Re: [PATCH 2/2] security.capability: fix conversions on getxattr
"Serge E. Hallyn" writes:
> On Fri, Jan 29, 2021 at 04:55:29PM -0600, Eric W. Biederman wrote:
>> "Serge E. Hallyn" writes:
>>
>> > On Thu, Jan 28, 2021 at 02:19:13PM -0600, Eric W. Biederman wrote:
>> >> "Serge E. Hallyn" writes:
>> >>
>> >> > On Tue, Jan 19, 2021 at 07:34:49PM -0600, Eric W. Biederman wrote:
>> >> >> Miklos Szeredi writes:
>> >> >>
>> >> >> > If a capability is stored on disk in v2 format
>> >> >> > cap_inode_getsecurity() will
>> >> >> > currently return in v2 format unconditionally.
>> >> >> >
>> >> >> > This is wrong: v2 cap should be equivalent to a v3 cap with zero
>> >> >> > rootid,
>> >> >> > and so the same conversions performed on it.
>> >> >> >
>> >> >> > If the rootid cannot be mapped v3 is returned unconverted. Fix this
>> >> >> > so
>> >> >> > that both v2 and v3 return -EOVERFLOW if the rootid (or the owner of
>> >> >> > the fs
>> >> >> > user namespace in case of v2) cannot be mapped in the current user
>> >> >> > namespace.
>> >> >>
>> >> >> This looks like a good cleanup.
>> >> >
>> >> > Sorry, I'm not following. Why is this a good cleanup? Why should
>> >> > the xattr be shown as faked v3 in this case?
>> >>
>> >> If the reader is in _user_ns. If the filesystem was mounted in a
>> >> user namespace. Then the reader looses the information that the
>> >
>> > Can you be more precise about "filesystem was mounted in a user namespace"?
>> > Is this a FUSE thing, the fs is marked as being mounted in a non-init
>> > userns?
>> > If that's a possible case, then yes that must be represented as v3. Using
>> > is_v2header() may be the simpler way to check for that, but the more
>> > accurate
>> > check would be "is it v2 header and mounted by init_user_ns".
>>
>> I think the filesystems current relevant are fuse,overlayfs,ramfs,tmpfs.
>>
>> > Basically yes, in as many cases as possible we want to just give a v2
>> > cap because more userspace knows what to do with that, but a
>> > non-init-userns
>> > mounted fs which provides a v2 fscap should have it represented as v3 cap
>> > with rootid being the kuid that owns the userns.
>>
>> That is the case we that is being fixed in the patch.
>>
>> > Or am I still thinking wrongly? Wouldn't be entirely surprised :)
>>
>> No you got it.
>
> So then can we make faking a v3 gated on whether
> sb->s_user_ns != _user_ns ?
Sort of.
What Miklos's patch implements is always treating a v2 cap xattr on disk
as v3 internally.
> if (is_v2header((size_t) ret, cap)) {
> root = 0;
> } else if (is_v3header((size_t) ret, cap)) {
> nscap = (struct vfs_ns_cap_data *) tmpbuf;
> root = le32_to_cpu(nscap->rootid);
> } else {
> size = -EINVAL;
> goto out_free;
> }
Then v3 is returned if:
> /* If the root kuid maps to a valid uid in current ns, then return
>* this as a nscap. */
> mappedroot = from_kuid(current_user_ns(), kroot);
> if (mappedroot != (uid_t)-1 && mappedroot != (uid_t)0) {
After that we verify that the fs capability can be seen by the caller
as a v2 cap xattr with:
> > if (!rootid_owns_currentns(kroot)) {
> > size = -EOVERFLOW;
> > goto out_free;
Anything that passes that test and does not encounter a memory
allocation error is returned as a v2.
...
Which in practice does mean that if sb->s_user_ns != _user_ns,
then mappedroot != 0, and is returned as a v3.
The rest of the logic takes care of all of the other crazy silly
combinations. Like a user namespace that identity maps uid 0,
and then mounts a filesystem.
