On 2017-10-16 00:35, Hauke Mehrtens wrote: > Hi, > > I added support for kernel 4.9 to the ar71xx target some days ago and it > was tested by multiple people. > > The only regression in kernel 4.9 compared to 4.4, I am aware of was > this performance regression reported by Sven Schönhoff: > http://lists.infradead.org/pipermail/lede-dev/2017-October/009251.html > I haven't found the time to investigate this problem, it would be nice > if some one could look into this. > > If there are no other complains I would like to switch the default > kernel of the ar71xx target to 4.9 next weekend, are there any > objections to this plan? Hi Hauke,
I did some testing and tracked down the source of this regression. There was a recent commit that added GRO support to the ag71xx ethernet driver. This causes a regression on both 4.4 and 4.9, however it is more severe on 4.9. I've reverted this change, and now 4.9 actually performs slightly better than 4.4. Let's update ar71xx to 4.9 soon :) - Felix _______________________________________________ Lede-dev mailing list Lede-dev@lists.infradead.org http://lists.infradead.org/mailman/listinfo/lede-dev
