Red Hat Bugzilla – Bug 415091
hald-addon-storage polling every 2 seconds should not cause 20 wakeups / second from libata
Last modified: 2013-03-05 22:53:58 EST
Description of problem:
Powertop reveals 20 interrupts from libata per second when there is no hard disk
activity. It may be relevant that there are two (ide) optical drives on this
system. I understand that hald-addon-storage polls the cd/dvd drives every
couple of seconds, but this still seems like too many.
Some data points:
* Killing hald-addon-storage from powertop causes libata wakeups to drop to
near-zero (0.2/second over 15 seconds).
* Inserting an audio cd in each drive and then quitting kscd so that both drives
are locked results in only around 2 interrupts/second from libata.
* As an aside, this is better than kernel 126.96.36.199-49 which had typically 40
wakeups/second and around 3 wakeups / second even after killing hald-addon-storage.
Based on the attached strace, below, I suspect that hald-addon-storage is
polling the two optical drives in alternate seconds. Something like bug #204969
comment #1 may help synchronize the times that the two optical drives are
polled. This would still leave the problem of hald-addon-storage causing 20
wakeups every time it polls is - whether this is due to some inefficiency or
can't be helped, I don't know.
Version-Release number of selected component (if applicable):
Created attachment 280311 [details]
strace -tt of both hald-addon-storage processes with empty drives
The strace output I've attached seems to actually have both drives being polled
in the same second, unlike an earlier test I did not record, and contrary to my
claim above. Still, there may be a way to better synchronize the polling or
make it more efficient in terms of interrupts.
Created attachment 280331 [details]
strace -tt of both hald-addon-storage processes with locked drives
*** This bug has been marked as a duplicate of 204969 ***