Eric
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
On Fri, Jan 29, 2021 at 04:55:29PM -0600, Eric W. Biederman wrote: > "Serge E. Hallyn" writes: > > > On Thu, Jan 28, 2021 at 02:19:13PM -0600, Eric W. Biederman wrote: > >> "Serge E. Hallyn" writes: > >> > >> > On Tue, Jan 19, 2021 at 07:34:49PM -0600, Eric W. Biederman wrote: > >> >> Miklos Szeredi writes: > >> >> > >> >> > If a capability is stored on disk in v2 format > >> >> > cap_inode_getsecurity() will > >> >> > currently return in v2 format unconditionally. > >> >> > > >> >> > This is wrong: v2 cap should be equivalent to a v3 cap with zero > >> >> > rootid, > >> >> > and so the same conversions performed on it. > >> >> > > >> >> > If the rootid cannot be mapped v3 is returned unconverted. Fix this > >> >> > so > >> >> > that both v2 and v3 return -EOVERFLOW if the rootid (or the owner of > >> >> > the fs > >> >> > user namespace in case of v2) cannot be mapped in the current user > >> >> > namespace. > >> >> > >> >> This looks like a good cleanup. > >> > > >> > Sorry, I'm not following. Why is this a good cleanup? Why should > >> > the xattr be shown as faked v3 in this case? > >> > >> If the reader is in _user_ns. If the filesystem was mounted in a > >> user namespace. Then the reader looses the information that the > > > > Can you be more precise about "filesystem was mounted in a user namespace"? > > Is this a FUSE thing, the fs is marked as being mounted in a non-init > > userns? > > If that's a possible case, then yes that must be represented as v3. Using > > is_v2header() may be the simpler way to check for that, but the more > > accurate > > check would be "is it v2 header and mounted by init_user_ns". > > I think the filesystems current relevant are fuse,overlayfs,ramfs,tmpfs. > > > Basically yes, in as many cases as possible we want to just give a v2 > > cap because more userspace knows what to do with that, but a non-init-userns > > mounted fs which provides a v2 fscap should have it represented as v3 cap > > with rootid being the kuid that owns the userns. > > That is the case we that is being fixed in the patch. > > > Or am I still thinking wrongly? Wouldn't be entirely surprised :) > > No you got it. So then can we make faking a v3 gated on whether sb->s_user_ns != _user_ns ?
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
On Fri, Jan 29, 2021 at 05:11:53PM -0600, Eric W. Biederman wrote:
> "Serge E. Hallyn" writes:
>
> > On Thu, Jan 28, 2021 at 08:44:26PM +0100, Miklos Szeredi wrote:
> >> On Thu, Jan 28, 2021 at 6:09 PM Serge E. Hallyn wrote:
> >> >
> >> > On Tue, Jan 19, 2021 at 07:34:49PM -0600, Eric W. Biederman wrote:
> >> > > Miklos Szeredi writes:
> >> > >
> >> > > > if (!rootid_owns_currentns(kroot)) {
> >> > > > - kfree(tmpbuf);
> >> > > > - return -EOPNOTSUPP;
> >> > > > + size = -EOVERFLOW;
> >> >
> >> > Why this change? Christian (cc:d) noticed that this is a user visible
> >> > change.
> >> > Without this change, if you are in a userns which has different rootid,
> >> > the
> >> > EOVERFLOW tells vfs_getxattr to vall back to __vfs_getxattr() and so you
> >> > can
> >> > see the v3 capability with its rootid.
> >> >
> >> > With this change, you instead just get EOVERFLOW.
> >>
> >> Why would the user want to see nonsense (in its own userns) rootid and
> >> what would it do with it?
> >
> > They would know that the data is there.
>
> But an error of -EOVERFLOW still indicates data is there.
> You just don't get the data because it can not be represented.
Ok - and this happens *after* the check for whether the rootid to maps
into the current ns.
That sounds reasonable, thanks.
> >> Please give an example where an untranslatable rootid would make any
> >> sense at all to the user.
> >
> > I may have accidentally, from init_user_ns, as uid 1000, set an
> > fscap with rootid 11 instead of 10, and wonder why the
> > cap is not working in the container where 10 is root.
>
> Getting -EOVERFLOW when attempting to read the cap from inside
> the user namespace will immediately tell you what is wrong. The rootid
> does not map.
>
> That is how all the non-mapping situations are handled. Either
> -EOVERFLOW or returning INVALID_UID/the unmapped user id aka nobody.
>
> The existing code is wrong because it returns a completely untranslated
> uid, which is completely non-sense.
>
> An argument could be made for returning a rootid of 0x aka
> INVALID_UID in a v3 cap xattr when the rootid can not be mapped. I
> think that is what we do with posix_acls that contain ids that don't
> map. My sense is returning -EOVERFLOW inside the container and
> returning the v3 cap xattr outside the container will most quickly get
> the problem diagnosed, and will be the most likely to not cause
> problems.
>
> If there is a good case for returning a v3 cap with rootid of 0x
> instead of -EOVERFLOW we can do that. Right now I don't see anything
> that would be compelling in either direction.
>
> Eric
>
>
>
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
"Serge E. Hallyn" writes:
> On Thu, Jan 28, 2021 at 08:44:26PM +0100, Miklos Szeredi wrote:
>> On Thu, Jan 28, 2021 at 6:09 PM Serge E. Hallyn wrote:
>> >
>> > On Tue, Jan 19, 2021 at 07:34:49PM -0600, Eric W. Biederman wrote:
>> > > Miklos Szeredi writes:
>> > >
>> > > > if (!rootid_owns_currentns(kroot)) {
>> > > > - kfree(tmpbuf);
>> > > > - return -EOPNOTSUPP;
>> > > > + size = -EOVERFLOW;
>> >
>> > Why this change? Christian (cc:d) noticed that this is a user visible
>> > change.
>> > Without this change, if you are in a userns which has different rootid, the
>> > EOVERFLOW tells vfs_getxattr to vall back to __vfs_getxattr() and so you
>> > can
>> > see the v3 capability with its rootid.
>> >
>> > With this change, you instead just get EOVERFLOW.
