Re: [PATCH 4.9 00/75] 4.9.74-stable review

[Neal Cardwell]

On Tue, Jan 2, 2018 at 1:32 PM, David Miller <davem@xxxxxxxxxxxxx> wrote:
> From: Neal Cardwell <ncardwell@xxxxxxxxxx>
> Date: Tue, 2 Jan 2018 11:57:59 -0500
>
>> On Mon, Jan 1, 2018 at 9:31 AM, Greg Kroah-Hartman
>> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>>> This is the start of the stable review cycle for the 4.9.74 release.
>>> There are 75 patches in this series, all will be posted as a response
>>> to this one.  If anyone has any issues with these being applied, please
>>> let me know.
>>>
>>> Responses should be made by Wed Jan  3 14:00:03 UTC 2018.
>>> Anything received after that time might be too late.
>>>
>>> The whole patch series can be found in one patch at:
>>>         kernel.org/pub/linux/kernel/v4.x/stable-review/patch-4.9.74-rc1.gz
>>> or in the git tree and branch at:
>>>   git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable-rc.git linux-4.9.y
>>> and the diffstat can be found below.
>>
>> Hi Greg,
>>
>> In looking at the 4.9 and 4.14 patches yesterday, I noticed there were
>> two TCP BBR fixes that made it into 4.14 but not 4.9. Doing an
>> inventory of the TCP BBR fixes, AFAICT we have:
>>
>> c589e69b508d tcp_bbr: record "full bw reached" decision in new
>> full_bw_reached bit
>>  - in 4.9 and 4.14 (great)
>>
>> 2f6c498e4f15 tcp_bbr: reset full pipe detection on loss recovery undo
>>   - in 4.14 (but not 4.9)
>>
>> 600647d467c6 tcp_bbr: reset long-term bandwidth sampling on loss recovery undo
>>   - in 4.14 (but not 4.9)
>>
>> Lacking the second and third patches in 4.9 will not cause any new
>> problems, but it will miss out on some nice fixes. If it's possible to
>> get  2f6c498e4f15 and 600647d467c6 either into 4.9.74 or 4.9.75, I
>> would be very grateful.
>
> These were not straight-forward to backport and I felt the risk outweighed
> the gains.
>
> If you want to do the backport yourself and you feel confident in it,
> feel free.

Thanks, Greg and David. Looks like these 2 patches will cherry-pick
cleanly if cherry-picked in the following sequence, on top of
4.9.74-rc1, which already has 6c9e73ef9aa7 ("tcp_bbr: record "full bw
reached" decision in new full_bw_reached bit"):

$ git checkout linux-stable-rc/linux-4.9.y

$ git cherry-pick 2f6c498e4f15
Performing inexact rename detection: 100% (17803152/17803152), done.
[detached HEAD 0982234c57e1] tcp_bbr: reset full pipe detection on
loss recovery undo
 Date: Thu Dec 7 12:43:31 2017 -0500
 1 file changed, 4 insertions(+)

$ git cherry-pick 600647d467c6
Performing inexact rename detection: 100% (17803152/17803152), done.
[detached HEAD 7e866eccd083] tcp_bbr: reset long-term bandwidth
sampling on loss recovery undo
 Date: Thu Dec 7 12:43:32 2017 -0500
 1 file changed, 1 insertion(+)

$ git log --oneline --decorate | head -3
7e866eccd083 (HEAD) tcp_bbr: reset long-term bandwidth sampling on
loss recovery undo
0982234c57e1 tcp_bbr: reset full pipe detection on loss recovery undo
79070be7f1ae (linux-stable-rc/linux-4.9.y) Linux 4.9.74-rc1

I verified that this compiles without warnings, and boots, and BBR works.

Shall I prepare another version of these 2 patches, or do we think
this recipe will be sufficient? (Sorry I am not more familiar with the
backport-to-stable process.)

Thanks!
neal