From: Mark Salyzyn Subject: Re: [PATCH] ovl: filter of trusted xattr results in audit Date: Mon, 7 Oct 2019 09:42:08 -0700 Message-ID: <7c610f92-5e1f-32ef-0a60-ed47ea999fe3@android.com> References: <20191007160918.29504-1-salyzyn@android.com> <20191007161616.GA988623@kroah.com> <20191007161725.GB988623@kroah.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20191007161725.GB988623@kroah.com> Content-Language: en-GB Sender: linux-kernel-owner@vger.kernel.org To: Greg Kroah-Hartman Cc: linux-kernel@vger.kernel.org, kernel-team@android.com, linux-security-module@vger.kernel.org, stable@vger.kernel.org, Miklos Szeredi , linux-unionfs@vger.kernel.org List-Id: linux-unionfs@vger.kernel.org On 10/7/19 9:17 AM, Greg Kroah-Hartman wrote: > On Mon, Oct 07, 2019 at 06:16:16PM +0200, Greg Kroah-Hartman wrote: >> On Mon, Oct 07, 2019 at 09:09:16AM -0700, Mark Salyzyn wrote: >>> When filtering xattr list for reading, presence of trusted xattr >>> results in a security audit log. However, if there is other content >>> no errno will be set, and if there isn't, the errno will be -ENODATA >>> and not -EPERM as is usually associated with a lack of capability. >>> The check does not block the request to list the xattrs present. >>> >>> Switch to has_capability_noaudit to reflect a more appropriate check. >>> >>> Signed-off-by: Mark Salyzyn >>> Cc: linux-security-module@vger.kernel.org >>> Cc: kernel-team@android.com >>> Cc: Greg Kroah-Hartman >>> Cc: stable@vger.kernel.org # v3.18 >>> Fixes: upstream a082c6f680da ("ovl: filter trusted xattr for non-admin") >>> Fixes: 3.18 4bcc9b4b3a0a ("ovl: filter trusted xattr for non-admin") >>> --- >>> Replaced ns_capable_noaudit with 3.18.y tree specific >>> has_capability_noaudit present in original submission to kernel.org >>> commit 5c2e9f346b815841f9bed6029ebcb06415caf640 >>> ("ovl: filter of trusted xattr results in audit") >>> >>> fs/overlayfs/inode.c | 3 ++- >>> 1 file changed, 2 insertions(+), 1 deletion(-) >>> >>> diff --git a/fs/overlayfs/inode.c b/fs/overlayfs/inode.c >>> index a01ec1836a72..1175efa5e956 100644 >>> --- a/fs/overlayfs/inode.c >>> +++ b/fs/overlayfs/inode.c >>> @@ -265,7 +265,8 @@ static bool ovl_can_list(const char *s) >>> return true; >>> >>> /* Never list trusted.overlay, list other trusted for superuser only */ >>> - return !ovl_is_private_xattr(s) && capable(CAP_SYS_ADMIN); >>> + return !ovl_is_private_xattr(s) && >>> + has_capability_noaudit(current, CAP_SYS_ADMIN); >>> } >>> >>> ssize_t ovl_listxattr(struct dentry *dentry, char *list, size_t size) >>> -- >>> 2.23.0.581.g78d2f28ef7-goog >>> >> Thanks for the backport, this one worked! > I spoke too soon: > > ERROR: "has_capability_noaudit" [fs/overlayfs/overlay.ko] undefined! > > That function isn't exported for modules :( > > greg k-h Now what is the playbook, we have three options in order of preference: 1) #ifdef MODULE use capable() to preserve API, add a short comment about the side effects if overlayfs is used as a module. 2) export has_capability_nodaudit (proc and oom_kill use it, and are both built-in only), but affect the 3.18 API at near EOL. AFAIK no one wants that? 3) Do nothing more. Make this a distro concern only. Leave this posted as a back-port for the record, but never merged, for those that are _interested_ and declare 3.18 stable as noisy for sepolicy and overlayfs under some usage patterns with few user space mitigation unless they explicitly take this back-port into their tree (eg: android common kernel) if used built-in. This way, in 3.18.y at least the module and built-in version behave the _same_ in stable. Looking for feedback. Sincerely -- Mark Salyzyn