| Message ID | 20241117090514.9386-1-stsp2@yandex.ru (mailing list archive) |
|---|---|
| State | Deferred |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next] tun: fix group permission check | expand |
Stas Sergeev wrote: > Currently tun checks the group permission even if the user have matched. > Besides going against the usual permission semantic, this has a > very interesting implication: if the tun group is not among the > supplementary groups of the tun user, then effectively no one can > access the tun device. CAP_SYS_ADMIN still can, but its the same as > not setting the tun ownership. > > This patch relaxes the group checking so that either the user match > or the group match is enough. This avoids the situation when no one > can access the device even though the ownership is properly set. > > Also I simplified the logic by removing the redundant inversions: > tun_not_capable() --> !tun_capable() > > Signed-off-by: Stas Sergeev <stsp2@yandex.ru> This behavior goes back through many patches to commit 8c644623fe7e: [NET]: Allow group ownership of TUN/TAP devices. Introduce a new syscall TUNSETGROUP for group ownership setting of tap devices. The user now is allowed to send packages if either his euid or his egid matches the one specified via tunctl (via -u or -g respecitvely). If both, gid and uid, are set via tunctl, both have to match. The choice evidently was on purpose. Even if indeed non-standard.
Willem de Bruijn wrote: > Stas Sergeev wrote: > > Currently tun checks the group permission even if the user have matched. > > Besides going against the usual permission semantic, this has a > > very interesting implication: if the tun group is not among the > > supplementary groups of the tun user, then effectively no one can > > access the tun device. CAP_SYS_ADMIN still can, but its the same as > > not setting the tun ownership. > > > > This patch relaxes the group checking so that either the user match > > or the group match is enough. This avoids the situation when no one > > can access the device even though the ownership is properly set. > > > > Also I simplified the logic by removing the redundant inversions: > > tun_not_capable() --> !tun_capable() > > > > Signed-off-by: Stas Sergeev <stsp2@yandex.ru> > > This behavior goes back through many patches to commit 8c644623fe7e: > > [NET]: Allow group ownership of TUN/TAP devices. > > Introduce a new syscall TUNSETGROUP for group ownership setting of tap > devices. The user now is allowed to send packages if either his euid or > his egid matches the one specified via tunctl (via -u or -g > respecitvely). If both, gid and uid, are set via tunctl, both have to > match. > > The choice evidently was on purpose. Even if indeed non-standard. I should clarify that I'm not against bringing this file in line with normal user/group behavior. Just want to give anyone a chance to speak up if they disagree and/or recall why the code was originally written as it is.
17.11.2024 18:04, Willem de Bruijn пишет: > Stas Sergeev wrote: >> Currently tun checks the group permission even if the user have matched. >> Besides going against the usual permission semantic, this has a >> very interesting implication: if the tun group is not among the >> supplementary groups of the tun user, then effectively no one can >> access the tun device. CAP_SYS_ADMIN still can, but its the same as >> not setting the tun ownership. >> >> This patch relaxes the group checking so that either the user match >> or the group match is enough. This avoids the situation when no one >> can access the device even though the ownership is properly set. >> >> Also I simplified the logic by removing the redundant inversions: >> tun_not_capable() --> !tun_capable() >> >> Signed-off-by: Stas Sergeev <stsp2@yandex.ru> > This behavior goes back through many patches to commit 8c644623fe7e: > > [NET]: Allow group ownership of TUN/TAP devices. > > Introduce a new syscall TUNSETGROUP for group ownership setting of tap > devices. The user now is allowed to send packages if either his euid or > his egid matches the one specified via tunctl (via -u or -g > respecitvely). If both, gid and uid, are set via tunctl, both have to > match. > > The choice evidently was on purpose. Even if indeed non-standard. So what would you suggest? Added Guido Guenther <agx@sigxcpu.org> to CC for an opinion. The main problem here is that by setting user and group properly, you end up with device inaccessible by anyone, unless the user belongs to the tun group. I don't think someone wants to set up inaccessible devices, so this property doesn't seem useful. OTOH if the user does have that group in his list, then, AFAICT, adding such group to tun changes nothing: neither limits nor extends the scope. If you had group already set and you set also user, then you limit the scope, but its the same as just setting user alone. So I really can't think of any valid usage scenario of setting both tun user and tun group.
