Re: [Xen-devel] [RFC PATCH] xen-netback: making the bandwidth limiter runtime settable

[Imre Palik <imrep.amz@xxxxxxxxx>]

On 03/17/15 12:17, Wei Liu wrote:
> On Fri, Mar 13, 2015 at 01:51:05PM +0100, Imre Palik wrote:
>> From: "Palik, Imre" <imrep@xxxxxxxxx>
>>
>> With the current netback, the bandwidth limiter's parameters are only
>> settable during vif setup time.  This patch register a watch on them, and
>> thus makes them runtime changeable.
>>
>> When the watch fires, the timer is reset.  The timer's mutex is used for
>> fencing the change.
>>
> 
> I think this is a valid idea.  Just that this commit message is not
> complete. It doesn't describe everything this patch does.
> 
>> Cc: Anthony Liguori <aliguori@xxxxxxxxxx>
>> Signed-off-by: Imre Palik <imrep@xxxxxxxxx>
>> ---
> [...]
>>      queue->rx_queue_max = XENVIF_RX_QUEUE_BYTES;
>> diff --git a/drivers/net/xen-netback/netback.c 
>> b/drivers/net/xen-netback/netback.c
>> index cab9f52..bcc1880 100644
>> --- a/drivers/net/xen-netback/netback.c
>> +++ b/drivers/net/xen-netback/netback.c
>> @@ -641,7 +641,7 @@ static void tx_add_credit(struct xenvif_queue *queue)
>>      queue->remaining_credit = min(max_credit, max_burst);
>>  }
>>  
>> -static void tx_credit_callback(unsigned long data)
>> +void xenvif_tx_credit_callback(unsigned long data)
> 
> Please keep this function static.

The trouble with that, is that now I am initialising credit_timeout.function in
drivers/net/xen-netback/interface.c .

The reason for the change is that mod_timer_pending() hits a BUG() if
the timeout function is not initialised.

> 
> And say in the commit message you change tx_credit_callback to a better
> name.
> 
>>  {
>>      struct xenvif_queue *queue = (struct xenvif_queue *)data;
>>      tx_add_credit(queue);
>> @@ -1170,7 +1170,7 @@ static bool tx_credit_exceeded(struct xenvif_queue 
>> *queue, unsigned size)
>>              queue->credit_timeout.data     =
>>                      (unsigned long)queue;
>>              queue->credit_timeout.function =
>> -                    tx_credit_callback;
>> +                    xenvif_tx_credit_callback;
>>              mod_timer(&queue->credit_timeout,
>>                        next_credit);
> [...]
>> @@ -594,13 +597,9 @@ static void xen_net_read_rate(struct xenbus_device *dev,
>>      unsigned long b, u;
>>      char *ratestr;
>>  
>> -    /* Default to unlimited bandwidth. */
>> -    *bytes = ~0UL;
>> -    *usec = 0;
>> -
>>      ratestr = xenbus_read(XBT_NIL, dev->nodename, "rate", NULL);
>>      if (IS_ERR(ratestr))
>> -            return;
>> +            goto reset;
>>  
>>      s = ratestr;
>>      b = simple_strtoul(s, &e, 10);
>> @@ -612,15 +611,21 @@ static void xen_net_read_rate(struct xenbus_device 
>> *dev,
>>      if ((s == e) || (*e != '\0'))
>>              goto fail;
>>  
>> +    kfree(ratestr);
>> +
>>      *bytes = b;
>>      *usec = u;
>>  
>> -    kfree(ratestr);
>>      return;
>>  
>> - fail:
>> +fail:
>>      pr_warn("Failed to parse network rate limit. Traffic unlimited.\n");
>>      kfree(ratestr);
>> +
>> +reset:
>> +    /* Default to unlimited bandwidth. */
>> +    *bytes = ~0UL;
>> +    *usec = 0;
>>  }
>>  
> 
> Any reason you modify this function? It is still doing the exact same
> thing, right?

These changes made sense before the channel support, but not anymore.
I will roll them back.

> 
>>  static int xen_net_read_mac(struct xenbus_device *dev, u8 mac[])
>> @@ -645,6 +650,59 @@ static int xen_net_read_mac(struct xenbus_device *dev, 
>> u8 mac[])
>>      return 0;
>>  }
>>  
>> +static void xen_net_rate_changed(struct xenbus_watch *watch,
>> +                            const char **vec, unsigned int len)
>> +{
>> +    struct xenvif *vif = container_of(watch, struct xenvif, credit_watch);
>> +    struct xenbus_device *dev = xenvif_to_xenbus_device(vif);
>> +    unsigned long   credit_bytes;
>> +    unsigned long   credit_usec;
>> +    unsigned int queue_index;
>> +
>> +    xen_net_read_rate(dev, &credit_bytes, &credit_usec);
>> +    for (queue_index = 0; queue_index < vif->num_queues; queue_index++) {
>> +            struct xenvif_queue *queue = &vif->queues[queue_index];
>> +
>> +            queue->credit_bytes = credit_bytes;
>> +            queue->credit_usec = credit_usec;
>> +            if (!mod_timer_pending(&queue->credit_timeout, jiffies) &&
>> +                    queue->remaining_credit > queue->credit_bytes) {
>> +                    queue->remaining_credit = queue->credit_bytes;
>> +            }
>> +    }
>> +}
>> +
>> +static int xen_register_watchers(struct xenbus_device *dev, struct xenvif 
>> *vif)
>> +{
>> +    int err = 0;
>> +    char *node;
>> +    unsigned maxlen = strlen(dev->nodename) + sizeof("/rate");
>> +
>> +    node = kmalloc(maxlen, GFP_KERNEL);
>> +    if (!node)
>> +            return -ENOMEM;
>> +    sprintf(node, "%s/rate", dev->nodename);
> 
> Please use snprintf. (Though I can see using sprintf is fine here but I
> want the code to be a bit future proof.)
> 
>> +    vif->credit_watch.node = node;
>> +    vif->credit_watch.callback = xen_net_rate_changed;
>> +    err = register_xenbus_watch(&vif->credit_watch);
>> +    if (err) {
>> +            pr_err("Failed to set watcher %s\n", vif->credit_watch.node);
>> +            kfree(node);
>> +            vif->credit_watch.node = 0;
>> +            vif->credit_watch.callback = 0;
> 
> Please use NULL here.
> 
> Wei.
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel