We cannot use rcu_dereference_bh safely in netpoll_rx as we may
be called with IRQs disabled. We could however simply disable
IRQs as that too causes BH to be disabled and is safe in either
case.
Thanks to John Linville for discovering this bug and providing
a patch.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
As xfrm_compile_policy runs within a read_lock, we cannot use
GFP_KERNEL for memory allocations.
Reported-by: Luca Tettamanti <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This reverts commit 8b9cfdca9c.
This patch needs to wait for the HWRNG API to start using void *
for priv first.
Signed-off-by: Herbert Xu <[email protected]>
On Thu, Aug 05, 2010 at 07:01:21PM -0700, Linus Torvalds wrote:
> On Thu, Aug 5, 2010 at 6:40 PM, Herbert Xu <[email protected]> wrote:
> >
> > -config CRYPTO_MANAGER_TESTS
> > - bool "Run algolithms' self-tests"
> > - default y
> > - depends on CRYPTO_MANAGER2
> > +config CRYPTO_MANAGER_DISABLE_TESTS
> > + bool "Disable run-time self tests"
> > + depends on CRYPTO_MANAGER2 && EMBEDDED
>
> Why do you still want to force-enable those tests? I was going to
> complain about the "default y" anyway, now I'm _really_ complaining,
> because you've now made it impossible to disable those tests. Why?
As requested, this patch sets the default to y and removes the
EMBEDDED dependency.
Signed-off-by: Herbert Xu <[email protected]>
This patch fixes a serious bug in the test disabling patch where
it can cause an spurious load of the cryptomgr module even when
it's compiled in.
It also negates the test disabling option so that its absence
causes tests to be enabled.
The Kconfig option is also now behind EMBEDDED.
Signed-off-by: Herbert Xu <[email protected]>
Currently you cannot disable multicast snooping while a device is
down. There is no good reason for this restriction and this patch
removes it.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
On the bridge TX path we're leaking an skb when br_multicast_rcv
returns an error.
Reported-by: David Lamparter <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Mark Wagner reported OOM symptoms when sending UDP traffic over
a macvtap link to a kvm receiver.
This appears to be caused by the fact that macvtap packet queues
are unlimited in length. This means that if the receiver can't
keep up with the rate of flow, then we will hit OOM. Of course
it gets worse if the OOM killer then decides to kill the receiver.
This patch imposes a cap on the packet queue length, in the same
way as the tuntap driver, using the device TX queue length.
Please note that macvtap currently has no way of giving congestion
notification, that means the software device TX queue cannot be
used and packets will always be dropped once the macvtap driver
queue fills up.
This shouldn't be a great problem for the scenario where macvtap
is used to feed a kvm receiver, as the traffic is most likely
external in origin so congestion notification can't be applied
anyway.
Of course, if anybody decides to complain about guest-to-guest
UDP packet loss down the track, then we may have to revisit this.
Incidentally, this patch also fixes a real memory leak when
macvtap_get_queue fails.
Chris Wright noticed that for this patch to work, we need a
non-zero TX queue length. This patch includes his work to change
the default macvtap TX queue length to 500.
Reported-by: Mark Wagner <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Acked-by: Chris Wright <[email protected]>
Acked-by: Arnd Bergmann <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The new netpoll code in bridging contains use-after-free bugs
that are non-trivial to fix.
This patch fixes this by removing the code that uses skbs after
they're freed.
As a consequence, this means that we can no longer call bridge
from the netpoll path, so this patch also removes the controller
function in order to disable netpoll.
Signed-off-by: Herbert Xu <[email protected]>
Thanks,
Signed-off-by: David S. Miller <[email protected]>
The bridge protocol lives dangerously by having incestuous relations
with the IP stack. In this instance an abomination has been created
where a bogus IPCB area from a bridged packet leads to a crash in
the IP stack because it's interpreted as IP options.
This patch papers over the problem by clearing the IPCB area in that
particular spot. To fix this properly we'd also need to parse any
IP options if present but I'm way too lazy for that.
Signed-off-by: Herbert Xu <[email protected]>
Cheers,
Signed-off-by: David S. Miller <[email protected]>
On Tue, Jul 06, 2010 at 08:48:35AM +0800, Herbert Xu wrote:
>
> bridge: Restore NULL check in br_mdb_ip_get
Resend with proper attribution.
bridge: Restore NULL check in br_mdb_ip_get
Somewhere along the line the NULL check in br_mdb_ip_get went
AWOL, causing crashes when we receive an IGMP packet with no
multicast table allocated.
