* [PATCH v2 1/3] uacce: supports device isolation feature
2022-06-14 12:23 [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature Kai Ye
@ 2022-06-14 12:23 ` Kai Ye
2022-06-14 12:23 ` [PATCH v2 2/3] Documentation: add a isolation strategy vfs node for uacce Kai Ye
` (5 subsequent siblings)
6 siblings, 0 replies; 9+ messages in thread
From: Kai Ye @ 2022-06-14 12:23 UTC (permalink / raw)
To: gregkh, herbert
Cc: linux-crypto, linux-accelerators, linux-kernel, linuxarm,
zhangfei.gao, wangzhou1, yekai13
UACCE add the hardware error isolation API. Users can configure
the error frequency threshold by this vfs node. This API interface
certainly supports the configuration of user protocol strategy. Then
parse it inside the device driver. UACCE only reports the device
isolate state. When the error frequency is exceeded, the device
will be isolated. The isolation strategy should be defined in each
driver module.
Signed-off-by: Kai Ye <yekai13@huawei.com>
Reviewed-by: Zhou Wang <wangzhou1@hisilicon.com>
---
drivers/misc/uacce/uacce.c | 37 +++++++++++++++++++++++++++++++++++++
include/linux/uacce.h | 16 +++++++++++++---
2 files changed, 50 insertions(+), 3 deletions(-)
diff --git a/drivers/misc/uacce/uacce.c b/drivers/misc/uacce/uacce.c
index b6219c6bfb48..525623215132 100644
--- a/drivers/misc/uacce/uacce.c
+++ b/drivers/misc/uacce/uacce.c
@@ -346,12 +346,47 @@ static ssize_t region_dus_size_show(struct device *dev,
uacce->qf_pg_num[UACCE_QFRT_DUS] << PAGE_SHIFT);
}
+static ssize_t isolate_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct uacce_device *uacce = to_uacce_device(dev);
+
+ return sysfs_emit(buf, "%d\n", uacce->ops->get_isolate_state(uacce));
+}
+
+static ssize_t isolate_strategy_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct uacce_device *uacce = to_uacce_device(dev);
+
+ return sysfs_emit(buf, "%s\n", uacce->isolate_strategy);
+}
+
+static ssize_t isolate_strategy_store(struct device *dev,
+ struct device_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct uacce_device *uacce = to_uacce_device(dev);
+ int ret;
+
+ if (!buf || sizeof(buf) > UACCE_MAX_ISOLATE_STRATEGY_LEN)
+ return -EINVAL;
+
+ memcpy(uacce->isolate_strategy, buf, strlen(buf));
+
+ ret = uacce->ops->isolate_strategy_write(uacce, buf);
+
+ return ret ? ret : count;
+}
+
static DEVICE_ATTR_RO(api);
static DEVICE_ATTR_RO(flags);
static DEVICE_ATTR_RO(available_instances);
static DEVICE_ATTR_RO(algorithms);
static DEVICE_ATTR_RO(region_mmio_size);
static DEVICE_ATTR_RO(region_dus_size);
+static DEVICE_ATTR_RO(isolate);
+static DEVICE_ATTR_RW(isolate_strategy);
static struct attribute *uacce_dev_attrs[] = {
&dev_attr_api.attr,
@@ -360,6 +395,8 @@ static struct attribute *uacce_dev_attrs[] = {
&dev_attr_algorithms.attr,
&dev_attr_region_mmio_size.attr,
&dev_attr_region_dus_size.attr,
+ &dev_attr_isolate.attr,
+ &dev_attr_isolate_strategy.attr,
NULL,
};
diff --git a/include/linux/uacce.h b/include/linux/uacce.h
index 48e319f40275..0f7668bfa645 100644
--- a/include/linux/uacce.h
+++ b/include/linux/uacce.h
@@ -8,6 +8,7 @@
#define UACCE_NAME "uacce"
#define UACCE_MAX_REGION 2
#define UACCE_MAX_NAME_SIZE 64
+#define UACCE_MAX_ISOLATE_STRATEGY_LEN 256
struct uacce_queue;
struct uacce_device;
@@ -30,6 +31,8 @@ struct uacce_qfile_region {
* @is_q_updated: check whether the task is finished
* @mmap: mmap addresses of queue to user space
* @ioctl: ioctl for user space users of the queue
+ * @get_isolate_state: get the device state after set the isolate strategy
+ * @isolate_strategy_store: stored the isolate strategy to the device
*/
struct uacce_ops {
int (*get_available_instances)(struct uacce_device *uacce);
@@ -43,6 +46,8 @@ struct uacce_ops {
struct uacce_qfile_region *qfr);
long (*ioctl)(struct uacce_queue *q, unsigned int cmd,
unsigned long arg);
+ enum uacce_dev_state (*get_isolate_state)(struct uacce_device *uacce);
+ int (*isolate_strategy_write)(struct uacce_device *uacce, const char *buf);
};
/**
@@ -57,6 +62,12 @@ struct uacce_interface {
