On Tue, Oct 20, 2020 at 01:49:22AM +0200, Alexandr Nedvedicky wrote:
> Currently we need to keep pf_rm_rule() under both locks. The function
> might be calling pf_tag_unref(), pf_dynaddr_remove()... which alter lists,
> which are currently supposed to be protected by
Hello Klemens,
> > the change is fairly large, but mostly mechanical.
> Relocating malloc(9) and pool(9) seems good but other pf_*() calls are
> now locked inconsistently after your diff.
>
> Given that only reason about "allocations" this is either an oversight
> and should be fixed or you
On Fri, Oct 16, 2020 at 11:37:22AM +0200, Alexandr Nedvedicky wrote:
> I've just found a forgotten diff in my tree. The diff pushes the NET_LCOK()
> further down in PF driver ioctl() path. The idea is to avoid sleeping while
> holding a NET_LOCK(). this typically may happen when we need to
Hello,
I've just found a forgotten diff in my tree. The diff pushes the NET_LCOK()
further down in PF driver ioctl() path. The idea is to avoid sleeping while
holding a NET_LOCK(). this typically may happen when we need to allocate
memory. The diff is the first step as it takes care of