This patch restores it and ensures all br_mdb_*_get functions
use it.
Reported-by: Frank Arnold <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Thanks,
Signed-off-by: David S. Miller <[email protected]>
It has been reported that the new UFO software fallback path
fails under certain conditions with NFS. I tracked the problem
down to the generation of UFO packets that are smaller than the
MTU. The software fallback path simply discards these packets.
This patch fixes the problem by not generating such packets on
the UFO path.
Signed-off-by: Herbert Xu <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch adds UFO to the list of GSO features with a software
fallback. This allows UFO to be used even if the hardware does
not support it.
In particular, this allows us to test the UFO fallback, as it
has been reported to not work in some cases.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Now that RCU debugging checks for matching rcu_dereference calls
and rcu_read_lock, we need to use the correct primitives or face
nasty warnings.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The version of br_netpoll_send_skb used when netpoll is off is
missing a const thus causing a warning.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The bridge multicast patches introduced an OOM crash in the forward
path, when deliver_clone fails to clone the skb.
Reported-by: Mark Wagner <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
There are multiple problems with the newly added netpoll support:
1) Use-after-free on each netpoll packet.
2) Invoking unsafe code on netpoll/IRQ path.
3) Breaks when netpoll is enabled on the underlying device.
This patch fixes all of these problems. In particular, we now
allocate proper netpoll structures for each underlying device.
We only allow netpoll to be enabled on the bridge when all the
devices underneath it support netpoll. Once it is enabled, we
do not allow non-netpoll devices to join the bridge (until netpoll
is disabled again).
This allows us to do away with the npinfo juggling that caused
problem number 1.
Incidentally this patch fixes number 2 by bypassing unsafe code
such as multicast snooping and netfilter.
Reported-by: Qianfeng Zhang <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch adds the helper netpoll_tx_running for use within
ndo_start_xmit. It returns non-zero if ndo_start_xmit is being
invoked by netpoll, and zero otherwise.
This is currently implemented by simply looking at the hardirq
count. This is because for all non-netpoll uses of ndo_start_xmit,
IRQs must be enabled while netpoll always disables IRQs before
calling ndo_start_xmit.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch adds the functions __netpoll_setup/__netpoll_cleanup
which is designed to be called recursively through ndo_netpoll_seutp.
They must be called with RTNL held, and the caller must initialise
np->dev and ensure that it has a valid reference count.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch adds ndo_netpoll_setup as the initialisation primitive
to complement ndo_netpoll_cleanup.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
As it stands, netpoll_setup and netpoll_cleanup have no locking
protection whatsoever. So chaos ensures if two entities try to
perform them on the same device.
This patch adds RTNL to the equation. The code has been rearranged so
that bits that do not need RTNL protection are now moved to the top of
netpoll_setup.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The use of RCU in netpoll is incorrect in a number of places:
1) The initial setting is lacking a write barrier.
2) The synchronize_rcu is in the wrong place.
3) Read barriers are missing.
4) Some places are even missing rcu_read_lock.
5) npinfo is zeroed after freeing.
This patch fixes those issues. As most users are in BH context,
this also converts the RCU usage to the BH variant.
Signed-off-by: Herbert Xu <[email protected]>
Acked-by: Paul E. McKenney <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Now that netpoll always zaps npinfo we no longer need to do it
in bridge.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Since we have to NULL npinfo regardless of whether there is a
ndo_netpoll_cleanup, it makes sense to do this unconditionally
in netpoll_cleanup rather than having every driver do it by
themselves.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The PCOMP Kconfig entry current allows the following combination
which is illegal:
ZLIB=y
PCOMP=y
ALGAPI=m
ALGAPI2=y
MANAGER=m
MANAGER2=m
This patch fixes this by adding PCOMP2 so that PCOMP can select
ALGAPI to propagate the setting to MANAGER2.
Signed-off-by: Herbert Xu <[email protected]>
Currently we disallow GSO packets on the IPv6 forward path.
This patch fixes this.
Note that I discovered that our existing GSO MTU checks (e.g.,
IPv4 forwarding) are buggy in that they skip the check altogether,
when they really should be checking gso_size + header instead.
I have also been lazy here in that I haven't bothered to segment
the GSO packet by hand before generating an ICMP message. Someone
should add that to be 100% correct.
