OSDN Git Service
Lorenzo Colitti [Tue, 29 Jul 2014 09:26:21 +0000 (18:26 +0900)]
Stop copying directly-connected routes to the main table, #2.
For a long time we have thought that copying directly-connected
routes to the main table was necessary to add gatewayed routes
to other routing tables. However, this is not necessary when the
directly-connected routes are properly created with "scope link"
as we do in http://ag/513100 .
Delete the copying code, but don't delete the rule that looks up
the main table or the code that dumps it. The main table is used
for things like cell networking, because the RIL emulates cell
networks, which are actually point-to-point, as directly
connected broadcast subnets (e.g., a /30 or a /27) with a fake
default gateway. The directly-connected route that covers the
fake default gateway is implicitly created by adding the IP
address, but it's in the main table, so we can't add the default
route without looking up the main table.
Change-Id: I93bd4764ac75fdcc98fa4206c601524100d53fc3
Lorenzo Colitti [Tue, 29 Jul 2014 09:25:44 +0000 (09:25 +0000)]
Revert "Stop copying directly-connected routes to the main table."
This reverts commit
2f5ea0e99e9a436cd43901b1772b77a410a62f8d.
Change-Id: I1fe1df0249714cb650a34fae56476236ac0108e3
Lorenzo Colitti [Tue, 29 Jul 2014 06:46:56 +0000 (15:46 +0900)]
Stop copying directly-connected routes to the main table.
For a long time we have thought that copying directly-connected
routes to the main table was necessary to add gatewayed routes
to other routing tables. However, this is not necessary when the
directly-connected routes are properly created with "scope link"
as we do in http://ag/513100 .
Delete the copying code, but keep dumping the main table in
bugreports, so we can see if third-party code such as RIL
daemons is putting anything in it.
Change-Id: Iddd531daaf9881ffd82f0a4b4f6cc857ce8788fd
Sreeram Ramachandran [Fri, 18 Jul 2014 20:03:47 +0000 (13:03 -0700)]
Add "scope link" when adding routes without a nexthop.
This is consistent with what /sbin/ip does, and it makes
recursive lookups on secondary tables work even when the main
table is empty.
This was originally proposed by Sreeram as http://ag/506456 to
fix a VPN issue, but then abandoned because it did not fix that
particular problem.
Bug:
16628572
Change-Id: I85753389c683ae2127b21af722463a35a33b60eb
Sreeram Ramachandran [Mon, 28 Jul 2014 18:53:50 +0000 (18:53 +0000)]
Merge "Prohibit address families by default unless a VPN explicitly allows them." into lmp-dev
Sreeram Ramachandran [Sun, 27 Jul 2014 01:43:25 +0000 (18:43 -0700)]
Prohibit address families by default unless a VPN explicitly allows them.
Bug:
15972465
Change-Id: I3278d94536fefacc86390c1ba4231680f7be8589
Sreeram Ramachandran [Wed, 23 Jul 2014 20:27:31 +0000 (13:27 -0700)]
Implement support for bypassable VPNs.
Bypassable VPNs grab all traffic by default (just like secure VPNs), but:
+ They allow all apps to choose other networks using the multinetwork APIs.
If these other networks are insecure ("untrusted"), they will enforce that the
app holds the necessary permissions, such as CHANGE_NETWORK_STATE.
+ They support consistent routing. If an app has an existing connection over
some other network when the bypassable VPN comes up, it's not interrupted.
Bug:
15347374
Change-Id: Iaee9c6f6fa8103215738570d2b65d3fcf10343f3
Sreeram Ramachandran [Mon, 28 Jul 2014 18:52:15 +0000 (18:52 +0000)]
Merge "Fix fwmark handling for bypassable VPNs and DNS." into lmp-dev
Sreeram Ramachandran [Fri, 25 Jul 2014 02:04:32 +0000 (19:04 -0700)]
Fix fwmark handling for bypassable VPNs and DNS.
This is a significant change to the way fwmarks are handled for two purposes:
1. Bypassable VPN.
This was introduced in http://ag/510058 and had an issue that if there's a
default network, it would always be used in connect(), so the bypassable VPN
wouldn't get any traffic. This CL fixes that issue by using the bypassable
VPN's NetId in connect(). See the comments in the code for more details.
2. DNS.
The previous DNS code (specifically, getNetworkForUser()) had two problems:
+ Even if a user asks for a NetId they have permission for, we'd always use
the user's VPN if they were subject to one. So, for example, a system IMS
app that brings up the mobile network in the presence of a VPN would still
have its DNS queries sent over the VPN, instead of mobile as desired.
