Re: [PATCH] treewide: const qualify ctl_tables where applicable

[Daniel Xu]

Hi Joel,

On 1/10/25 05:32, Joel Granados wrote:
> On Thu, Jan 09, 2025 at 01:38:33PM +0000, bot+bpf-ci@xxxxxxxxxx wrote:
>> Dear patch submitter,
>>
>> CI has tested the following submission:
>> Status:     FAILURE
>> Name:       treewide: const qualify ctl_tables where applicable
>> Patchwork:  https://patchwork.kernel.org/project/netdevbpf/list/?series=923743&state=*
>> Matrix:     https://github.com/kernel-patches/bpf/actions/runs/12690795270
>>
>> Failed jobs:
>> build-x86_64-gcc: https://github.com/kernel-patches/bpf/actions/runs/12690795270/job/35372434718
>> build-x86_64-llvm-17: https://github.com/kernel-patches/bpf/actions/runs/12690795270/job/35372434997
>> build-x86_64-llvm-17-O2: https://github.com/kernel-patches/bpf/actions/runs/12690795270/job/35372435294
>> build-x86_64-llvm-18: https://github.com/kernel-patches/bpf/actions/runs/12690795270/job/35372435638
>> build-x86_64-llvm-18-O2: https://github.com/kernel-patches/bpf/actions/runs/12690795270/job/35372435949
> We can't make watchdog_hardlockup_sysctl const here because it is
> changing the ctl_talbe.mode to 0644 if watchdog_hardlockup_available is
> true. I'll remove this sysctl array from my patchset to move forward
> with the general constification, but I still don't fully understand the
> need for the modification of the permissions.
>
> My main question is: Cant we just leave the permissions as they where
> originally (before the this commit [1])? The problem touched by [1] is
> when the user writes to nmi_watchdog and watchdog_hardlockup_available
> is false, they will receive a -ENOTSUPP error from proc_nmi_watchdog.
> But wont they get an error anyway if they try to write to a read-only
> file? Does this fix target some specific user-space application?
>
> I have added the original to:/from: contacts from [1]. Please correct me
> if I have miss-read the situation.
>
> Best
>
> [1] https://lore.kernel.org/lkml/20230526184139.1.I0d75971cc52a7283f495aac0bd5c3041aadc734e@changeid/

That seems like a reasonable question. Note BPF CI is off-list by 
default, so I've CC'd linux-kernel.

Thanks,

Daniel