On Fri, 04 Aug 2023 17:50:02 +0200, Mark Brown wrote: > > On Fri, Aug 04, 2023 at 05:40:20PM +0200, Takashi Iwai wrote: > > Mark Brown wrote: > > > > Hrm, do systems typically filter out debug logs? Most of the systems I > > > work on do still log debug logs, they just only include them in dmesg by > > > default. > > > I thought flipping the dynamic debug flag needed for showing them up > > with dev_dbg() and co. > > That's an optional feature, I guess perhaps distro kernels will tend to > turn it on but not the defconfigs I'm generally using so I've not been > seeing it. Ah yes, most distros set CONFIG_DYNAMIC_DEBUG, but defconfig is a different set. > > > There are ratelimited printk APIs which we could use, though > > > they're not quite ratelimited enough for this I think. > > > Yes, this is printed not too frequently but just repeatedly. > > > IMHO, dev_err_once() would be a good compromise. > > I can imagine that being *super* frustrating when trying to actually > develop configurations since you only get told once per boot that > there's something missing. dev_dbg() did seem like a sensible choice > here, even with dynamic debug you can at least turn it on. Perhaps both > a dev_err_once() and a dev_dbg() so you get a hint that there's > something wrong in the logs and can still get ongoing information when > developing your configuration. +1, this sounds like a solution. thanks, Takashi