+ Any user could perform DNS over any valid network, even one they didn't
have permissions for, as long as they weren't subject to a VPN. So, for
example, an app could use the DNS servers of a different profile's VPN.
This CL fixes those problems. See getNetworkForDns() for more details.
The two pieces above are inter-related. Previously, we never set the explicit
bit from the DNS code. But we need to do that if the user asks for a network
explicitly, for two reasons:
o So that the DNS query is really restricted to that network and doesn't
fallthrough to the default network.
o So that the heuristic described in ON_CONNECT works in all cases. I.e., if the
DNS proxy's connect() request comes in with the explicit bit NOT set, we know
that the NetId can only be either the default network or a VPN.
This CL is not intended to be robust against race conditions. In general, very
little of the netd code is resilient. A separate effort needs to be undertaken
to carefully audit all the code and logic to guard against things like:
* A VPN being established between calls to getNetworkForDns() and connect().
* State changes between multiple calls to NetworkController from clients such as
FwmarkServer and DnsProxyListener.
* Routing rules / iptables rules being set up in a less-than-ideal order.
* ... etc.
Bug:
15347374
Change-Id: I5baad9168c4f4f3ef4129e07234b4bf24b0d8ba2
Sreeram Ramachandran [Wed, 23 Jul 2014 05:23:20 +0000 (22:23 -0700)]
Implement the fallthrough rule to support split tunnel VPNs.
Change-Id: Ibc48caedb5954c6b12bfa553d978bab56c4b09aa
Sreeram Ramachandran [Wed, 23 Jul 2014 01:51:06 +0000 (18:51 -0700)]
Implement the rule to forward traffic to local networks via VPNs.
Change-Id: I4bffb2ce854a6fa7d4c0d35b97f123d91c6a84d6
Sreeram Ramachandran [Wed, 23 Jul 2014 01:16:44 +0000 (18:16 -0700)]
Cosmetic: Move some functions around to match rule priority order.
Change-Id: I81af639b66ac66272e0fae2d53119de1bfa12e69
Sreeram Ramachandran [Tue, 22 Jul 2014 23:44:37 +0000 (16:44 -0700)]
Remove the oif=iface clause when tethering.
The rule is meant to be "iif=inputIface => lookup <table_for_outputIface>", not
"iif=foo oif=bar => lookup <table_for_bar>" (the latter would never match
because packets coming in on the input interface will not have any output
interface binding).
Bug:
16242255
Change-Id: I98a2a8ab90765aee833134297b58d4f7a212036f
Sreeram Ramachandran [Sun, 20 Jul 2014 06:22:33 +0000 (23:22 -0700)]
Cleanup: Delete dead code.
Bug:
15413389
Change-Id: I315468832ef18ffc84174e54774ab63b86d284dc
Sreeram Ramachandran [Tue, 22 Jul 2014 19:40:36 +0000 (12:40 -0700)]
Try to undo iptables rules if a command fails.
As per review comments at:
https://googleplex-android-review.git.corp.google.com/#/c/504240/5/server/NatController.cpp@194
Bug:
15413694
Bug:
15413741
Change-Id: I99e2f83792fa81498447e4c03f5f8be1d2c43b7d
Sreeram Ramachandran [Sat, 19 Jul 2014 07:22:15 +0000 (00:22 -0700)]
Flush routing rules and add an "unreachable" rule on Init().
Without the flush, rules accumulate duplicates when netd is restarted due to a
runtime restart. Nothing functionally wrong with having duplicates; it just
makes the output of "ip rule" look as though something went wrong in the system.
Time to add the unreachable rule, to suss out issues with corner cases. With the
flush, there's no more a naked "from main" rule that we need to protect by
adding the unreachable rule. But it's a good idea to add the unreachable rule
anyway, in case somebody comes along and adds a rule below it later.
Change-Id: I975b2221868b7f5366bd7cf60937a82fb4b75913
Sreeram Ramachandran [Tue, 15 Jul 2014 23:20:28 +0000 (16:20 -0700)]
Fix WiFi-Direct and Tethering.
A LocalNetwork object now always exists in the NetworkController, with a fixed
NetId that's guaranteed not to collide with NetIds created by the framework.
When routes are added on an interface tracked by the LocalNetwork, they are
added to a fixed "local_network" table.
When NAT is enabled, we add a special "iif -> oif" tethering rule.
Bug:
15413694
Bug:
15413741
Change-Id: I36effc438d5ac193a77174493bf196cb68a5b97a
Sreeram Ramachandran [Sat, 12 Jul 2014 00:06:12 +0000 (17:06 -0700)]
Don't reset a VPN's NetId in the connect() shim.
