Hi folks, On 25/02/2024 16:39, Jens Axboe wrote: > On 2/24/24 5:58 PM, Pavel Begunkov wrote: >> On 2/24/24 18:51, Jens Axboe wrote: >>> On 2/24/24 8:31 AM, Pavel Begunkov wrote: >>>> On 2/24/24 05:07, David Wei wrote: >>>>> Currently we unconditionally account time spent waiting for events in CQ >>>>> ring as iowait time. >>>>> >>>>> Some userspace tools consider iowait time to be CPU util/load which can >>>>> be misleading as the process is sleeping. High iowait time might be >>>>> indicative of issues for storage IO, but for network IO e.g. socket >>>>> recv() we do not control when the completions happen so its value >>>>> misleads userspace tooling. >>>>> >>>>> This patch gates the previously unconditional iowait accounting behind a >>>>> new IORING_REGISTER opcode. By default time is not accounted as iowait, >>>>> unless this is explicitly enabled for a ring. Thus userspace can decide, >>>>> depending on the type of work it expects to do, whether it wants to >>>>> consider cqring wait time as iowait or not. >>>> >>>> I don't believe it's a sane approach. I think we agree that per >>>> cpu iowait is a silly and misleading metric. I have hard time to >>>> define what it is, and I'm sure most probably people complaining >>>> wouldn't be able to tell as well. Now we're taking that metric >>>> and expose even more knobs to userspace. >>> >>> For sure, it's a stupid metric. But at the same time, educating people >>> on this can be like talking to a brick wall, and it'll be years of doing >>> that before we're making a dent in it. Hence I do think that just >>> exposing the knob and letting the storage side use it, if they want, is >>> the path of least resistance. I'm personally not going to do a crusade >>> on iowait to eliminate it, I don't have the time for that. I'll educate >> >> Exactly my point but with a different conclusion. The path of least > > I think that's because I'm a realist, and you are an idealist ;-) > >> resistance is to have io_uring not accounted to iowait. That's how >> it was so nobody should complain about it, you don't have to care about >> it at all, you don't have to educate people on iowait when it comes up >> with in the context of that knob, and you don't have to educate folks >> on what this knob is and wtf it's there, and we're not pretending that >> it works when it's not. > > I don't think anyone cares about iowait going away for waiting on events > with io_uring, but some would very much care about losing the cpufreq > connection which is why it got added in the first place. If we can > trivially do that without iowait, then we should certainly just do that > and call it good. THAT is the main question to answer, in form of a > patch. I commented on Jens' patch regarding iowait and iowait_acct, which is probably the path of least resistance for that specific issue, but let me expand a bit on the cpufreq iowait connection problem. cpufreq iowait handling and cpuidle iowait handling I would consider vastly different in that respect and it seems to me that improving cpufreq should be feasible effort (if it only catches 90% of scenarios at first then so be it). I'm thinking something of a in_iowait_boost (or the opposite in_iowait_queue_full full meaning reasonably non-empty). The current behaviour of boosting CPU frequency on anything is just very unfortunate and for io_uring in particular destroys all of the power-savings it could have due to reduced CPU usage. (Just to be clear, current iowait boosting is not just broken in io_uring, but rather everywhere, particularly because of the lack of definition what iowait even means and when it should be set. Thus boosting on any iowait seen is like taking a sledgehammer to crack a nut.) I'm happy to propose some io_uring patch (that is probably nonsensical because of a lot of reasons) or test whatever ideas you have. Something like if pending_requests > device_queue_size: Don't boost would be an improvement from what I can tell. And I know I'm heavily reducing io_uring to block IO here and am aware of how wrong that is, but storage device IO was the original story that got iowait boosting introduced in the first place. If you have any cpufreq (or rather DVFS) related issues with anything else io_uring like networking I'll give reproducing that a shot as well. Would love to hear your thoughts! Kind Regards, Christian
