 |
 |
Hi Andy On Tuesday 02 January 2007 11:22 pm, Andrew Hanushevsky wrote: > Hi Jan, > > In the xroot world, the server assumes that it owns all the files. This > is normally a correct assumption. yes, but we are testing stuff :-). > The problem here is that file somehow > has managed to get group read only and the server is not a memebre of the > group. No, the server is a member of the group. It is run by the user who owns the file and who is also member of the group that has read access to the file. At least under linux, access to the file is denied if I as the owner have no read access but my group has read access. > You are correct that StatGen > should be return the true permission relative to the server. However, the > more pointed question is how the file got into the state that the server > couldn't read it. In a perfect (xrootd-only) world this should not happen, which IMHO means it will happen in the real world, even if we were just setting up a provocative test case to check that our xrootd client software returns the correct permissions. Cheers, Andreas > > Andy > > P.S. As far as OS differences, there are some relative to group > permissions, depending on whether you adopt the BSD model or the SVR4 > model. In practice, that should not matter if the files are owned by > the server. > > On Tue, 2 Jan 2007, Jan Erik Sundermann wrote: > > Hi, > > > > we are having trouble with the permissions returned by the server for the > > stat request. The circumstances are the following: > > > > * the server is run by user aUser on a linux 2.6 box > > * the client is run by aUser on the same host as the server > > * the file in question is readable by the group users which aUser belongs > > to (on the actual UNIX file system) > > * the file is not readable by the user aUser who is the owner of the > > file: > > > > ----r----- 1 aUser users 5932809 2006-04-30 18:01 Error.root > > > > The server returns that the file is readable, but the actual access to > > the file is denied by the OS. This is the result of the treatment of the > > different access privileges for users/group/others in > > XrdXrootdProtocol::StatGen. > > > > Should this be fixed or are different OSs treating the different > > privileges with different priority, so that a coherent treatment is > > difficult? > > > > Cheers, > > > > Jan Erik, > > Andreas