Reported-by: Ralf Baechle <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
When the cls_cgroup module is not loaded, task_cls_classid will
return an uninitialised classid instead of zero.
Reported-by: Stephen Rothwell <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
There is a typo in cgroup_cls_state when cls_cgroup is built-in.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch makes tun update its socket classid every time we
inject a packet into the network stack. This is so that any
updates made by the admin to the process writing packets to
tun is effected.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Up until now cls_cgroup has relied on fetching the classid out of
the current executing thread. This runs into trouble when a packet
processing is delayed in which case it may execute out of another
thread's context.
Furthermore, even when a packet is not delayed we may fail to
classify it if soft IRQs have been disabled, because this scenario
is indistinguishable from one where a packet unrelated to the
current thread is processed by a real soft IRQ.
In fact, the current semantics is inherently broken, as a single
skb may be constructed out of the writes of two different tasks.
A different manifestation of this problem is when the TCP stack
transmits in response of an incoming ACK. This is currently
unclassified.
As we already have a concept of packet ownership for accounting
purposes in the skb->sk pointer, this is a natural place to store
the classid in a persistent manner.
This patch adds the cls_cgroup classid in struct sock, filling up
an existing hole on 64-bit :)
The value is set at socket creation time. So all sockets created
via socket(2) automatically gains the ID of the thread creating it.
Whenever another process touches the socket by either reading or
writing to it, we will change the socket classid to that of the
process if it has a valid (non-zero) classid.
For sockets created on inbound connections through accept(2), we
inherit the classid of the original listening socket through
sk_clone, possibly preceding the actual accept(2) call.
In order to minimise risks, I have not made this the authoritative
classid. For now it is only used as a backup when we execute
with soft IRQs disabled. Once we're completely happy with its
semantics we can use it as the sole classid.
Footnote: I have rearranged the error path on cls_group module
creation. If we didn't do this, then there is a window where
someone could create a tc rule using cls_group before the cgroup
subsystem has been registered.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
When GRO produces fraglist entries, and the resulting skb hits
an interface that is incapable of TSO but capable of FRAGLIST,
we end up producing a bogus packet with gso_size non-zero.
This was reported in the field with older versions of KVM that
did not set the TSO bits on tuntap.
This patch fixes that.
Reported-by: Igor Zhang <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The macro CRYPTO_MINALIGN is not meant to be used directly. This
patch replaces it with crypto_tfm_ctx_alignment.
Signed-off-by: Herbert Xu <[email protected]>
This patch ensures that all places that schedule the DAD timer
look at the address state in a safe manner before scheduling the
timer. This ensures that we don't end up with pending timers
after deleting an address.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch makes use of the new POSTDAD state. This prevents
a race between DAD completion and failure.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch makes use of the new state_lock to synchronise between
updates to the ifa state. This fixes the issue where a remotely
triggered address deletion (through DAD failure) coincides with a
local administrative address deletion, causing certain actions to
be performed twice incorrectly.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch replaces the boolean dead flag on inet6_ifaddr with
a state enum. This allows us to roll back changes when deleting
an address according to whether DAD has completed or not.
This patch only adds the state field and does not change the logic.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
When Steffen originally wrote the authenc async hash patch, he
correctly had EINPROGRESS checks in place so that we did not invoke
the original completion handler with it.
Unfortuantely I told him to remove it before the patch was applied.
As only MAY_BACKLOG request completion handlers are required to
handle EINPROGRESS completions, those checks are really needed.
This patch restores them.
Reported-by: Sebastian Andrzej Siewior <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
My recent patch to remove the open-coded checksum sequence in
tcp_v6_send_response broke it as we did not set the transport
header pointer on the new packet.
Actually, there is code there trying to set the transport
header properly, but it sets it for the wrong skb ('skb'
instead of 'buff').
This bug was introduced by commit
a8fdf2b331 ("ipv6: Fix
tcp_v6_send_response(): it didn't set skb transport header")
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
inet: Remove unused send_check length argument
This patch removes the unused length argument from the send_check
function in struct inet_connection_sock_af_ops.
Signed-off-by: Herbert Xu <[email protected]>
Tested-by: Yinghai <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
tcp: Handle CHECKSUM_PARTIAL for SYNACK packets for IPv6
This patch moves the common code between tcp_v6_send_check and
tcp_v6_gso_send_check into a new function __tcp_v6_send_check.
It then uses the new function in tcp_v6_send_synack as well as
tcp_v6_send_response so that they handle CHECKSUM_PARTIAL properly.
