Bug 1321875

Summary:	blktrace attempts to start threads on offline CPUs
Product:	Red Hat Enterprise Linux 6	Reporter:	Milos Malik <mmalik>
Component:	blktrace	Assignee:	Eric Sandeen <esandeen>
Status:	CLOSED WONTFIX	QA Contact:	BaseOS QE Security Team <qe-baseos-security>
Severity:	medium	Docs Contact:
Priority:	medium
Version:	6.8	CC:	hannsj_uhl
Target Milestone:	rc
Target Release:	---
Hardware:	ppc64
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2016-05-19 17:40:04 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Milos Malik 2016-03-29 10:00:09 UTC

Description of problem:
 * either blktrace guesses the number of CPUs incorrectly or it is not able to run on more 32 CPUs

Version-Release number of selected component (if applicable):
RHEL-6.7
blktrace-1.0.1-7.el6.ppc64
kernel-2.6.32-573.el6.ppc64
kernel-bootwrapper-2.6.32-573.el6.ppc64
kernel-devel-2.6.32-573.el6.ppc64
kernel-firmware-2.6.32-573.el6.noarch
kernel-headers-2.6.32-573.el6.ppc64

How reproducible:
always

Steps to Reproduce:
# dmesg | grep -i cpus
Partition configured for 64 cpus.
Initializing cgroup subsys cpuset
Brought up 32 CPUs
Node 0 CPUs: 0-31
# mount -t debugfs debugfs /sys/kernel/debug/
#blktrace -w 1 -d /dev/sda 

Actual results:
FAILED to start thread on CPU 32: 22/Invalid argument
FAILED to start thread on CPU 33: 22/Invalid argument
FAILED to start thread on CPU 34: 22/Invalid argument
FAILED to start thread on CPU 35: 22/Invalid argument
FAILED to start thread on CPU 36: 22/Invalid argument
FAILED to start thread on CPU 37: 22/Invalid argument
FAILED to start thread on CPU 38: 22/Invalid argument
FAILED to start thread on CPU 39: 22/Invalid argument
FAILED to start thread on CPU 40: 22/Invalid argument
FAILED to start thread on CPU 41: 22/Invalid argument
FAILED to start thread on CPU 42: 22/Invalid argument
FAILED to start thread on CPU 43: 22/Invalid argument
FAILED to start thread on CPU 44: 22/Invalid argument
FAILED to start thread on CPU 45: 22/Invalid argument
FAILED to start thread on CPU 46: 22/Invalid argument
FAILED to start thread on CPU 47: 22/Invalid argument
FAILED to start thread on CPU 48: 22/Invalid argument
FAILED to start thread on CPU 49: 22/Invalid argument
FAILED to start thread on CPU 50: 22/Invalid argument
FAILED to start thread on CPU 51: 22/Invalid argument
FAILED to start thread on CPU 52: 22/Invalid argument
FAILED to start thread on CPU 53: 22/Invalid argument
FAILED to start thread on CPU 54: 22/Invalid argument
FAILED to start thread on CPU 55: 22/Invalid argument
FAILED to start thread on CPU 56: 22/Invalid argument
FAILED to start thread on CPU 57: 22/Invalid argument
FAILED to start thread on CPU 58: 22/Invalid argument
FAILED to start thread on CPU 59: 22/Invalid argument
FAILED to start thread on CPU 60: 22/Invalid argument
FAILED to start thread on CPU 61: 22/Invalid argument
FAILED to start thread on CPU 62: 22/Invalid argument
FAILED to start thread on CPU 63: 22/Invalid argument

Expected results:
# cat /sys/devices/system/cpu/online
0-31
# cat /sys/devices/system/cpu/offline 
32-1023

* blktrace runs on all CPUs which are online and it doesn't complain about the others

Comment 3 Eric Sandeen 2016-03-29 16:06:39 UTC

So the root problem is that it is trying to start threads on offline CPUs?

Comment 4 Eric Sandeen 2016-03-29 16:23:09 UTC

commit 80c4041b2e7a7d5afb75df563bf51bb27773c095
Author: Abutalib Aghayev <agayev>
Date:   Tue Feb 9 08:17:50 2016 -0700

    blktrace: Use number of online CPUs
    
    Currently, blktrace uses _SC_NPROCESSORS_CONF to find out the number of
    CPUs.  This is a problem, because if you reduce the number of online
    CPUs by passing kernel parameter maxcpus, then blktrace fails to start
    with the error:
    
    FAILED to start thread on CPU 4: 22/Invalid argument
    FAILED to start thread on CPU 5: 22/Invalid argument
    ...
    
    The attached patch fixes it to use _SC_NPROCESSORS_ONLN.
    
    Signed-off-by: Jens Axboe <axboe>

Comment 5 Karel Srot 2016-04-01 06:28:00 UTC

Postponing to rhel-6.9 as it is too late to address this bug in rhel-6.8.

Comment 6 Eric Sandeen 2016-05-19 17:40:04 UTC

Sadly I don't think we'll be able to ship this fix at this point in rhel6.