>>
>> Why would the user want to see nonsense (in its own userns) rootid and
>> what would it do with it?
>
> They would know that the data is there.
But an error of -EOVERFLOW still indicates data is there.
You just don't get the data because it can not be represented.
>> Please give an example where an untranslatable rootid would make any
>> sense at all to the user.
>
> I may have accidentally, from init_user_ns, as uid 1000, set an
> fscap with rootid 11 instead of 10, and wonder why the
> cap is not working in the container where 10 is root.
Getting -EOVERFLOW when attempting to read the cap from inside
the user namespace will immediately tell you what is wrong. The rootid
does not map.
That is how all the non-mapping situations are handled. Either
-EOVERFLOW or returning INVALID_UID/the unmapped user id aka nobody.
The existing code is wrong because it returns a completely untranslated
uid, which is completely non-sense.
An argument could be made for returning a rootid of 0x aka
INVALID_UID in a v3 cap xattr when the rootid can not be mapped. I
think that is what we do with posix_acls that contain ids that don't
map. My sense is returning -EOVERFLOW inside the container and
returning the v3 cap xattr outside the container will most quickly get
the problem diagnosed, and will be the most likely to not cause
problems.
If there is a good case for returning a v3 cap with rootid of 0x
instead of -EOVERFLOW we can do that. Right now I don't see anything
that would be compelling in either direction.
Eric
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
"Serge E. Hallyn" writes: > On Thu, Jan 28, 2021 at 02:19:13PM -0600, Eric W. Biederman wrote: >> "Serge E. Hallyn" writes: >> >> > On Tue, Jan 19, 2021 at 07:34:49PM -0600, Eric W. Biederman wrote: >> >> Miklos Szeredi writes: >> >> >> >> > If a capability is stored on disk in v2 format cap_inode_getsecurity() >> >> > will >> >> > currently return in v2 format unconditionally. >> >> > >> >> > This is wrong: v2 cap should be equivalent to a v3 cap with zero rootid, >> >> > and so the same conversions performed on it. >> >> > >> >> > If the rootid cannot be mapped v3 is returned unconverted. Fix this so >> >> > that both v2 and v3 return -EOVERFLOW if the rootid (or the owner of >> >> > the fs >> >> > user namespace in case of v2) cannot be mapped in the current user >> >> > namespace. >> >> >> >> This looks like a good cleanup. >> > >> > Sorry, I'm not following. Why is this a good cleanup? Why should >> > the xattr be shown as faked v3 in this case? >> >> If the reader is in _user_ns. If the filesystem was mounted in a >> user namespace. Then the reader looses the information that the > > Can you be more precise about "filesystem was mounted in a user namespace"? > Is this a FUSE thing, the fs is marked as being mounted in a non-init userns? > If that's a possible case, then yes that must be represented as v3. Using > is_v2header() may be the simpler way to check for that, but the more accurate > check would be "is it v2 header and mounted by init_user_ns". I think the filesystems current relevant are fuse,overlayfs,ramfs,tmpfs. > Basically yes, in as many cases as possible we want to just give a v2 > cap because more userspace knows what to do with that, but a non-init-userns > mounted fs which provides a v2 fscap should have it represented as v3 cap > with rootid being the kuid that owns the userns. That is the case we that is being fixed in the patch. > Or am I still thinking wrongly? Wouldn't be entirely surprised :) No you got it. Eric
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
Miklos Szeredi writes: > On Thu, Jan 28, 2021 at 9:24 PM Eric W. Biederman > wrote: > >> >> From our previous discussions I would also argue it would be good >> if there was a bypass that skipped all conversions if the reader >> and the filesystem are in the same user namespace. >> > > That's however just an optimization (AFAICS) that only makes sense if > it helps a read world workload. I'm not convinced that that's the > case. It is definitely a different issue. >From previous conversations with Serge, there is a concern with a sysadmin wanting to see what is actually on disk. In case there are bugs that care about the different layout. Just passing everything through when no translation is necessary will allow that kind of diagnosis. As your patch demonstrates we already have had bugs in this area so being able to get at the raw data may help people if they get into a situation where bugs matter. Eric
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
On Thu, Jan 28, 2021 at 9:24 PM Eric W. Biederman wrote: > > From our previous discussions I would also argue it would be good > if there was a bypass that skipped all conversions if the reader > and the filesystem are in the same user namespace. > That's however just an optimization (AFAICS) that only makes sense if it helps a read world workload. I'm not convinced that that's the case. Thanks, Miklos
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
"Serge E. Hallyn" writes:
> On Tue, Jan 19, 2021 at 07:34:49PM -0600, Eric W. Biederman wrote:
>> Miklos Szeredi writes:
>>
>> > If a capability is stored on disk in v2 format cap_inode_getsecurity() will
>> > currently return in v2 format unconditionally.
>> >
>> > This is wrong: v2 cap should be equivalent to a v3 cap with zero rootid,
>> > and so the same conversions performed on it.