Signed-off-by: Herbert Xu <[email protected]>
Tested-by: Yinghai <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
tcp: Handle CHECKSUM_PARTIAL for SYNACK packets for IPv4
This patch moves the common code between tcp_v4_send_check and
tcp_v4_gso_send_check into a new function __tcp_v4_send_check.
It then uses the new function in tcp_v4_send_synack so that it
handles CHECKSUM_PARTIAL properly.
Signed-off-by: Herbert Xu <[email protected]>
Tested-by: Yinghai <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
As we will set ip_summed to CHECKSUM_NONE when necessary in
nfqnl_mangle, there is no need to zap CHECKSUM_COMPLETE in
nfqnl_build_packet_message.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: Patrick McHardy <[email protected]>
As we will set ip_summed to CHECKSUM_NONE when necessary in
ipq_mangle_ipv6, there is no need to zap CHECKSUM_COMPLETE in
ipq_build_packet_message.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: Patrick McHardy <[email protected]>
While doing yet another audit on ip_summed I noticed ip_queue
calling skb_checksum_help unnecessarily. As we will set ip_summed
to CHECKSUM_NONE when necessary in ipq_mangle_ipv4, there is no
need to zap CHECKSUM_COMPLETE in ipq_build_packet_message.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: Patrick McHardy <[email protected]>
The IGMP3 report parsing is looking at the wrong address for
group records. This patch fixes it.
Reported-by: Banyeer <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The xfrm state genid only needs to be matched against the copy
saved in xfrm_dst. So we don't need a global genid at all. In
fact, we don't even need to initialise it.
Based on observation by Timo Teräs.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
ide-atapi requeues requests without holding the queue lock.
This patch fixes it by using ide_requeue_and_plug.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
I noticed that my KVM virtual machines were experiencing IDE
issues resulting in processes stuck on waiting for buffers to
complete.
The root cause is of course race conditions in the ancient qemu
backend that I'm using. However, the fact that the guest isn't
recovering is a bug.
I've tracked it down to the change made last year to dequeue
requests at the start rather than at the end in the IDE layer.
commit 8f6205cd57
Author: Tejun Heo <[email protected]>
Date: Fri May 8 11:53:59 2009 +0900
ide: dequeue in-flight request
The problem is that the function ide_dma_timeout_retry does not
requeue the current request, causing one request to be lost for
each DMA timeout.
This patch fixes this by requeueing the request.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
As the only path leading to ip6_dst_check makes an indirect call
through dst->ops, dst cannot be NULL in ip6_dst_check.
This patch removes this check in case it misleads people who
come across this code.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Since all callers of br_mdb_ip_get need to check whether the
hash table is NULL, this patch moves the check into the function.
This fixes the two callers (query/leave handler) that didn't
check it.
Reported-by: Michael Braun <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
If we are managing IPv6 addresses using DHCP, it would be nice
for user-space to be notified if an address configured through
DHCP fails DAD. Otherwise user-space would have to poll to see
whether DAD succeeds.
This patch uses the existing notification mechanism and simply
hooks it into the DAD failure code path.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Thanks to Paul McKenny for pointing out that it is incorrect to use
synchronize_rcu_bh to ensure that pending callbacks have completed.
Instead we should use rcu_barrier_bh.
Reported-by: Paul E. McKenney <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
As Paul McKenney correctly pointed out, __br_mdb_ip_get needs
to use the RCU list walking primitive in order to work correctly
on platforms where data-dependency ordering is not guaranteed.
Reported-by: Paul E. McKenney <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
When I merged the bundle creation code, I introduced a bogus
flowi value in the bundle. Instead of getting from the caller,
it was instead set to the flow in the route object, which is
totally different.
The end result is that the bundles we created never match, and
we instead end up with an ever growing bundle list.
Thanks to Jamal for find this problem.
Reported-by: Jamal Hadi Salim <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Acked-by: Steffen Klassert <[email protected]>
Acked-by: Jamal Hadi Salim <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
USB: Move hcd free_dev call into usb_disconnect
I found a way to oops the kernel:
1. Open a USB device through devio.
2. Remove the hcd module in the host kernel.
3. Close the devio file descriptor.
The problem is that closing the file descriptor does usb_release_dev
as it is the last reference. usb_release_dev then tries to invoke
the hcd free_dev function (or rather dereferencing the hcd driver
struct). This causes an oops as the hcd driver has already been
unloaded so the struct is gone.
