Skip to content

cloud-kernel
cloud-kernel

openanolis / cloud-kernel
1 年多 前同步成功

通知 160
Star 36
Fork 7
cloud-kernel
cloud-kernel
切换分支/标签
  1. 05 2月, 2013 6 次提交
    • D
      sctp: sctp_close: fix release of bindings for deferred call_rcu's · 8c98653f
      由 Daniel Borkmann 提交于 
      It seems due to RCU usage, i.e. within SCTP's address binding list,
a, say, ``behavioral change'' was introduced which does actually
not conform to the RFC anymore. In particular consider the following
(fictional) scenario to demonstrate this:

  do:
    Two SOCK_SEQPACKET-style sockets are opened (S1, S2)
    S1 is bound to 127.0.0.1, port 1024 [server]
    S2 is bound to 127.0.0.1, port 1025 [client]
    listen(2) is invoked on S1
    From S2 we call one sendmsg(2) with msg.msg_name and
       msg.msg_namelen parameters set to the server's
       address
    S1, S2 are closed
    goto do

The first pass of this loop passes successful, while the second round
fails during binding of S1 (address still in use). What is happening?
In the first round, the initial handshake is being done, and, at the
time close(2) is called on S1, a non-graceful shutdown is performed via
ABORT since in S1's receive queue an unprocessed packet is present,
thus stating an error condition. This can be considered as a correct
behavior.

During close also all bound addresses are freed, thus nothing *must*
be active anymore. In reference to RFC2960:

  After checking the Verification Tag, the receiving endpoint shall
  remove the association from its record, and shall report the
  termination to its upper layer. (9.1 Abort of an Association)

Also, no half-open states are supported, thus after an ungraceful
shutdown, we leave nothing behind. However, this seems not to be
happening though. In a real-world scenario, this is exactly where
it breaks the lksctp-tools functional test suite, *for instance*:

  ./test_sockopt
  test_sockopt.c  1 PASS : getsockopt(SCTP_STATUS) on a socket with no assoc
  test_sockopt.c  2 PASS : getsockopt(SCTP_STATUS)
  test_sockopt.c  3 PASS : getsockopt(SCTP_STATUS) with invalid associd
  test_sockopt.c  4 PASS : getsockopt(SCTP_STATUS) with NULL associd
  test_sockopt.c  5 BROK : bind: Address already in use

The underlying problem is that sctp_endpoint_destroy() hasn't been
triggered yet while the next bind attempt is being done. It will be
triggered eventually (but too late) by sctp_transport_destroy_rcu()
after one RCU grace period:

  sctp_transport_destroy()
    sctp_transport_destroy_rcu() ----.
      sctp_association_put() [*]  <--+--> sctp_packet_free()
        sctp_association_destroy()          [...]
          sctp_endpoint_put()                 skb->destructor
            sctp_endpoint_destroy()             sctp_wfree()
              sctp_bind_addr_free()               sctp_association_put() [*]

Thus, we move out the condition with sctp_association_put() as well as
the sctp_packet_free() invocation and the issue can be solved. We also
better free the SCTP chunks first before putting the ref of the association.

With this patch, the example above (which simulates a similar scenario
as in the implementation of this test case) and therefore also the test
suite run successfully through. Tested by myself.

Cc: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: NDaniel Borkmann <dborkman@redhat.com>
Acked-by: NVlad Yasevich <vyasevich@gmail.com>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      8c98653f
    • V
      cxgb3: Update VLAN extraction stats in the GRO path · 72073ad2
      由 Vipul Pandya 提交于 
      Signed-off-by: NDivy Le Ray <divy@chelsio.com>
Signed-off-by: NVipul Pandya <vipul@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      72073ad2
    • G
      netns: bond: allow unprivileged users to control bond device · 387ff911
      由 Gao feng 提交于 
      reduce the permission check of bond device's ioctl.
allow the userns root to control the bond device.
Signed-off-by: NGao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      387ff911
    • G
      netns: bridge: allow unprivileged users add/delete mdb entry · e4d343ea
      由 Gao feng 提交于 
      since the mdb table is belong to bridge device,and the
bridge device can only be seen in one netns.
So it's safe to allow unprivileged user which is the
creator of userns and netns to modify the mdb table.
Signed-off-by: NGao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      e4d343ea
    • G
      netns: ebtable: allow unprivileged users to operate ebtables · bb12b8b2
      由 Gao feng 提交于 
      ebt_table is a private resource of netns, operating ebtables
in one netns will not affect other netns, we can allow the
creator user of userns and netns to change the ebtables.
Signed-off-by: NGao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      bb12b8b2
    • G
      netns: fdb: allow unprivileged users to add/del fdb entries · c5c35108
      由 Gao feng 提交于 
      Right now,only ixgdb,macvlan,vxlan and bridge implement
fdb_add/fdb_del operations.

