| Jamal Hadi Salim | d892afe | 2007-10-26 02:49:09 -0700 | [diff] [blame] | 1 | |
| 2 | The "enviromental" rules for authors of any new tc actions are: |
| 3 | |
| 4 | 1) If you stealeth or borroweth any packet thou shalt be branching |
| 5 | from the righteous path and thou shalt cloneth. |
| 6 | |
| 7 | For example if your action queues a packet to be processed later |
| 8 | or intentionaly branches by redirecting a packet then you need to |
| 9 | clone the packet. |
| 10 | There are certain fields in the skb tc_verd that need to be reset so we |
| 11 | avoid loops etc. A few are generic enough so much so that skb_act_clone() |
| 12 | resets them for you. So invoke skb_act_clone() rather than skb_clone() |
| 13 | |
| 14 | 2) If you munge any packet thou shalt call pskb_expand_head in the case |
| 15 | someone else is referencing the skb. After that you "own" the skb. |
| 16 | You must also tell us if it is ok to munge the packet (TC_OK2MUNGE), |
| 17 | this way any action downstream can stomp on the packet. |
| 18 | |
| 19 | 3) dropping packets you dont own is a nono. You simply return |
| 20 | TC_ACT_SHOT to the caller and they will drop it. |
| 21 | |
| 22 | The "enviromental" rules for callers of actions (qdiscs etc) are: |
| 23 | |
| 24 | *) thou art responsible for freeing anything returned as being |
| 25 | TC_ACT_SHOT/STOLEN/QUEUED. If none of TC_ACT_SHOT/STOLEN/QUEUED is |
| 26 | returned then all is great and you dont need to do anything. |
| 27 | |
| 28 | Post on netdev if something is unclear. |
| 29 | |