[Bug 517488] Review Request: vhostmd - Virtualization host metrics daemon

[bugzilla at redhat.com]

Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=517488





--- Comment #1 from Matthew Booth <mbooth at redhat.com>  2009-10-07 12:11:27 EDT ---
MUST items:
*** rpmlint reports errors shown above [1]
Package is named according to the package naming guidelines
spec file name matches the base package name
Package meets the packaging guidelines
Package is licensed with a Fedora approved license
License field in spec file matches actual license
Spec file is written in American English
Spec file is legible
*** Upstream sources are not available [2]
Package successfully builds on i686 and x86_64
*** Package uses ExclusiveArch [3]
All build dependencies are listed in BuildRequires
Upstream does not contain localisation
ldconfig is called in %post and %postun for vm-dump-metrics
Package does not bundle copies of system libraries
Package is not relocatable
Package owns all directories that it creates
No files are listed more than once in %files
Permissions on files are set properly
Package has a %clean section containing rm -rf $RPM_BUILD_ROOT
Package uses macros consistently
Package contains only code
Package does not include 'large documentation files'
%doc files are not required for correct function
Header files are in -devel package
Static libraries are not built
Package does not include pkgconfig files (but autoconf still uses pkgconfig)
Bare .so file is in -devel pacakge
-devel subpackage requires parent subpackage
Package does not contain .la libtool archives
Package does not contain a gui application
Package does not own files or directories owned by other packages
%install runs rm -rf $RPM_BUILD_ROOT
All filenames are valid UTF-8

SHOULD items:
Source package includes license text
*** Description and summary sections do not contain translations (not blocker)
Package builds in mock (see koji build in original submission)
Package compiles and builds for supported architectures (but see [3])
*** Package caused system to become unstable requiring reboot [4]
Scriptlets are sane
Subpackage does not need to require base package in this case
Package does not contain pkgconfig files
Package does not have file dependencies

Misc:
Did you consider any of the other approaches to removing the use of rpath in
the spec file?

Notes:
[1] If the 2 DTDs are genuinely not configuration files, they should be
removed. I note that they're already in %doc, which seems appropriate.

[2] It's not clear how the tarball was obtained or can be reproduced exactly
from source. I'm not able to verify this.

[3] I'm taking ExclusiveArch here to be semantically equivalent to
ExcludeArch. I note that xen (which is a dependency), also has an
ExclusiveArch. libvirt has a solution to this which involves not building
against xen where it is not available. Is this possible in this package?

[4] Starting vhostmd on my F-11 host flooded /var/log/messages with:
Oct  7 14:51:06 mbooth libvirtd: 14:51:06.616: error :
qemudDomainLookupByID:3028 : Domain not found: no domain with matching id 0
Oct  7 14:51:06 mbooth libvirtd: 14:51:06.619: error :
qemudDomainLookupByName:3080 : Domain not found: no domain with matching name
'0'
Oct  7 14:53:51 mbooth vhostmd: Error retrieving metric UsedMem
Oct  7 14:53:53 mbooth vhostmd: Failed to collect vm metrics during update

and the following 2 AVC messages:
Oct  7 14:42:33 mbooth kernel: type=1400 audit(1254922953.004:17101): avc:
denied  { read write } for  pid=32590 comm="virsh"
path="/dev/shm/vhostmd/disk0"
dev=tmpfs ino=2946820 scontext=unconfined_u:system_r:xm_t:s0
tcontext=unconfined_u:object_r:tmpfs_t:s0 tclass=file
Oct  7 14:42:33 mbooth kernel: type=1400 audit(1254922953.012:17102): avc:
denied  { getsched } for  pid=32590 comm="virsh"
scontext=unconfined_u:system_r:xm_t:s0 tcontext=unconfined_u:system_r:xm_t:s0
tclass=process

The system became sluggish after only a few seconds, and unresponsive
requiring a reboot shortly after that. This may be a problem with SELinux
policy or the default configuration file. Either way, I'd be uncomfortable
shipping the package with this as the default behaviour. Is this still a
problem on rawhide? I didn't test there.

-----

Rich,

Can you respond to Misc and 4 Notes above first?

Matt

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.