On Mon, 4 Sept 2023 at 22:54, Roberto A. Foglietta <[email protected]> wrote: > > On Mon, 4 Sept 2023 at 21:52, Roberto A. Foglietta > <[email protected]> wrote: > > > > On Mon, 4 Sept 2023 at 18:53, Bastian Bittorf <[email protected]> wrote: > > > > > > can you please give us a hint, > > > what makes your system special? > > > maybe several 1000 processes? > > > > Hi Bastian, > > > > on that system there are 8 Arm 64bit cores 4x1.8GHz + 4x2.0Ghz put > > under conservative frequency governors usually sleeping but sometimes > > running at their lowest frequency 300MHz and I have no clue what's > > going to make my system "special". Someone says that it is the man > > behind the screen but they are rare while almost others ask for > > statistics. Do you wish to receive statistics? ;-) > > s/behind/in front of/ but usually behind because I use the network > rather than a physical keyboard to be connected to the device. >
It happens only when a script is executed by telnetd (a busybox app). However, an ash login shell spawned by that script parented by telnetd does not show the issue. The busybox is configured to not fork apps, compiled with -02 and ash for speed. The telnetd is set to run on CPU0 only by taskset. The busybox configuration is here: https://github.com/robang74/sailfish-os-busybox/blob/e03f839f9ce267279e18c6b7b9bc313e74e84106/rpm/busybox-static.config The issue about sleep 1 seems fixed by now. Some default values that I found in the kernel scheduler configuration were completely out of range, IMHO. Now, the maximum drift on sleep 1 that I have seen in stats is below 2%. In others nights stats less than 5%. I do not trust this stats very much because I have not stressed-out the system but the original system was not even able to set CPUs to sleep without the need to awake them immediately after and so on, so on, so on... Best regards, R- _______________________________________________ busybox mailing list [email protected] http://lists.busybox.net/mailman/listinfo/busybox