>> >
>> > If the rootid cannot be mapped v3 is returned unconverted. Fix this so
>> > that both v2 and v3 return -EOVERFLOW if the rootid (or the owner of the fs
>> > user namespace in case of v2) cannot be mapped in the current user
>> > namespace.
>>
>> This looks like a good cleanup.
>
> Sorry, I'm not following. Why is this a good cleanup? Why should
> the xattr be shown as faked v3 in this case?
If the reader is in _user_ns. If the filesystem was mounted in a
user namespace. Then the reader looses the information that the
capability xattr only applies to a subset of user namespaces.
A trivial place where this would be important is if userspace was to
copy the file and the associated capability xattr to another
filesystem, that is mounted differently.
>From our previous discussions I would also argue it would be good
if there was a bypass that skipped all conversions if the reader
and the filesystem are in the same user namespace.
> A separate question below.
>
>> I do wonder how well this works with stacking. In particular
>> ovl_xattr_set appears to call vfs_getxattr without overriding the creds.
>> What the purpose of that is I haven't quite figured out. It looks like
>> it is just a probe to see if an xattr is present so maybe it is ok.
>>
>> Acked-by: "Eric W. Biederman"
>>
>> >
>> > Signed-off-by: Miklos Szeredi
>> > ---
>> > security/commoncap.c | 67
>> > 1 file changed, 43 insertions(+), 24 deletions(-)
>> >
>> > diff --git a/security/commoncap.c b/security/commoncap.c
>> > index baccd871..c9d99f8f4c82 100644
>> > --- a/security/commoncap.c
>> > +++ b/security/commoncap.c
>> > @@ -371,10 +371,11 @@ int cap_inode_getsecurity(struct inode *inode, const
>> > char *name, void **buffer,
>> > {
>> >int size, ret;
>> >kuid_t kroot;
>> > + __le32 nsmagic, magic;
>> >uid_t root, mappedroot;
>> >char *tmpbuf = NULL;
>> >struct vfs_cap_data *cap;
>> > - struct vfs_ns_cap_data *nscap;
>> > + struct vfs_ns_cap_data *nscap = NULL;
>> >struct dentry *dentry;
>> >struct user_namespace *fs_ns;
>> >
>> > @@ -396,46 +397,61 @@ int cap_inode_getsecurity(struct inode *inode, const
>> > char *name, void **buffer,
>> >fs_ns = inode->i_sb->s_user_ns;
>> >cap = (struct vfs_cap_data *) tmpbuf;
>> >if (is_v2header((size_t) ret, cap)) {
>> > - /* If this is sizeof(vfs_cap_data) then we're ok with the
>> > - * on-disk value, so return that. */
>> > - if (alloc)
>> > - *buffer = tmpbuf;
>> > - else
>> > - kfree(tmpbuf);
>> > - return ret;
>> > - } else if (!is_v3header((size_t) ret, cap)) {
>> > - kfree(tmpbuf);
>> > - return -EINVAL;
>> > + root = 0;
>> > + } else if (is_v3header((size_t) ret, cap)) {
>> > + nscap = (struct vfs_ns_cap_data *) tmpbuf;
>> > + root = le32_to_cpu(nscap->rootid);
>> > + } else {
>> > + size = -EINVAL;
>> > + goto out_free;
>> >}
>> >
>> > - nscap = (struct vfs_ns_cap_data *) tmpbuf;
>> > - root = le32_to_cpu(nscap->rootid);
>> >kroot = make_kuid(fs_ns, root);
>> >
>> >/* If the root kuid maps to a valid uid in current ns, then return
>> > * this as a nscap. */
>> >mappedroot = from_kuid(current_user_ns(), kroot);
>> >if (mappedroot != (uid_t)-1 && mappedroot != (uid_t)0) {
>> > + size = sizeof(struct vfs_ns_cap_data);
>> >if (alloc) {
>> > - *buffer = tmpbuf;
>> > + if (!nscap) {
>> > + /* v2 -> v3 conversion */
>> > + nscap = kzalloc(size, GFP_ATOMIC);
>> > + if (!nscap) {
>> > + size = -ENOMEM;
>> > + goto out_free;
>> > + }
>> > + nsmagic = VFS_CAP_REVISION_3;
>> > + magic = le32_to_cpu(cap->magic_etc);
>> > + if (magic & VFS_CAP_FLAGS_EFFECTIVE)
>> > + nsmagic |= VFS_CAP_FLAGS_EFFECTIVE;
>> > + memcpy(>data, >data, sizeof(__le32)
>> > * 2 * VFS_CAP_U32);
>> > + nscap->magic_etc = cpu_to_le32(nsmagic);
>> > + } else {
>> > + /* use allocated v3 buffer */
>> > + tmpbuf = NULL;
>> > + }
>> >nscap->rootid = cpu_to_le32(mappedroot);
>> > -
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
On Tue, Jan 19, 2021 at 07:34:49PM -0600, Eric W. Biederman wrote:
> Miklos Szeredi writes:
>
> > If a capability is stored on disk in v2 format cap_inode_getsecurity() will
> > currently return in v2 format unconditionally.
