Re: nfsd: add the ability to enable use of RWF_DONTCACHE for all nfsd IO

[Chuck Lever <chuck.lever@oracle.com>]

On 2/21/25 10:46 AM, Jeff Layton wrote:
> On Fri, 2025-02-21 at 10:39 -0500, Chuck Lever wrote:
>> On 2/21/25 10:02 AM, Mike Snitzer wrote:
>>> On Thu, Feb 20, 2025 at 01:17:42PM -0500, Chuck Lever wrote:
>>>> * It might be argued that putting these experimental tunables under /sys
>>>>   eliminates the support longevity question, since there aren't strict
>>>>   rules about removing files under /sys.
>>>
>>> Right, I do think a sysfs knob (that defaults to disabled, requires
>>> user opt-in) is a pretty useful and benign means to expose
>>> experimental functionality.
>>
>> Seems like we want to figure out a blessed way to add this kind of
>> experimental "hidden" tunable in a way that can be easily removed
>> once we have the answers we need.
>>
>> I'd really like to keep the documented administrative interface as
>> straightforward as possible, but I agree that having a way to
>> experiment is valuable.
> 
> We do have this fancy new netlink interface that is extensible.
> 
> You could extend the "threads" call to add a server-wide boolean for
> this and then extend nfsdctl to set that value in some cases.

If we don't have other options that are more familiar outside of
the NFSD world (someone mentioned debugfs), then adding a netlink
operation that is explicitly documented as "access to experimental
temporary features that no-one can rely on existing" seems like a good
alternative. We can make our own rules there.


-- 
Chuck Lever