Hi Jim, On Wed, Sep 20, 2017 at 11:17:06AM -0600, Jim Fehlig wrote:

On 09/20/2017 08:57 AM, Jim Fehlig wrote:

...

On 09/20/2017 12:51 AM, Guido Günther wrote:

...

Hi Jim, On Mon, Sep 18, 2017 at 02:06:13PM -0600, Jim Fehlig wrote:

...

Kernel 4.13 introduced finer-grained ptrace checks

https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?h=v4.13.2&id=290f458a4f16f9cf6cb6562b249e69fe1c3c3a07

When Apparmor is enabled and libvirtd is confined, attempting to start a domain fails

virsh start test error: Failed to start domain test error: internal error: child reported: Kernel does not provide mount         namespace: Permission denied

The audit log contains

type=AVC msg=audit(1505466699.828:534): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=6621 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="/usr/sbin/libvirtd"

It seems access to /proc/<pid>/tasks already requires trace permissions.

...

It was also noticed that simply connecting to libvirtd (e.g. virsh list) resulted in the following entries in the audit log

type=AVC msg=audit(1505755799.975:65): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1418 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined" type=AVC msg=audit(1505755799.976:66): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1418 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined"

Both Apparmor denials can be fixed by adding ptrace rules to the libvirtd profile. The new rules only grant trace permission.

I'm seeing the same denials with 4.13 (4.13.1-1~exp1 (2017-09-11) in Debian) but the proposed profile change does not fix the vm start issue for me. I can't tell why atm, will have to look into this in more detail at the WE.

I have other problems when running with 'security_default_confined = 1' in qemu.conf, but the changes allow starting unconfined domains.

Cedric remembered this old thread

https://www.redhat.com/archives/libvir-list/2014-October/msg00011.html

Some of those changes have been merged, but the ptrace, dbus, signal, etc. have not. I used Stefan's changes to the libvirtd profile but still see the same issue with confined domains

I dug a bit further in that thread to find Stefan's most recent version of the patches

https://www.redhat.com/archives/libvir-list/2014-October/msg00556.html

I took the ptrace, dbus, signal, etc. changes out of patch 2 and used the attached patch to successfully start confined domains.

Since a few years have passed, I'm not sure if patch 1 is still relevant. IIUC, it allows to conditionalize profile content based on apparmor version, which patch 2 uses to add some stuff if version >= 2.9. 2.9 has been out for a while...

Great, this helps a lot!

From e3bb609812776b30acfc0349b25b2e4d539c45c2 Mon Sep 17 00:00:00 2001 From: Jim Fehlig <jfehlig@suse.com> Date: Mon, 18 Sep 2017 13:41:26 -0600 Subject: [PATCH] apparmor: support ptrace checks

Kernel 4.13 introduced finer-grained ptrace checks

https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?h=v4.13.2&id=290f458a4f16f9cf6cb6562b249e69fe1c3c3a07

When Apparmor is enabled and libvirtd is confined, attempting to start a domain fails

virsh start test error: Failed to start domain test error: internal error: child reported: Kernel does not provide mount namespace: Permission denied

The audit log contains

type=AVC msg=audit(1505466699.828:534): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=6621 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="/usr/sbin/libvirtd"

It was also noticed that simply connecting to libvirtd (e.g. virsh list) resulted in the following entries in the audit log

type=AVC msg=audit(1505755799.975:65): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1418 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined" type=AVC msg=audit(1505755799.976:66): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1418 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined"

Both Apparmor denials can be fixed by supporting ptrace in the libvirtd, qemu, and lxc profiles. While at it, also add support for dbus, signal, and unix.

