Comment # 31
on bug 6898
from Michael Mann
(In reply to comment #30)
> We can keep lowering the default memory limit, but the real
> problem is that the dissector is allocating all the memory for reassembly
> 'up front' on the first packet, based on the length field in that packet. Is
> there any reason it can't use the reassembly.c mechanism that only allocates
> enough for each received fragment until the entire packet is ready to
> reassemble?
> Or something along the lines of Michael's suggestion wrapping
> tcp_dissect_pdus?
I tried to go down that road but gave up. There was something about the
reassembly that didn't quite fit, like an "extra" byte or a "skipped" byte that
didn't make it very compatible with existing reassembly APIs.
You are receiving this mail because:
- You are watching all bug changes.