hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OWS6 CVE: NA

------------------------------------------------------------------

Platform devices which supports power control are often required to be power off/on together with the devices in the same power domain. However, there isn't a generic driver that support the power control logic of these devices.

ACPI container seems to be a good place to hold these control logic. Add platform devices in the same power domain in a ACPI container, we can easily get the locality information about these devices and can moniter the power of these devices in the same power domain together.

This patch provide three userspace control interface to control the power of devices together in the container: - state: Echo online to state to power up the devices in the container and then online these devices which will be triggered by BIOS. Echo offline to the state to offline and eject the child devices in the container which are ejectable. - pxms: show the pxms of devices which are present in the container.

In our scenario, we need to control the power of HBM memory devices which can be power consuming and will only be used in some specialized scenarios, such as HPC. HBM memory devices in a socket are in the same power domain, and should be power off/on together. We have come up with an idea that put these power control logic in a specialized driver, but ACPI container seems to be a more generic place to hold these control logic.

Signed-off-by: Zhang Zekun zhangzekun11@huawei.com --- drivers/base/container.c | 3 + drivers/soc/hisilicon/Kconfig | 12 ++ drivers/soc/hisilicon/Makefile | 2 + drivers/soc/hisilicon/hisi_hbmdev.c | 240 ++++++++++++++++++++++++++ drivers/soc/hisilicon/hisi_internal.h | 31 ++++ 5 files changed, 288 insertions(+) create mode 100644 drivers/soc/hisilicon/hisi_hbmdev.c create mode 100644 drivers/soc/hisilicon/hisi_internal.h

diff --git a/drivers/base/container.c b/drivers/base/container.c index 1ba42d2d3532..055dfa8daff1 100644 --- a/drivers/base/container.c +++ b/drivers/base/container.c @@ -30,6 +30,9 @@ struct bus_type container_subsys = { .online = trivial_online, .offline = container_offline, }; +#if IS_ENABLED(CONFIG_HISI_HBMDEV) +EXPORT_SYMBOL_GPL(container_subsys); +#endif

void __init container_dev_init(void) { diff --git a/drivers/soc/hisilicon/Kconfig b/drivers/soc/hisilicon/Kconfig index 0ab688af308f..3e5cf16be223 100644 --- a/drivers/soc/hisilicon/Kconfig +++ b/drivers/soc/hisilicon/Kconfig @@ -3,6 +3,18 @@ menu "Hisilicon SoC drivers" depends on ARCH_HISI || COMPILE_TEST

