Re: [CHECKER] 12 potential leaks in kernel 2.5.69

[Jean Tourrilhes <jt@bougret.hpl.hp.com>]

Ted Kremenek wrote :
> 
> linux-2.5.69/net/irda/af_irda.c (lines 868-911)
> [BUG/LEAK, kfree_skb not called on error path]

	Fixed in the mega "memory-leak" patch that I sent to Jeff a
few days ago :
	http://marc.theaimsgroup.com/?l=linux-kernel&m=105286497718003&w=2

> linux-2.5.69/drivers/net/wireless/wavelan.c (lines 3012-3041)
> [BUG/LEAK: skb_padto may return new address.  Note certain what
>            the exact semantics are, but skb_padto returns possibly
>            a new skb.  It also may free the skb pointer passed to
>            it, meaning the calling function may have a dangling reference.]
>
>     printk(KERN_DEBUG "%s: ->wavelan_packet_xmit(0x%X)\n", dev->name,
>            (unsigned) skb);
> #endif
> 
>     if (skb->len < ETH_ZLEN) {
> Start --->
>         skb = skb_padto(skb, ETH_ZLEN);
> 
>     ... DELETED 23 lines ...
> 
>         printk(KERN_INFO "skb has next\n");
> #endif
> 
>     /* Write packet on the card */
>     if(wv_packet_write(dev, skb->data, skb->len))
> Error --->
>         return 1;    /* We failed */
> 
>     dev_kfree_skb(skb);

	This is very yucky. The memory leak is easy to fix, but the
dandling reference is *very* serious. And I don't see how to fix that
without either changing the behaviour of skb_padto or the semantic of
the xmit API.
	Alan, would you mind thinking 2sec about this one ?
	Thanks...

	Jean