| Message ID | 20220222112326.15070-1-xiangxia.m.yue@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | net: sched: allow user to select txqueue | expand |
On Tue, 22 Feb 2022 19:23:24 +0800 xiangxia.m.yue@gmail.com wrote: > From: Tonghao Zhang <xiangxia.m.yue@gmail.com> > > Patch 1 allow user to select txqueue in clsact hook. > Patch 2 support skbhash, classid, cpuid to select txqueue. Jamal, you had feedback on the previous version, does this one look good?
On 2022-03-02 14:24, Jakub Kicinski wrote: > On Tue, 22 Feb 2022 19:23:24 +0800 xiangxia.m.yue@gmail.com wrote: >> From: Tonghao Zhang <xiangxia.m.yue@gmail.com> >> >> Patch 1 allow user to select txqueue in clsact hook. >> Patch 2 support skbhash, classid, cpuid to select txqueue. > > Jamal, you had feedback on the previous version, > does this one look good? No it doesnt. Maybe i wasnt clear enough to Tonghao because they sent out a patch right after that discussion which didnt address my last comments. Here's my take: Out of the three options in 2/2 that Tonghao showed - I agree with Cong on the last two (cpu and class): that user space can correctly set policy without needing the specified mechanisms being added to the kernel. skbhash otoh is meaningful. Patch 1 is useful irrespective of patch 2. cheers, jamal
On Thu, Mar 3, 2022 at 7:33 PM Jamal Hadi Salim <jhs@mojatatu.com> wrote: > > On 2022-03-02 14:24, Jakub Kicinski wrote: > > On Tue, 22 Feb 2022 19:23:24 +0800 xiangxia.m.yue@gmail.com wrote: > >> From: Tonghao Zhang <xiangxia.m.yue@gmail.com> > >> > >> Patch 1 allow user to select txqueue in clsact hook. > >> Patch 2 support skbhash, classid, cpuid to select txqueue. > > > > Jamal, you had feedback on the previous version, > > does this one look good? > > No it doesnt. > Maybe i wasnt clear enough to Tonghao because they sent out > a patch right after that discussion which didnt address my > last comments. Here's my take: > > Out of the three options in 2/2 that Tonghao showed - I agree with > Cong on the last two (cpu and class): that user space can correctly set > policy without needing the specified mechanisms being added to the > kernel. Sorry, I miss that. I will remove the cpu and class hash in next version. > skbhash otoh is meaningful. > > Patch 1 is useful irrespective of patch 2. > > cheers, > jamal
From: Tonghao Zhang <xiangxia.m.yue@gmail.com> Patch 1 allow user to select txqueue in clsact hook. Patch 2 support skbhash, classid, cpuid to select txqueue. Tonghao Zhang (2): net: sched: use queue_mapping to pick tx queue net: sched: support hash/classid/cpuid selecting tx queue include/linux/netdevice.h | 3 + include/linux/rtnetlink.h | 1 + include/net/tc_act/tc_skbedit.h | 1 + include/uapi/linux/tc_act/tc_skbedit.h | 8 +++ net/core/dev.c | 31 +++++++++- net/sched/act_skbedit.c | 84 ++++++++++++++++++++++++-- 6 files changed, 122 insertions(+), 6 deletions(-)