Subject: Re: [RFC PATCH] access(2) vs. SELinux From: Eric Paris To: Stephen Smalley Cc: Chad Sellers , Joshua Brindle , selinux@tycho.nsa.gov, James Morris , Eric Paris , Daniel J Walsh In-Reply-To: <1240256506.1678.11.camel@localhost.localdomain> References: <1240251074.2562.80.camel@localhost.localdomain> <1240256506.1678.11.camel@localhost.localdomain> Content-Type: text/plain Date: Mon, 20 Apr 2009 18:24:25 -0400 Message-Id: <1240266265.2562.411.camel@localhost.localdomain> Mime-Version: 1.0 Sender: owner-selinux@tycho.nsa.gov List-Id: selinux@tycho.nsa.gov On Mon, 2009-04-20 at 15:41 -0400, Stephen Smalley wrote: > On Mon, 2009-04-20 at 14:11 -0400, Eric Paris wrote: > > On Mon, 2009-04-20 at 13:35 -0400, Chad Sellers wrote: > > > On 4/20/09 1:21 PM, "Eric Paris" wrote: > > > > > > > > > > > I'm going to spend a couple minutes and try to prototype an all in > > > > kernel solution such that on access() we will check the permission: > > > > > > > > allow process_t file_t:file read > > > > > > > > if there is a dontaudit rule for EITHER "read" or "access_read" we will > > > > NOT print an avc denial. > > > > > > > > on the actual usage we will ONLY check for the "read" dontaudit. > > > > > > > That sounds like a better solution than anything we've talked about so far. > > > It would mean the policy author would only have to worry about the access_* > > > perms if they cared about the issue they're designed to solve. Hopefully you > > > can find an acceptable way to pull it off in the kernel. I'm rooting for > > > you! > > > > So James, Stephen what do we think of something like this? Lot of code > > bloat and pain on the access+denial path, but shouldn't hurt us any in > > the hot case... > > > > --- > > diff --git a/fs/open.c b/fs/open.c > > index 377eb25..485cfd8 100644 > > --- a/fs/open.c > > +++ b/fs/open.c > > @@ -493,7 +493,7 @@ SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) > > goto out_path_release; > > } > > > > - res = inode_permission(inode, mode | MAY_ACCESS); > > + res = inode_access_permission(inode, mode | MAY_ACCESS); > > I had forgotten that there is already a MAY_ACCESS flag defined and used > by the vfs. So why can't we just pass that on to the security modules > and use that to distinguish access(2) from open(2)? Looks like it gets > used on chdir(2) and chroot(2) as well, which is unfortunate, and I > don't quite understand why. Maybe we need to look back at the origins > of the MAY_ACCESS flag and its rationale. Actually since both of those should be S_ISDIR we should still be able to tell in the case we care about. Looks like there was a MAY_CHDIR flag, but viro realized having 2 flags that didn't overlap was useless... a110343f0d6d41f68b7cf8c00b57a3172c67f816 -Eric -- This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to majordomo@tycho.nsa.gov with the words "unsubscribe selinux" without quotes as the message.