchset adds sysfs files to "/sys/devices/hv_gpci/interface/" ofhv_gpcipmu driver, which will expose system topology information usingH_GET_PERF_COUNTER_INFOhcall. The added sysfs files are available for power10 and above platforms and needs root access to read the data.[...]
Applied to powerpc/next.
[01/10] powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show processor bus topology information
https://git.kernel.org/powerpc/c/71f1c39647d8c9d4d54a861ec81f1ff17544bcb6 [02/10] docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document processor_bus_topology sysfs interface file https://git.kernel.org/powerpc/c/9caf9e2b8bae58e39501f6fb1fc9189009538ccc [03/10] powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show processor config information https://git.kernel.org/powerpc/c/1a160c2a13c66c9ad47436e73c821f3d26818733 [04/10] docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document processor_config sysfs interface file https://git.kernel.org/powerpc/c/3255171d3e861b14e4248a5c676099819cadfb6d [05/10] powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity domain via virtual processor information https://git.kernel.org/powerpc/c/71a7ccb478fcfe1495bcabf4972d859b24264df7 [06/10] docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document affinity_domain_via_virtual_processor sysfs interface file https://git.kernel.org/powerpc/c/cc89ff3491b61cebc90b3394eb6b36635173d0dd [07/10] powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity domain via domain information https://git.kernel.org/powerpc/c/a69a57cac1ec8995bb0b571dfccc3fe2f046719a [08/10] docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document affinity_domain_via_domain sysfs interface file https://git.kernel.org/powerpc/c/ab7e991286e729b0018722591c04eb2fd31771b0 [09/10] powerpc/hv_gpci: Add sysfs file inside hv_gpci device to show affinity domain via partition information https://git.kernel.org/powerpc/c/a15e0d6a6929e737f71578ed4b05531fed5a96e8 [10/10] docs: ABI: sysfs-bus-event_source-devices-hv_gpci: Document affinity_domain_via_partition sysfs interface file https://git.kernel.org/powerpc/c/8df99066940b6c82a0851f13adf653827dc524f7
cheers