Resolves: https://bugzilla.suse.com/show_bug.cgi?id=1058847 Signed-off-by: Jim Fehlig <jfehlig@suse.com> --- examples/apparmor/libvirt-lxc | 3 +++ examples/apparmor/libvirt-qemu | 3 +++ examples/apparmor/usr.sbin.libvirtd | 6 ++++++ 3 files changed, 12 insertions(+)

diff --git a/examples/apparmor/libvirt-lxc b/examples/apparmor/libvirt-lxc index 4bfb503aa..0db137de0 100644 --- a/examples/apparmor/libvirt-lxc +++ b/examples/apparmor/libvirt-lxc @@ -3,6 +3,9 @@ #include <abstractions/base>

umount, + dbus, + signal, + ptrace,

Can't we get a long with a ptrace (tracedby) peer=/usr/sbin/libvirtd, here too?

# ignore DENIED message on / remount deny mount options=(ro, remount) -> /, diff --git a/examples/apparmor/libvirt-qemu b/examples/apparmor/libvirt-qemu index dcfb1a598..6a4a2335a 100644 --- a/examples/apparmor/libvirt-qemu +++ b/examples/apparmor/libvirt-qemu @@ -170,6 +170,9 @@ @{PROC}/device-tree/** r, /sys/firmware/devicetree/** r,

+ signal (receive) peer=/usr/sbin/libvirtd, + ptrace (tracedby) peer=/usr/sbin/libvirtd,

I would have expected to need the ptrace part but don't (see below).

+ # for gathering information about available host resources /sys/devices/system/cpu/ r, /sys/devices/system/node/ r, diff --git a/examples/apparmor/usr.sbin.libvirtd b/examples/apparmor/usr.sbin.libvirtd index acb59e071..9aadba411 100644 --- a/examples/apparmor/usr.sbin.libvirtd +++ b/examples/apparmor/usr.sbin.libvirtd @@ -36,6 +36,12 @@ network inet6 dgram, network packet dgram, network packet raw, + network netlink, + + dbus bus=system, + signal, + ptrace,

^^^^^^^ This single line is enough to make things work for me on 4.13. AFAIK dbus mediation is not upstream yet and I think unix socket and signal support is neither. Should we drop these for now (the syntax and behaviour might change while things are being upsreamed)? Regarding the ptrace changes in libvirt-{lxc,qemu}. They're not needed here but if you need them to run domains I'm fine (although it would be great to know which denials you were seeing). Cheers, -- Guido

+ unix,

# Very lenient profile for libvirtd since we want to first focus on confining # the guests. Guests will have a very restricted profile. -- 2.14.1

Hi, On Fri, Sep 22, 2017 at 10:29:22AM -0500, Jamie Strandboge wrote:

On Fri, 2017-09-22 at 14:52 +0200, Guido Günther wrote:

...

...

+ ptrace,

^^^^^^^

This single line is enough to make things work for me on 4.13. AFAIK dbus mediation is not upstream yet and I think unix socket and signal support is neither. Should we drop these for now (the syntax and behaviour might change while things are being upsreamed)?

Note that if you are upstreaming profile changes for ptrace, you may as well add them for signal and dbus because an apparmor parser that can understand 'ptrace' can understand the other two. The parser is designed to deal with kernels that don't have the full set of apparmor capabilities. The policy syntax for all of these rules should not change as part of upstreaming dbus and unix.

'unix' is probably ok to add because support for it was added to the parser in devel releases of AppArmor within 6 months of ptrace and signal. 'dbus', 'ptrace', 'signal' and 'unix' were officially introduced in 2.9[1]. By adding 'ptrace' you are saying AppArmor 2.9 is required, therefore, the other 3 are parseable.

The assumption then is that upstreaming the kernel side will never affect the parser, that's fine. Thanks for the clarification! That makes the libvirt-qemu and usr.sbin.libvirtd parts look good. What I don't understand yet is why we have in libvirt-lxc:

diff --git a/examples/apparmor/libvirt-lxc b/examples/apparmor/libvirt-lxc index 4bfb503aa..0db137de0 100644 --- a/examples/apparmor/libvirt-lxc +++ b/examples/apparmor/libvirt-lxc @@ -3,6 +3,9 @@ #include <abstractions/base>

umount, + dbus, + signal, + ptrace,

# ignore DENIED message on / remount deny mount options=(ro, remount) -> /,

Why does it need dbus and why can't we go with a restricted ptrace and signal specifications like in the libvirt-qemu case:

