Control: severity -1 normal Control: tag -1 upstream On Sat, 2018-03-24 at 04:51 +0100, Tomasz Warniełło wrote: > Package: src:linux > Version: 4.9.82-1+deb9u3 > Severity: important > File: /lib/modules/4.9.0-6-amd64/kernel/drivers/mmc/host/rtsx_usb_sdmmc.ko > > Dear Maintainer, > > I observed high load in idle system in `top`. The load was ca. 2.00, the CPU > i5. > I checked i/o and net and didn't find load. There were always 3 processes in > the D state: > two kworkers and rtsx_usb_ms_1. I did `rmmod rtsx_usb_ms`, which didn't help, > but then I did `rmmod rtsx_usb_sdmmc` and the load went down to 0.00. > > The situation is also described here: > https://unix.stackexchange.com/questions/433052/high-load-on-i5-with-no-visible-reason/433192#433192 [...]
D state means uninterruptible sleep (i.e. they are waiting for something and can't be interrupted by signals). For historical reasons Linux counts tasks in D state the same as tasks in R state when calculating CPU load, but it doesn't mean they're using any CPU time. These drivers should probably be using schedule_timeout_idle() to avoid the confusing behaviour. (I already made that change in one place in rtsx_usb_ms.) Ben. -- Ben Hutchings When in doubt, use brute force. - Ken Thompson
signature.asc
Description: This is a digitally signed message part
