Xen Security Advisory 90 - Linux netback crash trying to disable due to malformed packet
Xen.org security team <[email protected]>
| Newsgroups | gmane.comp.emulators.xen.announce |
|---|---|
| Message-ID | <E1WS4V3-0001Xh-Se__5922.7590951652$1395667279$gmane$org@xenbits.xen.org> |
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Xen Security Advisory XSA-90
Linux netback crash trying to disable due to malformed packet
ISSUE DESCRIPTION
=================
When Linux's netback sees a malformed packet, it tries to disable the
interface which serves the misbehaving frontend.
This involves taking a mutex, which might sleep. But in recent
versions of Linux the guest transmit path is handled by NAPI in
softirq context, where sleeping is not allowed. The end result is
that the backend domain (often, Dom0) crashes with "scheduling while
atomic".
IMPACT
======
Malicious guest administrators can cause denial of service. If driver
domains are not in use, the impact is a host crash.
VULNERABLE SYSTEMS
==================
This bug affects systems using Linux as the driver domain, including
non-disaggregated systems using Linux as dom0.
Only versions of Linux whose netback uses NAPI are affected. In Linux
mainline this is all versions of Linux containing git changeset
b3f980bd82, which was introduced between Linux 3.11 and 3.12-rc1.
Systems using a different OS as dom0 (eg, NetBSD, Solaris) are not
vulnerable.
Both x86 and ARM systems are affected.
MITIGATION
==========
Using driver domains may limit the scope of the denial of service, and
may make it possible to resume service without restarting guests (by
restarting the driver domain). Advice on reconfiguring a system to
use driver domains is beyond the reasonable scope of this advisory.
In the case of an x86 HVM guest, the exploit can be prevented by
disabling the PV IO paths; normally this would come with a substantial
performance cost, and it may involve reconfiguring the guest as well
as the host. This is not recommended.
NOTE REGARDING LACK OF EMBARGO
==============================
This bug was publicly reported on xen-devel, before it was appreciated
that there was a security problem. The public mailing list thread
nevertheless contains information strongly suggestive of a security
bug, and a different security bug (with CVE) is suggested as seeming
"similar".
For these reasons we (the Xen Project Security Team) have concluded
that the presence of this bug, as a security problem, is not (any
longer) a secret.
CREDITS
=======
This issue was discovered as a bug by Török Edwin and analysed by
Wei Liu of Citrix.
RESOLUTION
==========
Applying the appropriate attached patch resolves this issue.
$ sha256sum xsa90*.patch
07341ffb7f577d32510602797a08009eade817009b425a124413ee743bdb6f05 xsa90.patch
$
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
iQEcBAEBAgAGBQJTMCxRAAoJEIP+FMlX6CvZaAEIAIIVfNdz3CwFYbiSwa51RJ3L
YFarP71/0EjNJKSaRwf6EQjDNnApqq6ep4+WKFvlMbm515jyQXp6mAbb8ffqnLUQ
2SDOlQXOpbnZrJrgo4YcT5ru8ZusauYz36TkFVcXBmcKWq29KoUARo5zG7YGyh9H
aWajaZs6RQPv3QE8IInNSP0oitRQZg/5xAW+Lz4Kn8xpO/IJuYW3ROH6JQcFF67H
r7xVAzxjrNQ3P5mN0iiOkQYK39PqhwGUhWaa6JlejsjUgU1nsGIBOHH+ISCaZrtL
e/6XK3awaDiu1dAL4Py1SdhPiA0sTeqA3bf6ARd7ymoIFqGuxrqYlupcUKTupjE=
=LrLN
-----END PGP SIGNATURE-----
_______________________________________________
Xen-announce mailing list
[email protected]
http://lists.xen.org/xen-announce
xsa90.patch
(application/octet-stream, 4 KB)