Change-Id: I0cc6c0e221a40c9100c8f4c0c5e761fce3f9b0ae
Sreeram Ramachandran [Fri, 11 Jul 2014 23:27:02 +0000 (16:27 -0700)]
Introduce a SELECT_FOR_USER fwmarkd command.
This lets SYSTEM apps mark a socket with the network that would apply to an
arbitrary UID. I.e., either the VPN that applies to that user, or the default
network, if there's no such VPN.
This command will be used by system apps that proxy stuff for a user, so that
they can route the same way that user would have. Examples of such system apps
are the DnsProxyListener, MediaServer and DownloadManager.
The "explicit" bit is NOT set, so that if the user's VPN is a split tunnel, the
route lookup will fall-through to the default network as desired.
The "protect" bit is set, so that the socket bypasses any VPN applicable to the
system app itself. Note that even if the uid being proxied for is also subject
to the same VPN, this still works because the relevant rule doesn't care about
the protected bit (see modifyVpnSystemPermissionRule() in RouteController.cpp).
Change-Id: I4d501e5214b127f4ae9eaeb7befb1751cd102308
Sreeram Ramachandran [Fri, 11 Jul 2014 23:01:33 +0000 (16:01 -0700)]
Cosmetic: Remove unnecessary constructor.
Change-Id: Ib1de6e8988c4728e962da8607d0ecd98d6121e0c
Sreeram Ramachandran [Fri, 11 Jul 2014 22:29:49 +0000 (22:29 +0000)]
Merge "Add default routing tables to rt_tables."
Sreeram Ramachandran [Fri, 11 Jul 2014 18:45:14 +0000 (11:45 -0700)]
Add default routing tables to rt_tables.
This is so that these tables also get dumped by bugreport / dumpstate.
Change-Id: Ia5ed8a23911d7b34ba964934dec42849f4a73824
Sreeram Ramachandran [Fri, 11 Jul 2014 16:10:20 +0000 (09:10 -0700)]
Fix tethering in the case of a regular upstream connection.
Fixes tethering via Ethernet, Bluetooth and WiFi (hotspot).
Tethering when the upstream has a DUN-specific APN is likely still broken
(untested).
For now, assign a fixed NetId (a hack) until we can change the framework to
create a valid NetworkAgent and all that jazz.
Bug:
15968336
Bug:
14988803
Change-Id: Idcf4d492d9329a9c87913e27be6dd835a792bea2
Sreeram Ramachandran [Wed, 9 Jul 2014 14:39:30 +0000 (07:39 -0700)]
Cosmetic: Move constants into file scope.
Change-Id: Ic1398a5867693bb8f1e588e727869ec6a4574432
Sreeram Ramachandran [Tue, 8 Jul 2014 05:11:37 +0000 (22:11 -0700)]
Add symbolic table names for routing table numbers.
This makes the output from "ip rule" much more readable.
Companion changes are in AOSP.
Change-Id: I69deb1a64d5d6647470823405bf0cc55b24b22de
Sreeram Ramachandran [Sun, 6 Jul 2014 00:15:14 +0000 (17:15 -0700)]
Add full support for UIDs in VPNs.
Major:
+ Implement the functions mentioned in http://go/android-multinetwork-routing
correctly, including handling accept(), connect(), setNetworkForSocket()
and protect() and supporting functions like canUserSelectNetwork().
+ Eliminate the old code path of getting/setting UID ranges through
SecondaryTableController (which is currently unused) and mUidMap.
Minor:
+ Rename some methods/variables for clarity and consistency.
+ Moved some methods in .cpp files to match declaration order in the .h files.
Bug:
15409918
Change-Id: Ic6ce3646c58cf645db0d9a53cbeefdd7ffafff93
Sreeram Ramachandran [Thu, 3 Jul 2014 19:20:48 +0000 (12:20 -0700)]
Update routing rules.
As per the latest changes to: http://go/android-multinetwork-routing
Functional changes:
+ Add explicit=NO to the implicit network rules, though it's a no-op.
+ Remove most of the UID=0 (kernel access) rules since they are no longer
needed, except in one case to allow access to a VPN.
+ Add the explicit, protect and permissions bits to the incoming packet mark.
+ VPNs now don't need an implicit network rule.
+ Modifying network permissions now modifies the incoming packet mark as well.
Cosmetic changes:
+ Renamed the legacy tables to match their permissions (SYSTEM and NETWORK).