+config HISI_HBMDEV + tristate "add extra support for hbm memory device" + depends on ACPI_HOTPLUG_MEMORY + select ACPI_CONTAINER + help + This driver add extra supports for memory devices. The driver + provides methods for userpace to control the power of memory + devices in a container. + + To compile this driver as a module, choose M here: + the module will be called hisi_hbmdev. + config KUNPENG_HCCS tristate "HCCS driver on Kunpeng SoC" depends on ACPI diff --git a/drivers/soc/hisilicon/Makefile b/drivers/soc/hisilicon/Makefile index 226e747e70d6..4d5794b150b3 100644 --- a/drivers/soc/hisilicon/Makefile +++ b/drivers/soc/hisilicon/Makefile @@ -1,2 +1,4 @@ # SPDX-License-Identifier: GPL-2.0-only obj-$(CONFIG_KUNPENG_HCCS) += kunpeng_hccs.o + +obj-$(CONFIG_HISI_HBMDEV) += hisi_hbmdev.o diff --git a/drivers/soc/hisilicon/hisi_hbmdev.c b/drivers/soc/hisilicon/hisi_hbmdev.c new file mode 100644 index 000000000000..0a0977956042 --- /dev/null +++ b/drivers/soc/hisilicon/hisi_hbmdev.c @@ -0,0 +1,240 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) Huawei Technologies Co., Ltd. 2023. All rights reserved. + */ + +#include <linux/kobject.h> +#include <linux/module.h> +#include <linux/nodemask.h> +#include <linux/acpi.h> +#include <linux/container.h> + +#include "hisi_internal.h" + +#define ACPI_MEMORY_DEVICE_HID "PNP0C80" +#define ACPI_GENERIC_CONTAINER_DEVICE_HID "PNP0A06" + +struct cdev_node { + struct device *dev; + struct list_head clist; +}; + +struct memory_dev { + struct kobject *memdev_kobj; + struct cdev_node cdev_list; +}; + +static struct memory_dev *mdev; + +static int get_pxm(struct acpi_device *acpi_device, void *arg) +{ + acpi_handle handle = acpi_device->handle; + nodemask_t *mask = arg; + unsigned long long sta; + acpi_status status; + int nid; + + status = acpi_evaluate_integer(handle, "_STA", NULL, &sta); + if (ACPI_SUCCESS(status) && (sta & ACPI_STA_DEVICE_ENABLED)) { + nid = acpi_get_node(handle); + if (nid >= 0) + node_set(nid, *mask); + } + + return 0; +} + +static ssize_t pxms_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct acpi_device *adev = ACPI_COMPANION(dev); + nodemask_t mask; + + nodes_clear(mask); + acpi_dev_for_each_child(adev, get_pxm, &mask); + + return sysfs_emit(buf, "%*pbl\n", + nodemask_pr_args(&mask)); +} +static DEVICE_ATTR_RO(pxms); + +static int memdev_power_on(struct acpi_device *adev) +{ + acpi_handle handle = adev->handle; + acpi_status status; + + status = acpi_evaluate_object(handle, "_ON", NULL, NULL); + if (ACPI_FAILURE(status)) { + acpi_handle_warn(handle, "Power on failed (0x%x)\n", status); + return -ENODEV; + } + + return 0; +} + +static int eject_device(struct acpi_device *acpi_device, void *not_used) +{ + acpi_object_type unused; + acpi_status status; + + status = acpi_get_type(acpi_device->handle, &unused); + if (ACPI_FAILURE(status) || !acpi_device->flags.ejectable) + return -ENODEV; + + get_device(&acpi_device->dev); + status = acpi_hotplug_schedule(acpi_device, ACPI_OST_EC_OSPM_EJECT); + if (ACPI_SUCCESS(status)) + return 0; + + put_device(&acpi_device->dev); + acpi_evaluate_ost(acpi_device->handle, ACPI_OST_EC_OSPM_EJECT, + ACPI_OST_SC_NON_SPECIFIC_FAILURE, NULL); + + return status == AE_NO_MEMORY ? -ENOMEM : -EAGAIN; +} + +static int memdev_power_off(struct acpi_device *adev) +{ + return acpi_dev_for_each_child(adev, eject_device, NULL); +} + +static ssize_t state_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct acpi_device *adev = ACPI_COMPANION(dev); + const int type = online_type_from_str(buf); + int ret = -EINVAL; + + switch (type) { + case STATE_ONLINE: + ret = memdev_power_on(adev); + break; + case STATE_OFFLINE: + ret = memdev_power_off(adev); + break; + default: + break; + } + + if (ret) + return ret; + + return count; +} +static DEVICE_ATTR_WO(state); + +static int hbmdev_find(struct acpi_device *adev, void *arg) +{ + const char *hid = acpi_device_hid(adev); + bool *found = arg; + + if (!strcmp(hid, ACPI_MEMORY_DEVICE_HID)) { + *found = true; + return -1; + } + + return 0; +} + +static bool has_hbmdev(struct device *dev) +{ + struct acpi_device *adev = ACPI_COMPANION(dev); + const char *hid = acpi_device_hid(adev); + bool found = false; + + if (strcmp(hid, ACPI_GENERIC_CONTAINER_DEVICE_HID)) + return found; + + acpi_dev_for_each_child(adev, hbmdev_find, &found); + + return found; +} + +static int container_add(struct device *dev, void *data) +{ + struct cdev_node *cnode; + + if (!has_hbmdev(dev)) + return 0; + + cnode = kmalloc(sizeof(struct cdev_node), GFP_KERNEL); + if (!cnode) + return -ENOMEM; + + cnode->dev = dev; + list_add_tail(&cnode->clist, &mdev->cdev_list.clist); + + return 0; +} + +static void container_remove(void) +{ + struct cdev_node *cnode, *tmp; + + list_for_each_entry_safe(cnode, tmp, &mdev->cdev_list.clist, clist) { + device_remove_file(cnode->dev, &dev_attr_state); + device_remove_file(cnode->dev, &dev_attr_pxms); + list_del(&cnode->clist); + kfree(cnode); + } +} + +static int container_init(void) +{ + struct cdev_node *cnode; + + INIT_LIST_HEAD(&mdev->cdev_list.clist); + + if (bus_for_each_dev(&container_subsys, NULL, NULL, container_add)) { + container_remove(); + return -ENOMEM; + } + + if (list_empty(&mdev->cdev_list.clist)) + return -ENODEV; + + list_for_each_entry(cnode, &mdev->cdev_list.clist, clist) { + device_create_file(cnode->dev, &dev_attr_state); + device_create_file(cnode->dev, &dev_attr_pxms); + } + + return 0; +} + + +static int __init mdev_init(void) +{ + int ret; + + mdev = kzalloc(sizeof(struct memory_dev), GFP_KERNEL); + if (!mdev) + return -ENOMEM; + + ret = container_init(); + if (ret) { + kfree(mdev); + return ret; + } + + mdev->memdev_kobj = kobject_create_and_add("hbm_memory", kernel_kobj); + if (!mdev->memdev_kobj) { + container_remove(); + kfree(mdev); + return -ENOMEM; + } + + return ret; +} +module_init(mdev_init); + +static void __exit mdev_exit(void) +{ + container_remove(); + kobject_put(mdev->memdev_kobj); + kfree(mdev); +} +module_exit(mdev_exit); + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Zhang Zekun zhangzekun11@huawei.com"); diff --git a/drivers/soc/hisilicon/hisi_internal.h b/drivers/soc/hisilicon/hisi_internal.h new file mode 100644 index 000000000000..5345174f6b84 --- /dev/null +++ b/drivers/soc/hisilicon/hisi_internal.h @@ -0,0 +1,31 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (C) Huawei Technologies Co., Ltd. 2023. All rights reserved. + */ + +#ifndef _HISI_INTERNAL_H +#define _HISI_INTERNAL_H + +enum { + STATE_ONLINE, + STATE_OFFLINE, +}; + +static const char *const online_type_to_str[] = { + [STATE_ONLINE] = "online", + [STATE_OFFLINE] = "offline", +}; + +static inline int online_type_from_str(const char *str) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(online_type_to_str); i++) { + if (sysfs_streq(str, online_type_to_str[i])) + return i; + } + + return -EINVAL; +} + +#endif