From: Wei Liu <[email protected]> Date: Mon, 17 Mar 2014 11:52:53 +0000 Subject: [PATCH RFC] xen-netback: disable rogue vif in kthread context When netback discovers frontend is sending malformed packet it will disables the interface which serves that frontend. However disabling a network interface involving taking a mutex which cannot be done in softirq context, so we need to defer this process to kthread context. This patch does the following: 1. introduce a flag to indicate the interface is disabled. 2. check that flag in TX path, don't do any work if it's true. 3. check that flag in RX path, turn off that interface if it's true. The reason to disable it in RX path is because RX uses kthread. After this change the behavior of netback is still consistent -- it won't do any TX work for a rogue frontend, and the interface will be eventually turned off. Also change a "continue" to "break" after xenvif_fatal_tx_err, as it doesn't make sense to continue processing packets if frontend is rogue. Signed-off-by: Wei Liu <[email protected]> Acked-by: Ian Campbell <[email protected]> --- drivers/net/xen-netback/common.h | 5 +++++ drivers/net/xen-netback/interface.c | 9 +++++++++ drivers/net/xen-netback/netback.c | 14 ++++++++++++-- 3 files changed, 26 insertions(+), 2 deletions(-) diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-netback/common.h index ae413a2..4bf5b33 100644 --- a/drivers/net/xen-netback/common.h +++ b/drivers/net/xen-netback/common.h @@ -113,6 +113,11 @@ struct xenvif { domid_t domid; unsigned int handle; + /* Is this interface disabled? True when backend discovers + * frontend is rogue. + */ + bool disabled; + /* Use NAPI for guest TX */ struct napi_struct napi; /* When feature-split-event-channels = 0, tx_irq = rx_irq. */ diff --git a/drivers/net/xen-netback/interface.c b/drivers/net/xen-netback/interface.c index 301cc03..234f1c8 100644 --- a/drivers/net/xen-netback/interface.c +++ b/drivers/net/xen-netback/interface.c @@ -62,6 +62,13 @@ static int xenvif_poll(struct napi_struct *napi, int budget) struct xenvif *vif = container_of(napi, struct xenvif, napi); int work_done; + /* This vif is rogue, we pretend we've used up all budget to + * deschedule it from NAPI. But this interface will be turned + * off in thread context later. + */ + if (unlikely(vif->disabled)) + return budget; + work_done = xenvif_tx_action(vif, budget); if (work_done < budget) { @@ -321,6 +328,8 @@ struct xenvif *xenvif_alloc(struct device *parent, domid_t domid, vif->ip_csum = 1; vif->dev = dev; + vif->disabled = false; + vif->credit_bytes = vif->remaining_credit = ~0UL; vif->credit_usec = 0UL; init_timer(&vif->credit_timeout); diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c index 438d0c0..94e7261 100644 --- a/drivers/net/xen-netback/netback.c +++ b/drivers/net/xen-netback/netback.c @@ -655,7 +655,7 @@ static void xenvif_tx_err(struct xenvif *vif, static void xenvif_fatal_tx_err(struct xenvif *vif) { netdev_err(vif->dev, "fatal error; disabling device\n"); - xenvif_carrier_off(vif); + vif->disabled = true; } static int xenvif_count_requests(struct xenvif *vif, @@ -1126,7 +1126,7 @@ static unsigned xenvif_tx_build_gops(struct xenvif *vif, int budget) vif->tx.sring->req_prod, vif->tx.req_cons, XEN_NETIF_TX_RING_SIZE); xenvif_fatal_tx_err(vif); - continue; + break; } work_to_do = RING_HAS_UNCONSUMED_REQUESTS(&vif->tx); @@ -1549,6 +1549,16 @@ int xenvif_kthread(void *data) wait_event_interruptible(vif->wq, rx_work_todo(vif) || kthread_should_stop()); + + /* This frontend is found to be rogue, disable it in + * kthread context. Currently this is only set when + * netback finds out frontend sends malformed packet, + * but we cannot disable the interface in softirq + * context so we defer it here. + */ + if (unlikely(vif->disabled && netif_carrier_ok(vif->dev))) + xenvif_carrier_off(vif); + if (kthread_should_stop()) break; -- 1.7.10.4