On 11/18/24 22:40, Willem de Bruijn wrote: > Willem de Bruijn wrote: >> Stas Sergeev wrote: >>> Currently tun checks the group permission even if the user have matched. >>> Besides going against the usual permission semantic, this has a >>> very interesting implication: if the tun group is not among the >>> supplementary groups of the tun user, then effectively no one can >>> access the tun device. CAP_SYS_ADMIN still can, but its the same as >>> not setting the tun ownership. >>> >>> This patch relaxes the group checking so that either the user match >>> or the group match is enough. This avoids the situation when no one >>> can access the device even though the ownership is properly set. >>> >>> Also I simplified the logic by removing the redundant inversions: >>> tun_not_capable() --> !tun_capable() >>> >>> Signed-off-by: Stas Sergeev <stsp2@yandex.ru> >> >> This behavior goes back through many patches to commit 8c644623fe7e: >> >> [NET]: Allow group ownership of TUN/TAP devices. >> >> Introduce a new syscall TUNSETGROUP for group ownership setting of tap >> devices. The user now is allowed to send packages if either his euid or >> his egid matches the one specified via tunctl (via -u or -g >> respecitvely). If both, gid and uid, are set via tunctl, both have to >> match. >> >> The choice evidently was on purpose. Even if indeed non-standard. > > I should clarify that I'm not against bringing this file in line with > normal user/group behavior. > > Just want to give anyone a chance to speak up if they disagree and/or > recall why the code was originally written as it is. I think we can't accept a behaviour changing patch this late in the cycle. If an agreement is reached it should be reposted after the merge window. /P
19.11.2024 13:51, Paolo Abeni пишет: > I think we can't accept a behaviour changing patch this late in the > cycle. If an agreement is reached it should be reposted after the merge > window. > > /P Noted, thanks.
stsp wrote: > 17.11.2024 18:04, Willem de Bruijn пишет: > > Stas Sergeev wrote: > >> Currently tun checks the group permission even if the user have matched. > >> Besides going against the usual permission semantic, this has a > >> very interesting implication: if the tun group is not among the > >> supplementary groups of the tun user, then effectively no one can > >> access the tun device. CAP_SYS_ADMIN still can, but its the same as > >> not setting the tun ownership. > >> > >> This patch relaxes the group checking so that either the user match > >> or the group match is enough. This avoids the situation when no one > >> can access the device even though the ownership is properly set. > >> > >> Also I simplified the logic by removing the redundant inversions: > >> tun_not_capable() --> !tun_capable() > >> > >> Signed-off-by: Stas Sergeev <stsp2@yandex.ru> > > This behavior goes back through many patches to commit 8c644623fe7e: > > > > [NET]: Allow group ownership of TUN/TAP devices. > > > > Introduce a new syscall TUNSETGROUP for group ownership setting of tap > > devices. The user now is allowed to send packages if either his euid or > > his egid matches the one specified via tunctl (via -u or -g > > respecitvely). If both, gid and uid, are set via tunctl, both have to > > match. > > > > The choice evidently was on purpose. Even if indeed non-standard. > > So what would you suggest? > Added Guido Guenther <agx@sigxcpu.org> to CC > for an opinion. > The main problem here is that by > setting user and group properly, you > end up with device inaccessible by > anyone, unless the user belongs to > the tun group. I don't think someone > wants to set up inaccessible devices, > so this property doesn't seem useful. > OTOH if the user does have that group > in his list, then, AFAICT, adding such > group to tun changes nothing: neither > limits nor extends the scope. > If you had group already set and you > set also user, then you limit the scope, > but its the same as just setting user alone. > So I really can't think of any valid usage > scenario of setting both tun user and tun > group. Understood. If no one comments before the window reopens, I think it's fine to just resubmit.
diff --git a/drivers/net/tun.c b/drivers/net/tun.c index 9a0f6eb32016..d35b6a48d138 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c @@ -574,14 +574,18 @@ static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, return ret; } -static inline bool tun_not_capable(struct tun_struct *tun) +static inline bool tun_capable(struct tun_struct *tun) { const struct cred *cred = current_cred(); struct net *net = dev_net(tun->dev); - return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || - (gid_valid(tun->group) && !in_egroup_p(tun->group))) && - !ns_capable(net->user_ns, CAP_NET_ADMIN); + if (ns_capable(net->user_ns, CAP_NET_ADMIN)) + return 1; + if (uid_valid(tun->owner) && uid_eq(cred->euid, tun->owner)) + return 1; + if (gid_valid(tun->group) && in_egroup_p(tun->group)) + return 1; + return 0; } static void tun_set_real_num_queues(struct tun_struct *tun) @@ -2778,7 +2782,7 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) !!(tun->flags & IFF_MULTI_QUEUE)) return -EINVAL; - if (tun_not_capable(tun)) + if (!tun_capable(tun)) return -EPERM; err = security_tun_dev_open(tun->security); if (err < 0)
Currently tun checks the group permission even if the user have matched. Besides going against the usual permission semantic, this has a very interesting implication: if the tun group is not among the supplementary groups of the tun user, then effectively no one can access the tun device. CAP_SYS_ADMIN still can, but its the same as not setting the tun ownership. This patch relaxes the group checking so that either the user match or the group match is enough. This avoids the situation when no one can access the device even though the ownership is properly set. Also I simplified the logic by removing the redundant inversions: tun_not_capable() --> !tun_capable() Signed-off-by: Stas Sergeev <stsp2@yandex.ru> CC: Willem de Bruijn <willemdebruijn.kernel@gmail.com> CC: Jason Wang <jasowang@redhat.com> CC: Andrew Lunn <andrew+netdev@lunn.ch> CC: "David S. Miller" <davem@davemloft.net> CC: Eric Dumazet <edumazet@google.com> CC: Jakub Kicinski <kuba@kernel.org> CC: Paolo Abeni <pabeni@redhat.com> CC: netdev@vger.kernel.org CC: linux-kernel@vger.kernel.org --- drivers/net/tun.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-)