Nikolay Aleksandrov
2020-Oct-08 10:18 UTC
[Bridge] [PATCH net] bridge: Netlink interface fix.
On Wed, 2020-10-07 at 14:49 +0000, Nikolay Aleksandrov wrote:> On Wed, 2020-10-07 at 12:07 +0000, Henrik Bjoernlund wrote: > > This commit is correcting NETLINK br_fill_ifinfo() to be able to > > handle 'filter_mask' with multiple flags asserted. > > > > Fixes: 36a8e8e265420 ("bridge: Extend br_fill_ifinfo to return MPR status") > > > > Signed-off-by: Henrik Bjoernlund <henrik.bjoernlund at microchip.com> > > Reviewed-by: Horatiu Vultur <horatiu.vultur at microchip.com> > > Suggested-by: Nikolay Aleksandrov <nikolay at nvidia.com> > > Tested-by: Horatiu Vultur <horatiu.vultur at microchip.com> > > --- > > net/bridge/br_netlink.c | 26 +++++++++++--------------- > > 1 file changed, 11 insertions(+), 15 deletions(-) > > > > The patch looks good, please don't separate the Fixes tag from the others. > Acked-by: Nikolay Aleksandrov <nikolay at nvidia.com> >TBH, this does change a user facing api (the attribute nesting), but I think in this case it's acceptable due to the format being wrong and MRP being new, so I doubt anyone is yet dumping it mixed with vlan filter_mask and checking for two identical attributes, i.e. in the old/broken case parsing the attributes into a table would hide one of them and you'd have to walk over all attributes to catch that.
On Thu, 8 Oct 2020 10:18:09 +0000 Nikolay Aleksandrov wrote:> On Wed, 2020-10-07 at 14:49 +0000, Nikolay Aleksandrov wrote: > > On Wed, 2020-10-07 at 12:07 +0000, Henrik Bjoernlund wrote: > > > This commit is correcting NETLINK br_fill_ifinfo() to be able to > > > handle 'filter_mask' with multiple flags asserted. > > > > > > Fixes: 36a8e8e265420 ("bridge: Extend br_fill_ifinfo to return MPR status") > > > > > > Signed-off-by: Henrik Bjoernlund <henrik.bjoernlund at microchip.com> > > > Reviewed-by: Horatiu Vultur <horatiu.vultur at microchip.com> > > > Suggested-by: Nikolay Aleksandrov <nikolay at nvidia.com> > > > Tested-by: Horatiu Vultur <horatiu.vultur at microchip.com> > > > --- > > > net/bridge/br_netlink.c | 26 +++++++++++--------------- > > > 1 file changed, 11 insertions(+), 15 deletions(-) > > > > > > > The patch looks good, please don't separate the Fixes tag from the others. > > Acked-by: Nikolay Aleksandrov <nikolay at nvidia.com> > > > > TBH, this does change a user facing api (the attribute nesting), but I think > in this case it's acceptable due to the format being wrong and MRP being new, so > I doubt anyone is yet dumping it mixed with vlan filter_mask and checking for > two identical attributes, i.e. in the old/broken case parsing the attributes > into a table would hide one of them and you'd have to walk over all attributes > to catch that.To be clear - this changes the uAPI as far as 5.9-rcs are concerned. So if this change was to hit 5.9 final there would be no uAPI breakage, right?
Nikolay Aleksandrov
2020-Oct-08 17:54 UTC
[Bridge] [PATCH net] bridge: Netlink interface fix.
On Thu, 2020-10-08 at 10:09 -0700, Jakub Kicinski wrote:> On Thu, 8 Oct 2020 10:18:09 +0000 Nikolay Aleksandrov wrote: > > On Wed, 2020-10-07 at 14:49 +0000, Nikolay Aleksandrov wrote: > > > On Wed, 2020-10-07 at 12:07 +0000, Henrik Bjoernlund wrote: > > > > This commit is correcting NETLINK br_fill_ifinfo() to be able to > > > > handle 'filter_mask' with multiple flags asserted. > > > > > > > > Fixes: 36a8e8e265420 ("bridge: Extend br_fill_ifinfo to return MPR status") > > > > > > > > Signed-off-by: Henrik Bjoernlund <henrik.bjoernlund at microchip.com> > > > > Reviewed-by: Horatiu Vultur <horatiu.vultur at microchip.com> > > > > Suggested-by: Nikolay Aleksandrov <nikolay at nvidia.com> > > > > Tested-by: Horatiu Vultur <horatiu.vultur at microchip.com> > > > > --- > > > > net/bridge/br_netlink.c | 26 +++++++++++--------------- > > > > 1 file changed, 11 insertions(+), 15 deletions(-) > > > > > > > > > > The patch looks good, please don't separate the Fixes tag from the others. > > > Acked-by: Nikolay Aleksandrov <nikolay at nvidia.com> > > > > > > > TBH, this does change a user facing api (the attribute nesting), but I think > > in this case it's acceptable due to the format being wrong and MRP being new, so > > I doubt anyone is yet dumping it mixed with vlan filter_mask and checking for > > two identical attributes, i.e. in the old/broken case parsing the attributes > > into a table would hide one of them and you'd have to walk over all attributes > > to catch that. > > To be clear - this changes the uAPI as far as 5.9-rcs are concerned. > So if this change was to hit 5.9 final there would be no uAPI breakage, > right?Yes, correct.