On Fri, Jan 24, 2020 at 2:15 PM Arnaud Patard <apatard(a)hupstream.com> wrote:
Christian Ehrhardt <christian.ehrhardt(a)canonical.com> writes:
> On Thu, Dec 5, 2019 at 6:25 PM Arnaud Patard <apatard(a)hupstream.com>
wrote:
>
>> When emulating smartcard with host certificates, qemu needs to
>> be able to read the certificates files. Add necessary code to
>> add the smartcard certificates file path to the apparmor profile.
>>
>> Passthrough support has been tested with spicevmc and remote-viewer.
>>
>> v2:
>> - Fix CodingStyle
>> - Add support for 'host' case.
>> - Add a comment to mention that the passthrough case doesn't need
>> some configuration
>> - Use one rule with '{,*}' instead of two rules.
>>
>> Signed-off-by: Arnaud Patard <apatard(a)hupstream.com>
>> Index: libvirt/src/security/virt-aa-helper.c
>> ===================================================================
>> --- libvirt.orig/src/security/virt-aa-helper.c
>> +++ libvirt/src/security/virt-aa-helper.c
>> @@ -1271,6 +1271,39 @@ get_files(vahControl * ctl)
>> }
>> }
>>
>> + for (i = 0; i < ctl->def->nsmartcards; i++) {
>> + virDomainSmartcardDefPtr sc = ctl->def->smartcards[i];
>> + virDomainSmartcardType sc_type = sc->type;
>> + char *sc_db = (char *)VIR_DOMAIN_SMARTCARD_DEFAULT_DATABASE;
>> + if (sc->data.cert.database)
>> + sc_db = sc->data.cert.database;
>> + switch (sc_type) {
>> + /*
>> + * Note: At time of writing, to get this working, qemu
>> seccomp sandbox has
>> + * to be disabled or the host must be running QEMU with
commit
>> + * 9a1565a03b79d80b236bc7cc2dbce52a2ef3a1b8.
>> + * It's possibly due to
>> libcacard:vcard_emul_new_event_thread(), which calls
>> + * PR_CreateThread(), which calls {g,s}etpriority(). And
>> resourcecontrol seccomp
>> + * filter forbids it (cf src/qemu/qemu_command.c which
seems
>> to always use
>> + * resourcecontrol=deny).
>> + */
>> + case VIR_DOMAIN_SMARTCARD_TYPE_HOST:
>> + virBufferAddLit(&buf, "
\"/etc/pki/nssdb/{,*}\"
rk,\n");
>>
>
> That path matches the examples in libvirt/qemu and also the fedora nss
> package
> [root@fedora~]# rpm -qf /etc/pki/nssdb/
> nss-3.47.0-2.fc29.x86_64
> [root@fedora ~]# ll /etc/pki/nssdb/
> total 8
> -rw-r--r-- 1 root root 65536 Jan 6 2017 cert8.db
> -rw-r--r-- 1 root root 9216 Jan 6 2017 cert9.db
> -rw-r--r-- 1 root root 16384 Jan 6 2017 key3.db
> -rw-r--r-- 1 root root 11264 Jan 6 2017 key4.db
> -rw-r--r-- 1 root root 451 Oct 23 11:23 pkcs11.txt
> -rw-r--r-- 1 root root 16384 Jan 6 2017 secmod.db
>
> But on Debian/Ubuntu the paths are slightly different.
> root@x:~# dpkg -L libnss3-nssdb
> [...]
> /var/lib/nssdb/key4.db
> /var/lib/nssdb/cert9.db
> /var/lib/nssdb/pkcs11.txt
> /var/lib/nssdb/secmod.db
>
> Therefore I'd ask you to add that path as well here.
>
I don't remember seeing this /var/lib/nssdb path on my Debian. I don't
even find the libnss3-nssdb package. Is it Ubuntu-specific and
documented somewhere ?
Hmm, I was just checking paths on a multitude of containers and which
package it belongs.
After you wondered I checked more in detail - the package and the /var...
path are only in older releases.
So please feel free to ignore that part of my question.
I can add this path, I've no problem with that. Nevertheless,
I'm
wondering that if it's really distribution specific (like done only on
Ubuntu), wouldn't it be better to have the distribution patch libvirt to
add this path ?
>
> + break;
>> + case VIR_DOMAIN_SMARTCARD_TYPE_HOST_CERTIFICATES:
>> + virBufferAsprintf(&buf, " \"%s/{,*}\"
rk,\n", sc_db);
>> + break;
>>
>
>>From
https://libvirt.org/formatdomain.html#elementsSmartcard
> "An additional sub-element <database> can specify the absolute path to
an
> alternate directory ... if not present, it defaults to /etc/pki/nssdb."
> That is in "VIR_DOMAIN_SMARTCARD_DEFAULT_DATABASE".
> Have you tested if sc_db is actually set to
> "VIR_DOMAIN_SMARTCARD_DEFAULT_DATABASE"
> in that case?
> If it is e.g. undefined we need to check for that and add
> "VIR_DOMAIN_SMARTCARD_DEFAULT_DATABASE"
> instead.
I remember testing the case with certificates inside /etc/pki/nssdb and
not specifying the <database/> element at the early version of the
patch.
>
> Furthermore actually this lets us define:
> <smartcard mode='host-certificates'>
> <certificate>cert1</certificate>
> <certificate>cert2</certificate>
> <certificate>cert3</certificate>
> <database>/etc/pki/nssdb/</database>
> </smartcard>
>
> There could be two guests using rather different cert[1-3] and there is
no
> need letting them cross read right?
> So instead of
> virBufferAsprintf(&buf, " \"%s/{,*}\" rk,\n", sc_db);
> maybe something like this is safer:
> iterate on certs-that-are-defined => cert
> virBufferAsprintf(&buf, " \"%s/%s\" rk,\n", sc_db,
cert);
Not sure to understand. The certificates are inside a SQLite database,
for instance:
$ ls /etc/pki/nssdb
cert9.db key4.db pkcs11.txt
How your proposed change would be valid, since there's no specific file
for the certificate (for instance /etc/pki/nssdb/cert1). Is apparmor
able to filter the access of the content of a SQLite database ?
Hmm - yeah I thought that would be files, but you are right "...provide
three NSS certificate names residing in a database..."
The reason I was looking for an alternative is that I'm just generally a
bit averse to most rules that end with *.
Once we went as far as generating them (instead of adding them to the
general template) you'd hope that we can make only explicit paths.
But it seems if people need separation they have to do so via the paths and
that is it.
Thanks Arnaud for the clarifications that helped me to get the nss details
I was missing.
Acked-by: Christian Ehrhardt <christian.ehrhardt(a)canonical.com>
Arnaud
--
Christian Ehrhardt
Staff Engineer, Ubuntu Server
Canonical Ltd