From: "Rafael J. Wysocki" rafael.j.wysocki@intel.com

mainline inclusion from mainline-v5.19-rc1 commit 10fa1b2cdc899ab471000968af56215bf3c90d8e category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I67QNJ CVE: NA

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?...

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

When walking the children of an ACPI device, take extra care to avoid using to_acpi_device() on the ones that are not ACPI devices, because that may lead to out-of-bounds access and memory corruption.

While at it, make the function passed to acpi_dev_for_each_child() take a struct acpi_device pointer argument (instead of a struct device one), so it is more straightforward to use.

Fixes: b7dd6298db81 ("ACPI: PM: Introduce acpi_dev_power_up_children_with_adr()") Reported-by: kernel test robot oliver.sang@intel.com BugLink: https://lore.kernel.org/lkml/20220420064725.GB16310@xsang-OptiPlex-9020/ Signed-off-by: Rafael J. Wysocki rafael.j.wysocki@intel.com Reviewed-by: Mika Westerberg mika.westerberg@linux.intel.com Signed-off-by: Zhang Zekun zhangzekun11@huawei.com --- drivers/acpi/bus.c | 24 ++++++++++++++++++++++-- include/acpi/acpi_bus.h | 2 +- 2 files changed, 23 insertions(+), 3 deletions(-)

diff --git a/drivers/acpi/bus.c b/drivers/acpi/bus.c index d8a6b32a17ad..de8185585f46 100644 --- a/drivers/acpi/bus.c +++ b/drivers/acpi/bus.c @@ -963,10 +963,30 @@ struct bus_type acpi_bus_type = { .uevent = acpi_device_uevent, };

+struct acpi_dev_walk_context { + int (*fn)(struct acpi_device *, void *); + void *data; +}; + +static int acpi_dev_for_one_check(struct device *dev, void *context) +{ + struct acpi_dev_walk_context *adwc = context; + + if (dev->bus != &acpi_bus_type) + return 0; + + return adwc->fn(to_acpi_device(dev), adwc->data); +} + int acpi_dev_for_each_child(struct acpi_device *adev, - int (*fn)(struct device *, void *), void *data) + int (*fn)(struct acpi_device *, void *), void *data) { - return device_for_each_child(&adev->dev, data, fn); + struct acpi_dev_walk_context adwc = { + .fn = fn, + .data = data, + }; + + return device_for_each_child(&adev->dev, &adwc, acpi_dev_for_one_check); }

/* -------------------------------------------------------------------------- diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h index d16a181d8a0e..fc0693c1b87e 100644 --- a/include/acpi/acpi_bus.h +++ b/include/acpi/acpi_bus.h @@ -495,7 +495,7 @@ extern int acpi_notifier_call_chain(struct acpi_device *, u32, u32); extern int register_acpi_notifier(struct notifier_block *); extern int unregister_acpi_notifier(struct notifier_block *); int acpi_dev_for_each_child(struct acpi_device *adev, - int (*fn)(struct device *, void *), void *data); + int (*fn)(struct acpi_device *, void *), void *data);

/* * External Functions

Offering HULK hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I67QNJ CVE: NA

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

Export the containers in the system, drivers can take the advantage of these containers to manipulate the devices in the container.

Signed-off-by: Zhang Zekun zhangzekun11@huawei.com --- drivers/acpi/container.c | 51 +++++++++++++++++++++++++++++++++++++++ include/linux/container.h | 8 ++++++ 2 files changed, 59 insertions(+)

diff --git a/drivers/acpi/container.c b/drivers/acpi/container.c index ccaa647ac3d4..fc9816b52537 100644 --- a/drivers/acpi/container.c +++ b/drivers/acpi/container.c @@ -11,6 +11,7 @@ */ #include <linux/acpi.h> #include <linux/container.h> +#include <linux/list.h>

#include "internal.h"

