Message
Leyi Rong
June 4, 2019, 5:41 a.m. UTC
Main changes: 1. Advanced switch rule support. 2. Add more APIs for tunnel management. 3. Add some minor features. 4. Code clean and bug fix. Leyi Rong (49): net/ice/base: add macro for rounding up net/ice/base: update standard extr seq to include DIR flag net/ice/base: add API to configure MIB net/ice/base: add more recipe commands net/ice/base: add funcs to create new switch recipe net/ice/base: programming a new switch recipe net/ice/base: replay advanced rule after reset net/ice/base: code for removing advanced rule net/ice/base: add lock around profile map list net/ice/base: save and post reset replay q bandwidth net/ice/base: rollback AVF RSS configurations net/ice/base: move RSS replay list net/ice/base: cache the data of set PHY cfg AQ in SW net/ice/base: refactor HW table init function net/ice/base: add compatibility check for package version net/ice/base: add API to init FW logging net/ice/base: use macro instead of magic 8 net/ice/base: move and redefine ice debug cq API net/ice/base: separate out control queue lock creation net/ice/base: add helper functions for PHY caching net/ice/base: added sibling head to parse nodes net/ice/base: add and fix debuglogs net/ice/base: add support for reading REPC statistics net/ice/base: move VSI to VSI group net/ice/base: forbid VSI to remove unassociated ucast filter net/ice/base: add some minor features net/ice/base: call out dev/func caps when printing net/ice/base: add some minor features net/ice/base: cleanup update link info net/ice/base: add rd64 support net/ice/base: track HW stat registers past rollover net/ice/base: implement LLDP persistent settings net/ice/base: check new FD filter duplicate location net/ice/base: correct UDP/TCP PTYPE assignments net/ice/base: calculate rate limit burst size correctly net/ice/base: add lock around profile map list net/ice/base: fix Flow Director VSI count net/ice/base: use more efficient structures net/ice/base: slightly code update net/ice/base: code clean up net/ice/base: cleanup ice flex pipe files net/ice/base: change how VMDq capability is wrapped net/ice/base: refactor VSI node sched code net/ice/base: add some minor new defines net/ice/base: add 16-byte Flex Rx Descriptor net/ice/base: add vxlan/generic tunnel management net/ice/base: enable additional switch rules net/ice/base: allow forward to Q groups in switch rule net/ice/base: changes for reducing ice add adv rule time drivers/net/ice/base/ice_adminq_cmd.h | 128 +- drivers/net/ice/base/ice_common.c | 618 ++++-- drivers/net/ice/base/ice_common.h | 32 +- drivers/net/ice/base/ice_controlq.c | 247 ++- drivers/net/ice/base/ice_controlq.h | 4 +- drivers/net/ice/base/ice_dcb.c | 74 +- drivers/net/ice/base/ice_dcb.h | 12 +- drivers/net/ice/base/ice_fdir.c | 11 +- drivers/net/ice/base/ice_fdir.h | 1 - drivers/net/ice/base/ice_flex_pipe.c | 1251 ++++++------ drivers/net/ice/base/ice_flex_pipe.h | 74 +- drivers/net/ice/base/ice_flex_type.h | 54 +- drivers/net/ice/base/ice_flow.c | 447 +++- drivers/net/ice/base/ice_flow.h | 26 +- drivers/net/ice/base/ice_lan_tx_rx.h | 31 +- drivers/net/ice/base/ice_nvm.c | 18 +- drivers/net/ice/base/ice_osdep.h | 23 + drivers/net/ice/base/ice_protocol_type.h | 7 +- drivers/net/ice/base/ice_sched.c | 214 +- drivers/net/ice/base/ice_sched.h | 24 +- drivers/net/ice/base/ice_switch.c | 2382 +++++++++++++++++++++- drivers/net/ice/base/ice_switch.h | 64 +- drivers/net/ice/base/ice_type.h | 95 +- drivers/net/ice/ice_ethdev.c | 4 +- 24 files changed, 4478 insertions(+), 1363 deletions(-)
Comments
Hi Leyi, On 6/4/19 7:41 AM, Leyi Rong wrote: > Main changes: > 1. Advanced switch rule support. > 2. Add more APIs for tunnel management. > 3. Add some minor features. > 4. Code clean and bug fix. In order to ease the review process, I think it would be much better to split this series in multiple ones, by features. Otherwise, it is more difficult to keep track if comments are taken into account in the next revision. Also, it is suggested to put the fixes first in the series to ease the backporting. Thanks, Maxime