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OWS6 CVE: NA

--------------------------------------------------

Provide additional memory device topology information in hisi_hbmdev. This memory topology informaton can provide useful information for userspace to select the closest memory device to a certain cpu.

Signed-off-by: Zhang Zekun zhangzekun11@huawei.com --- drivers/soc/hisilicon/Kconfig | 9 +++-- drivers/soc/hisilicon/hisi_hbmdev.c | 61 +++++++++++++++++++++++++++++ 2 files changed, 67 insertions(+), 3 deletions(-)

diff --git a/drivers/soc/hisilicon/Kconfig b/drivers/soc/hisilicon/Kconfig index 3e5cf16be223..6792f4a4f62b 100644 --- a/drivers/soc/hisilicon/Kconfig +++ b/drivers/soc/hisilicon/Kconfig @@ -8,9 +8,12 @@ config HISI_HBMDEV depends on ACPI_HOTPLUG_MEMORY select ACPI_CONTAINER help - This driver add extra supports for memory devices. The driver - provides methods for userpace to control the power of memory - devices in a container. + This driver add two extra supports for memory devices. The driver + provides methods for userpace to control the power of memory devices + in a container. Besides, it provides extra locality information + between cpus and memory devices for userspace, which can take + advantage of this functionality to select the closet memory device + to a certain cpu.

To compile this driver as a module, choose M here: the module will be called hisi_hbmdev. diff --git a/drivers/soc/hisilicon/hisi_hbmdev.c b/drivers/soc/hisilicon/hisi_hbmdev.c index 0a0977956042..a4b72d232e19 100644 --- a/drivers/soc/hisilicon/hisi_hbmdev.c +++ b/drivers/soc/hisilicon/hisi_hbmdev.c @@ -8,6 +8,9 @@ #include <linux/nodemask.h> #include <linux/acpi.h> #include <linux/container.h> +#include <linux/node.h> +#include <linux/arch_topology.h> +#include <linux/memory_hotplug.h>

#include "hisi_internal.h"

@@ -21,11 +24,67 @@ struct cdev_node {

struct memory_dev { struct kobject *memdev_kobj; + struct kobject *topo_kobj; struct cdev_node cdev_list; + nodemask_t cluster_cpumask[MAX_NUMNODES]; };

static struct memory_dev *mdev;

