Bug 1028119 - dbus timeout
Summary: dbus timeout
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: Red Hat Enterprise Linux 7
Classification: Red Hat
Component: tuned
Version: 7.0
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: rc
: ---
Assignee: Jaroslav Škarvada
QA Contact: Branislav Blaškovič
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2013-11-07 17:22 UTC by Jaroslav Škarvada
Modified: 2014-06-18 07:34 UTC (History)
5 users (show)

Fixed In Version: tuned-2.3.0-2.el7
Doc Type: Bug Fix
Doc Text:
Cause: There was an hidden race condition in the Tuned code since the beginning of the Tuned v2 development. Consequence: If the switch profile command was sent to the Tuned daemon quickly more times (e.g. twice) and the previous switch profile command wasn't yet finished, the race condition may occur and the deadlock may trigger. In the deadlocked state the Tuned daemon stopped responding to further commands and had to be killed by the operator. Fix: The Tuned code was modified for the deadlock not to occur any more. Result: Multiple quick requests can be processed now and the Tuned daemon still responds.
Clone Of:
Environment:
Last Closed: 2014-06-13 11:00:37 UTC
Target Upstream Version:

Attachments (Terms of Use)
Proposed fix (664 bytes, patch)
2013-11-07 22:34 UTC, Jaroslav Škarvada 		no flags Details | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Description Jaroslav Škarvada 2013-11-07 17:22:37 UTC 
Description of problem:
tuned-adm can timeout and tuned dies.

Version-Release number of selected component (if applicable):
tuned-2.3.0

How reproducible:
Sometimes

Steps to Reproduce:
1. Run the reproducer

Actual results:
dbus timeout

Expected results:
No timeout

Additional info:
Older tuned seems also affected.

Reproducer:

PROFILES=(balanced latency-performance powersave throughput-performance virtual-guest virtual-host)

for (( i=0; i < 100; i++ ))
do
  p=${PROFILES[$(( RANDOM % ${#PROFILES[@]} ))]}
  echo "iter: $i, profile: $p"
  tuned-adm profile $p
done
Comment 1 Jaroslav Škarvada 2013-11-07 17:24:51 UTC 
The source of this problem is currently unclear to me. I am working on it
Comment 2 Jaroslav Škarvada 2013-11-07 21:34:23 UTC 
I got it, there is race condition in the start/stop code. The race is there since the beginning of the tuned-2 development.
Comment 3 Jaroslav Škarvada 2013-11-07 21:36:53 UTC 
Simple reproducer:
tuned-adm profile balanced && tuned-adm profile balanced
Comment 4 Jaroslav Škarvada 2013-11-07 22:34:17 UTC 
Created attachment 821328 [details]
Proposed fix
Comment 7 Branislav Blaškovič 2013-11-19 10:25:14 UTC 
I cannot reproduce this with tuned-2.3.0-3.el7.noarch either.
Marking as VERIFIED.
Comment 8 Ludek Smid 2014-06-13 11:00:37 UTC 
This request was resolved in Red Hat Enterprise Linux 7.0.

Contact your manager or support representative in case you have further questions about the request.
Note You need to log in before you can comment on or make changes to this bug.