const struct uacce_ops *ops;
};
+enum uacce_dev_state {
+ UACCE_DEV_ERR = -1,
+ UACCE_DEV_NORMAL,
+ UACCE_DEV_ISOLATE,
+};
+
enum uacce_q_state {
UACCE_Q_ZOMBIE = 0,
UACCE_Q_INIT,
@@ -117,6 +128,7 @@ struct uacce_device {
struct list_head queues;
struct mutex queues_lock;
struct inode *inode;
+ char isolate_strategy[UACCE_MAX_ISOLATE_STRATEGY_LEN];
};
#if IS_ENABLED(CONFIG_UACCE)
@@ -125,7 +137,7 @@ struct uacce_device *uacce_alloc(struct device *parent,
struct uacce_interface *interface);
int uacce_register(struct uacce_device *uacce);
void uacce_remove(struct uacce_device *uacce);
-
+struct uacce_device *dev_to_uacce(struct device *dev);
#else /* CONFIG_UACCE */
static inline
@@ -140,8 +152,6 @@ static inline int uacce_register(struct uacce_device *uacce)
return -EINVAL;
}
-static inline void uacce_remove(struct uacce_device *uacce) {}
-
#endif /* CONFIG_UACCE */
#endif /* _LINUX_UACCE_H */
--
2.33.0
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH v2 2/3] Documentation: add a isolation strategy vfs node for uacce
2022-06-14 12:23 [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature Kai Ye
2022-06-14 12:23 ` [PATCH v2 1/3] uacce: " Kai Ye
@ 2022-06-14 12:23 ` Kai Ye
2022-06-14 12:23 ` [PATCH v2 3/3] crypto: hisilicon/qm - defining the device isolation strategy Kai Ye
` (4 subsequent siblings)
6 siblings, 0 replies; 9+ messages in thread
From: Kai Ye @ 2022-06-14 12:23 UTC (permalink / raw)
To: gregkh, herbert
Cc: linux-crypto, linux-accelerators, linux-kernel, linuxarm,
zhangfei.gao, wangzhou1, yekai13
Update documentation describing DebugFS that could help to
configure hard error frequency for users in th user space.
Signed-off-by: Kai Ye <yekai13@huawei.com>
---
Documentation/ABI/testing/sysfs-driver-uacce | 17 +++++++++++++++++
1 file changed, 17 insertions(+)
diff --git a/Documentation/ABI/testing/sysfs-driver-uacce b/Documentation/ABI/testing/sysfs-driver-uacce
index 08f2591138af..0c4226364182 100644
--- a/Documentation/ABI/testing/sysfs-driver-uacce
+++ b/Documentation/ABI/testing/sysfs-driver-uacce
@@ -19,6 +19,23 @@ Contact: linux-accelerators@lists.ozlabs.org
Description: Available instances left of the device
Return -ENODEV if uacce_ops get_available_instances is not provided
+What: /sys/class/uacce/<dev_name>/isolate_strategy
+Date: Jun 2022
+KernelVersion: 5.19
+Contact: linux-accelerators@lists.ozlabs.org
+Description: A vfs node that used to configures the hardware
+ error frequency. This frequency is abstract. Like once an hour
+ or once a day. The specific isolation strategy can be defined in
+ each driver module.
+
+What: /sys/class/uacce/<dev_name>/isolate
+Date: Jun 2022
+KernelVersion: 5.19
+Contact: linux-accelerators@lists.ozlabs.org
+Description: A vfs node that show the device isolated state. The value 0
+ means that the device is working. The value 1 means that the
+ device has been isolated.
+
What: /sys/class/uacce/<dev_name>/algorithms
Date: Feb 2020
KernelVersion: 5.7
--
2.33.0
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH v2 3/3] crypto: hisilicon/qm - defining the device isolation strategy
2022-06-14 12:23 [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature Kai Ye
2022-06-14 12:23 ` [PATCH v2 1/3] uacce: " Kai Ye
2022-06-14 12:23 ` [PATCH v2 2/3] Documentation: add a isolation strategy vfs node for uacce Kai Ye
@ 2022-06-14 12:23 ` Kai Ye
2022-06-14 12:23 ` [PATCH 1/3] uacce: supports device isolation feature Kai Ye
` (3 subsequent siblings)
6 siblings, 0 replies; 9+ messages in thread
From: Kai Ye @ 2022-06-14 12:23 UTC (permalink / raw)
To: gregkh, herbert
Cc: linux-crypto, linux-accelerators, linux-kernel, linuxarm,
zhangfei.gao, wangzhou1, yekai13
Define the device isolation strategy by the device driver. if the
AER error frequency exceeds the value of setting for a certain
period of time, The device will not be available in user space. The VF
device use the PF device isolation strategy. All the hardware errors
are processed by PF driver.
