On 2012-11-12 15:41, Nickolai Zeldovich wrote: > On Mon, Nov 12, 2012 at 4:37 AM, Jan Kiszka <jan.kis...@siemens.com> wrote: >> On 2012-11-12 01:59, Nickolai Zeldovich wrote: >>> LWIP can generate packets with a source of 0.0.0.0, which triggers an >>> assertion failure in arp_table_add(). Instead of crashing, simply return >>> to avoid adding an invalid ARP table entry. >> >> I would prefer to filter out such invalid packets at a different level. >> Did you analyzed which path it takes through the stack? > > The particular packet that crashed qemu for me was a gratuitous ARP, > though it looks like all three calls to arp_table_add() in arp_input() > can trigger this. > > Popping up one level, I'm not sure why arp_table_add() and > arp_table_search() need a special case for 0.0.0.0/8 in the first > place. I couldn't find any other code that assumes the ARP table > cannot contain 0.0.0.0/8 entries. Would anything break if the check > for 0.0.0.0/8 was removed from arp_table_add() and arp_table_search() > altogether?
0.0.0.0/8 are source-only, invalid as destination. So they have no place in the ARP table. OK, let's follow your path and filter them in arp_table_add. Just add the missing braces and resend. Jan -- Siemens AG, Corporate Technology, CT RTC ITP SDP-DE Corporate Competence Center Embedded Linux