> >
> > This is wrong: v2 cap should be equivalent to a v3 cap with zero rootid,
> > and so the same conversions performed on it.
> >
> > If the rootid cannot be mapped v3 is returned unconverted. Fix this so
> > that both v2 and v3 return -EOVERFLOW if the rootid (or the owner of the fs
> > user namespace in case of v2) cannot be mapped in the current user
> > namespace.
>
> This looks like a good cleanup.
Sorry, I'm not following. Why is this a good cleanup? Why should
the xattr be shown as faked v3 in this case?
A separate question below.
> I do wonder how well this works with stacking. In particular
> ovl_xattr_set appears to call vfs_getxattr without overriding the creds.
> What the purpose of that is I haven't quite figured out. It looks like
> it is just a probe to see if an xattr is present so maybe it is ok.
>
> Acked-by: "Eric W. Biederman"
>
> >
> > Signed-off-by: Miklos Szeredi
> > ---
> > security/commoncap.c | 67
> > 1 file changed, 43 insertions(+), 24 deletions(-)
> >
> > diff --git a/security/commoncap.c b/security/commoncap.c
> > index baccd871..c9d99f8f4c82 100644
> > --- a/security/commoncap.c
> > +++ b/security/commoncap.c
> > @@ -371,10 +371,11 @@ int cap_inode_getsecurity(struct inode *inode, const
> > char *name, void **buffer,
> > {
> > int size, ret;
> > kuid_t kroot;
> > + __le32 nsmagic, magic;
> > uid_t root, mappedroot;
> > char *tmpbuf = NULL;
> > struct vfs_cap_data *cap;
> > - struct vfs_ns_cap_data *nscap;
> > + struct vfs_ns_cap_data *nscap = NULL;
> > struct dentry *dentry;
> > struct user_namespace *fs_ns;
> >
> > @@ -396,46 +397,61 @@ int cap_inode_getsecurity(struct inode *inode, const
> > char *name, void **buffer,
> > fs_ns = inode->i_sb->s_user_ns;
> > cap = (struct vfs_cap_data *) tmpbuf;
> > if (is_v2header((size_t) ret, cap)) {
> > - /* If this is sizeof(vfs_cap_data) then we're ok with the
> > -* on-disk value, so return that. */
> > - if (alloc)
> > - *buffer = tmpbuf;
> > - else
> > - kfree(tmpbuf);
> > - return ret;
> > - } else if (!is_v3header((size_t) ret, cap)) {
> > - kfree(tmpbuf);
> > - return -EINVAL;
> > + root = 0;
> > + } else if (is_v3header((size_t) ret, cap)) {
> > + nscap = (struct vfs_ns_cap_data *) tmpbuf;
> > + root = le32_to_cpu(nscap->rootid);
> > + } else {
> > + size = -EINVAL;
> > + goto out_free;
> > }
> >
> > - nscap = (struct vfs_ns_cap_data *) tmpbuf;
> > - root = le32_to_cpu(nscap->rootid);
> > kroot = make_kuid(fs_ns, root);
> >
> > /* If the root kuid maps to a valid uid in current ns, then return
> > * this as a nscap. */
> > mappedroot = from_kuid(current_user_ns(), kroot);
> > if (mappedroot != (uid_t)-1 && mappedroot != (uid_t)0) {
> > + size = sizeof(struct vfs_ns_cap_data);
> > if (alloc) {
> > - *buffer = tmpbuf;
> > + if (!nscap) {
> > + /* v2 -> v3 conversion */
> > + nscap = kzalloc(size, GFP_ATOMIC);
> > + if (!nscap) {
> > + size = -ENOMEM;
> > + goto out_free;
> > + }
> > + nsmagic = VFS_CAP_REVISION_3;
> > + magic = le32_to_cpu(cap->magic_etc);
> > + if (magic & VFS_CAP_FLAGS_EFFECTIVE)
> > + nsmagic |= VFS_CAP_FLAGS_EFFECTIVE;
> > + memcpy(>data, >data, sizeof(__le32)
> > * 2 * VFS_CAP_U32);
> > + nscap->magic_etc = cpu_to_le32(nsmagic);
> > + } else {
> > + /* use allocated v3 buffer */
> > + tmpbuf = NULL;
> > + }
> > nscap->rootid = cpu_to_le32(mappedroot);
> > - } else
> > - kfree(tmpbuf);
> > - return size;
> > + *buffer = nscap;
> > + }
> > + goto out_free;
> > }
> >
> > if (!rootid_owns_currentns(kroot)) {
> > - kfree(tmpbuf);
> > - return -EOPNOTSUPP;
> > + size = -EOVERFLOW;
Why this change? Christian (cc:d) noticed that this is a user visible change.
Without this change, if you are in a userns which has different rootid, the
EOVERFLOW tells vfs_getxattr to vall back to __vfs_getxattr() and so you can
see the v3 capability with its rootid.
