Re: [PATCH 07/13] ACPI/IPMI: Add reference counting for ACPI IPMI transfers

[Rafael J. Wysocki]

On Tuesday, July 23, 2013 04:09:54 PM Lv Zheng wrote:
> This patch adds reference counting for ACPI IPMI transfers to tune the
> locking granularity of tx_msg_lock.
> 
> The acpi_ipmi_msg handling is re-designed using referece counting.
> 1. tx_msg is always unlinked before complete(), so that:
>    1.1. it is safe to put complete() out side of tx_msg_lock;
>    1.2. complete() can only happen once, thus smp_wmb() is not required.
> 2. Increasing the reference of tx_msg before calling
>    ipmi_request_settime() and introducing tx_msg_lock protected
>    ipmi_cancel_tx_msg() so that a complete() can happen in parellel with
>    tx_msg unlinking in the failure cases.
> 3. tx_msg holds the reference of acpi_ipmi_device so that it can be flushed
>    and freed in the contexts other than acpi_ipmi_space_handler().
> 
> The lockdep_chains shows all acpi_ipmi locks are leaf locks after the
> tuning:
> 1. ipmi_lock is always leaf:
>    irq_context: 0
>    [ffffffff81a943f8] smi_watchers_mutex
>    [ffffffffa06eca60] driver_data.ipmi_lock
>    irq_context: 0
>    [ffffffff82767b40] &buffer->mutex
>    [ffffffffa00a6678] s_active#103
>    [ffffffffa06eca60] driver_data.ipmi_lock
> 2. without this patch applied, lock used by complete() is held after
>    holding tx_msg_lock:
>    irq_context: 0
>    [ffffffff82767b40] &buffer->mutex
>    [ffffffffa00a6678] s_active#103
>    [ffffffffa06ecce8] &(&ipmi_device->tx_msg_lock)->rlock
>    irq_context: 1
>    [ffffffffa06ecce8] &(&ipmi_device->tx_msg_lock)->rlock
>    irq_context: 1
>    [ffffffffa06ecce8] &(&ipmi_device->tx_msg_lock)->rlock
>    [ffffffffa06eccf0] &x->wait#25
>    irq_context: 1
>    [ffffffffa06ecce8] &(&ipmi_device->tx_msg_lock)->rlock
>    [ffffffffa06eccf0] &x->wait#25
>    [ffffffff81e36620] &p->pi_lock
>    irq_context: 1
>    [ffffffffa06ecce8] &(&ipmi_device->tx_msg_lock)->rlock
>    [ffffffffa06eccf0] &x->wait#25
>    [ffffffff81e36620] &p->pi_lock
>    [ffffffff81e5d0a8] &rq->lock
> 3. with this patch applied, tx_msg_lock is always leaf:
>    irq_context: 0
>    [ffffffff82767b40] &buffer->mutex
>    [ffffffffa00a66d8] s_active#107
>    [ffffffffa07ecdc8] &(&ipmi_device->tx_msg_lock)->rlock
>    irq_context: 1
>    [ffffffffa07ecdc8] &(&ipmi_device->tx_msg_lock)->rlock
> 
> Signed-off-by: Lv Zheng <lv.zheng@xxxxxxxxx>
> Cc: Zhao Yakui <yakui.zhao@xxxxxxxxx>
> Reviewed-by: Huang Ying <ying.huang@xxxxxxxxx>
> ---
>  drivers/acpi/acpi_ipmi.c |  107 +++++++++++++++++++++++++++++++++-------------
>  1 file changed, 77 insertions(+), 30 deletions(-)
> 
> diff --git a/drivers/acpi/acpi_ipmi.c b/drivers/acpi/acpi_ipmi.c
> index 2a09156..0ee1ea6 100644
> --- a/drivers/acpi/acpi_ipmi.c
> +++ b/drivers/acpi/acpi_ipmi.c
> @@ -105,6 +105,7 @@ struct acpi_ipmi_msg {
>  	u8	data[ACPI_IPMI_MAX_MSG_LENGTH];
>  	u8	rx_len;
>  	struct acpi_ipmi_device *device;
> +	atomic_t	refcnt;

Again: kref, please?

