Bug 205647

Created attachment 286039 [details]
dmesg, sudo lspci -xxxx, .config, result of bisect

Marking as high severity as it directly impacts the user - the user will be wondering why they cannot connect to a wireless network.

Using Arch Linux
Using Linux next-20191122
Using Dell XPS 9370 (i7-8650U, Intel Wireless 8265 [8086:24fd])

Will attach in a zip file: dmesg, sudo lspci -xxxx, .config, result of bisect

At the end of the dmesg, you can see repeated failed attempts to connect - I could not get a single successful connection. On another kernel version (or same kernel version with the culprit patch reversed), on the same computer, the wireless comes straight up without any user intervention.

I have done a bisect. The result was:
"
6570bc79c0dfff0f228b7afd2de720fb4e84d61d is the first bad commit
commit 6570bc79c0dfff0f228b7afd2de720fb4e84d61d
Author: Alexander Lobakin <alobakin@dlink.ru>
Date:   Mon Oct 14 11:00:33 2019 +0300

    net: core: use listified Rx for GRO_NORMAL in napi_gro_receive()

....

:040000 040000 56eacee948f7d3f2e04b09d6af502116c2cdd492 96653f8053eb477d1c37afa3ffdea678ee2fa7bd M      net
"

I reverse applied the diff of this patch to a clean Linux next-20191122 and the regression disappeared.

I use Intel Wireless 8265 [8086:24fd] (iwlwifi.ko), but there were other network driver modules which were re-generated upon reversing the patch, also.

Can somebody else please reproduce / confirm this regression? I will notify the relevant people by replying to [0] in the mean time, but they probably should not revert it on the basis of testimony by a single person.

[0]
https://lore.kernel.org/lkml/7e68da00d7c129a8ce290229743beb3d@dlink.ru/