On Mon, Jul 12, 2021 at 11:41:06AM -0400, J. Bruce Fields wrote: > On Mon, Jul 12, 2021 at 10:02:47AM -0400, Vivek Goyal wrote: > > On Fri, Jul 09, 2021 at 04:10:16PM -0400, Bruce Fields wrote: > > > On Fri, Jul 9, 2021 at 1:59 PM Vivek Goyal <vgoyal@xxxxxxxxxx> wrote: > > > > nfs seems to have some issues. > > > > > > I'm not sure what the expected behavior is for nfs. All I have for > > > now is some generic troubleshooting ideas, sorry: > > > > > > > - I can set user.foo xattr on symlink and query it back using xattr name. > > > > > > > > getfattr -h -n user.foo foo-link.txt > > > > > > > > But when I try to dump all xattrs on this file, user.foo is being > > > > filtered out it looks like. Not sure why. > > > > > > Logging into the server and seeing what's set there could help confirm > > > whether it's the client or server that's at fault. (Or watching the > > > traffic in wireshark; there are GET/SET/LISTXATTR ops that should be > > > easy to spot.) > > > > > > > - I can't set "user.foo" xattr on a device node on nfs and I get > > > > "Permission denied". I am assuming nfs server is returning this. > > > > > > Wireshark should tell you whether it's the server or client doing that. > > > > > > The RFC is https://datatracker.ietf.org/doc/html/rfc8276, and I don't > > > see any explicit statement about what the server should do in the case > > > of symlinks or device nodes, but I do see "Any regular file or > > > directory may have a set of extended attributes", so that was clearly > > > the assumption. Also, NFS4ERR_WRONG_TYPE is listed as a possible > > > error return for the xattr ops. But on a quick skim I don't see any > > > explicit checks in the nfsd code, so I *think* it's just relying on > > > the vfs for any file type checks. > > > > Hi Bruce, > > > > Thanks for the response. I am just trying to do set a user.foo xattr on > > a device node on nfs. > > > > setfattr -n "user.foo" -v "bar" /mnt/nfs/test-dev > > > > and I get -EACCESS. > > > > I put some printk() statements and EACCESS is being returned from here. > > > > nfs4_xattr_set_nfs4_user() { > > if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) { > > if (!(cache.mask & NFS_ACCESS_XAWRITE)) { > > return -EACCES; > > } > > } > > } > > > > Value of cache.mask=0xd at the time of error. > > Looks like 0xd is what the server returns to access on a device node > with mode bits rw- for the caller. > > Commit c11d7fd1b317 "nfsd: take xattr bits into account for permission > checks" added the ACCESS_X* bits for regular files and directories but > not others. > > But you don't want to determine permission from the mode bits anyway, > you want it to depend on the owner, Thinking more about this part. Current implementation of my patch is effectively doing both the checks. It checks that you are owner or have CAP_FOWNER in xattr_permission() and then goes on to call inode_permission(). And that means file mode bits will also play a role. If caller does not have write permission on the file, it will be denied setxattr(). If I don't call inode_permission(), and just return 0 right away for file owner (for symlinks and special files), then just being owner is enough to write user.* xattr. And then even security modules will not get a chance to block that operation. IOW, if you are owner of a symlink or special file, you can write as many user.* xattr as you like and except quota does not look like anything else can block it. I am wondering if this approach is ok? > so I guess we should be calling > xattr_permission somewhere if we want that behavior. > > The RFC assumes user xattrs are for regular files and directories, > without, as far as I can tell, actually explicitly forbidding them on > other objects. We should also raise this with the working group if we > want to increase the chances that you'll get the behavior you want on > non-Linux servers. Ok. I am hoping once this patch merges in some form, then I can follow it up with relevant working group. > > The "User extended attributes" section of the xattr(7) man page will > need updating. Agreed. I will take care of that in a separate patch. Right now, I am not too sure if being owner should be the only check and I should skip calling inode_permission() entirely or not. Thanks Vivek > > --b. >
