On Mon, Jun 13, 2016 at 10:27 AM, Eric Dumazet wrote:
> On Mon, 2016-06-13 at 10:13 -0700, Eric Dumazet wrote:
>
>> Presumably we should commit changes on qdisc only if the whole
>> ->change() succeeded, but I guess lot of qdisc are buggy in this
>> respect.
I suppose so, it should work like a tr
On Mon, 2016-06-13 at 10:13 -0700, Eric Dumazet wrote:
> Presumably we should commit changes on qdisc only if the whole
> ->change() succeeded, but I guess lot of qdisc are buggy in this
> respect.
I'll post a v2 fixing this issue as well.
On Mon, 2016-06-13 at 09:28 -0700, Cong Wang wrote:
> On Sun, Jun 12, 2016 at 10:03 PM, Eric Dumazet wrote:
> > From: Eric Dumazet
> >
> > Now prio_init() can return -ENOMEM, it also has to make sure
> > any allocated qdisc are freed, since the caller (qdisc_create()) wont
> > call ->destroy() ha
On Sun, Jun 12, 2016 at 10:03 PM, Eric Dumazet wrote:
> From: Eric Dumazet
>
> Now prio_init() can return -ENOMEM, it also has to make sure
> any allocated qdisc are freed, since the caller (qdisc_create()) wont
> call ->destroy() handler for us.
But prio_tune() is called by ->change() too, so
From: Eric Dumazet
Now prio_init() can return -ENOMEM, it also has to make sure
any allocated qdisc are freed, since the caller (qdisc_create()) wont
call ->destroy() handler for us.
Fixes: cbdf45116478 ("net_sched: prio: properly report out of memory errors")
Signed-off-by: Eric Dumazet
Report