This patch tries to fix this by bringing the free_dev call earlier
and into usb_disconnect. I have verified that repeating the
above steps no longer crashes with this patch applied.
Signed-off-by: Herbert Xu <[email protected]>
Cc: stable <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>
As md5 now has export/import functions, it must set the attribute
statesize. Otherwise anything that relies on import/export may
fail as they will see a zero statesize.
Signed-off-by: Herbert Xu <[email protected]>
This patch allows the user to the IGMP parameters related to the
snooping function of the bridge. This includes various time
values and retransmission limits.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch allows the user to control the hash elasticity/max
parameters. The elasticity setting does not take effect until
the next new multicast group is added. At which point it is
checked and if after rehashing it still can't be satisfied then
snooping will be disabled.
The max setting on the other hand takes effect immediately. It
must be a power of two and cannot be set to a value less than the
current number of multicast group entries. This is the only way
to shrink the multicast hash.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch allows the user to disable IGMP snooping completely
through a sysfs toggle. It also allows the user to reenable
snooping when it has been automatically disabled due to hash
collisions. If the collisions have not been resolved however
the system will refuse to reenable snooping.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch allows the user to forcibly enable/disable ports as
having multicast routers attached. A port with a multicast router
will receive all multicast traffic.
The value 0 disables it completely. The default is 1 which lets
the system automatically detect the presence of routers (currently
this is limited to picking up queries), and 2 means that the port
will always receive all multicast traffic.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch finally hooks up the multicast snooping module to the
data path. In particular, all multicast packets passing through
the bridge are fed into the module and switched by it.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch hooks up the bridge start/stop and add/delete/disable
port functions to the new multicast module.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch adds code to perform selective multicast forwarding.
We forward multicast traffic to a set of ports plus all multicast
router ports. In order to avoid duplications among these two
sets of ports, we order all ports by the numeric value of their
pointers. The two lists are then walked in lock-step to eliminate
duplicates.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch adds the core functionality of IGMP snooping support
without actually hooking it up. So this patch should be a no-op
as far as the bridge's external behaviour is concerned.
All the new code and data is controlled by the Kconfig option
BRIDGE_IGMP_SNOOPING. A run-time toggle is also available.
The multicast switching is done using an hash table that is
lockless on the read-side through RCU. On the write-side the
new multicast_lock is used for all operations. The hash table
supports dynamic growth/rehashing.
The hash table will be rehashed if any chain length exceeds a
preset limit. If rehashing does not reduce the maximum chain
length then snooping will be disabled.
These features may be added in future (in no particular order):
* IGMPv3 source support
* Non-querier router detection
* IPv6
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch moves the main loop body in br_flood into the function
may_deliver. The code that clones an skb and delivers it is moved
into the deliver_clone function.
This allows this to be reused by the future multicast forward
function.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
this patch makes BR_INPUT_SKB_CB available on the xmit path so
that we could avoid passing the br pointer around for the purpose
of collecting device statistics.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
When the packet is delivered to the local bridge device we may
end up cloning it unnecessarily if no bridge port can receive
the packet in br_flood.
This patch avoids this by moving the skb_clone into br_flood.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch allows tail-call on the call to br_pass_frame_up
in br_handle_frame_finish. This is now possible because of the
previous patch to call br_pass_frame_up last.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
At the moment we deliver to the local bridge port via the function
br_pass_frame_up before all other ports. There is no requirement
for this.
For the purpose of IGMP snooping, it would be more convenient if
we did the local port last. Therefore this patch rearranges the
bridge input processing so that the local bridge port gets to see
the packet last (if at all).
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
xfrm_state_clone calls kfree instead of xfrm_state_put to free
a failed state. Depending on the state of the failed state, it
can cause leaks to things like module references.
All states should be freed by xfrm_state_put past the point of
xfrm_init_state.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
When ipcomp_tunnel_attach fails we will call ipcomp_destroy twice.
This may lead to double-frees on certain structures.
As there is no reason to explicitly call ipcomp_destroy, this patch
removes it from ipcomp*.c and lets the standard xfrm_state destruction
take place.
This is based on the discovery and patch by Alexey Dobriyan.
Tested-by: Alexey Dobriyan <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Currently we treat IGMPv3 reports as if it were an IGMPv2/v1 report.
