> -----Original Message-----
> From: netdev-ow...@vger.kernel.org [mailto:netdev-ow...@vger.kernel.org] On
> Behalf Of Arnd Bergmann
> Sent: Monday, February 25, 2019 6:46 AM
> To: Kirsher, Jeffrey T <jeffrey.t.kirs...@intel.com>
> Cc: David Miller <da...@davemloft.net>; Brandeburg, Jesse
> <jesse.brandeb...@intel.com>; Networking <netdev@vger.kernel.org>;
> nhor...@redhat.com; sassm...@redhat.com; jogre...@redhat.com
> Subject: Re: [net-next 05/14] i40e: fix up 32 bit timespec references
>
> On Wed, Jul 26, 2017 at 12:33 PM Jeff Kirsher
> <jeffrey.t.kirs...@intel.com> wrote:
> >
> > From: Jesse Brandeburg <jesse.brandeb...@intel.com>
> >
> > As it turns out there was only a small set of errors
> > on 32 bit, and we just needed to be using the right calls
> > for dealing with timespec64 variables.
>
> I just stumbled over code added by this older patch, and can't make sense
> of the commit description here. Was this an attempt to fix a bug, or
> just a cleanup?
>
> >
> > - then = ns_to_timespec64(delta);
> > mutex_lock(&pf->tmreg_lock);
> >
> > i40e_ptp_read(pf, &now);
> > - now = timespec64_add(now, then);
> > + timespec64_add_ns(&now, delta);
> > i40e_ptp_write(pf, (const struct timespec64 *)&now);
>
> The problem I noticed here is that 'delta' is a user provided 64-bit
> number from clock_adjtime(), and timespec64_add_ns() performs uses
> a repeated addition instead of a div/mod pair. When the number
> is large, we may end up adding a single second 8 billion times,
> which may take a while even on a fast CPU.
>
It looked like the timespec64_add_ns does a div/mod pair...? Or am I
mis-reading how the function is implemented? Quite probably.
Either way, the code is incorrect, because timespec64_add_ns doesn't actually
work with signed values. A negative delta actually ends up resulting in a
significant positive addition. Woops!
> Should the commit 0ac30ce43323 ("i40e: fix up 32 bit timespec
> references") just be reverted?
>
Yea, let's revert it.
Thanks,
Jake
> Arnd
