| |
| NET2Debugged 1.24 README |
| ------------------------ |
| |
| Major Changes |
| |
| o PLIP driver sort of works |
| o UDP and RAW have been partially rewritten for speed |
| o Internals heavily cleaned up, and memory monitoring of network |
| memory is now done. (On shift-scroll-lock) |
| o ARP should now not generate garbage |
| o Using MSG_PEEK can't cause race conditions and crashes |
| o Support for bootp clients. |
| o Supports RFC931 TAP authd |
| o NFS problems with certain types of network configuration are |
| fixed. |
| o Doesn't forward packets for other subnet (can cause packet storms) |
| o TCP won't ack rst frames causing packet storms (especially with |
| Lan workplace for DOS). |
| o Numerous fixes for solidity |
| o Verify_area used properly. |
| o MSG_PEEK is faster again |
| o Minor TCP fixes. Hopefully no more TCP lockups (ha!) |
| o Donald's promiscuous mode. Go forth and write protocol analysers... |
| ------------------------------------------------------------------------- |
| NOTE: |
| Drivers for this stack set must be using alloc_skb() not just |
| kmalloc. If you get millions of 'non sk_buff...' errors please check the |
| driver you are using. All Donald's drivers know about this. If you have |
| a problem driver replace all cases of |
| |
| .. =(struct sk_buff *)kmalloc(sizeof(struct sk_buff)+... |
| |
| With |
| ..=alloc_skb(sizeof(struct sk_buff)+... |
| |
| And if it uses kfree_s on the packet change that to use kfree_skbmem(). |
| |
| ------------------------------------------------------------------------- |
| Bug fixes and improvements for this section of the code should be mailed to |
| iiitac@pyr.swan.ac.uk. |
| |
| |
| Alan |
