Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <1566461871-21992-1-git-send-email-ayal@mellanox.com>
 <20190822140635.GH13020@lunn.ch> <20190822174037.GA18030@splinter>
In-Reply-To: <20190822174037.GA18030@splinter>
From:   Arnd Bergmann <arnd@arndb.de>
Date:   Thu, 22 Aug 2019 21:11:42 +0200
Message-ID: <CAK8P3a2mQHvQKzWSKofBPdzFDTZq9oJkDHqR2PR85Hswocy45g@mail.gmail.com>
Subject: Re: [net] devlink: Add method for time-stamp on reporter's dump
To:     Ido Schimmel <idosch@idosch.org>
Cc:     Andrew Lunn <andrew@lunn.ch>, Aya Levin <ayal@mellanox.com>,
        "David S. Miller" <davem@davemloft.net>,
        Jiri Pirko <jiri@mellanox.com>,
        Networking <netdev@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Aug 22, 2019 at 7:40 PM Ido Schimmel <idosch@idosch.org> wrote:
> On Thu, Aug 22, 2019 at 04:06:35PM +0200, Andrew Lunn wrote:
> > On Thu, Aug 22, 2019 at 11:17:51AM +0300, Aya Levin wrote:
> > > When setting the dump's time-stamp, use ktime_get_real in addition to
> > > jiffies. This simplifies the user space implementation and bypasses
> > > some inconsistent behavior with translating jiffies to current time.
> >
> > Is this year 2038 safe? I don't know enough about this to answer the
> > question myself.
>
> Good point. 'struct timespec' is not considered year 2038 safe and
> unfortunately I recently made the mistake of using it to communicate
> timestamps to user space over netlink. :/ The code is still in net-next,
> so I will fix it while I can.
>
> Arnd, would it be acceptable to use 'struct __kernel_timespec' instead?

The in-kernel representation should just use 'timespec64' if you need
separate seconds and nanoseconds, you can convert that to
__kernel_timespec while copying to user space.

However, please consider two other points:

- for simplicity, the general recommendation is to use 64-bit nanoseconds
  without separate seconds for timestamps
- instead of CLOCK_REALTIME, you could use CLOCK_MONOTONIC
  timestamps that are not affected by clock_settime() or leap second jumps.

      Arnd