linux-stable.git/net/sctp/protocol.c, branch linux-3.10.y

sctp: fix race on protocol/netns initialization

2015-10-01T10:07:38+00:00

[ Upstream commit 8e2d61e0aed2b7c4ecb35844fe07e0b2b762dee4 ]

Consider sctp module is unloaded and is being requested because an user
is creating a sctp socket.

During initialization, sctp will add the new protocol type and then
initialize pernet subsys:

        status = sctp_v4_protosw_init();
        if (status)
                goto err_protosw_init;

        status = sctp_v6_protosw_init();
        if (status)
                goto err_v6_protosw_init;

        status = register_pernet_subsys(&sctp_net_ops);

The problem is that after those calls to sctp_v{4,6}_protosw_init(), it
is possible for userspace to create SCTP sockets like if the module is
already fully loaded. If that happens, one of the possible effects is
that we will have readers for net->sctp.local_addr_list list earlier
than expected and sctp_net_init() does not take precautions while
dealing with that list, leading to a potential panic but not limited to
that, as sctp_sock_init() will copy a bunch of blank/partially
initialized values from net->sctp.

The race happens like this:

     CPU 0                           |  CPU 1
  socket()                           |
   __sock_create                     | socket()
    inet_create                      |  __sock_create
     list_for_each_entry_rcu(        |
        answer, &inetsw[sock->type], |
        list) {                      |   inet_create
      /* no hits */                  |
     if (unlikely(err)) {            |
      ...                            |
      request_module()               |
      /* socket creation is blocked  |
       * the module is fully loaded  |
       */                            |
       sctp_init                     |
        sctp_v4_protosw_init         |
         inet_register_protosw       |
          list_add_rcu(&p->list,     |
                       last_perm);   |
                                     |  list_for_each_entry_rcu(
                                     |     answer, &inetsw[sock->type],
        sctp_v6_protosw_init         |     list) {
                                     |     /* hit, so assumes protocol
                                     |      * is already loaded
                                     |      */
                                     |  /* socket creation continues
                                     |   * before netns is initialized
                                     |   */
        register_pernet_subsys       |

Simply inverting the initialization order between
register_pernet_subsys() and sctp_v4_protosw_init() is not possible
because register_pernet_subsys() will create a control sctp socket, so
the protocol must be already visible by then. Deferring the socket
creation to a work-queue is not good specially because we loose the
ability to handle its errors.

So, as suggested by Vlad, the fix is to split netns initialization in
two moments: defaults and control socket, so that the defaults are
already loaded by when we register the protocol, while control socket
initialization is kept at the same moment it is today.

Fixes: 4db67e808640 ("sctp: Make the address lists per network namespace")
Signed-off-by: Vlad Yasevich 
Signed-off-by: Marcelo Ricardo Leitner 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

sctp: reset flowi4_oif parameter on route lookup

2014-05-31T04:52:16+00:00

[ Upstream commit 85350871317a5adb35519d9dc6fc9e80809d42ad ]

commit 813b3b5db83 (ipv4: Use caller's on-stack flowi as-is
in output route lookups.) introduces another regression which
is very similar to the problem of commit e6b45241c (ipv4: reset
flowi parameters on route connect) wants to fix:
Before we call ip_route_output_key() in sctp_v4_get_dst() to
get a dst that matches a bind address as the source address,
we have already called this function previously and the flowi
parameters have been initialized including flowi4_oif, so when
we call this function again, the process in __ip_route_output_key()
will be different because of the setting of flowi4_oif, and we'll
get a networking device which corresponds to the inputted flowi4_oif
as the output device, this is wrong because we'll never hit this
place if the previously returned source address of dst match one
of the bound addresses.

To reproduce this problem, a vlan setting is enough:
  # ifconfig eth0 up
  # route del default
  # vconfig add eth0 2
  # vconfig add eth0 3
  # ifconfig eth0.2 10.0.1.14 netmask 255.255.255.0
  # route add default gw 10.0.1.254 dev eth0.2
  # ifconfig eth0.3 10.0.0.14 netmask 255.255.255.0
  # ip rule add from 10.0.0.14 table 4
  # ip route add table 4 default via 10.0.0.254 src 10.0.0.14 dev eth0.3
  # sctp_darn -H 10.0.0.14 -P 36422 -h 10.1.4.134 -p 36422 -s -I
You'll detect that all the flow are routed to eth0.2(10.0.1.254).