Signed-off-by: Kai Ye <yekai13@huawei.com>
---
drivers/crypto/hisilicon/qm.c | 157 +++++++++++++++++++++++++++++++---
include/linux/hisi_acc_qm.h | 9 ++
2 files changed, 152 insertions(+), 14 deletions(-)
diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
index ad83c194d664..47c41fa52693 100644
--- a/drivers/crypto/hisilicon/qm.c
+++ b/drivers/crypto/hisilicon/qm.c
@@ -12,7 +12,6 @@
#include <linux/pm_runtime.h>
#include <linux/seq_file.h>
#include <linux/slab.h>
-#include <linux/uacce.h>
#include <linux/uaccess.h>
#include <uapi/misc/uacce/hisi_qm.h>
#include <linux/hisi_acc_qm.h>
@@ -417,6 +416,16 @@ struct hisi_qm_resource {
struct list_head list;
};
+/**
+ * struct qm_hw_err - structure of describes the device err
+ * @list: hardware error list
+ * @tick_stamp: timestamp when the error occurred
+ */
+struct qm_hw_err {
+ struct list_head list;
+ unsigned long long tick_stamp;
+};
+
struct hisi_qm_hw_ops {
int (*get_vft)(struct hisi_qm *qm, u32 *base, u32 *number);
void (*qm_db)(struct hisi_qm *qm, u16 qn,
@@ -3278,6 +3287,7 @@ static int hisi_qm_uacce_get_queue(struct uacce_device *uacce,
qp->event_cb = qm_qp_event_notifier;
qp->pasid = arg;
qp->is_in_kernel = false;
+ atomic_inc(&qm->uacce_ref);
return 0;
}
@@ -3285,7 +3295,9 @@ static int hisi_qm_uacce_get_queue(struct uacce_device *uacce,
static void hisi_qm_uacce_put_queue(struct uacce_queue *q)
{
struct hisi_qp *qp = q->priv;
+ struct hisi_qm *qm = qp->qm;
+ atomic_dec(&qm->uacce_ref);
hisi_qm_cache_wb(qp->qm);
hisi_qm_release_qp(qp);
}
@@ -3410,6 +3422,102 @@ static long hisi_qm_uacce_ioctl(struct uacce_queue *q, unsigned int cmd,
return 0;
}
+/**
+ * qm_hw_err_isolate() - Try to isolate the uacce device with its VFs
+ * @qm: The qm which we want to configure.
+ *
+ * according to user's configuration of isolation strategy. Warning: this
+ * API should be called while there is no user on the device, or the users
+ * on this device are suspended by slot resetting preparation of PCI AER.
+ */
+static int qm_hw_err_isolate(struct hisi_qm *qm)
+{
+ struct qm_hw_err *err, *tmp, *hw_err;
+ struct qm_err_isolate *isolate;
+ u32 count = 0;
+
+ isolate = &qm->isolate_data;
+
+#define SECONDS_PER_HOUR 3600
+
+ /* All the hw errs are processed by PF driver */
+ if (qm->uacce->is_vf || atomic_read(&isolate->is_isolate) ||
+ !isolate->hw_err_isolate_hz)
+ return 0;
+
+ hw_err = kzalloc(sizeof(*hw_err), GFP_ATOMIC);
+ if (!hw_err)
+ return -ENOMEM;
+ hw_err->tick_stamp = jiffies;
+ list_for_each_entry_safe(err, tmp, &qm->uacce_hw_errs, list) {
+ if ((hw_err->tick_stamp - err->tick_stamp) / HZ >
+ SECONDS_PER_HOUR) {
+ list_del(&err->list);
+ kfree(err);
+ } else {
+ count++;
+ }
+ }
+ list_add(&hw_err->list, &qm->uacce_hw_errs);
+
+ if (count >= isolate->hw_err_isolate_hz)
+ atomic_set(&isolate->is_isolate, 1);
+
+ return 0;
+}
+
+static void qm_hw_err_destroy(struct hisi_qm *qm)
+{
+ struct qm_hw_err *err, *tmp;
+
+ list_for_each_entry_safe(err, tmp, &qm->uacce_hw_errs, list) {
+ list_del(&err->list);
+ kfree(err);
+ }
+}
+
+static enum uacce_dev_state hisi_qm_get_isolate_state(struct uacce_device *uacce)
+{
+ struct hisi_qm *qm = uacce->priv;
+ struct hisi_qm *pf_qm;
+
+ if (uacce->is_vf) {
+ pf_qm = pci_get_drvdata(pci_physfn(qm->pdev));
+ qm->isolate_data.is_isolate = pf_qm->isolate_data.is_isolate;
+ }
+
+ return atomic_read(&qm->isolate_data.is_isolate) ?
+ UACCE_DEV_ISOLATE : UACCE_DEV_NORMAL;
+}
+
+static int hisi_qm_isolate_strategy_write(struct uacce_device *uacce,
+ const char *buf)
+{
+ struct hisi_qm *qm = uacce->priv;
+ unsigned long val = 0;
+
+#define MAX_ISOLATE_STRATEGY 65535
+
+ if (atomic_read(&qm->uacce_ref))
+ return -EBUSY;
+
+ /* must be set by PF */
+ if (atomic_read(&qm->isolate_data.is_isolate) || uacce->is_vf)
+ return -EINVAL;
+
+ if (kstrtoul(buf, 0, &val) < 0)
+ return -EINVAL;
+
+ if (val > MAX_ISOLATE_STRATEGY)
+ return -EINVAL;
+
+ qm->isolate_data.hw_err_isolate_hz = val;
+ dev_info(&qm->pdev->dev,
+ "the value of isolate_strategy is set to %lu.\n", val);
+
+ return 0;
+}
+
static const struct uacce_ops uacce_qm_ops = {
.get_available_instances = hisi_qm_get_available_instances,
.get_queue = hisi_qm_uacce_get_queue,
