Re: [RFC 00/12] module: avoid userspace pressure on unwanted allocations

[David Hildenbrand <david@redhat.com>]

On 21.03.23 20:32, David Hildenbrand wrote:
> On 20.03.23 22:27, Luis Chamberlain wrote:
>> On Mon, Mar 20, 2023 at 02:23:36PM -0700, Luis Chamberlain wrote:
>>> On Mon, Mar 20, 2023 at 10:15:23PM +0100, David Hildenbrand wrote:
>>>> Not able to reproduce with 20230319-module-alloc-opts so far (2 tries).
>>>
>>> Oh wow, so to clarify, it boots OK?
>>>
>>
>> Now that we know that tree works, I'm curious also now if you can
>> confirm just re-ordering the patches still works (it should)
>>
>> https://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux.git/log/?h=20230319-module-alloc-opts-adjust
>>
> 
> So far no vmap errors booting the debug/kasan kernel (2 tries).
> 
>> And although it's *probably* just noise, but I'm very curious how much,
>> if any difference there is if you just revert "module: use
>> list_add_tail_rcu() when adding module".
> 
> Dito, no vmap errors booting the debug/kasan kernel (2 tries).
> 
>>
>> The data on that commit log is pretty small as I have a low end system,
>> and I'm not yet done beating the hell out of a system with stress-ng,
>> but getting some data froma  pretty large system would be great.
>> Specially if this series seems to prove fixing boot on them.
> 
> 2x booting RHEL9.1 on a !debug kernel. Something went wrong with kdump in 2 runs (think I
> had to delete the kdump initrd to make space for another kernel), but we can just mostly
> ignore that. I wanted to rerun with kdump disabled completely, but I'm out of time for today.
> 
> 
> 1) v6.3-rc1:
> 
> #1
> 
> Startup finished in 25.354s (kernel) + 7.662s (initrd) + 1min 8.805s (userspace) = 1min 41.822s
> multi-user.target reached after 29.186s in userspace
> 
> 47.178s kdump.service
> 14.898s tuned.service
> 11.394s chrony-wait.service
>    7.486s systemd-udev-settle.service
>    7.334s NetworkManager-wait-online.service
>    2.908s initrd-switch-root.service
>    2.451s smartd.service
>    2.316s dracut-initqueue.service
>    2.057s polkit.service
>    1.290s NetworkManager.service
>    1.046s cups.service
>    ...
> 
> #2
> 
> Startup finished in 25.375s (kernel) + 7.497s (initrd) + 29.428s (userspace) = 1min 2.301s
> multi-user.target reached after 29.392s in userspace
> 
> 14.552s tuned.service
>    9.410s chrony-wait.service
>    8.126s systemd-udev-settle.service
>    7.502s NetworkManager-wait-online.service
>    2.871s initrd-switch-root.service
>    2.401s kdump.service
>    2.297s polkit.service
>    2.116s dracut-initqueue.service
>    2.027s smartd.service
>    1.262s NetworkManager.service
>    1.102s cups.service
>    1.011s sshd.service
>    ...
> 
> 
> 2) 20230319-module-alloc-opts-adjust + revert of list_add_tail_rcu():
> 
> #1
> 
> Startup finished in 25.441s (kernel) + 7.285s (initrd) + 1min 7.644s (userspace) = 1min 40.371s
> multi-user.target reached after 27.213s in userspace
> 
> 47.232s kdump.service
> 14.235s tuned.service
>    8.220s chrony-wait.service
>    7.444s NetworkManager-wait-online.service
>    5.986s systemd-udev-settle.service
>    2.881s initrd-switch-root.service
>    2.236s smartd.service
>    1.899s dracut-initqueue.service
>    1.812s polkit.service
>    1.554s NetworkManager.service
>    1.140s ModemManager.service
>    ...
> 
> #2
> 
> Startup finished in 25.377s (kernel) + 7.271s (initrd) + 28.247s (userspace) = 1min 897ms
> multi-user.target reached after 28.210s in userspace
> 
> 15.435s tuned.service
> 11.365s chrony-wait.service
>    7.512s NetworkManager-wait-online.service
>    5.962s systemd-udev-settle.service
>    2.889s initrd-switch-root.service
>    2.846s smartd.service
>    2.819s kdump.service
>    2.228s polkit.service
>    1.872s dracut-initqueue.service
>    1.312s NetworkManager.service
>    1.152s ModemManager.service
>    1.011s sshd.service
>    ...
> 
> 3) 20230319-module-alloc-opts-adjust:
> 
> 
> #1
> 
> Startup finished in 25.320s (kernel) + 7.192s (initrd) + 1min 6.511s (userspace) = 1min 39.024s
> multi-user.target reached after 28.205s in userspace
> 
> 46.307s kdump.service
> 14.199s tuned.service
> 13.358s chrony-wait.service
>    7.468s NetworkManager-wait-online.service
>    6.329s systemd-udev-settle.service
>    2.910s initrd-switch-root.service
>    2.752s smartd.service
>    2.142s polkit.service
>    1.909s dracut-initqueue.service
>    1.210s NetworkManager.service
>    1.041s ModemManager.service
>     925ms sshd.service
>    ...
> 
> #2
> 
> Startup finished in 25.368s (kernel) + 7.303s (initrd) + 1min 6.897s (userspace) = 1min 39.569s
> multi-user.target reached after 27.326s in userspace
> 
> 45.626s kdump.service
> 14.707s tuned.service
> 13.246s chrony-wait.service
>    7.428s NetworkManager-wait-online.service
>    6.507s systemd-udev-settle.service
>    3.038s initrd-switch-root.service
>    3.001s smartd.service
>    2.057s polkit.service
>    1.746s dracut-initqueue.service
>    1.221s NetworkManager.service
>    ...
> 
> 
> I think we primarily only care about systemd-udev-settle.service.
> 
> That is fastest without the rcu patch (~6s), compared to with the rcu
> patch (~6.5s) and with stock (~7.5s -- 8s).
> 
> Looks like dracut-initqueue also might be a bit faster with your changes, but
> maybe it's mostly noise (would have to do more runs).
> 
> So maybe drop that rcu patch? But of course, there could be other scenarios where it's
> helpful ...

Are there any other things you would like me to measure/test? I'll have 
to hand back that test machine soonish.

-- 
Thanks,

David / dhildenb