Webserver Woes++ (coldfire mcf5235 + rtems-4.7.1 + GoAhead)

[Ian Caddy]

Hi Chris,


CWolfe at motioncontrol.org wrote:
> I believe the problem is solved, though I don't know why yet. I changed the number of 
> maximum_file_descriptors from 100 to 200, and now the server doesn't crash.
> 

That is good, but...

> 
> On 10 Jul 2007 at 9:36, Ian Caddy wrote:
> 
>> Hi Christopher,
>>
>> Do you have enough mbuf and mbuf cluster space allocated?
> 64K
>> Can you provide your network configuration parameters 
>> (rtems_bsdnet_config) for your system?
> static struct rtems_bsdnet_ifconfig netdriver_config = {
> 	RTEMS_BSP_NETWORK_DRIVER_NAME,		/* name */
> 	RTEMS_BSP_NETWORK_DRIVER_ATTACH,	/* attach function */
> 	NULL,				                      /* No more interfaces */
> 	"192.168.x.x",		              /* IP address */
> 	"255.255.255.0",		              /* IP net mask */
> 	ethernet_address,                 /* Ethernet hardware address */
> 	0				                          /* Use default driver parameters */
> };
> 
> /*
>  * Network configuration
>  */
> struct rtems_bsdnet_config rtems_bsdnet_config = {
> 	&netdriver_config,
> 	NULL,
> 	1,			/* Default network task priority */
>   64*1024,			/* Default mbuf capacity */

I would also look at increasing your mbuf count if possible.  The two 
sorts of buffers used in the system are mbufs and mbuf clusters.  Now a 
cluster is only used if a packet is greater than a certain size, from 
memory, it is around 250 bytes.  For any smaller packet, and mbuf is used.

The other thing is the mbuf and mbuf cluster do not share their memory 
pool, if you run out of one or the other then things tend to get stuck.

regards,

Ian Caddy

-- 
Ian Caddy
Goanna Technologies Pty Ltd
+61 8 9221 1860