Received-SPF: pass (google.com: domain of linux-crypto-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 23 Jun 2022 11:05:23 +0200
From:   Greg KH <gregkh@linuxfoundation.org>
To:     Kai Ye <yekai13@huawei.com>
Cc:     herbert@gondor.apana.org.au, linux-crypto@vger.kernel.org,
        linux-accelerators@lists.ozlabs.org, linux-kernel@vger.kernel.org,
        linuxarm@huawei.com, zhangfei.gao@linaro.org,
        wangzhou1@hisilicon.com
Subject: Re: [PATCH v4 3/3] crypto: hisilicon/qm - defining the device
 isolation strategy
Message-ID: <YrQs06SXaqwaoAOY@kroah.com>
References: <20220623061452.40732-1-yekai13@huawei.com>
 <20220623061452.40732-4-yekai13@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220623061452.40732-4-yekai13@huawei.com>
Precedence: bulk

On Thu, Jun 23, 2022 at 02:14:52PM +0800, Kai Ye wrote:
> Define the device isolation strategy by the device driver. The
> user configures a frequency value by uacce interface. If the
> slot reset frequency exceeds the value of setting for a certain
> period of time, the device will not be available in user space.
> This frequency is an abstract number of times that can be
> considered to occur in a time window. The time window can be set
> to one hour or one day. 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 | 177 +++++++++++++++++++++++++++++++---
>  include/linux/hisi_acc_qm.h   |   9 ++
>  2 files changed, 174 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
> index ad83c194d664..f92cf20fc84e 100644
> --- a/drivers/crypto/hisilicon/qm.c
> +++ b/drivers/crypto/hisilicon/qm.c
> @@ -417,6 +417,16 @@ struct hisi_qm_resource {
>  	struct list_head list;
>  };
>  
> +/**
> + * struct qm_hw_err - structure of describes the device err
> + * @list: hardware error list
> + * @timestamp: timestamp when the error occurred
> + */
> +struct qm_hw_err {
> +	struct list_head list;
> +	unsigned long long timestamp;
> +};
> +
>  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,
> @@ -3410,6 +3420,125 @@ 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
> + * according to user's configuration of isolation strategy. Warning: this
> + * API should be called while there the users on this device are suspended
> + * by slot resetting preparation of PCI AER.
> + * @qm: the uacce device
> + */
> +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;
> +
> +	mutex_lock(&isolate->isolate_lock);
> +	hw_err->timestamp = jiffies;
> +	list_for_each_entry_safe(err, tmp, &qm->uacce_hw_errs, list) {
> +		if ((hw_err->timestamp - err->timestamp) / HZ >
> +		    SECONDS_PER_HOUR) {
> +			list_del(&err->list);
> +			kfree(err);
> +		} else {
> +			count++;
> +		}
> +	}
> +	list_add(&hw_err->list, &qm->uacce_hw_errs);
> +	mutex_unlock(&isolate->isolate_lock);
> +
> +	if (count >= isolate->hw_err_isolate_hz)
> +		atomic_set(&isolate->is_isolate, 1);

Why is this an atomic value?  You can change it after it has been set,
and it can be modified after it is read.  A normal 'bool' would work
exactly the same.


> +
> +	return 0;
> +}
> +
> +static void qm_hw_err_destroy(struct hisi_qm *qm)
> +{
> +	struct qm_hw_err *err, *tmp;
> +
> +	mutex_lock(&qm->isolate_data.isolate_lock);
> +	list_for_each_entry_safe(err, tmp, &qm->uacce_hw_errs, list) {
> +		list_del(&err->list);
> +		kfree(err);
> +	}
> +	mutex_unlock(&qm->isolate_data.isolate_lock);
> +}
> +
> +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));
> +	else
> +		pf_qm = qm;
> +
> +	return atomic_read(&pf_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, size_t len)
> +{
> +	struct hisi_qm *qm = uacce->priv;
> +	unsigned long val;
> +
> +#define MAX_ISOLATE_STRATEGY	65535

What is this value?  Please document it better.

> +
> +	/* Must be set by PF */
> +	if (uacce->is_vf) {
> +		dev_info(&qm->pdev->dev, "the isolation strategy must be set by PF.\n");

Do not let userspace spam the kernel log for no good reason.

And why is this an info?

> +		return -EINVAL;

This looks like an error.

> +	}
> +
> +	if (atomic_read(&qm->isolate_data.is_isolate))
> +		return -EINVAL;

What happens if this changes right after reading it?

> +
> +	if (kstrtoul(buf, 0, &val) < 0)
> +		return -EINVAL;
> +
> +	if (val > MAX_ISOLATE_STRATEGY)
> +		return -EINVAL;
> +
> +	qm->isolate_data.hw_err_isolate_hz = val;
> +
> +	/* After the policy is updated, need to reset the hardware err list */
> +	qm_hw_err_destroy(qm);
> +
> +	return 0;

Don't you need to return the number of bytes read, not 0?

> +}
> +
> +static int hisi_qm_isolate_strategy_read(struct uacce_device *uacce, char *buf)
> +{
> +	struct hisi_qm *qm = uacce->priv;
> +	struct hisi_qm *pf_qm;
> +	unsigned long val;
> +
> +	if (uacce->is_vf) {
> +		pf_qm = pci_get_drvdata(pci_physfn(qm->pdev));
> +		val = pf_qm->isolate_data.hw_err_isolate_hz;
> +	} else {
> +		val = qm->isolate_data.hw_err_isolate_hz;
> +	}
> +
> +	return sysfs_emit(buf, "%lu\n", val);
> +}
> +
>  static const struct uacce_ops uacce_qm_ops = {
>  	.get_available_instances = hisi_qm_get_available_instances,
>  	.get_queue = hisi_qm_uacce_get_queue,
> @@ -3419,8 +3548,22 @@ static const struct uacce_ops uacce_qm_ops = {
>  	.mmap = hisi_qm_uacce_mmap,
>  	.ioctl = hisi_qm_uacce_ioctl,
>  	.is_q_updated = hisi_qm_is_q_updated,
> +	.get_isolate_state = hisi_qm_get_isolate_state,
> +	.isolate_strategy_write = hisi_qm_isolate_strategy_write,
> +	.isolate_strategy_read = hisi_qm_isolate_strategy_read,
>  };
>  
> +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 +3576,8 @@ static int qm_alloc_uacce(struct hisi_qm *qm)
>  	};
>  	int ret;
>  
> +	INIT_LIST_HEAD(&qm->uacce_hw_errs);
> +	mutex_init(&qm->isolate_data.isolate_lock);
>  	ret = strscpy(interface.name, dev_driver_string(&pdev->dev),
>  		      sizeof(interface.name));
>  	if (ret < 0)
> @@ -3446,8 +3591,7 @@ 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 +5253,12 @@ 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 +5586,25 @@ 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 +6402,7 @@ 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..44454150c205 100644
> --- a/include/linux/hisi_acc_qm.h
> +++ b/include/linux/hisi_acc_qm.h
> @@ -271,6 +271,13 @@ struct hisi_qm_poll_data {
>  	u16 *qp_finish_id;
>  };
>  
> +struct qm_err_isolate {
> +	struct mutex isolate_lock;
> +	/* user cfg freq which triggers isolation */
> +	u32 hw_err_isolate_hz;
> +	atomic_t is_isolate;

Again, why is this an atomic value?

thanks,

greg k-h