+ signal (receive) peer=/usr/sbin/libvirtd, + ptrace (tracedby) peer=/usr/sbin/libvirtd,

maybe having to add /usr/lib/libvirt/libvirt_lxc into the mix. (I'd check myself but due to the lack of ustream support I can't). Maybe just an omission? Cheers, -- Guido

Hi, On Fri, Sep 22, 2017 at 11:30:39AM -0500, Jamie Strandboge wrote:

On Fri, 2017-09-22 at 17:46 +0200, Guido Günther wrote:

...

...

What I don't understand yet is why we have in libvirt-lxc:

...

diff --git a/examples/apparmor/libvirt-lxc b/examples/apparmor/libvirt-lxc index 4bfb503aa..0db137de0 100644 --- a/examples/apparmor/libvirt-lxc +++ b/examples/apparmor/libvirt-lxc @@ -3,6 +3,9 @@ #include <abstractions/base>

umount, + dbus, + signal, + ptrace,

# ignore DENIED message on / remount deny mount options=(ro, remount) -> /,

Why does it need dbus and why can't we go with a restricted ptrace and signal specifications like in the libvirt-qemu case:

...

+ signal (receive) peer=/usr/sbin/libvirtd, + ptrace (tracedby) peer=/usr/sbin/libvirtd,

maybe having to add /usr/lib/libvirt/libvirt_lxc into the mix. (I'd check myself but due to the lack of ustream support I can't). Maybe just an omission?

I'm not sure how the skew got there, but suspect it was simply that they were developed at different times, perhaps on different distributions initially where is might've looked consistent with the submitter's (perhaps distro-patched) libvirt-qemu.

All that said, in the qemu case, we know that qemu is launching a kernel and that accesses within the guest are mediated by whatever happens to be running in there and libvirt-qemu policy is for hypervisor. For libvirt-lxc, the host's kernel is mediating the access for the guest, so if that kernel has, for example, signal mediation, since the container is confined by apparmor it will need signal rules. Ideally we wouldn't use a bare signal rule, but instead something like:

signal peer=@{profile_name}, signal (receive) peer=unconfined, signal (receive) peer=/usr/sbin/libvirtd,

As it happens, the base abstraction in AppArmor 2.9+ has the first two rules, so, yes, we should be able to do as you suggest for signal and just have the last one.

The base abstraction has similar rules for unix, but I don't think we need a 'unix (receive) peer=/usr/sbin/libvirtd,' in libvirt-lxc unless there is something in the guest that libvirtd can connect to in some manner.

As for ptrace, it is a little different because the base abstraction in 2.9+ has this:

ptrace (tracedby),

So this rule is not actually needed:

ptrace (tracedby) peer=/usr/sbin/libvirtd,

The base abstraction has nothing to say about dbus (there are other abstractions that can be used, but libvirt-lxc is not using them). dbus rules are different than the other three because it isn't the kernel that is enforcing them, but dbus-daemon (dbus-daemon uses libapparmor to query the kernel to ask if an access of a connecting process is allowed or not by asking if the the policy that is loaded for the connecting process' security label (eg, profile name) allows it, and then dbus-daemon will do the actual allow/deny. Therefore, if dbus-daemon in the guest has apparmor support (and it is allowed to use sysfs for the host's apparmor within the container's mount namespace), then dbus- daemon will need some dbus rules. I don't know if libvirt-lxc exposing enough of sysfs for this to work all work, but if it did I *suspect* (but haven't tested this myself) that all that would be needed would be (the parentheses are needed for this rule):

dbus peer=(label=@{profile_name}),

since everything in the container will run under the same profile name, ie, that of the container. Like with 'unix', I don't think we'd need an extra rule for 'dbus (receive) peer=(label=/usr/sbin/libvirtd),' unless there is something in the guest that libvirtd can connect to in some manner.