these operations only operate the private data of net
device. So allowing the unprivileged users who creates
the userns and netns to add/del fdb entries will do no
harm to other netns.
Signed-off-by: NGao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      c5c35108
  3. 04 2月, 2013 1 次提交
  5. 03 2月, 2013 4 次提交
    • D
      Merge branch 'delete-wanrouter' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux · 33397a71
      由 David S. Miller 提交于 
      Paul Gortmaker says:

====================
The removal of wanrouter code was originally listed in the (now
gone) feature removal file since May 2012, and an RFC of the
deletion was posted[1] in late 2012.  The overall concept was given
an OK, but defconfig contamination, build failures, etc. meant that
it didn't quite make it into mainline for 3.8.

Since that time, Dan discovered (via code audit) a runtime bug that
proves nobody has been using this for over four years[2].  With that
new information, I think it makes sense for someone to follow through
on Joe's original RFC and get this done for the 3.9 release.

In addition to resolving the build failures of the RFC by keeping
stub headers, this also splits the change into two parts, just like
the token ring removal did.  Part #1 decouples the mainline kernel
from the expired subsystem, and part #2 does the large scale
deletion of the subsystem content.

The advantage of the above, is that a "git blame" will never lead
you to a 4000+ line deletion commit.  The large scale deletion will
never show up in a "git blame" and hence the same advantages that we
get from the "--irreversible-delete" in the review stage of "git
format-patch" are also embedded into the git history itself.  This
may seem like a moot point to some, but for those who spend a
considerable amount of time data mining in the git history, this is
probably worth doing.

I have done build tests of all[mod/yes]config for both the stage 1
(Makefile and Kconfig) and stage 2 (full driver delete) as a sanity
check, and the issues with the previously posted RFC should be gone.

Speaking of "--irreversible-delete" -- these patches were created
with that option, so if you want to use them locally, you are going
to have to pull (location below) the content instead of doing a
"git am" of the mailed out content.

[1] http://patchwork.ozlabs.org/patch/198794/
[2] http://www.spinics.net/lists/netdev/msg218670.html
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      33397a71
    • D
      qlcnic: silence false positive overflow warning · 7bc48646
      由 Dan Carpenter 提交于 
      We actually store the MAC address as well as the board_name here.  The
longest board_name is 75 characters so there is more than enough room
to hold the 17 character MAC and the ": " divider.  But making this
buffer larger silences a static checker warning.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Acked-By: NJitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      7bc48646
    • M
      bnx2x: Force link UP when the interface is in LOOPBACK mode · 7614fe88
      由 Mahesh Bandewar 提交于 
      When the interface does not have carrier but when it's put into
loopback mode (for tests), it does not make sense to not have
the carrier. So force it!
Signed-off-by: NMahesh Bandewar <maheshb@google.com>
Acked-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      7614fe88
    • D
      Merge branch 'intel' · daef2262
      由 David S. Miller 提交于 
      Jeff Kirsher says:

====================
This series contains updates to ixgbe and e1000e.  The ixgbe patches are
a mix of fixes, cleanup and added functionality.  The first fix is for
traffic classes, where if the mapping has changed reset the NIC.  The other
ixgbe fix resolves an issue where the device lookup neglected to do a
pci_dev_put() to decrement the device reference count.

The ixgbe cleanup was done by Josh, where the auto-negotiation variables
were renamed/cleaned up and refactored.

The remaining patches are from Bruce to do additional cleanup on e1000e as
well as bump the driver version.  Most notably is the cleanup to use the
kernel IEEE MII definitions where possible instead of the local MII
definitions.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
      daef2262
  7. 01 2月, 2013 24 次提交
  9. 31 1月, 2013 5 次提交