+ Renamed most functions and methods for clarity and consistency.
+ Renamed and adjusted some ule priorities.
+ Move most rule modifications into their own functions, to prevent brittle
reliance on the previous state of the fwmark/mask variables.
Change-Id: I958a7e158ee918d5254de606fcfa55fe23327438
Sreeram Ramachandran [Sat, 5 Jul 2014 19:31:05 +0000 (12:31 -0700)]
Fix permissions handling.
+ Rename the permissions as per: http://go/android-multinetwork-routing
+ Make the SYSTEM permission explicitly include NETWORK.
+ Grant the SYSTEM permission to system UIDs by default, but allow the framework
to override them if necessary.
+ Move the "string to permission" parsing to CommandListener.cpp, thus allowing
us to get rid of Permission.cpp.
+ There's no need to support multiple permissions string arguments, so tighten
that up.
Change-Id: I73d51b5e2f44a97e6d5ab5943ff198cebfbcc0c4
Sreeram Ramachandran [Sat, 5 Jul 2014 18:00:55 +0000 (11:00 -0700)]
Only add permissions in accept(); do not try to enforce anything.
Previously, we were enforcing that the user has access to the network over which
the socket is accepted. This has two problems:
1. We weren't handling the loopback interface ('lo') correctly. It's not part of
any network, so the NetId in the fwmark is NETID_UNSET. In
NetworkController::isUserPermittedOnNetwork(), we would fail to find a valid
network, and so we would return false.
2. We have decided that in fact, we don't want to enforce this even for other
interfaces, due to the SYN-ACK problem. See the comments and discussion at:
http://go/android-multinetwork-routing
Bug:
16079376
Change-Id: I3f07f67d06dd7b48bfdfa5616ee22e098e31f8af
Sreeram Ramachandran [Thu, 3 Jul 2014 01:06:34 +0000 (18:06 -0700)]
Fix regressions in flushRoutes().
http://ag/486276 left out a comma, causing the command to be "routeflush" (which
is invalid), instead of the correct "route flush".
http://ag/495068 changed the order, causing the interface to be removed from the
interfaceToIndex map before we are done using it in modifyPerNetworkRules().
Change-Id: I65af0b5763b13e47b48e7e2b81d243dc19e1a03b
Sreeram Ramachandran [Wed, 2 Jul 2014 21:49:33 +0000 (14:49 -0700)]
Minor style change.
Change-Id: Ib1f89bff55506cf3881047dd876adbf97c789423
Sreeram Ramachandran [Wed, 2 Jul 2014 17:09:49 +0000 (10:09 -0700)]
Keep track of users allowed to call protect() explicitly.
This is an API change between ConnectivityService and Netd.
The ownerUid was meant for this purpose, but it's insufficient, as apps need to
call protect() _before_ they create a VPN.
Bug:
15409918
Change-Id: If804aa106002e96d5ffb623d32db35fd76928367
Sreeram Ramachandran [Wed, 2 Jul 2014 02:01:56 +0000 (19:01 -0700)]
Cosmetic: Add a way to query the type of a network.
This is a cosmetic change, i.e., there's no change in functionality.
This is the poor man's RTTI. It turns out that maintaining separate lists (or
maps) of the different types of networks gets burdensome pretty quickly
(especially in an upcoming CL where we add functions like
canUserSelectNetwork()).
Change-Id: If5250c0fc106045f681d0fd71278b793addbe1e3
Sreeram Ramachandran [Tue, 1 Jul 2014 21:30:30 +0000 (14:30 -0700)]
Fix routing rules.
Fix the rules in RouteController as per the latest iteration of the routing
design: http://go/android-multinetwork-routing. Changes include:
+ The legacy tables have to be global. So remove the UID being passed in to
modifyRoute() and remove the associated TODOs.
+ Add UID=0 rules to let the kernel access routes on privileged networks.
+ Add a UID=0 clause to the directly-connected hack, thus fixing the TODO.
+ Add the privileged_legacy table just above the legacy table, when overriding
the default network. (The same table remains added at the top of the rule
chain, to override VPNs, but only for CONNECTIVITY_INTERNAL-privileged apps).
Other cosmetic changes:
+ Update the names and values of the rule priorities.
+ Move the legacy table IDs to the .h file in anticipation of using them from
bugreport / dump commands.
+ Make 'action' the first parameter consistently.
Change-Id: I6634a19ddc8062b2ef55d926c7892fff8c586106
Sreeram Ramachandran [Tue, 1 Jul 2014 22:49:20 +0000 (15:49 -0700)]
Don't use %zu for uid_t. It's always unsigned int, so %u is correct.
