* James Morris (jmorrisat_private) wrote: > On Thu, 13 Sep 2001, Chris Wright wrote: > > > * Chris Wright (chrisat_private) wrote: > > > > > > Finally, changed the skb_ops->alloc/clone docs to reflect that they > > > may be called from an interrupt. > > > > Well, much of the networking code can run under interrupt -- I'm not sure > it's something we need to document here. yeah, i wanted to start docuementing this. i see no reason to leave this information out of the docs. > > > hmm, considering the return type for skb_ops->clone is void, i think it is > > safe to say we are explicitly not allowing memory allocation during clone. > > (which makes sense to me). > > It's possible that someone might want to do a new allocation here. > > > should we say _anything_ here? should we > > consider supporting a new blob allocation in clone (and return an int)? > > > > We should probably cater for allocation failures for both clone() and > copy(). I'll start working on a patch. i've already got started one, see attached. (it includes the documentation changes). and i don't believe we need to allocate for the copy cases. the copy cases happen after a call to skb_alloc. the clone case happens after only a skb_head_from_pool/kmem_cache_alloc allocation (which, of course doesn't call our skb_ops->alloc hook). what do you think? -chris
This archive was generated by hypermail 2b30 : Thu Sep 13 2001 - 17:01:54 PDT