Re: xt_u32 20070605 (kernel)

[Patrick McHardy <kaber@trash.net>]

Jan Engelhardt wrote:
>>>+static bool xt_u32_match(const struct sk_buff *skb,
>>>+			 const struct net_device *in,
>>>+			 const struct net_device *out,
>>>+			 const struct xt_match *match, const void *matchinfo,
>>>+			 int offset, unsigned int protoff, bool *hotdrop)
>>>+{
>>>+	const struct xt_u32 *data = matchinfo;
>>>+	bool ret;
>>>+
>>>+	spin_lock_bh(&xt_u32_lock);
>>>+	ret = xt_u32_match_it(data, skb);
>>
>>Do we really need this wrapper?
> 
> 
> I thought "yes". Otherwise, I'd need a spin_unlock_bh at every return in
> xt_u32_match_it(), and also would need to fiddle data->invert too.
> With two functions (_match and _match_it), this is done nicely IMHO.

OK.

>>>+MODULE_PARM_DESC(buffer_size, "Buffer size to hold a packet "
>>>+	"(default: 65536 bytes)");
>>
>>No parameter for this. We need 64k.
> 
> 
> If no interface in a system has an MTU >17k (to be determined by
> the system's owner), why need 64?


Again, TSO. Long-term we want to move away from packet copying, so
I don't want to add any module parameters for this.