@@ -23,6 +24,52 @@ static const struct acpi_device_id container_device_ids[] = {

#ifdef CONFIG_ACPI_CONTAINER

+struct cdev_node *cdev_list; +EXPORT_SYMBOL_GPL(cdev_list); + +static void cdev_list_init(void) +{ + cdev_list = kmalloc(sizeof(struct cdev_node), GFP_KERNEL); + if (!cdev_list) { + pr_debug("container list init failed\n"); + return; + } + + INIT_LIST_HEAD(&cdev_list->clist); +} + +static void cdev_list_del_node(struct acpi_device *adev) +{ + struct cdev_node *cnode, *tmp; + + if (!cdev_list) + return; + + list_for_each_entry_safe(cnode, tmp, &cdev_list->clist, clist) { + if (cnode->dev == &adev->dev) { + list_del(&cnode->clist); + kfree(cnode); + } + } +} + +static void cdev_list_add_node(struct acpi_device *adev) +{ + struct cdev_node *cnode; + + if (!cdev_list) + return; + + cnode = kmalloc(sizeof(struct cdev_node), GFP_KERNEL); + if (!cnode) { + pr_debug("container list add cdev_node failed\n"); + return; + } + + cnode->dev = &adev->dev; + list_add_tail(&cnode->clist, &cdev_list->clist); +} + static int acpi_container_offline(struct container_dev *cdev) { struct acpi_device *adev = ACPI_COMPANION(&cdev->dev); @@ -67,6 +114,8 @@ static int container_device_attach(struct acpi_device *adev, return ret; } adev->driver_data = dev; + cdev_list_add_node(adev); + return 1; }

@@ -74,6 +123,7 @@ static void container_device_detach(struct acpi_device *adev) { struct device *dev = acpi_driver_data(adev);

+ cdev_list_del_node(adev); adev->driver_data = NULL; if (dev) device_unregister(dev); @@ -100,6 +150,7 @@ static struct acpi_scan_handler container_handler = { void __init acpi_container_init(void) { acpi_scan_add_handler(&container_handler); + cdev_list_init(); }

#else diff --git a/include/linux/container.h b/include/linux/container.h index 2566a1baa736..1f5fa722777f 100644 --- a/include/linux/container.h +++ b/include/linux/container.h @@ -24,4 +24,12 @@ static inline struct container_dev *to_container_dev(struct device *dev) return container_of(dev, struct container_dev, dev); }

+/* drivers/acpi/container.c */ +extern struct cdev_node *cdev_list; + +struct cdev_node { + struct device *dev; + struct list_head clist; +}; + #endif /* _LINUX_CONTAINER_H */

Offering: HULK hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I67QNJ 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 --- v3: - move the common code to hisi_internal.h

drivers/soc/Kconfig | 1 + drivers/soc/Makefile | 1 + drivers/soc/hisilicon/Kconfig | 19 +++ drivers/soc/hisilicon/Makefile | 3 + drivers/soc/hisilicon/hisi_hbmdev.c | 166 ++++++++++++++++++++++++++ drivers/soc/hisilicon/hisi_internal.h | 31 +++++ 6 files changed, 221 insertions(+) create mode 100644 drivers/soc/hisilicon/Kconfig create mode 100644 drivers/soc/hisilicon/Makefile create mode 100644 drivers/soc/hisilicon/hisi_hbmdev.c create mode 100644 drivers/soc/hisilicon/hisi_internal.h

diff --git a/drivers/soc/Kconfig b/drivers/soc/Kconfig index 425ab6f7e375..f7c59b063321 100644 --- a/drivers/soc/Kconfig +++ b/drivers/soc/Kconfig @@ -23,5 +23,6 @@ source "drivers/soc/versatile/Kconfig" source "drivers/soc/xilinx/Kconfig" source "drivers/soc/zte/Kconfig" source "drivers/soc/kendryte/Kconfig" +source "drivers/soc/hisilicon/Kconfig"