>  };
>  
>  /* IPMI request/response buffer per ACPI 4.0, sec 5.5.2.4.3.2 */
> @@ -195,22 +196,47 @@ static struct acpi_ipmi_device *acpi_ipmi_get_selected_smi(void)
>  	return ipmi_device;
>  }
>  
> -static struct acpi_ipmi_msg *acpi_alloc_ipmi_msg(struct acpi_ipmi_device *ipmi)
> +static struct acpi_ipmi_msg *ipmi_msg_alloc(void)
>  {
> +	struct acpi_ipmi_device *ipmi;
>  	struct acpi_ipmi_msg *ipmi_msg;
> -	struct pnp_dev *pnp_dev = ipmi->pnp_dev;
>  
> +	ipmi = acpi_ipmi_get_selected_smi();
> +	if (!ipmi)
> +		return NULL;
>  	ipmi_msg = kzalloc(sizeof(struct acpi_ipmi_msg), GFP_KERNEL);
> -	if (!ipmi_msg)	{
> -		dev_warn(&pnp_dev->dev, "Can't allocate memory for ipmi_msg\n");
> +	if (!ipmi_msg) {
> +		acpi_ipmi_dev_put(ipmi);
>  		return NULL;
>  	}
> +	atomic_set(&ipmi_msg->refcnt, 1);
>  	init_completion(&ipmi_msg->tx_complete);
>  	INIT_LIST_HEAD(&ipmi_msg->head);
>  	ipmi_msg->device = ipmi;
> +
>  	return ipmi_msg;
>  }
>  
> +static struct acpi_ipmi_msg *
> +acpi_ipmi_msg_get(struct acpi_ipmi_msg *tx_msg)
> +{
> +	if (tx_msg)
> +		atomic_inc(&tx_msg->refcnt);
> +	return tx_msg;
> +}
> +
> +static void ipmi_msg_release(struct acpi_ipmi_msg *tx_msg)
> +{
> +	acpi_ipmi_dev_put(tx_msg->device);
> +	kfree(tx_msg);
> +}
> +
> +static void acpi_ipmi_msg_put(struct acpi_ipmi_msg *tx_msg)
> +{
> +	if (tx_msg && atomic_dec_and_test(&tx_msg->refcnt))
> +		ipmi_msg_release(tx_msg);
> +}
> +
>  #define		IPMI_OP_RGN_NETFN(offset)	((offset >> 8) & 0xff)
>  #define		IPMI_OP_RGN_CMD(offset)		(offset & 0xff)
>  static int acpi_format_ipmi_request(struct acpi_ipmi_msg *tx_msg,
> @@ -300,7 +326,7 @@ static void acpi_format_ipmi_response(struct acpi_ipmi_msg *msg,
>  
>  static void ipmi_flush_tx_msg(struct acpi_ipmi_device *ipmi)
>  {
> -	struct acpi_ipmi_msg *tx_msg, *temp;
> +	struct acpi_ipmi_msg *tx_msg;
>  	unsigned long flags;
>  
>  	/*
> @@ -311,16 +337,46 @@ static void ipmi_flush_tx_msg(struct acpi_ipmi_device *ipmi)
>  	 */
>  	while (atomic_read(&ipmi->refcnt) > 1) {
>  		spin_lock_irqsave(&ipmi->tx_msg_lock, flags);
> -		list_for_each_entry_safe(tx_msg, temp,
> -					 &ipmi->tx_msg_list, head) {
> +		while (!list_empty(&ipmi->tx_msg_list)) {
> +			tx_msg = list_first_entry(&ipmi->tx_msg_list,
> +						  struct acpi_ipmi_msg,
> +						  head);
> +			list_del(&tx_msg->head);
> +			spin_unlock_irqrestore(&ipmi->tx_msg_lock, flags);
> +
>  			/* wake up the sleep thread on the Tx msg */
>  			complete(&tx_msg->tx_complete);
> +			acpi_ipmi_msg_put(tx_msg);
> +			spin_lock_irqsave(&ipmi->tx_msg_lock, flags);
>  		}
>  		spin_unlock_irqrestore(&ipmi->tx_msg_lock, flags);
> +
>  		schedule_timeout_uninterruptible(msecs_to_jiffies(1));
>  	}
>  }
>  
> +static void ipmi_cancel_tx_msg(struct acpi_ipmi_device *ipmi,
> +			       struct acpi_ipmi_msg *msg)
> +{
> +	struct acpi_ipmi_msg *tx_msg;
> +	int msg_found = 0;

Use bool?

> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&ipmi->tx_msg_lock, flags);
> +	list_for_each_entry(tx_msg, &ipmi->tx_msg_list, head) {
> +		if (msg == tx_msg) {
> +			msg_found = 1;
> +			break;
> +		}
> +	}
> +	if (msg_found)
> +		list_del(&tx_msg->head);

The list_del() can be done when you set msg_found.

