Re: [B.A.T.M.A.N.] Batman gateway lock ups

[Sven Eckelmann <sven.eckelmann@gmx.de>]

[-- Attachment #1: Type: text/plain, Size: 1289 bytes --]

On Wednesday 10 September 2008 00:45:07 Sven Eckelmann wrote:
> [...]
>     /* v0 and v1 holds pointer to new allocated struct */
>     /* a3 holds "zero" - like t0 */
>     9a8: sw	s1,8(v0) /* store pointer to client_data in new data buffer */
>     9ac: lw	v0,36(a3) /* v0 gets free_client_list.next -> lets call it
>                            next_element */
>     /* shouldn't be another instruction between load and usage of the
>        register? - like a nop */
>     9b0: sw	v0,0(v1) /* store next_element in tmp_entry.list.next */
>     9b4: sw	v1,4(v0) /* store pointer to tmp_entry next_element.prev */
>     9b8: sw	v1,36(a3) /* store pointer to tmp_entry in freeclient_list.next
> */
>     9bc: j	9d4 <cleanup_module+0x524>
>     9c0: sw	t0,4(v1) /* saves freeclient_list in tmp_entry.prev << if this
>                           would not be executed in parallel, we would get
>                           wrong data here, but because we are using mips it
>                           must be executed */
> [...]
Ok, searched inside the official mips32 instruction set reference and both 
questionable instructions are defined in a hard way without any unpredictable 
or undefined remarks. So they should be working fine.

Best regards
	Sven Eckelmann

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 197 bytes --]