Signed-off-by: Xufeng Zhang 
Signed-off-by: Julian Anastasov 
Acked-by: Vlad Yasevich 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

net: sctp: remove cast for kmalloc/kzalloc return value

2013-03-18T13:16:00+00:00

remove cast for kmalloc/kzalloc return value.

Signed-off-by: Zhang Yanfei 
Acked-by: Neil Horman 
Cc: Vlad Yasevich 
Cc: Sridhar Samudrala 
Cc: Neil Horman 
Cc: Andrew Morton 
Cc: linux-sctp@vger.kernel.org
Signed-off-by: Jiri Kosina

sctp: make sctp_addr_wq_timeout_handler static

2012-12-29T04:32:36+00:00

Fix sparse warning about local function that should be static.

Signed-off-by: Stephen Hemminger 
Signed-off-by: David S. Miller

sctp: Change defaults on cookie hmac selection

2012-12-16T01:14:38+00:00

Recently I posted commit 3c68198e75 which made selection of the cookie hmac
algorithm selectable.  This is all well and good, but Linus noted that it
changes the default config:
http://marc.info/?l=linux-netdev&m=135536629004808&w=2

I've modified the sctp Kconfig file to reflect the recommended way of making
this choice, using the thermal driver example specified, and brought the
defaults back into line with the way they were prior to my origional patch

Also, on Linus' suggestion, re-adding ability to select default 'none' hmac
algorithm, so we don't needlessly bloat the kernel by forcing a non-none
default.  This also led me to note that we won't honor the default none
condition properly because of how sctp_net_init is encoded.  Fix that up as
well.

Tested by myself (allbeit fairly quickly).  All configuration combinations seems
to work soundly.

Signed-off-by: Neil Horman 
CC: David Miller 
CC: Linus Torvalds 
CC: Vlad Yasevich 
CC: linux-sctp@vger.kernel.org
Signed-off-by: David S. Miller

sctp: Fix compiler warning when CONFIG_DEBUG_SECTION_MISMATCH=y

2012-12-07T17:57:13+00:00

WARNING: net/sctp/sctp.o(.text+0x72f1): Section mismatch in reference
from the function sctp_net_init() to the function
.init.text:sctp_proc_init()
The function sctp_net_init() references
the function __init sctp_proc_init().
This is often because sctp_net_init lacks a __init
annotation or the annotation of sctp_proc_init is wrong.

And put __net_init after 'int' for sctp_proc_init - as it is done
everywhere else in the sctp-stack.

Signed-off-by: Christoph Paasch 
Acked-by: Neil Horman 
Acked-by: Vlad Yasevich 
Signed-off-by: David S. Miller

sctp: Make hmac algorithm selection for cookie generation dynamic

2012-10-26T06:22:18+00:00

Currently sctp allows for the optional use of md5 of sha1 hmac algorithms to
generate cookie values when establishing new connections via two build time
config options.  Theres no real reason to make this a static selection.  We can
add a sysctl that allows for the dynamic selection of these algorithms at run
time, with the default value determined by the corresponding crypto library
availability.
This comes in handy when, for example running a system in FIPS mode, where use
of md5 is disallowed, but SHA1 is permitted.

Note: This new sysctl has no corresponding socket option to select the cookie
hmac algorithm.  I chose not to implement that intentionally, as RFC 6458
contains no option for this value, and I opted not to pollute the socket option
namespace.

Change notes:
v2)
	* Updated subject to have the proper sctp prefix as per Dave M.
	* Replaced deafult selection options with new options that allow
	  developers to explicitly select available hmac algs at build time
	  as per suggestion by Vlad Y.

Signed-off-by: Neil Horman 
CC: Vlad Yasevich 
CC: "David S. Miller" 
CC: netdev@vger.kernel.org
Acked-by: Vlad Yasevich 
Signed-off-by: David S. Miller

sctp: Make sysctl tunables per net

2012-08-15T06:32:16+00:00

Signed-off-by: "Eric W. Biederman" 
Signed-off-by: David S. Miller

sctp: Push struct net down into sctp_in_scope

2012-08-15T06:30:37+00:00

struct net will be needed shortly when the tunables are made per network
namespace.

Signed-off-by: "Eric W. Biederman" 
Signed-off-by: David S. Miller

sctp: Add infrastructure for per net sysctls

2012-08-15T06:30:37+00:00

Start with an empty sctp_net_table that will be populated as the various
tunable sysctls are made per net.

Signed-off-by: "Eric W. Biederman" 
Acked-by: Vlad Yasevich 
Signed-off-by: David S. Miller