@@ -3418,9 +3526,22 @@ static const struct uacce_ops uacce_qm_ops = {
.stop_queue = hisi_qm_uacce_stop_queue,
.mmap = hisi_qm_uacce_mmap,
.ioctl = hisi_qm_uacce_ioctl,
+ .get_isolate_state = hisi_qm_get_isolate_state,
.is_q_updated = hisi_qm_is_q_updated,
+ .isolate_strategy_write = hisi_qm_isolate_strategy_write,
};
+static void qm_remove_uacce(struct hisi_qm *qm)
+{
+ struct uacce_device *uacce = qm->uacce;
+
+ if (qm->use_sva) {
+ qm_hw_err_destroy(qm);
+ uacce_remove(uacce);
+ qm->uacce = NULL;
+ }
+}
+
static int qm_alloc_uacce(struct hisi_qm *qm)
{
struct pci_dev *pdev = qm->pdev;
@@ -3433,6 +3554,7 @@ static int qm_alloc_uacce(struct hisi_qm *qm)
};
int ret;
+ INIT_LIST_HEAD(&qm->uacce_hw_errs);
ret = strscpy(interface.name, dev_driver_string(&pdev->dev),
sizeof(interface.name));
if (ret < 0)
@@ -3446,8 +3568,7 @@ static int qm_alloc_uacce(struct hisi_qm *qm)
qm->use_sva = true;
} else {
/* only consider sva case */
- uacce_remove(uacce);
- qm->uacce = NULL;
+ qm_remove_uacce(qm);
return -EINVAL;
}
@@ -5109,6 +5230,12 @@ static int qm_controller_reset_prepare(struct hisi_qm *qm)
return ret;
}
+ if (qm->use_sva) {
+ ret = qm_hw_err_isolate(qm);
+ if (ret)
+ pci_err(pdev, "failed to isolate hw err!\n");
+ }
+
ret = qm_wait_vf_prepare_finish(qm);
if (ret)
pci_err(pdev, "failed to stop by vfs in soft reset!\n");
@@ -5436,19 +5563,24 @@ static int qm_controller_reset(struct hisi_qm *qm)
ret = qm_soft_reset(qm);
if (ret) {
pci_err(pdev, "Controller reset failed (%d)\n", ret);
- qm_reset_bit_clear(qm);
- return ret;
+ goto err_reset;
}
ret = qm_controller_reset_done(qm);
- if (ret) {
- qm_reset_bit_clear(qm);
- return ret;
- }
+ if (ret)
+ goto err_reset;
pci_info(pdev, "Controller reset complete\n");
-
return 0;
+
+err_reset:
+ pci_err(pdev, "Controller reset failed (%d)\n", ret);
+ qm_reset_bit_clear(qm);
+
+ /* if resetting fails, isolate the device */
+ if (qm->use_sva && !qm->uacce->is_vf)
+ atomic_set(&qm->isolate_data.is_isolate, 1);
+ return ret;
}
/**
@@ -6246,10 +6378,7 @@ int hisi_qm_init(struct hisi_qm *qm)
err_free_qm_memory:
hisi_qm_memory_uninit(qm);
err_alloc_uacce:
- if (qm->use_sva) {
- uacce_remove(qm->uacce);
- qm->uacce = NULL;
- }
+ qm_remove_uacce(qm);
err_irq_register:
qm_irq_unregister(qm);
err_pci_init:
diff --git a/include/linux/hisi_acc_qm.h b/include/linux/hisi_acc_qm.h
index 116e8bd68c99..c17fd6de8551 100644
--- a/include/linux/hisi_acc_qm.h
+++ b/include/linux/hisi_acc_qm.h
@@ -8,6 +8,7 @@
#include <linux/iopoll.h>
#include <linux/module.h>
#include <linux/pci.h>
+#include <linux/uacce.h>
#define QM_QNUM_V1 4096
#define QM_QNUM_V2 1024
@@ -271,6 +272,11 @@ struct hisi_qm_poll_data {
u16 *qp_finish_id;
};
+struct qm_err_isolate {
+ u32 hw_err_isolate_hz; /* user cfg freq which triggers isolation */
+ atomic_t is_isolate;
+};
+
struct hisi_qm {
enum qm_hw_ver ver;
enum qm_fun_type fun_type;
@@ -335,6 +341,9 @@ struct hisi_qm {
struct qm_shaper_factor *factor;
u32 mb_qos;
u32 type_rate;
+ struct list_head uacce_hw_errs;
+ atomic_t uacce_ref; /* reference of the uacce */
+ struct qm_err_isolate isolate_data;
};
struct hisi_qp_status {
--
2.33.0
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH 1/3] uacce: supports device isolation feature
2022-06-14 12:23 [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature Kai Ye
` (2 preceding siblings ...)
2022-06-14 12:23 ` [PATCH v2 3/3] crypto: hisilicon/qm - defining the device isolation strategy Kai Ye
@ 2022-06-14 12:23 ` Kai Ye
2022-06-14 12:23 ` [PATCH 2/3] Documentation: add a isolation strategy vfs node for uacce Kai Ye
` (2 subsequent siblings)
6 siblings, 0 replies; 9+ messages in thread
From: Kai Ye @ 2022-06-14 12:23 UTC (permalink / raw)
To: gregkh, herbert
Cc: linux-crypto, linux-accelerators, linux-kernel, linuxarm,
zhangfei.gao, wangzhou1, yekai13
UACCE add the hardware error isolation API. Users can configure
the error frequency threshold by this vfs node. This API interface
certainly supports the configuration of user protocol strategy. Then
parse it inside the device driver. UACCE only reports the device
isolate state. When the error frequency is exceeded, the device
will be isolated. The isolation strategy should be defined in each
driver module.
