conditions of dump isn't limit in nla, so if you replace cb->data with
cb->nla,
sometimes it's hard to implement user's dump().

2018-07-23 19:09 GMT+08:00 Pablo Neira Ayuso <[email protected]>:

> On Mon, Jul 23, 2018 at 12:47:14PM +0200, Florian Westphal wrote:
> > Shaochun Chen points out we leak dumper filter state allocations
> > stored in dump_control->data in case there is an error before netlink
> sets
> > cb_running (after which ->done will be called at some point).
> >
> > In order to fix this, add .start functions and do the allocations
> > there.
> >
> > ->done is going to clean up, and in case error occurs before
> > ->start invocation no cleanups need to be done anymore.
>
> LGTM. Thanks Florian
>
> BTW, probably we can add cb->nla with strict typing instead, so we
> don't need to use cb->data. This cb->nla would be only valid from
> .start().
>
> Then, set cb->nla to NULL after .start() is called. Just to make sure
> people don't make wrong assumptions and try to access this from the
> classic dump path?
>
> Just an idea.
>

Reply via email to