Re: [Xen-devel] [BUG]: [PATCH added] Incorrect queue counters in netback interface

[Juergen Gross <jgross@xxxxxxxx>]

On 23/12/16 15:24, Mart van Santen wrote:
> Hello Xen Devel,
> 
> We encountered an issue with one of our instances pushing a lot of
> (network) data, resulting in incorrect bandwidth reporting. We do poll
> interface traffic every 5 minutes for every interface of our instances
> (from Dom0) and store those values for reporting/graphing. At some point
> we discovered reports of machines pushing a lot of data, were incorrect
> and were going down in chucks of about 4 GB, while you expect this is an
> always increasing number (or wraps around a 32 or 64 bit counter)
> 
> After futher analysing the problem and the source code we discovered
> that the counters of the queues in the code are handled as 32-bit
> integers, while the total bytes tx/rx for the while interface are
> counted as 64-bit longs.
> 
> 
> Attached a patch to solve the problem.

...

> --- a/drivers/net/xen-netback/common.h        2016-12-22 15:41:07.785535748 
> +0000
> +++ b/drivers/net/xen-netback/common.h        2016-12-23 13:08:18.123080064 
> +0000
> @@ -119,10 +119,10 @@
>        * A subset of struct net_device_stats that contains only the
>        * fields that are updated in netback.c for each queue.
>        */
> -     unsigned int rx_bytes;
> -     unsigned int rx_packets;
> -     unsigned int tx_bytes;
> -     unsigned int tx_packets;
> +     unsigned long rx_bytes;
> +     unsigned long rx_packets;
> +     unsigned long tx_bytes;
> +     unsigned long tx_packets;

So on 32 bit kernels you'll still encounter the same problem.

Please use the type "u64" as (nearly) everywhere else.

You might want to read Documentation/SubmittingPatches in the Linux
kernel source tree.


Juergen

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