Signed-off-by: Kai Ye <yekai13@huawei.com>
Reviewed-by: Zhou Wang <wangzhou1@hisilicon.com>
---
drivers/misc/uacce/uacce.c | 51 ++++++++++++++++++++++++++++++++++++++
include/linux/uacce.h | 15 ++++++++++-
2 files changed, 65 insertions(+), 1 deletion(-)
diff --git a/drivers/misc/uacce/uacce.c b/drivers/misc/uacce/uacce.c
index b6219c6bfb48..4d9d9aeb145a 100644
--- a/drivers/misc/uacce/uacce.c
+++ b/drivers/misc/uacce/uacce.c
@@ -12,6 +12,20 @@ static dev_t uacce_devt;
static DEFINE_MUTEX(uacce_mutex);
static DEFINE_XARRAY_ALLOC(uacce_xa);
+static int cdev_get(struct device *dev, void *data)
+{
+ struct uacce_device *uacce;
+ struct device **t_dev = data;
+
+ uacce = container_of(dev, struct uacce_device, dev);
+ if (uacce->parent == *t_dev) {
+ *t_dev = dev;
+ return 1;
+ }
+
+ return 0;
+}
+
static int uacce_start_queue(struct uacce_queue *q)
{
int ret = 0;
@@ -346,12 +360,47 @@ static ssize_t region_dus_size_show(struct device *dev,
uacce->qf_pg_num[UACCE_QFRT_DUS] << PAGE_SHIFT);
}
+static ssize_t isolate_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct uacce_device *uacce = to_uacce_device(dev);
+
+ return sysfs_emit(buf, "%d\n", uacce->ops->get_isolate_state(uacce));
+}
+
+static ssize_t isolate_strategy_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct uacce_device *uacce = to_uacce_device(dev);
+
+ return sysfs_emit(buf, "%s\n", uacce->isolate_strategy);
+}
+
+static ssize_t isolate_strategy_store(struct device *dev,
+ struct device_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct uacce_device *uacce = to_uacce_device(dev);
+ int ret;
+
+ if (!buf || sizeof(buf) > UACCE_MAX_ISOLATE_STRATEGY_LEN)
+ return -EINVAL;
+
+ memcpy(uacce->isolate_strategy, buf, strlen(buf));
+
+ ret = uacce->ops->isolate_strategy_write(uacce, buf);
+
+ return ret > 0 ? count : ret;
+}
+
static DEVICE_ATTR_RO(api);
static DEVICE_ATTR_RO(flags);
static DEVICE_ATTR_RO(available_instances);
static DEVICE_ATTR_RO(algorithms);
static DEVICE_ATTR_RO(region_mmio_size);
static DEVICE_ATTR_RO(region_dus_size);
+static DEVICE_ATTR_RO(isolate);
+static DEVICE_ATTR_RW(isolate_strategy);
static struct attribute *uacce_dev_attrs[] = {
&dev_attr_api.attr,
@@ -360,6 +409,8 @@ static struct attribute *uacce_dev_attrs[] = {
&dev_attr_algorithms.attr,
&dev_attr_region_mmio_size.attr,
&dev_attr_region_dus_size.attr,
+ &dev_attr_isolate.attr,
+ &dev_attr_isolate_strategy.attr,
NULL,
};
diff --git a/include/linux/uacce.h b/include/linux/uacce.h
index 48e319f40275..e00a43a07e4b 100644
--- a/include/linux/uacce.h
+++ b/include/linux/uacce.h
@@ -8,6 +8,7 @@
#define UACCE_NAME "uacce"
#define UACCE_MAX_REGION 2
#define UACCE_MAX_NAME_SIZE 64
+#define UACCE_MAX_ISOLATE_STRATEGY_LEN 256
struct uacce_queue;
struct uacce_device;
@@ -30,6 +31,8 @@ struct uacce_qfile_region {
* @is_q_updated: check whether the task is finished
* @mmap: mmap addresses of queue to user space
* @ioctl: ioctl for user space users of the queue
+ * @get_isolate_state: get the device state after set the isolate strategy
+ * @isolate_strategy_store: stored the isolate strategy to the device
*/
struct uacce_ops {
int (*get_available_instances)(struct uacce_device *uacce);
@@ -43,6 +46,8 @@ struct uacce_ops {
struct uacce_qfile_region *qfr);
long (*ioctl)(struct uacce_queue *q, unsigned int cmd,
unsigned long arg);
+ enum uacce_dev_state (*get_isolate_state)(struct uacce_device *uacce);
+ int (*isolate_strategy_write)(struct uacce_device *uacce, const char *buf);
};
/**
@@ -57,6 +62,12 @@ struct uacce_interface {
const struct uacce_ops *ops;
};
+enum uacce_dev_state {
+ UACCE_DEV_ERR = -1,
+ UACCE_DEV_NORMAL,
+ UACCE_DEV_ISOLATE,
+};
+
enum uacce_q_state {
UACCE_Q_ZOMBIE = 0,
UACCE_Q_INIT,
@@ -99,6 +110,7 @@ struct uacce_queue {
* @dev: dev of the uacce
* @priv: private pointer of the uacce
* @queues: list of queues
+ * @ref: reference of the uacce
* @queues_lock: lock for queues list
* @inode: core vfs
*/
@@ -117,6 +129,7 @@ struct uacce_device {
struct list_head queues;
struct mutex queues_lock;
struct inode *inode;
+ char isolate_strategy[UACCE_MAX_ISOLATE_STRATEGY_LEN];
};
#if IS_ENABLED(CONFIG_UACCE)
@@ -125,7 +138,7 @@ struct uacce_device *uacce_alloc(struct device *parent,
struct uacce_interface *interface);
int uacce_register(struct uacce_device *uacce);
void uacce_remove(struct uacce_device *uacce);
-
+struct uacce_device *dev_to_uacce(struct device *dev);
#else /* CONFIG_UACCE */
static inline
--
2.33.0
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH 2/3] Documentation: add a isolation strategy vfs node for uacce
2022-06-14 12:23 [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature Kai Ye
` (3 preceding siblings ...)
2022-06-14 12:23 ` [PATCH 1/3] uacce: supports device isolation feature Kai Ye
@ 2022-06-14 12:23 ` Kai Ye
2022-06-14 12:23 ` [PATCH 3/3] crypto: hisilicon/qm - defining the device isolation strategy Kai Ye
2022-06-14 12:34 ` [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature yekai(A)
6 siblings, 0 replies; 9+ messages in thread
From: Kai Ye @ 2022-06-14 12:23 UTC (permalink / raw)
To: gregkh, herbert
Cc: linux-crypto, linux-accelerators, linux-kernel, linuxarm,
zhangfei.gao, wangzhou1, yekai13
Update documentation describing DebugFS that could help to
configure hard error frequency for users in th user space.
Signed-off-by: Kai Ye <yekai13@huawei.com>
---
Documentation/ABI/testing/sysfs-driver-uacce | 17 +++++++++++++++++
1 file changed, 17 insertions(+)
diff --git a/Documentation/ABI/testing/sysfs-driver-uacce b/Documentation/ABI/testing/sysfs-driver-uacce
index 08f2591138af..0c4226364182 100644
--- a/Documentation/ABI/testing/sysfs-driver-uacce
+++ b/Documentation/ABI/testing/sysfs-driver-uacce
@@ -19,6 +19,23 @@ Contact: linux-accelerators@lists.ozlabs.org
Description: Available instances left of the device
Return -ENODEV if uacce_ops get_available_instances is not provided
+What: /sys/class/uacce/<dev_name>/isolate_strategy
+Date: Jun 2022
+KernelVersion: 5.19
+Contact: linux-accelerators@lists.ozlabs.org
+Description: A vfs node that used to configures the hardware
+ error frequency. This frequency is abstract. Like once an hour
+ or once a day. The specific isolation strategy can be defined in
+ each driver module.
+
+What: /sys/class/uacce/<dev_name>/isolate
+Date: Jun 2022
+KernelVersion: 5.19
+Contact: linux-accelerators@lists.ozlabs.org
+Description: A vfs node that show the device isolated state. The value 0
+ means that the device is working. The value 1 means that the
+ device has been isolated.
+
What: /sys/class/uacce/<dev_name>/algorithms
Date: Feb 2020
KernelVersion: 5.7
--
2.33.0
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH 3/3] crypto: hisilicon/qm - defining the device isolation strategy
2022-06-14 12:23 [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature Kai Ye
` (4 preceding siblings ...)
2022-06-14 12:23 ` [PATCH 2/3] Documentation: add a isolation strategy vfs node for uacce Kai Ye
@ 2022-06-14 12:23 ` Kai Ye
2022-06-14 12:34 ` [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature yekai(A)
6 siblings, 0 replies; 9+ messages in thread
From: Kai Ye @ 2022-06-14 12:23 UTC (permalink / raw)
To: gregkh, herbert
Cc: linux-crypto, linux-accelerators, linux-kernel, linuxarm,
zhangfei.gao, wangzhou1, yekai13
Define the device isolation strategy by the device driver. if the
AER error frequency exceeds the value of setting for a certain
period of time, The device will not be available in user space. The VF
device use the PF device isolation strategy. All the hardware errors
are processed by PF driver.
Signed-off-by: Kai Ye <yekai13@huawei.com>
---
drivers/crypto/hisilicon/qm.c | 155 +++++++++++++++++++++++++++++++---
include/linux/hisi_acc_qm.h | 9 ++
2 files changed, 152 insertions(+), 12 deletions(-)
diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
index b4ca2eb034d7..391d83929ad2 100644
--- a/drivers/crypto/hisilicon/qm.c
+++ b/drivers/crypto/hisilicon/qm.c
@@ -12,7 +12,6 @@
#include <linux/pm_runtime.h>
#include <linux/seq_file.h>
#include <linux/slab.h>
-#include <linux/uacce.h>
#include <linux/uaccess.h>
#include <uapi/misc/uacce/hisi_qm.h>
#include <linux/hisi_acc_qm.h>
@@ -417,6 +416,16 @@ struct hisi_qm_resource {
struct list_head list;
};
+/**
+ * struct qm_hw_err - structure of describes the device err
+ * @list: hardware error list
+ * @tick_stamp: timestamp when the error occurred
+ */
+struct qm_hw_err {
+ struct list_head list;
+ unsigned long long tick_stamp;
+};
+
struct hisi_qm_hw_ops {
int (*get_vft)(struct hisi_qm *qm, u32 *base, u32 *number);
void (*qm_db)(struct hisi_qm *qm, u16 qn,
@@ -3265,6 +3274,7 @@ static int hisi_qm_uacce_get_queue(struct uacce_device *uacce,
qp->event_cb = qm_qp_event_notifier;
qp->pasid = arg;
qp->is_in_kernel = false;
+ atomic_inc(&qm->uacce_ref);
return 0;
}
@@ -3272,7 +3282,9 @@ static int hisi_qm_uacce_get_queue(struct uacce_device *uacce,
static void hisi_qm_uacce_put_queue(struct uacce_queue *q)
{
struct hisi_qp *qp = q->priv;
+ struct hisi_qm *qm = qp->qm;
+ atomic_dec(&qm->uacce_ref);
hisi_qm_cache_wb(qp->qm);
hisi_qm_release_qp(qp);
}
@@ -3397,6 +3409,102 @@ static long hisi_qm_uacce_ioctl(struct uacce_queue *q, unsigned int cmd,
return 0;
}
+/**
+ * qm_hw_err_isolate() - Try to isolate the uacce device with its VFs
+ * @qm: The qm which we want to configure.