> -----Original Message----- > From: Maxime Coquelin [mailto:maxime.coquelin@redhat.com] > Sent: Wednesday, June 5, 2019 12:56 AM > To: Rong, Leyi <leyi.rong@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com> > Cc: dev@dpdk.org > Subject: Re: [dpdk-dev] [PATCH 00/49] shared code update > > Hi Leyi, > > On 6/4/19 7:41 AM, Leyi Rong wrote: > > Main changes: > > 1. Advanced switch rule support. > > 2. Add more APIs for tunnel management. > > 3. Add some minor features. > > 4. Code clean and bug fix. > > In order to ease the review process, I think it would be much better to split this series in multiple ones, by features. > Otherwise, it is more difficult to keep track if comments are taken into account in the next revision. > > Also, it is suggested to put the fixes first in the series to ease the backporting. > > Thanks, > Maxime +Paul, Hello Maxime, Thanks for all your constructive comments, but we do the same process for the CVL shared code update to DPDK upstream on the previous release. This series of patches are extracted/reorganized/squashed from the ND released packages, which the shared code difference between 1905 and 1908 can be more than 200 commits from the original shared code repo. IMHO, there might be some reasons for take all these patches into one patchset. - the patchset try to keeps the history order as the commits in the original shared code repo. - the relatively behind patch in the patchset may have dependency on the front patches. - it's difficult to split this series into multiple ones, since the patches are irregular and squashed. Best Regards, Leyi Rong
Hi Leyi, On 6/6/19 7:44 AM, Rong, Leyi wrote: > >> -----Original Message----- >> From: Maxime Coquelin [mailto:maxime.coquelin@redhat.com] >> Sent: Wednesday, June 5, 2019 12:56 AM >> To: Rong, Leyi <leyi.rong@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com> >> Cc: dev@dpdk.org >> Subject: Re: [dpdk-dev] [PATCH 00/49] shared code update >> >> Hi Leyi, >> >> On 6/4/19 7:41 AM, Leyi Rong wrote: >>> Main changes: >>> 1. Advanced switch rule support. >>> 2. Add more APIs for tunnel management. >>> 3. Add some minor features. >>> 4. Code clean and bug fix. >> >> In order to ease the review process, I think it would be much better to split this series in multiple ones, by features. >> Otherwise, it is more difficult to keep track if comments are taken into account in the next revision. >> >> Also, it is suggested to put the fixes first in the series to ease the backporting. >> >> Thanks, >> Maxime > > +Paul, > > Hello Maxime, > Thanks for all your constructive comments, but we do the same process for the CVL shared code update to DPDK upstream on the previous release. > This series of patches are extracted/reorganized/squashed from the ND released packages, which the shared code difference between 1905 and 1908 can be more than 200 commits from the original shared code repo. > > IMHO, there might be some reasons for take all these patches into one patchset. > - the patchset try to keeps the history order as the commits in the original shared code repo. > - the relatively behind patch in the patchset may have dependency on the front patches. > - it's difficult to split this series into multiple ones, since the patches are irregular and squashed. On the other hand, it means we should just apply the series without even reviewing it as it would not be taken into account. I personally think this is not a sane practice. This is not the case of this series, which is almost good to me except some missing errors handling, but imagine there is a security issue in it, should we just apply it as-is not to diverge from your internal code base and wait for the next shared code release to get the fix? Note that my comment is not intended at Intel drivers specifically, as it seems a common practice to have the base driver not reviewed and applied as-is. Best regards, Maxime > > > Best Regards, > Leyi Rong >