With this change, you
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
Hi Miklos,
I love your patch! Perhaps something to improve:
[auto build test WARNING on security/next-testing]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url:
https://github.com/0day-ci/linux/commits/Miklos-Szeredi/capability-conversion-fixes/20210120-152933
base:
https://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security.git
next-testing
config: x86_64-randconfig-s022-20210120 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-15) 9.3.0
reproduce:
# apt-get install sparse
# sparse version: v0.6.3-208-g46a52ca4-dirty
#
https://github.com/0day-ci/linux/commit/bcf70adf8bcc3e52cb1b262ae2e1d9154da75097
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review
Miklos-Szeredi/capability-conversion-fixes/20210120-152933
git checkout bcf70adf8bcc3e52cb1b262ae2e1d9154da75097
# save the attached .config to linux build tree
make W=1 C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=x86_64
If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot
"sparse warnings: (new ones prefixed by >>)"
>> security/commoncap.c:424:41: sparse: sparse: incorrect type in assignment
>> (different base types) @@ expected restricted __le32 [usertype] nsmagic
>> @@ got int @@
security/commoncap.c:424:41: sparse: expected restricted __le32
[usertype] nsmagic
security/commoncap.c:424:41: sparse: got int
>> security/commoncap.c:425:39: sparse: sparse: incorrect type in assignment
>> (different base types) @@ expected restricted __le32 [usertype] magic @@
>> got unsigned int [usertype] @@
security/commoncap.c:425:39: sparse: expected restricted __le32
[usertype] magic
security/commoncap.c:425:39: sparse: got unsigned int [usertype]
security/commoncap.c:426:37: sparse: sparse: restricted __le32 degrades to
integer
security/commoncap.c:427:49: sparse: sparse: invalid assignment: |=
security/commoncap.c:427:49: sparse:left side has type restricted __le32
security/commoncap.c:427:49: sparse:right side has type int
security/commoncap.c:429:52: sparse: sparse: cast from restricted __le32
security/commoncap.c:455:31: sparse: sparse: incorrect type in assignment
(different base types) @@ expected restricted __le32 [usertype] magic @@
got int @@
security/commoncap.c:455:31: sparse: expected restricted __le32
[usertype] magic
security/commoncap.c:455:31: sparse: got int
security/commoncap.c:456:33: sparse: sparse: incorrect type in assignment
(different base types) @@ expected restricted __le32 [usertype] nsmagic @@
got unsigned int [usertype] @@
security/commoncap.c:456:33: sparse: expected restricted __le32
[usertype] nsmagic
security/commoncap.c:456:33: sparse: got unsigned int [usertype]
security/commoncap.c:457:29: sparse: sparse: restricted __le32 degrades to
integer
security/commoncap.c:458:39: sparse: sparse: invalid assignment: |=
security/commoncap.c:458:39: sparse:left side has type restricted __le32
security/commoncap.c:458:39: sparse:right side has type int
security/commoncap.c:460:42: sparse: sparse: cast from restricted __le32
security/commoncap.c:1281:41: sparse: sparse: dubious: !x | y
vim +424 security/commoncap.c
357
358 /*
359 * getsecurity: We are called for security.* before any attempt to read
the
360 * xattr from the inode itself.
361 *
362 * This gives us a chance to read the on-disk value and convert it. If
we
363 * return -EOPNOTSUPP, then vfs_getxattr() will call the i_op handler.
364 *
365 * Note we are not called by vfs_getxattr_alloc(), but that is only
called
366 * by the integrity subsystem, which really wants the unconverted
values -
367 * so that's good.
368 */
369 int cap_inode_getsecurity(struct inode *inode, const char *name, void
**buffer,
370bool alloc)
371 {
372 int size, ret;
373 kuid_t kroot;
374 __le32 nsmagic, magic;
375 uid_t root, mappedroot;
376 char *tmpbuf = NULL;
377 struct vfs_cap_data *cap;
378 struct vfs_ns_cap_data *nscap = NULL;
379 struct dentry *dentry;
380 struct user_namespace *fs_ns;
381
382 if (strcmp(name, "capability") != 0)
383 return -EOPNOTSUPP;
384
385 dentry = d_find_any_alias(inode);
386 if (!dentry)
387 return -EINVAL;
388
389 size = sizeof(struct vfs_ns_cap_data);
390 ret = (int) vfs_getxattr_alloc(dentry, XATTR_NAME_CAPS,
391 , size, GFP_NOFS);
392
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