+ *
+ * according to user's configuration of isolation strategy. Warning: this
+ * API should be called while there is no user on the device, or the users
+ * on this device are suspended by slot resetting preparation of PCI AER.
+ */
+static int qm_hw_err_isolate(struct hisi_qm *qm)
+{
+ struct qm_hw_err *err, *tmp, *hw_err;
+ struct qm_err_isolate *isolate;
+ u32 count = 0;
+
+ isolate = &qm->isolate_data;
+
+#define SECONDS_PER_HOUR 3600
+
+ /* All the hw errs are processed by PF driver */
+ if (qm->uacce->is_vf || atomic_read(&isolate->is_isolate) ||
+ !isolate->hw_err_isolate_hz)
+ return 0;
+
+ hw_err = kzalloc(sizeof(*hw_err), GFP_ATOMIC);
+ if (!hw_err)
+ return -ENOMEM;
+ hw_err->tick_stamp = jiffies;
+ list_for_each_entry_safe(err, tmp, &qm->uacce_hw_errs, list) {
+ if ((hw_err->tick_stamp - err->tick_stamp) / HZ >
+ SECONDS_PER_HOUR) {
+ list_del(&err->list);
+ kfree(err);
+ } else {
+ count++;
+ }
+ }
+ list_add(&hw_err->list, &qm->uacce_hw_errs);
+
+ if (count >= isolate->hw_err_isolate_hz)
+ atomic_set(&isolate->is_isolate, 1);
+
+ return 0;
+}
+
+static void qm_hw_err_destroy(struct hisi_qm *qm)
+{
+ struct qm_hw_err *err, *tmp;
+
+ list_for_each_entry_safe(err, tmp, &qm->uacce_hw_errs, list) {
+ list_del(&err->list);
+ kfree(err);
+ }
+}
+
+static enum uacce_dev_state hisi_qm_get_isolate_state(struct uacce_device *uacce)
+{
+ struct hisi_qm *qm = uacce->priv;
+ struct hisi_qm *pf_qm;
+
+ if (uacce->is_vf) {
+ pf_qm = pci_get_drvdata(pci_physfn(qm->pdev));
+ qm->isolate_data.is_isolate = pf_qm->isolate_data.is_isolate;
+ }
+
+ return atomic_read(&qm->isolate_data.is_isolate) ?
+ UACCE_DEV_ISOLATE : UACCE_DEV_NORMAL;
+}
+
+static int hisi_qm_isolate_strategy_write(struct uacce_device *uacce,
+ const char *buf)
+{
+ struct hisi_qm *qm = uacce->priv;
+ unsigned long val = 0;
+
+#define MAX_ISOLATE_STRATEGY 65535
+
+ if (atomic_read(&qm->uacce_ref))
+ return -EBUSY;
+
+ /* must be set by PF */
+ if (atomic_read(&qm->isolate_data.is_isolate) || uacce->is_vf)
+ return -EINVAL;
+
+ if (kstrtoul(buf, 0, &val) < 0)
+ return -EINVAL;
+
+ if (val > MAX_ISOLATE_STRATEGY)
+ return -EINVAL;
+
+ qm->isolate_data.hw_err_isolate_hz = val;
+ dev_info(&qm->pdev->dev,
+ "the value of isolate_strategy is set to %lu.\n", val);
+
+ return 0;
+}
+
static const struct uacce_ops uacce_qm_ops = {
.get_available_instances = hisi_qm_get_available_instances,
.get_queue = hisi_qm_uacce_get_queue,
@@ -3405,9 +3513,22 @@ static const struct uacce_ops uacce_qm_ops = {
.stop_queue = hisi_qm_uacce_stop_queue,
.mmap = hisi_qm_uacce_mmap,
.ioctl = hisi_qm_uacce_ioctl,
+ .get_isolate_state = hisi_qm_get_isolate_state,
.is_q_updated = hisi_qm_is_q_updated,
+ .isolate_strategy_write = hisi_qm_isolate_strategy_write,
};
+static void qm_remove_uacce(struct hisi_qm *qm)
+{
+ struct uacce_device *uacce = qm->uacce;
+
+ if (qm->use_sva) {
+ qm_hw_err_destroy(qm);
+ uacce_remove(uacce);
+ qm->uacce = NULL;
+ }
+}
+
static int qm_alloc_uacce(struct hisi_qm *qm)
{
struct pci_dev *pdev = qm->pdev;
@@ -3420,6 +3541,7 @@ static int qm_alloc_uacce(struct hisi_qm *qm)
};
int ret;
+ INIT_LIST_HEAD(&qm->uacce_hw_errs);
ret = strscpy(interface.name, dev_driver_string(&pdev->dev),
sizeof(interface.name));
if (ret < 0)
@@ -3433,8 +3555,7 @@ static int qm_alloc_uacce(struct hisi_qm *qm)
qm->use_sva = true;
} else {
/* only consider sva case */
- uacce_remove(uacce);
- qm->uacce = NULL;
+ qm_remove_uacce(qm);
return -EINVAL;
}
@@ -5074,6 +5195,12 @@ static int qm_controller_reset_prepare(struct hisi_qm *qm)
return ret;
}
+ if (qm->use_sva) {
+ ret = qm_hw_err_isolate(qm);
+ if (ret)
+ pci_err(pdev, "failed to isolate hw err!\n");
+ }
+
ret = qm_wait_vf_prepare_finish(qm);