+static ssize_t memory_locality_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *buf) +{ + int i, count = 0; + + for (i = 0; i < MAX_NUMNODES; i++) { + if (hotplug_mdev[i] != NULL && !nodes_empty(mdev->cluster_cpumask[i])) { + count += sysfs_emit_at(buf, count, "%d %*pbl\n", i, + nodemask_pr_args(&mdev->cluster_cpumask[i])); + } + } + + return count; +} + +static struct kobj_attribute memory_locality_attribute = + __ATTR(memory_locality, 0444, memory_locality_show, NULL); + +static void memory_topo_init(void) +{ + int ret, nid, cluster_id, cpu; + struct acpi_device *adev; + nodemask_t mask; + + for (nid = 0; nid < MAX_NUMNODES; nid++) { + if (!hotplug_mdev[nid]) + continue; + + adev = hotplug_mdev[nid]; + ret = fwnode_property_read_u32(acpi_fwnode_handle(adev), + "cluster-id", &cluster_id); + if (ret < 0) { + pr_debug("Failed to read cluster id\n"); + return; + } + + nodes_clear(mask); + for_each_possible_cpu(cpu) { + if (topology_cluster_id(cpu) == cluster_id) + node_set(cpu, mask); + } + mdev->cluster_cpumask[nid] = mask; + } + + mdev->topo_kobj = kobject_create_and_add("memory_topo", mdev->memdev_kobj); + if (!mdev->topo_kobj) + return; + + ret = sysfs_create_file(mdev->topo_kobj, &memory_locality_attribute.attr); + if (ret) + kobject_put(mdev->topo_kobj); +} + static int get_pxm(struct acpi_device *acpi_device, void *arg) { acpi_handle handle = acpi_device->handle; @@ -224,6 +283,7 @@ static int __init mdev_init(void) return -ENOMEM; }

+ memory_topo_init(); return ret; } module_init(mdev_init); @@ -232,6 +292,7 @@ static void __exit mdev_exit(void) { container_remove(); kobject_put(mdev->memdev_kobj); + kobject_put(mdev->topo_kobj); kfree(mdev); } module_exit(mdev_exit);

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OWRG CVE: NA

----------------------------------

Hbm memory device add support for acls hot repair. The patch add two methods for userpace: - query a paddr if it support acls repair - repair a paddr in hbm memory device

The feature of ACLS hot repair can help to fix a memory error from userspace by passing through the error physical address to HBM hardware.

Signed-off-by: Zhang Zekun zhangzekun11@huawei.com --- drivers/soc/hisilicon/Kconfig | 10 +++ drivers/soc/hisilicon/hisi_hbmdev.c | 134 ++++++++++++++++++++++++++++ 2 files changed, 144 insertions(+)

diff --git a/drivers/soc/hisilicon/Kconfig b/drivers/soc/hisilicon/Kconfig index 6d3067440c54..0a07d0e266a5 100644 --- a/drivers/soc/hisilicon/Kconfig +++ b/drivers/soc/hisilicon/Kconfig @@ -29,6 +29,16 @@ config HISI_HBMCACHE To compile the driver as a module, choose M here: the module will be called hisi_hbmcache.

+config HISI_HBMDEV_ACLS + bool "Add support for HISI ACLS repair" + depends on HISI_HBMDEV + help + Add ACLS support for hbm device, which can be used to query and + repair hardware error in HBM devices. This feature need to work with + hardware firmwares. + + If not sure say no. + config KUNPENG_HCCS tristate "HCCS driver on Kunpeng SoC" depends on ACPI diff --git a/drivers/soc/hisilicon/hisi_hbmdev.c b/drivers/soc/hisilicon/hisi_hbmdev.c index a4b72d232e19..e60fe1c14c0f 100644 --- a/drivers/soc/hisilicon/hisi_hbmdev.c +++ b/drivers/soc/hisilicon/hisi_hbmdev.c @@ -11,6 +11,7 @@ #include <linux/node.h> #include <linux/arch_topology.h> #include <linux/memory_hotplug.h> +#include <linux/mm.h>

#include "hisi_internal.h"