This is broken as IGMPv3 reports are formatted differently. So we
end up suppressing a bogus multicast group (which should be harmless
as long as the leading reserved field is zero).
In fact, IGMPv3 does not allow membership report suppression so
we should simply ignore IGMPv3 membership reports as a host.
This patch does exactly that. I kept the case statement for it
so people won't accidentally add it back thinking that we overlooked
this case.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch removes the unused age_list member from the net_bridge
structure.
Signed-off-by: Herbert Xu <[email protected]>
Acked-by: Stephen Hemminger <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The previous changeset left behind an unused inode variable.
This patch removes it.
Reported-by: Stephen Rothwell <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
As the padlock driver for SHA uses a software fallback to perform
partial hashing, it must implement custom import/export functions.
Otherwise hmac which depends on import/export for prehashing will
not work with padlock-sha.
Reported-by: Wolfgang Walter <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
I have seen RX stalls on a machine that experienced a suspected
OOM. After the stall, the RX buffer is empty on the guest side
and there are exactly 16 entries available on the host side. As
the number of entries is less than that required by a maximal
skb, the host cannot proceed.
The guest did not have a refill job scheduled.
My diagnosis is that an OOM had occured, with the delayed refill
job scheduled. The job was able to allocate at least one skb, but
not enough to overcome the minimum required by the host to proceed.
As the refill job would only reschedule itself if it failed completely
to allocate any skbs, this would lead to an RX stall.
The following patch removes this stall possibility by always
rescheduling the refill job until the ring is totally refilled.
Testing has shown that the RX stall no longer occurs whereas
previously it would occur within a day.
Signed-off-by: Herbert Xu <[email protected]>
Acked-by: Rusty Russell <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
When the loop terminates with size == 0 in rng_dev_read we will
unlock the rng mutex twice.
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
When we've merged skb's with page frags, and subsequently receive
a trailer skb (< MSS) that is not completely non-linear (this can
occur on Intel NICs if the packet size falls below the threshold),
GRO ends up producing an illegal GSO skb with a frag_list.
This is harmless unless the skb is then forwarded through an
interface that requires software GSO, whereupon the GSO code
will BUG.
This patch detects this case in GRO and avoids merging the
trailer skb.
Reported-by: Mark Wagner <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch rearranges the SIT DF bit handling using the new IPIP DF
code. The only externally visible effect should be the case where
PMTU is enabled and the MTU is exactly 1280 bytes. In this case the
previous code would send packets out with DF off while the new code
would set the DF bit. This is inline with RFC 4213.
Signed-off-by: Herbert Xu <[email protected]>
Thanks,
Signed-off-by: David S. Miller <[email protected]>
RFC 2003 requires the outer header to have DF set if DF is set
on the inner header, even when PMTU discovery is off for the
tunnel. Our implementation does exactly that.
For this to work properly the IPIP gateway also needs to engate
in PMTU when the inner DF bit is set. As otherwise the original
host would not be able to carry out its PMTU successfully since
part of the path is only visible to the gateway.
Unfortunately when the tunnel PMTU discovery setting is off, we
do not collect the necessary soft state, resulting in blackholes
when the original host tries to perform PMTU discovery.
This problem is not reproducible on the IPIP gateway itself as
the inner packet usually has skb->local_df set. This is not
correctly cleared (an unrelated bug) when the packet passes
through the tunnel, which allows fragmentation to occur. For
hosts behind the IPIP gateway it is readily visible with a simple
ping.
This patch fixes the problem by performing PMTU discovery for
all packets with the inner DF bit set, regardless of the PMTU
discovery setting on the tunnel itself.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Old gases don't have a clue what pshufb stands for so we have
to hard-code it for now.
Reported-by: Andrew Morton <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Nathan Neulinger noticed that gretap devices get their MAC address
from the local IP address, which results in invalid MAC addresses
half of the time.
This is because gretap is still using the tunnel netdev ops rather
than the correct tap netdev ops struct.
This patch also fixes changelink to not clobber the MAC address
for the gretap case.
Signed-off-by: Herbert Xu <[email protected]>
Acked-by: Stephen Hemminger <[email protected]>
Tested-by: Nathan Neulinger <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
This patch tries to print out more information when we hit the
MSG_PEEK bug in tcp_recvmsg. It's been around since at least
2005 and it's about time that we finally fix it.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The PadLock hardware requires the output buffer for SHA to be
128-bit aligned. We currentply place the buffer on the stack,
and ask gcc to align it to 128 bits. That doesn't work on i386
because the kernel stack is only aligned to 32 bits. This patch
changes the code to align the buffer by hand so that the hardware
doesn't fault on unaligned buffers.