if (ret)
pci_err(pdev, "failed to stop by vfs in soft reset!\n");
@@ -5401,19 +5528,24 @@ static int qm_controller_reset(struct hisi_qm *qm)
ret = qm_soft_reset(qm);
if (ret) {
pci_err(pdev, "Controller reset failed (%d)\n", ret);
- qm_reset_bit_clear(qm);
- return ret;
+ goto err_reset;
}
ret = qm_controller_reset_done(qm);
- if (ret) {
- qm_reset_bit_clear(qm);
- return ret;
- }
+ if (ret)
+ goto err_reset;
pci_info(pdev, "Controller reset complete\n");
-
return 0;
+
+err_reset:
+ pci_err(pdev, "Controller reset failed (%d)\n", ret);
+ qm_reset_bit_clear(qm);
+
+ /* if resetting fails, isolate the device */
+ if (qm->use_sva && !qm->uacce->is_vf)
+ atomic_set(&qm->isolate_data.is_isolate, 1);
+ return ret;
}
/**
@@ -6186,8 +6318,7 @@ int hisi_qm_init(struct hisi_qm *qm)
err_alloc_uacce:
if (qm->use_sva) {
- uacce_remove(qm->uacce);
- qm->uacce = NULL;
+ qm_remove_uacce(qm);
}
err_irq_register:
qm_irq_unregister(qm);
diff --git a/include/linux/hisi_acc_qm.h b/include/linux/hisi_acc_qm.h
index 6cabafffd0dd..c090aaaf9974 100644
--- a/include/linux/hisi_acc_qm.h
+++ b/include/linux/hisi_acc_qm.h
@@ -8,6 +8,7 @@
#include <linux/iopoll.h>
#include <linux/module.h>
#include <linux/pci.h>
+#include <linux/uacce.h>
#define QM_QNUM_V1 4096
#define QM_QNUM_V2 1024
@@ -265,6 +266,11 @@ struct hisi_qm_list {
void (*unregister_from_crypto)(struct hisi_qm *qm);
};
+struct qm_err_isolate {
+ u32 hw_err_isolate_hz; /* user cfg freq which triggers isolation */
+ atomic_t is_isolate;
+};
+
struct hisi_qm {
enum qm_hw_ver ver;
enum qm_fun_type fun_type;
@@ -329,6 +335,9 @@ struct hisi_qm {
struct qm_shaper_factor *factor;
u32 mb_qos;
u32 type_rate;
+ struct list_head uacce_hw_errs;
+ atomic_t uacce_ref; /* reference of the uacce */
+ struct qm_err_isolate isolate_data;
};
struct hisi_qp_status {
--
2.33.0
^ permalink raw reply related [flat|nested] 9+ messages in thread
* Re: [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature
2022-06-14 12:23 [PATCH v2 0/3] crypto: hisilicon - supports device isolation feature Kai Ye
` (5 preceding siblings ...)
2022-06-14 12:23 ` [PATCH 3/3] crypto: hisilicon/qm - defining the device isolation strategy Kai Ye
@ 2022-06-14 12:34 ` yekai(A)
6 siblings, 0 replies; 9+ messages in thread
From: yekai(A) @ 2022-06-14 12:34 UTC (permalink / raw)
To: gregkh, herbert
Cc: linux-crypto, linux-accelerators, linux-kernel, linuxarm,
zhangfei.gao, wangzhou1
On 2022/6/14 20:23, Kai Ye wrote:
> Add the hardware error isolation feature for ACC. Defines a driver debugfs
> node that used to configures the hardware error frequency. When the error
> frequency is exceeded, the device will be isolated. The isolation strategy
> can be defined in each driver module. e.g. Defining the isolation strategy
> for ACC, if the AER error frequency exceeds the value of setting for a
> certain period of time, The device will not be available in user space. The
> VF device use the PF device isolation strategy. as well as the isolation
> strategy should not be set during device use.
>
> changes v1->v2:
> 1、deleted dev_to_uacce api.
> 2、add vfs node doc.
> 3、move uacce->ref to driver.
>
> Kai Ye (3):
> uacce: supports device isolation feature
> Documentation: add a isolation strategy vfs node for uacce
> crypto: hisilicon/qm - defining the device isolation strategy
>
> Documentation/ABI/testing/sysfs-driver-uacce | 17 ++
> drivers/crypto/hisilicon/qm.c | 157 +++++++++++++++++--
> drivers/misc/uacce/uacce.c | 37 +++++
> include/linux/hisi_acc_qm.h | 9 ++
> include/linux/uacce.h | 16 +-
> 5 files changed, 219 insertions(+), 17 deletions(-)
>
please ignore this patchset.
thanks
Kai
^ permalink raw reply [flat|nested] 9+ messages in thread