| Message ID | 20230428094232.528745-1-amusil@redhat.com |
|---|---|
| Headers | show |
| Series | Prevent race in packet buffering | expand |
On 4/28/23 11:42, Ales Musil wrote: > The series is split into 5 patches. > > It aims to simplify the packet buffering > process and encapsulate it inside mac-learn > module. This is done through refactoring > and preparation. The last patch then fixes the > race condition in packet buffering by utilizing > I-P node. > Hi Ales, I reviewed the series and I left some minor comments for patches 1-4. If it's fine by you I can go ahead and fold those in and apply them to the main branch. That should make it easier to respin a new version. Patch 5, on the other hand, needs some more re-design I think. I'm afraid the performance cost it introduces is too high. Please let me know if you're ok with me applying the first 4 patches. Regards, Dumitru > The difference between v4 and v5 is that the > first patch was devided even further into four > patches, the last two patches from v4 are squashed > into v5 as this their separation didn't have any > meaningful value. > > Ales Musil (5): > mac-learn: Make the mac_binding struct more flexible > ovn-util: Add helper function to parse ip into mapped ipv6 > pinctrl: Simplify packet buffering structure > controller: Move the packet buffering into mac-learn > controller: Prevent race in packet buffering > > controller/mac-learn.c | 244 ++++++++++++++++++++---- > controller/mac-learn.h | 88 +++++++-- > controller/ovn-controller.c | 200 +++++++++++++++++++- > controller/pinctrl.c | 356 ++++++++---------------------------- > controller/pinctrl.h | 6 +- > lib/ovn-util.c | 12 ++ > lib/ovn-util.h | 2 + > northd/northd.c | 32 +--- > utilities/ovn-trace.c | 5 +- > 9 files changed, 583 insertions(+), 362 deletions(-) >
On Tue, May 9, 2023 at 3:40 PM Dumitru Ceara <dceara@redhat.com> wrote: > On 4/28/23 11:42, Ales Musil wrote: > > The series is split into 5 patches. > > > > It aims to simplify the packet buffering > > process and encapsulate it inside mac-learn > > module. This is done through refactoring > > and preparation. The last patch then fixes the > > race condition in packet buffering by utilizing > > I-P node. > > > > Hi Ales, > > I reviewed the series and I left some minor comments for patches 1-4. > If it's fine by you I can go ahead and fold those in and apply them to > the main branch. That should make it easier to respin a new version. > > Patch 5, on the other hand, needs some more re-design I think. I'm > afraid the performance cost it introduces is too high. > > Please let me know if you're ok with me applying the first 4 patches. > > Regards, > Dumitru > I'm fine with the suggestions for 1-4, I'll take a look at different way that wouldn't have a huge performance impact. Thanks, Ales > > > The difference between v4 and v5 is that the > > first patch was devided even further into four > > patches, the last two patches from v4 are squashed > > into v5 as this their separation didn't have any > > meaningful value. > > > > Ales Musil (5): > > mac-learn: Make the mac_binding struct more flexible > > ovn-util: Add helper function to parse ip into mapped ipv6 > > pinctrl: Simplify packet buffering structure > > controller: Move the packet buffering into mac-learn > > controller: Prevent race in packet buffering > > > > controller/mac-learn.c | 244 ++++++++++++++++++++---- > > controller/mac-learn.h | 88 +++++++-- > > controller/ovn-controller.c | 200 +++++++++++++++++++- > > controller/pinctrl.c | 356 ++++++++---------------------------- > > controller/pinctrl.h | 6 +- > > lib/ovn-util.c | 12 ++ > > lib/ovn-util.h | 2 + > > northd/northd.c | 32 +--- > > utilities/ovn-trace.c | 5 +- > > 9 files changed, 583 insertions(+), 362 deletions(-) > > > >
On 5/9/23 15:42, Ales Musil wrote: > On Tue, May 9, 2023 at 3:40 PM Dumitru Ceara <dceara@redhat.com> wrote: > >> On 4/28/23 11:42, Ales Musil wrote: >>> The series is split into 5 patches. >>> >>> It aims to simplify the packet buffering >>> process and encapsulate it inside mac-learn >>> module. This is done through refactoring >>> and preparation. The last patch then fixes the >>> race condition in packet buffering by utilizing >>> I-P node. >>> >> >> Hi Ales, >> >> I reviewed the series and I left some minor comments for patches 1-4. >> If it's fine by you I can go ahead and fold those in and apply them to >> the main branch. That should make it easier to respin a new version. >> >> Patch 5, on the other hand, needs some more re-design I think. I'm >> afraid the performance cost it introduces is too high. >> >> Please let me know if you're ok with me applying the first 4 patches. >> >> Regards, >> Dumitru >> > > I'm fine with the suggestions for 1-4, I'll take a look at different way > that wouldn't have a huge performance impact. > Cool, thanks! I pushed the first packets to the main branch. Regards, Dumitru