Re: [libvirt] [PATCH v5 3/3] security_dac: Favour ACLs over chown()

On 28.03.2013 12:12, Daniel P. Berrange wrote: > On Thu, Mar 28, 2013 at 11:38:04AM +0100, Michal Privoznik wrote: >> On 28.03.2013 10:46, Daniel P. Berrange wrote: >>> On Thu, Mar 21, 2013 at 05:50:49PM +0100, Michal Privoznik wrote: >>>> #define VIR_FROM_THIS VIR_FROM_SECURITY >>>> #define SECURITY_DAC_NAME "dac" >>>> +#define SECURITY_DAC_XATTR_OLD_ACL "trusted.libvirt.dac.oldACL" >>>> +#define SECURITY_DAC_XATTR_OLD_OWNER "trusted.libvirt.dac.oldOwner" >>>> +#define SECURITY_DAC_XATTR_REFCOUNT "trusted.libvirt.dac.refCount" >>> >>> IMHO we don't need the 'trusted.' prefix on these. >>> >>> Daniel >>> >> >> An XATTR has to have a prefix. We can choose from several prefixes. >> attr(5) says: >> >> Currently the security, system, trusted, and user extended attribute >> classes are defined as described below. Additional classes may be >> added in the future. >> >> security - should be used by kernel security modules, such as Security >> Enhanced Linux. As long as libvirt doesn't provide kernel module, we >> should not be polluting this prefix. >> >> system - used by the kernel to store system objects such as Access >> Control Lists and Capabilities. Again, we are not kernel. >> >> trusted - visible and accessible only to processes that have the >> CAP_SYS_ADMIN capability (the super user usually has this capability). >> Attributes in this class are used to implement mechanisms in user >> space (i.e., outside the kernel) which keep information in extended >> attributes to which ordinary processes should not have access. >> >> Note, that the three above can be touched only by root (or >> CAP_SYS_ADMIN'ed process). >> >> user - may be assigned to files and directories for storing arbitrary >> additional information such as the mime type, character set or encoding >> of a file. >> >> The user. can be manipulated by ordinary user. >> >> My rationale for not allowing ordinary user to play with our XATTRs is >> to prevent them chowning to arbitrary user. > > Ok, that makes more sense now. I wonder how portable this list of > prefixes is though - does anyone know if *BSD use the same conventions ? > > > Daniel > Aah. On BSD they support just 'system' and 'user': http://svnweb.freebsd.org/base/head/sys/sys/extattr.h?revision=184413&... Does it mean we should move from 'trusted' to 'system'? Or is conditional prefix ('trusted' on linux, 'system' on BSD) sufficient?