[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 9 Oct 2020 20:43:09 -0400
From: Willem de Bruijn <willemb@...gle.com>
To: Deepa Dinamani <deepa.kernel@...il.com>
Cc: Christian Eggers <ceggers@...i.de>,
"David S . Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>,
Christoph Hellwig <hch@....de>,
Linux Network Devel Mailing List <netdev@...r.kernel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Arnd Bergmann <arnd@...db.de>
Subject: Re: [PATCH net 1/2] socket: fix option SO_TIMESTAMPING_NEW
On Fri, Oct 9, 2020 at 8:30 PM Deepa Dinamani <deepa.kernel@...il.com> wrote:
>
> On Fri, Oct 9, 2020 at 3:32 AM Christian Eggers <ceggers@...i.de> wrote:
> >
> > The comparison of optname with SO_TIMESTAMPING_NEW is wrong way around,
> > so SOCK_TSTAMP_NEW will first be set and than reset again. Additionally
> > move it out of the test for SOF_TIMESTAMPING_RX_SOFTWARE as this seems
> > unrelated.
>
> The SOCK_TSTAMP_NEW is reset only in the case when
> SOF_TIMESTAMPING_RX_SOFTWARE is not set.
> Note that we only call sock_enable_timestamp() at that time.
>
> Why would SOCK_TSTAMP_NEW be relevant otherwise?
Other timestamps can be configured, such as hardware timestamps.
As the follow-on patch shows, there is also the issue of overlap
between SO_TIMESTAMP(NS) and SO_TIMESTAMPING.
Don't select OLD on timestamp disable, which may only disable
some of the ongoing timestamping.
Setting based on the syscall is simpler, too. __sock_set_timestamps
already uses for SO_TIMESTAMP(NS) the valbool approach I
suggest for SO_TIMESTAMPING.
The fallthrough can also be removed. My rough patch missed that.
Powered by blists - more mailing lists