Re: [Kimchi-devel] [PATCH] Set virt_use_nfs when NFS pool is added.

On Thu, 2014-04-03 at 10:22 +0800, Royce Lv wrote: > On 2014年04月03日 03:40, Aline Manera wrote: >> On 04/01/2014 03:24 AM, Royce Lv wrote: >>> On 2014年03月29日 05:20, Christy Perez wrote: >>>> selinux has a special boolean to make it easier for disk images >>>> to be stored on a remote NFS server. Set this to true when a user >>>> adds an NFS storage pool. >>>> >>>> Most virtualzation documentation recommends that this be set >>>> to true. For example: >>>> http://www.ovirt.org/Troubleshooting_NFS_Storage_Issues >>>> http://fedoraproject.org/wiki/How_to_debug_Virtualization_problems >>>> >>>> This will leave it set to true, even if >>>> the user removes NFS storage pools. It is not a security risk, and >>>> we should not set it to False in case it had already been set by the >>>> user for another non-kimchi use. >>>> >>>> Signed-off-by: Christy Perez <christy(a)linux.vnet.ibm.com&gt; >>>> --- >>>> src/kimchi/i18n.py | 2 ++ >>>> src/kimchi/model/storagepools.py | 5 +++++ >>>> 2 files changed, 7 insertions(+) >>>> >>>> diff --git a/src/kimchi/i18n.py b/src/kimchi/i18n.py >>>> index d45f607..8ade7d7 100644 >>>> --- a/src/kimchi/i18n.py >>>> +++ b/src/kimchi/i18n.py >>>> @@ -144,6 +144,8 @@ messages = { >>>> "KCHPOOL0034E": _("Unable to deactivate pool %(name)s as it is >>>> associated with some templates"), >>>> "KCHPOOL0035E": _("Unable to delete pool %(name)s as it is >>>> associated with some templates"), >>>> "KCHPOOL0036E": _("A volume group named '%(name)s' already >>>> exists. Please, choose another name to create the logical pool."), >>>> + "KCHPOOL0037E": _("Unable to set selinux bool virt_use_nfs for >>>> NFS pool usage. Depending on \ >>>> + your NFS config, this may prevent the pool >>>> from being used."), >>>> >>>> "KCHVOL0001E": _("Storage volume %(name)s already exists"), >>>> "KCHVOL0002E": _("Storage volume %(name)s does not exist in >>>> storage pool %(pool)s"), >>>> diff --git a/src/kimchi/model/storagepools.py >>>> b/src/kimchi/model/storagepools.py >>>> index 92b2496..d279ffa 100644 >>>> --- a/src/kimchi/model/storagepools.py >>>> +++ b/src/kimchi/model/storagepools.py >>>> @@ -126,6 +126,11 @@ class StoragePoolsModel(object): >>>> kimchi_log.error("Problem creating Storage Pool: %s", e) >>>> raise OperationFailed("KCHPOOL0007E", >>>> {'name': name, 'err': >>>> e.get_error_message()}) >>>> + if params['type'] == 'netfs': >>>> + output, error, returncode = run_command(['setsebool', >>>> '-P', >>>> + 'virt_use_nfs=1']) >>> 1. what about turn this on when start kimchi? Cause we just need to >>> enable this for the first time. >> I don't think it is good. >> If we modify it only on Kimchi startup and user or other application >> revert our changes - the user will not be able to create the NFS pool > Well, then what if other users change the firewall configuration? We > still cannot access kimchi through browser. And I can't find a reason > why user will intensionally toggle this off. > I can accept this if you think this is safer, still we need to handle > versions despite fedora/rhel. >>> 2. For Debian using apparmor, it does not have setsebool, I think >>> this need to be handled too. If a debian distro is using AppArmor, it won't be an issue like it is with SELinux. AppArmor is FS agnostic (see http://www.ibm.com/developerworks/security/library/l-selinux/index.html). So, I think that we can safely catch an exception here if there's no SELinux. If AppArmor is enabled instead of SELinux we don't need to do anything NFS-specific for libvirt. Does that address your concern?

>>>> + if error or returncode: >>>> + kimchi_log.error('KCHPOOL0037E') >>>> return name >>>> >>>> def _clean_scan(self, pool_name): >>> _______________________________________________ >>> Kimchi-devel mailing list >>> Kimchi-devel(a)ovirt.org >>> http://lists.ovirt.org/mailman/listinfo/kimchi-devel