On Wed, Apr 27, 2022 at 09:50:06AM -0700, Eric Dumazet wrote:
> On Wed, Apr 27, 2022 at 9:35 AM Willy Tarreau <w@1wt.eu> wrote:
> >
> > On Wed, Apr 27, 2022 at 12:07:14PM +0200, Willy Tarreau wrote:
> > > On Wed, Apr 27, 2022 at 05:56:41PM +0800, kernel test robot wrote:
> > > > Hi Willy,
> > > >
> > > > I love your patch! Yet something to improve:
> > > >
> > > > [auto build test ERROR on net/master]
> > > >
> > > > url:    https://github.com/intel-lab-lkp/linux/commits/Willy-Tarreau/insufficient-TCP-source-port-randomness/20220427-145651
> > > > base:   https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git 71cffebf6358a7f5031f5b208bbdc1cb4db6e539
> > > > config: i386-randconfig-r026-20220425 (https://download.01.org/0day-ci/archive/20220427/202204271705.VrWNPv7n-lkp@intel.com/config)
> > > > compiler: gcc-11 (Debian 11.2.0-20) 11.2.0
> > > > reproduce (this is a W=1 build):
> > > >         # https://github.com/intel-lab-lkp/linux/commit/01b26e522b598adf346b809075880feab3dcdc08
> > > >         git remote add linux-review https://github.com/intel-lab-lkp/linux
> > > >         git fetch --no-tags linux-review Willy-Tarreau/insufficient-TCP-source-port-randomness/20220427-145651
> > > >         git checkout 01b26e522b598adf346b809075880feab3dcdc08
> > > >         # save the config file
> > > >         mkdir build_dir && cp config build_dir/.config
> > > >         make W=1 O=build_dir ARCH=i386 SHELL=/bin/bash
> > > >
> > > > If you fix the issue, kindly add following tag as appropriate
> > > > Reported-by: kernel test robot <lkp@intel.com>
> > > >
> > > > All errors (new ones prefixed by >>):
> > > >
> > > >    ld: net/ipv4/inet_hashtables.o: in function `__inet_hash_connect':
> > > > >> inet_hashtables.c:(.text+0x187d): undefined reference to `__umoddi3'
> > >
> > > Argh! indeed, we spoke about using div_u64_rem() at the beginning and
> > > that one vanished over time. Will respin it.
> >
> > I fixed it, built it for i386 and x86_64, tested it on x86_64 and confirmed
> > that it still does what I need. The change is only this:
> >
> > -       offset = (READ_ONCE(table_perturb[index]) + (port_offset >> 32)) % remaining;
> > +       div_u64_rem(READ_ONCE(table_perturb[index]) + (port_offset >> 32), remaining, &offset);
> >
> > I'll send a v2 series in a few hours if there are no more comments.
> 
> We really do not need 33 bits here.
> 
> I would suggest using a 32bit divide.
> 
> offset = READ_ONCE(table_perturb[index]) + (port_offset >> 32));
> offset %= remaining;

Yeah much better indeed, I'll do that. Thanks Eric!

Willy