Hi Miklos, I love your patch! Perhaps something to improve: [auto build test WARNING on security/next-testing] [also build test WARNING on linux/master linus/master v5.11-rc4 next-20210120] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in https://git-scm.com/docs/git-format-patch] url: https://github.com/0day-ci/linux/commits/Miklos-Szeredi/capability-conversion-fixes/20210120-152933 base: https://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security.git next-testing config: xtensa-randconfig-s032-20210120 (attached as .config) compiler: xtensa-linux-gcc (GCC) 9.3.0 reproduce: wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross # apt-get install sparse # sparse version: v0.6.3-208-g46a52ca4-dirty # https://github.com/0day-ci/linux/commit/bcf70adf8bcc3e52cb1b262ae2e1d9154da75097 git remote add linux-review https://github.com/0day-ci/linux git fetch --no-tags linux-review Miklos-Szeredi/capability-conversion-fixes/20210120-152933 git checkout bcf70adf8bcc3e52cb1b262ae2e1d9154da75097 # save the attached .config to linux build tree COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=xtensa If you fix the issue, kindly add following tag as appropriate Reported-by: kernel test robot "sparse warnings: (new ones prefixed by >>)" security/commoncap.c:424:41: sparse: sparse: incorrect type in assignment (different base types) @@ expected restricted __le32 [usertype] nsmagic @@ got int @@ security/commoncap.c:424:41: sparse: expected restricted __le32 [usertype] nsmagic security/commoncap.c:424:41: sparse: got int >> security/commoncap.c:425:39: sparse: sparse: incorrect type in assignment >> (different base types) @@ expected restricted __le32 [usertype] magic @@ >> got unsigned int @@ security/commoncap.c:425:39: sparse: expected restricted __le32 [usertype] magic security/commoncap.c:425:39: sparse: got unsigned int security/commoncap.c:426:37: sparse: sparse: restricted __le32 degrades to integer security/commoncap.c:427:49: sparse: sparse: invalid assignment: |= security/commoncap.c:427:49: sparse:left side has type restricted __le32 security/commoncap.c:427:49: sparse:right side has type int security/commoncap.c:429:52: sparse: sparse: cast from restricted __le32 >> security/commoncap.c:429:52: sparse: sparse: incorrect type in argument 1 >> (different base types) @@ expected unsigned int [usertype] val @@ >> got restricted __le32 [usertype] nsmagic @@ security/commoncap.c:429:52: sparse: expected unsigned int [usertype] val security/commoncap.c:429:52: sparse: got restricted __le32 [usertype] nsmagic security/commoncap.c:429:52: sparse: sparse: cast from restricted __le32 security/commoncap.c:429:52: sparse: sparse: cast from restricted __le32 security/commoncap.c:429:52: sparse: sparse: cast from restricted __le32 security/commoncap.c:429:52: sparse: sparse: cast from restricted __le32 security/commoncap.c:455:31: sparse: sparse: incorrect type in assignment (different base types) @@ expected restricted __le32 [usertype] magic @@ got int @@ security/commoncap.c:455:31: sparse: expected restricted __le32 [usertype] magic security/commoncap.c:455:31: sparse: got int security/commoncap.c:456:33: sparse: sparse: incorrect type in assignment (different base types) @@ expected restricted __le32 [usertype] nsmagic @@ got unsigned int @@ security/commoncap.c:456:33: sparse: expected restricted __le32 [usertype] nsmagic security/commoncap.c:456:33: sparse: got unsigned int security/commoncap.c:457:29: sparse: sparse: restricted __le32 degrades to integer security/commoncap.c:458:39: sparse: sparse: invalid assignment: |= security/commoncap.c:458:39: sparse:left side has type restricted __le32 security/commoncap.c:458:39: sparse:right side has type int security/commoncap.c:460:42: sparse: sparse: cast from restricted __le32 security/commoncap.c:460:42: sparse: sparse: incorrect type in argument 1 (different base types) @@ expected unsigned int [usertype] val @@ got restricted __le32 [usertype] magic @@ security/commoncap.c:460:42: sparse: expected unsigned int [usertype] val security/commoncap.c:460:42: sparse: got restricted __le32 [usertype] magic security/commoncap.c:460:42: sparse: sparse: cast from restricted __le32 security/commoncap.c:460:42: sparse: sparse: cast from restricted __le32 security/commoncap.c:460:42: sparse: sparse: cast from restricted __le32 security/commoncap.c:460:42: sparse: sparse: cast from restricted __le32
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
On Wed, Jan 20, 2021 at 2:39 AM Eric W. Biederman wrote:
>
> Miklos Szeredi writes:
>
> > If a capability is stored on disk in v2 format cap_inode_getsecurity() will
> > currently return in v2 format unconditionally.
> >
> > This is wrong: v2 cap should be equivalent to a v3 cap with zero rootid,
> > and so the same conversions performed on it.
> >
> > If the rootid cannot be mapped v3 is returned unconverted. Fix this so
> > that both v2 and v3 return -EOVERFLOW if the rootid (or the owner of the fs
> > user namespace in case of v2) cannot be mapped in the current user
> > namespace.
>
> This looks like a good cleanup.
>
> I do wonder how well this works with stacking. In particular
> ovl_xattr_set appears to call vfs_getxattr without overriding the creds.
> What the purpose of that is I haven't quite figured out. It looks like
> it is just a probe to see if an xattr is present so maybe it is ok.
Yeah, it's checking in the removexattr case whether copy-up is needed
or not (i.e. if trying to remove a non-existent xattr, then no need to
copy up).
But for consistency it should also be wrapped in override creds.
Adding fix to this series.
I'll also audit for any remaining omissions. One known and documented
case is vfs_ioctl(FS_IOC_{[SG]ETFLAGS,FS[SG]ETXATTR}), but that
shouldn't be affected by user namespaces.