Hope this helps!

A lot! Since I can't really test dbus and signal rules at the moment I make sure too look at this again once they arrive in mainline kernels (in the hopefully not too distant future). Thanks again! -- Guido

On 09/22/2017 06:52 AM, Guido Günther wrote:

Hi Jim, On Wed, Sep 20, 2017 at 11:17:06AM -0600, Jim Fehlig wrote:

...

On 09/20/2017 08:57 AM, Jim Fehlig wrote:

...

On 09/20/2017 12:51 AM, Guido Günther wrote:

...

Hi Jim, On Mon, Sep 18, 2017 at 02:06:13PM -0600, Jim Fehlig wrote:

...

Kernel 4.13 introduced finer-grained ptrace checks

https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?h=v4.13.2&id=290f458a4f16f9cf6cb6562b249e69fe1c3c3a07

When Apparmor is enabled and libvirtd is confined, attempting to start a domain fails

virsh start test error: Failed to start domain test error: internal error: child reported: Kernel does not provide mount         namespace: Permission denied

The audit log contains

type=AVC msg=audit(1505466699.828:534): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=6621 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="/usr/sbin/libvirtd"

It seems access to /proc/<pid>/tasks already requires trace permissions.

...

It was also noticed that simply connecting to libvirtd (e.g. virsh list) resulted in the following entries in the audit log

type=AVC msg=audit(1505755799.975:65): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1418 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined" type=AVC msg=audit(1505755799.976:66): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1418 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined"

Both Apparmor denials can be fixed by adding ptrace rules to the libvirtd profile. The new rules only grant trace permission.

I'm seeing the same denials with 4.13 (4.13.1-1~exp1 (2017-09-11) in Debian) but the proposed profile change does not fix the vm start issue for me. I can't tell why atm, will have to look into this in more detail at the WE.

I have other problems when running with 'security_default_confined = 1' in qemu.conf, but the changes allow starting unconfined domains.

Cedric remembered this old thread

https://www.redhat.com/archives/libvir-list/2014-October/msg00011.html

Some of those changes have been merged, but the ptrace, dbus, signal, etc. have not. I used Stefan's changes to the libvirtd profile but still see the same issue with confined domains

I dug a bit further in that thread to find Stefan's most recent version of the patches

https://www.redhat.com/archives/libvir-list/2014-October/msg00556.html

I took the ptrace, dbus, signal, etc. changes out of patch 2 and used the attached patch to successfully start confined domains.

Since a few years have passed, I'm not sure if patch 1 is still relevant. IIUC, it allows to conditionalize profile content based on apparmor version, which patch 2 uses to add some stuff if version >= 2.9. 2.9 has been out for a while...

Great, this helps a lot!

...

From e3bb609812776b30acfc0349b25b2e4d539c45c2 Mon Sep 17 00:00:00 2001 From: Jim Fehlig <jfehlig@suse.com> Date: Mon, 18 Sep 2017 13:41:26 -0600 Subject: [PATCH] apparmor: support ptrace checks

Kernel 4.13 introduced finer-grained ptrace checks

https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?h=v4.13.2&id=290f458a4f16f9cf6cb6562b249e69fe1c3c3a07

When Apparmor is enabled and libvirtd is confined, attempting to start a domain fails

virsh start test error: Failed to start domain test error: internal error: child reported: Kernel does not provide mount namespace: Permission denied

The audit log contains

type=AVC msg=audit(1505466699.828:534): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=6621 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="/usr/sbin/libvirtd"

It was also noticed that simply connecting to libvirtd (e.g. virsh list) resulted in the following entries in the audit log

type=AVC msg=audit(1505755799.975:65): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1418 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined" type=AVC msg=audit(1505755799.976:66): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1418 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined"

Both Apparmor denials can be fixed by supporting ptrace in the libvirtd, qemu, and lxc profiles. While at it, also add support for dbus, signal, and unix.

