Re: [PATCH} Network interface for IPMI

[Corey Minyard <cminyard@mvista.com>]

Ok, I agree.  You are right, at least the ioctls can be fixed up.

-Corey

Andi Kleen wrote:

>Corey Minyard <cminyard@mvista.com> writes:
>
>  
>
>>I have decided
>>that the network interface for IPMI is a good thing, as the IPMI
>>device ioctls have pointers and require ugly hacks.  None should be
>>needed for the network interface.
>>    
>>
>
>That's a joke, right? 
> 
>
>  
>
>>+struct ipmi_sock_msg {
>>+	int                   recv_type;
>>+	long                  msgid;
>>    
>>
>        ^^^^^^^^^^^
>
>Of course long would need to be always emulated. Your patch
>shows exactly why packet based protocols for this are a bad
>idea. The problem is that people will get it wrong, and then
>it's nearly impossible to fix for a socket based protocol
>because read/write cannot be easily hooked
>(with ipsec we have exactly this problem already)
>
>ioctls at least can be fixed up. Please keep using them.
>
>  
>
>>+	int                   data_len;
>>+	unsigned char         data[0];
>>    
>>
>
>And I don't even want to know what's in there.
>
>Andrew, please don't apply this broken patch.
>
>-Andi
>
>  
>