@@ -25,6 +26,9 @@ struct cdev_node { struct memory_dev { struct kobject *memdev_kobj; struct kobject *topo_kobj; +#ifdef CONFIG_HISI_HBMDEV_ACLS + struct kobject *acls_kobj; +#endif struct cdev_node cdev_list; nodemask_t cluster_cpumask[MAX_NUMNODES]; }; @@ -85,6 +89,134 @@ static void memory_topo_init(void) kobject_put(mdev->topo_kobj); }

+#ifdef CONFIG_HISI_HBMDEV_ACLS +static struct acpi_device *paddr_to_acpi_device(u64 paddr) +{ + unsigned long pfn; + int nid; + + pfn = __phys_to_pfn(paddr); + if (!pfn_valid(pfn)) + return NULL; + + nid = pfn_to_nid(pfn); + if (nid < 0 && nid >= MAX_NUMNODES) + return NULL; + + return hotplug_mdev[nid]; +} + +static ssize_t acls_query_store(struct kobject *kobj, + struct kobj_attribute *attr, + const char *buf, size_t count) +{ + struct acpi_object_list arg_list; + struct acpi_device *adev; + union acpi_object obj; + acpi_status status; + u64 paddr, res; + + if (kstrtoull(buf, 16, &paddr)) + return -EINVAL; + + adev = paddr_to_acpi_device(paddr); + if (!adev) + return -EINVAL; + + obj.type = ACPI_TYPE_INTEGER; + obj.integer.value = paddr; + arg_list.count = 1; + arg_list.pointer = &obj; + + status = acpi_evaluate_integer(adev->handle, "AQRY", &arg_list, &res); + if (ACPI_FAILURE(status)) + return -ENODEV; + + /* AQRY will return a positive error code to represent error status */ + if (IS_ERR_VALUE(-res)) + return -res; + else if (res) + return -ENODEV; + + return count; +} + +static struct kobj_attribute acls_query_store_attribute = + __ATTR(acls_query, 0200, NULL, acls_query_store); + +static ssize_t acls_repair_store(struct kobject *kobj, + struct kobj_attribute *attr, + const char *buf, size_t count) +{ + struct acpi_object_list arg_list; + struct acpi_device *adev; + union acpi_object obj; + acpi_status status; + u64 paddr, res; + + if (kstrtoull(buf, 16, &paddr)) + return -EINVAL; + + adev = paddr_to_acpi_device(paddr); + if (!adev) + return -EINVAL; + + obj.type = ACPI_TYPE_INTEGER; + obj.integer.value = paddr; + arg_list.count = 1; + arg_list.pointer = &obj; + + status = acpi_evaluate_integer(adev->handle, "AREP", &arg_list, &res); + if (ACPI_FAILURE(status)) + return -ENODEV; + + /* AREP will return a positive error code to represent error status */ + if (IS_ERR_VALUE(-res)) + return -res; + else if (res) + return -ENODEV; + + return count; +} +static struct kobj_attribute acls_repair_store_attribute = + __ATTR(acls_repair, 0200, NULL, acls_repair_store); + +static struct attribute *acls_attrs[] = { + &acls_query_store_attribute.attr, + &acls_repair_store_attribute.attr, + NULL, +}; + +static struct attribute_group acls_attr_group = { + .attrs = acls_attrs, +}; + +static void acls_init(void) +{ + int ret = -ENOMEM; + + mdev->acls_kobj = kobject_create_and_add("acls", mdev->memdev_kobj); + if (!mdev->acls_kobj) + goto out; + + ret = sysfs_create_group(mdev->acls_kobj, &acls_attr_group); + if (ret) + kobject_put(mdev->acls_kobj); + +out: + if (ret) + pr_err("ACLS hot repair is not enabled\n"); +} + +static void acls_remove(void) +{ + kobject_put(mdev->acls_kobj); +} +#else +static void acls_init(void) {} +static void acls_remove(void) {} +#endif + static int get_pxm(struct acpi_device *acpi_device, void *arg) { acpi_handle handle = acpi_device->handle; @@ -284,6 +416,7 @@ static int __init mdev_init(void) }

memory_topo_init(); + acls_init(); return ret; } module_init(mdev_init); @@ -293,6 +426,7 @@ static void __exit mdev_exit(void) container_remove(); kobject_put(mdev->memdev_kobj); kobject_put(mdev->topo_kobj); + acls_remove(); kfree(mdev); } module_exit(mdev_exit);