Resolves: https://bugzilla.suse.com/show_bug.cgi?id=1058847 Signed-off-by: Jim Fehlig <jfehlig@suse.com> --- examples/apparmor/libvirt-lxc | 3 +++ examples/apparmor/libvirt-qemu | 3 +++ examples/apparmor/usr.sbin.libvirtd | 6 ++++++ 3 files changed, 12 insertions(+)

diff --git a/examples/apparmor/libvirt-lxc b/examples/apparmor/libvirt-lxc index 4bfb503aa..0db137de0 100644 --- a/examples/apparmor/libvirt-lxc +++ b/examples/apparmor/libvirt-lxc @@ -3,6 +3,9 @@ #include <abstractions/base>

umount, + dbus, + signal, + ptrace,

Can't we get a long with a

ptrace (tracedby) peer=/usr/sbin/libvirtd,

here too?

Probably. I simply took the changes from Stefan's latest patch and didn't test lxc.

...

# ignore DENIED message on / remount deny mount options=(ro, remount) -> /, diff --git a/examples/apparmor/libvirt-qemu b/examples/apparmor/libvirt-qemu index dcfb1a598..6a4a2335a 100644 --- a/examples/apparmor/libvirt-qemu +++ b/examples/apparmor/libvirt-qemu @@ -170,6 +170,9 @@ @{PROC}/device-tree/** r, /sys/firmware/devicetree/** r,

+ signal (receive) peer=/usr/sbin/libvirtd, + ptrace (tracedby) peer=/usr/sbin/libvirtd,

I would have expected to need the ptrace part but don't (see below).

Likewise for these. But after testing I've found they are not needed.

...

+ # for gathering information about available host resources /sys/devices/system/cpu/ r, /sys/devices/system/node/ r, diff --git a/examples/apparmor/usr.sbin.libvirtd b/examples/apparmor/usr.sbin.libvirtd index acb59e071..9aadba411 100644 --- a/examples/apparmor/usr.sbin.libvirtd +++ b/examples/apparmor/usr.sbin.libvirtd @@ -36,6 +36,12 @@ network inet6 dgram, network packet dgram, network packet raw, + network netlink, + + dbus bus=system, + signal, + ptrace,

^^^^^^^

This single line is enough to make things work for me on 4.13.

Same here, but it allows more than needed IMO. Although Jamie provided excellent info on dbus, signal, etc., I'm going to ignore those for now since I can't test on 4.13 kernel. I can however test ptrace and here is a summary of my findings. Using kernel 4.13, apparmor 2.11, and the current libvirt.git profiles, simply starting libvirtd results in the following denial type=AVC msg=audit(1506112085.645:954): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=6984 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="unconfined" Adding 'ptrace (trace) peer=unconfined,' allows starting libvirtd with no denials. But this rule is not enough to start unconfined domains, where I see the following denial type=AVC msg=audit(1506112301.227:1112): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=7498 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="/usr/sbin/libvirtd" Adding 'ptrace (trace) peer=/usr/sbin/libvirtd,' allows starting unconfined domains. But this is still not enough to start confined domains, where I see the following denials type=AVC msg=audit(1506112631.408:1312): apparmor="DENIED" operation="open" profile="virt-aa-helper" name="/etc/libnl/classid" pid=8283 comm="virt-aa-helper" requested_mask="r" denied_mask="r" fsuid=0 ouid=0 type=AVC msg=audit(1506112631.530:1319): apparmor="DENIED" operation="open" profile="virt-aa-helper" name="/etc/libnl/classid" pid=8289 comm="virt-aa-helper" requested_mask="r" denied_mask="r" fsuid=0 ouid=0 type=AVC msg=audit(1506112632.186:1324): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=8342 comm="libvirtd" requested_mask="trace" denied_mask="trace" peer="libvirt-66154842-e926-4f92-92f0-1c1bf61dd1ff" Finally, adding 'ptrace (trace) peer=(label=@{profile_name}),' allows starting confined domains. I'll send a V2 patch that only adds the above ptrace rules to the libvirtd profile. Regards, Jim