Thanks,
Miklos
Re: [PATCH 2/2] security.capability: fix conversions on getxattr
Miklos Szeredi writes:
> If a capability is stored on disk in v2 format cap_inode_getsecurity() will
> currently return in v2 format unconditionally.
>
> This is wrong: v2 cap should be equivalent to a v3 cap with zero rootid,
> and so the same conversions performed on it.
>
> If the rootid cannot be mapped v3 is returned unconverted. Fix this so
> that both v2 and v3 return -EOVERFLOW if the rootid (or the owner of the fs
> user namespace in case of v2) cannot be mapped in the current user
> namespace.
This looks like a good cleanup.
I do wonder how well this works with stacking. In particular
ovl_xattr_set appears to call vfs_getxattr without overriding the creds.
What the purpose of that is I haven't quite figured out. It looks like
it is just a probe to see if an xattr is present so maybe it is ok.
Acked-by: "Eric W. Biederman"
>
> Signed-off-by: Miklos Szeredi
> ---
> security/commoncap.c | 67
> 1 file changed, 43 insertions(+), 24 deletions(-)
>
> diff --git a/security/commoncap.c b/security/commoncap.c
> index baccd871..c9d99f8f4c82 100644
> --- a/security/commoncap.c
> +++ b/security/commoncap.c
> @@ -371,10 +371,11 @@ int cap_inode_getsecurity(struct inode *inode, const
> char *name, void **buffer,
> {
> int size, ret;
> kuid_t kroot;
> + __le32 nsmagic, magic;
> uid_t root, mappedroot;
> char *tmpbuf = NULL;
> struct vfs_cap_data *cap;
> - struct vfs_ns_cap_data *nscap;
> + struct vfs_ns_cap_data *nscap = NULL;
> struct dentry *dentry;
> struct user_namespace *fs_ns;
>
> @@ -396,46 +397,61 @@ int cap_inode_getsecurity(struct inode *inode, const
> char *name, void **buffer,
> fs_ns = inode->i_sb->s_user_ns;
> cap = (struct vfs_cap_data *) tmpbuf;
> if (is_v2header((size_t) ret, cap)) {
> - /* If this is sizeof(vfs_cap_data) then we're ok with the
> - * on-disk value, so return that. */
> - if (alloc)
> - *buffer = tmpbuf;
> - else
> - kfree(tmpbuf);
> - return ret;
> - } else if (!is_v3header((size_t) ret, cap)) {
> - kfree(tmpbuf);
> - return -EINVAL;
> + root = 0;
> + } else if (is_v3header((size_t) ret, cap)) {
> + nscap = (struct vfs_ns_cap_data *) tmpbuf;
> + root = le32_to_cpu(nscap->rootid);
> + } else {
> + size = -EINVAL;
> + goto out_free;
> }
>
> - nscap = (struct vfs_ns_cap_data *) tmpbuf;
> - root = le32_to_cpu(nscap->rootid);
> kroot = make_kuid(fs_ns, root);
>
> /* If the root kuid maps to a valid uid in current ns, then return
>* this as a nscap. */
> mappedroot = from_kuid(current_user_ns(), kroot);
> if (mappedroot != (uid_t)-1 && mappedroot != (uid_t)0) {
> + size = sizeof(struct vfs_ns_cap_data);
> if (alloc) {
> - *buffer = tmpbuf;
> + if (!nscap) {
> + /* v2 -> v3 conversion */
> + nscap = kzalloc(size, GFP_ATOMIC);
> + if (!nscap) {
> + size = -ENOMEM;
> + goto out_free;
> + }
> + nsmagic = VFS_CAP_REVISION_3;
> + magic = le32_to_cpu(cap->magic_etc);
> + if (magic & VFS_CAP_FLAGS_EFFECTIVE)
> + nsmagic |= VFS_CAP_FLAGS_EFFECTIVE;
> + memcpy(>data, >data, sizeof(__le32)
> * 2 * VFS_CAP_U32);
> + nscap->magic_etc = cpu_to_le32(nsmagic);
> + } else {
> + /* use allocated v3 buffer */
> + tmpbuf = NULL;
> + }
> nscap->rootid = cpu_to_le32(mappedroot);
> - } else
> - kfree(tmpbuf);
> - return size;
> + *buffer = nscap;
> + }
> + goto out_free;
> }
>
> if (!rootid_owns_currentns(kroot)) {
> - kfree(tmpbuf);
> - return -EOPNOTSUPP;
> + size = -EOVERFLOW;
> + goto out_free;
> }
>
> /* This comes from a parent namespace. Return as a v2 capability */
> size = sizeof(struct vfs_cap_data);
> if (alloc) {
> - *buffer = kmalloc(size, GFP_ATOMIC);
> - if (*buffer) {
> - struct vfs_cap_data *cap = *buffer;
> - __le32 nsmagic, magic;
> + if (nscap) {
> + /* v3 -> v2 conversion */
> + cap = kzalloc(size, GFP_ATOMIC);
> + if (!cap) {
> + size =