Reported-by: Séguier Régis <[email protected]>
Tested-by: Séguier Régis <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
We have a mechanism where newly registered algorithms of a higher
priority can displace existing instances that use a different
implementation of the same algorithm with a lower priority.
Unfortunately the same mechanism can cause a newly registered
algorithm to displace itself if it depends on an existing version
of the same algorithm.
This patch fixes this by keeping all algorithms that the newly
reigstered algorithm depends on, thus protecting them from being
removed.
Signed-off-by: Herbert Xu <[email protected]>
As struct skcipher_givcrypt_request includes struct crypto_request
at a non-zero offset, testing for NULL after converting the pointer
returned by crypto_dequeue_request does not work. This can result
in IPsec crashes when the queue is depleted.
This patch fixes it by doing the pointer conversion only when the
return value is non-NULL. In particular, we create a new function
__crypto_dequeue_request that does the pointer conversion.
Reported-by: Brad Bosch <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Recently we switched to using eseqiv on SMP machines in preference
over chainiv. However, eseqiv does not support stream ciphers so
they should still default to chainiv.
This patch applies the same check as done by eseqiv to weed out
the stream ciphers. In particular, all algorithms where the IV
size is not equal to the block size will now default to chainiv.
Signed-off-by: Herbert Xu <[email protected]>
Raw counter mode only works with chainiv, which is no longer
the default IV generator on SMP machines. This broke raw counter
mode as it can no longer instantiate as a givcipher.
This patch fixes it by always picking chainiv on raw counter
mode. This is based on the diagnosis and a patch by Huang
Ying.
Signed-off-by: Herbert Xu <[email protected]>
This reverts commit 215ccd6f55.
It causes CPRNG and everything selected by it to be built-in
whenever FIPS is enabled. The problem is that it is selecting
a tristate from a bool, which is usually not what is intended.
Signed-off-by: Herbert Xu <[email protected]>
As it is, parts of the ioctl runs under the RTNL and parts of
it do not. The unlocked section is still protected by the BKL,
but there can be subtle races. For example, Eric Biederman and
Paul Moore observed that if two threads tried to create two tun
devices on the same file descriptor, then unexpected results
may occur.
As there isn't anything in the ioctl that is expected to sleep
indefinitely, we can prevent this from occurring by extending
the RTNL lock coverage.
This also allows to get rid of the BKL.
Finally, I changed tun_get_iff to take a tun device in order to
avoid calling tun_put which would dead-lock as it also tries to
take the RTNL lock.
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
The aligned ctx helper was using a bogus alignment value thas was
one off the correct value. Fortunately the current users do not
require anything beyond the natural alignment of the platform so
this hasn't caused a problem.
This patch fixes that and also removes the unnecessary minimum
check since if the alignment is less than the natural alignment
then the subsequent ALIGN operation should be a noop.
Signed-off-by: Herbert Xu <[email protected]>
This patch uses crypto_shash_export/crypto_shash_import to prehash
ipad/opad to speed up hmac. This is partly based on a similar patch
by Steffen Klassert.
Signed-off-by: Herbert Xu <[email protected]>
This patch provides a default export/import function for all
shash algorithms. It simply copies the descriptor context as
is done by sha1_generic.
This in essence means that all existing shash algorithms now
support export/import. This is something that will be depended
upon in implementations such as hmac. Therefore all new shash
and ahash implementations must support export/import.
For those that cannot obtain a partial result, padlock-sha's
fallback model should be used so that a partial result is always
available.
Signed-off-by: Herbert Xu <[email protected]>
This patch adds export/import support to sha512-s390 (which includes
sha384-s390). The exported type is defined by struct sha512_state,
which is basically the entire descriptor state of sha512_generic.
Since sha512-s390 only supports a 64-bit byte count the import
function will reject anything that exceeds that.
Signed-off-by: Herbert Xu <[email protected]>
This patch replaces the 32-bit counters in sha512_generic with
64-bit counters. It also switches the bit count to the simpler
byte count.
Signed-off-by: Herbert Xu <[email protected]>
This patch renames struct sha512_ctx and exports it as struct
sha512_state so that other sha512 implementations can use it
as the reference structure for exporting their state.
Signed-off-by: Herbert Xu <[email protected]>
Herbert Xu