> +	spin_unlock_irqrestore(&ipmi->tx_msg_lock, flags);
> +
> +	if (msg_found)
> +		acpi_ipmi_msg_put(tx_msg);
> +}
> +
>  static void ipmi_msg_handler(struct ipmi_recv_msg *msg, void *user_msg_data)
>  {
>  	struct acpi_ipmi_device *ipmi_device = user_msg_data;
> @@ -343,12 +399,15 @@ static void ipmi_msg_handler(struct ipmi_recv_msg *msg, void *user_msg_data)
>  			break;
>  		}
>  	}
> +	if (msg_found)
> +		list_del(&tx_msg->head);
> +	spin_unlock_irqrestore(&ipmi_device->tx_msg_lock, flags);
>  
>  	if (!msg_found) {
>  		dev_warn(&pnp_dev->dev,
>  			 "Unexpected response (msg id %ld) is returned.\n",
>  			 msg->msgid);
> -		goto out_lock;
> +		goto out_msg;
>  	}
>  
>  	/* copy the response data to Rx_data buffer */
> @@ -360,14 +419,11 @@ static void ipmi_msg_handler(struct ipmi_recv_msg *msg, void *user_msg_data)
>  	}
>  	tx_msg->rx_len = msg->msg.data_len;
>  	memcpy(tx_msg->data, msg->msg.data, tx_msg->rx_len);
> -	/* tx_msg content must be valid before setting msg_done flag */
> -	smp_wmb();
>  	tx_msg->msg_done = 1;
>  
>  out_comp:
>  	complete(&tx_msg->tx_complete);
> -out_lock:
> -	spin_unlock_irqrestore(&ipmi_device->tx_msg_lock, flags);
> +	acpi_ipmi_msg_put(tx_msg);
>  out_msg:
>  	ipmi_free_recv_msg(msg);
>  }
> @@ -493,21 +549,17 @@ acpi_ipmi_space_handler(u32 function, acpi_physical_address address,
>  	if ((function & ACPI_IO_MASK) == ACPI_READ)
>  		return AE_TYPE;
>  
> -	ipmi_device = acpi_ipmi_get_selected_smi();
> -	if (!ipmi_device)
> +	tx_msg = ipmi_msg_alloc();
> +	if (!tx_msg)
>  		return AE_NOT_EXIST;
> -
> -	tx_msg = acpi_alloc_ipmi_msg(ipmi_device);
> -	if (!tx_msg) {
> -		status = AE_NO_MEMORY;
> -		goto out_ref;
> -	}
> +	ipmi_device = tx_msg->device;
>  
>  	if (acpi_format_ipmi_request(tx_msg, address, value) != 0) {
> -		status = AE_TYPE;
> -		goto out_msg;
> +		ipmi_msg_release(tx_msg);
> +		return AE_TYPE;
>  	}
>  
> +	acpi_ipmi_msg_get(tx_msg);
>  	spin_lock_irqsave(&ipmi_device->tx_msg_lock, flags);
>  	list_add_tail(&tx_msg->head, &ipmi_device->tx_msg_list);
>  	spin_unlock_irqrestore(&ipmi_device->tx_msg_lock, flags);
> @@ -518,21 +570,16 @@ acpi_ipmi_space_handler(u32 function, acpi_physical_address address,
>  				   NULL, 0, 0, 0);
>  	if (err) {
>  		status = AE_ERROR;
> -		goto out_list;
> +		goto out_msg;
>  	}
>  	rem_time = wait_for_completion_timeout(&tx_msg->tx_complete,
>  					       IPMI_TIMEOUT);
>  	acpi_format_ipmi_response(tx_msg, value, rem_time);
>  	status = AE_OK;
>  
> -out_list:
> -	spin_lock_irqsave(&ipmi_device->tx_msg_lock, flags);
> -	list_del(&tx_msg->head);
> -	spin_unlock_irqrestore(&ipmi_device->tx_msg_lock, flags);
>  out_msg:
> -	kfree(tx_msg);
> -out_ref:
> -	acpi_ipmi_dev_put(ipmi_device);
> +	ipmi_cancel_tx_msg(ipmi_device, tx_msg);
> +	acpi_ipmi_msg_put(tx_msg);
>  	return status;
>  }
>  
> 
-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/