> -----Original Message----- > From: linux-nfs-owner@xxxxxxxxxxxxxxx [mailto:linux-nfs- > owner@xxxxxxxxxxxxxxx] On Behalf Of Andrey Borzenkov > Sent: Wednesday, November 07, 2012 2:28 PM > To: J. Bruce Fields > Cc: linux-nfs@xxxxxxxxxxxxxxx > Subject: Re: Effective process GID is ignored when client creates file on NFS > > В Wed, 7 Nov 2012 14:13:36 -0500 > "J. Bruce Fields" <bfields@xxxxxxxxxxxx> пишет: > > > On Mon, Oct 29, 2012 at 06:09:29PM +0400, Andrey Borzenkov wrote: > > > I have met application that is badly broken when installed on NFS. > > > The reason is - it expects files to belong to specific group. It > > > switches to this group on startup (explicit setgid) and creates > > > files. But files come out as belonging to GID 0. > > > > > > I finally reduced it to this trivial script: > > > > > > === cut here === > > > #include <sys/types.h> > > > #include <sys/stat.h> > > > #include <fcntl.h> > > > #include <unistd.h> > > > > > > main() > > > { > > > int fd; > > > > > > setgid(107); > > > fd = open("bar", O_CREAT, 0666); > > > close(fd); > > > } > > > === cut here === > > > > > > On local storage file comes with GID 107; on NFS file comes with GID > > > 0. > > > > > > Linux is SLES10 SP3 with relatively old kernel: > > > 2.6.16.60-0.89.1-smp, server(s) are NetApp with different Data ONTAP > > > versions (7.x and 8.1.1 as the last). > > > > > > Client passes correct credentials (UID:0, GID:107), but does not > > > > Those are the credentials in the rpc header on the CREATE call? > > > > Yes. > > > > explicitly request file ownership in CREATE call (uid set_it - 0, > > > gid set_it - 0). > > > > The client shouldn't have to set the owner or group itself. > > > > So this is server behavior. > > > > Have you checked that the directory you're creating in doesn't have > > the sgid bit? > > Yes. It does not. > > > Or perhaps there's some other server configuration that > > causes this. > > It appears that server ignores passed group if UID == 0. It correctly creates > files if UID != 0. I got bug number from support but it is non-public and no > information is visible so far. I asked about possible workaround (or > undocumented options) but have not got any reply as yet. So is it possible that you forgot to turn off root squashing on the server before testing? Cheers Trond -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