endmenu diff --git a/drivers/soc/Makefile b/drivers/soc/Makefile index 36452bed86ef..68f186e00e44 100644 --- a/drivers/soc/Makefile +++ b/drivers/soc/Makefile @@ -29,3 +29,4 @@ obj-$(CONFIG_PLAT_VERSATILE) += versatile/ obj-y += xilinx/ obj-$(CONFIG_ARCH_ZX) += zte/ obj-$(CONFIG_SOC_KENDRYTE) += kendryte/ +obj-y += hisilicon/ diff --git a/drivers/soc/hisilicon/Kconfig b/drivers/soc/hisilicon/Kconfig new file mode 100644 index 000000000000..497787af004e --- /dev/null +++ b/drivers/soc/hisilicon/Kconfig @@ -0,0 +1,19 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Hisilicon SoC drivers +# +menu "Hisilicon SoC driver support" + +config HISI_HBMDEV + tristate "add extra support for hbm memory device" + depends on ACPI_HOTPLUG_MEMORY + select ACPI_CONTAINER + help + This driver add extra supports for memory devices. The driver + provides methods for userpace to control the power of memory + devices in a container. + + To compile this driver as a module, choose M here: + the module will be called hisi_hbmdev. + +endmenu diff --git a/drivers/soc/hisilicon/Makefile b/drivers/soc/hisilicon/Makefile new file mode 100644 index 000000000000..22e87acb1ab3 --- /dev/null +++ b/drivers/soc/hisilicon/Makefile @@ -0,0 +1,3 @@ +# SPDX-License-Identifier: GPL-2.0 + +obj-$(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..82943cd35fa2 --- /dev/null +++ b/drivers/soc/hisilicon/hisi_hbmdev.c @@ -0,0 +1,166 @@ +// 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" + +struct memory_dev { + struct kobject *memdev_kobj; +}; + +static struct memory_dev *mdev; + +static int get_pxm(struct acpi_device *acpi_device, void *arg) +{ + int nid; + unsigned long long sta; + acpi_handle handle; + nodemask_t *mask; + acpi_status status; + + mask = arg; + handle = acpi_device->handle; + + 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) +{ + nodemask_t mask; + struct acpi_device *adev; + + adev = to_acpi_device(dev); + 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_status status; + acpi_handle handle; + + handle = adev->handle; + 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 *d, struct device_attribute *attr, + const char *buf, size_t count) +{ + int ret; + struct acpi_device *adev; + const int online_type = online_type_from_str(buf); + + if (online_type < 0) + return -EINVAL; + + adev = to_acpi_device(d); + switch (online_type) { + case STATE_ONLINE: + ret = memdev_power_on(adev); + if (!ret) + return count; + break; + case STATE_OFFLINE: + ret = memdev_power_off(adev); + if (!ret) + return count; + break; + default: + return -EINVAL; + } + + return ret; +} +static DEVICE_ATTR_WO(state); + +static int __init mdev_init(void) +{ + struct cdev_node *cnode; + + mdev = kzalloc(sizeof(struct memory_dev), GFP_KERNEL); + if (!mdev) + return -ENOMEM; + + mdev->memdev_kobj = kobject_create_and_add("hbm_memory", kernel_kobj); + if (!mdev->memdev_kobj) { + kfree(mdev); + return -ENOMEM; + } + + list_for_each_entry(cnode, &cdev_list->clist, clist) { + device_create_file(cnode->dev, &dev_attr_state); + device_create_file(cnode->dev, &dev_attr_pxms); + } + + return 0; +} +module_init(mdev_init); + +static void __exit mdev_exit(void) +{ + struct cdev_node *cnode; + + list_for_each_entry(cnode, &cdev_list->clist, clist) { + device_remove_file(cnode->dev, &dev_attr_state); + device_remove_file(cnode->dev, &dev_attr_pxms); + } + + kobject_put(mdev->memdev_kobj); + kfree(mdev); +} +module_exit(mdev_exit); + +MODULE_LICENSE("GPL v2"); +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..f14596f58a05 --- /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 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

Offering: HULK hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I67QNJ 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 | 63 +++++++++++++++++++++++++++++ 2 files changed, 69 insertions(+), 3 deletions(-)

diff --git a/drivers/soc/hisilicon/Kconfig b/drivers/soc/hisilicon/Kconfig index 497787af004e..781e97b1b742 100644 --- a/drivers/soc/hisilicon/Kconfig +++ b/drivers/soc/hisilicon/Kconfig @@ -9,9 +9,12 @@ 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 82943cd35fa2..ba5fa940ba04 100644 --- a/drivers/soc/hisilicon/hisi_hbmdev.c +++ b/drivers/soc/hisilicon/hisi_hbmdev.c @@ -8,15 +8,75 @@ #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"

struct memory_dev { struct kobject *memdev_kobj; + struct kobject *topo_kobj; + 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; + + 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) +{ + struct acpi_device *adev; + int ret, nid, cluster_id, cpu; + 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) { int nid; @@ -144,6 +204,8 @@ static int __init mdev_init(void) device_create_file(cnode->dev, &dev_attr_pxms); }

+ memory_topo_init(); + return 0; } module_init(mdev_init); @@ -158,6 +220,7 @@ static void __exit mdev_exit(void) }

kobject_put(mdev->memdev_kobj); + kobject_put(mdev->topo_kobj); kfree(mdev); } module_exit(mdev_exit);