Change-Id: I5be1d479b524495037c2aedc8336c794d2698914
Sreeram Ramachandran [Tue, 1 Jul 2014 22:06:05 +0000 (15:06 -0700)]
Use %zu to printf size_t correctly on both 32-bit and 64-bit platforms.
Change-Id: I5223e574084fca47606b844d74a99a642c7d66be
Sreeram Ramachandran [Tue, 24 Jun 2014 01:54:27 +0000 (18:54 -0700)]
Add UID range support to VPNs.
This adds the necessary routing rules.
Future CLs will add the ability to select the right netId for connect(),
setNetworkForSocket(), DNS resolutions, etc.
Bug:
15409918
Change-Id: I88a67660d49cecda834dd72ab947fbfed250f09d
Sreeram Ramachandran [Mon, 23 Jun 2014 19:41:37 +0000 (12:41 -0700)]
Introduce VPN support.
This change sets up the basic routing rules for VPNs. It doesn't yet handle UID
ranges (that are meant to apply to the VPN) correctly. That's forthcoming in
other CLs.
Bug:
15409918
Change-Id: I284de04f176dcf6ba702361de6a614266256d04e
Sreeram Ramachandran [Mon, 23 Jun 2014 16:54:06 +0000 (09:54 -0700)]
Refactor: Encapsulate permissions and interfaces into a Network class.
Currently, there's a lot of logic in NetworkController surrounding events such
as interface addition/removal, network creation/destruction and default network
change, because these events are interwined. For example, adding an interface
means also adding a corresponding default network rule if the interface is being
added to the current default network.
When we introduce VPNs into this mix, things will get hairy real quick for all
this logic in NetworkController.
In this refactor, we introduce an abstract base class Network which supports
adding and removing interfaces. The main concrete implementation of this is
PhysicalNetwork, which allows setting permissions and "default network" state.
Since we've moved network permissions into the above class, and user permissions
into NetworkController, PermissionsController is unused and has been removed.
Also fix a few bugs in RouteController:
+ Use uidEnd correctly.
+ Check for all error cases in inet_pton.
+ Check the return value of android_fork_execvp() correctly.
+ The "return cmd1() && cmd2()" pattern is wrong. Rewrite that code.
Also (non-functional changes):
+ Remove instantiations of RouteController. It has static methods only.
+ Reorder some blocks in CommandListener so that the most frequent commands are
checked first.
+ Remove unused paramError() and clearNetworkPreference().
+ Change all return codes to int (negative errno) wherever applicable.
+ Add WARN_UNUSED_RESULT everywhere.
+ Cleanup some style in RouteController and NetworkController.
+ Use uid_t instead of unsigned for user IDs.
+ Add clearer log messages at the source of failures.
+ Add a check for when fwmark bits are set without corresponding mask bits.
Bug:
15409918
Change-Id: Ibba78b0850160f9f3d17d476f16331a6db0025d1
Sreeram Ramachandran [Fri, 27 Jun 2014 13:42:11 +0000 (06:42 -0700)]
Don't fail when adding a duplicate legacy route.
This only affects calls to requestRouteToHost() by apps. We still fail if the
framework itself tries to add a non-legacy duplicate route, since the framework
should know better (we can consider relaxing that too in the future).
Bug:
15925532
Change-Id: I9ee434277e462d570f88e6fe63a096e5ae41eee9
Lorenzo Colitti [Thu, 26 Jun 2014 04:51:10 +0000 (13:51 +0900)]
Convert rta_* and fra_* variable names to camelCase.
Change-Id: I0ca539ac4c54bb71b033f288fb4229afd71b7989
Lorenzo Colitti [Tue, 24 Jun 2014 18:20:29 +0000 (03:20 +0900)]
Support adding and removing UID rules via netlink.
Change-Id: Idae13bceda869261689260759084b8d6ef1ff639
Lorenzo Colitti [Mon, 23 Jun 2014 06:09:54 +0000 (15:09 +0900)]
Pass rule modification errors back to CommandListener.
Change-Id: If01334dccad8b6230648713a57fd58be180ac66b
Lorenzo Colitti [Fri, 20 Jun 2014 14:03:29 +0000 (23:03 +0900)]
Use netlink to add/delete rules as well as routes.
Also change the indentation of the rtattrs used in modifyIpRoute
to make it easier to see what attributes are being used and in
what sequence.
This change does not yet pass the errors back to CommandListener;
that is done in the next change in the series.
Change-Id: Ib2e174386c63cb0647d838d9c7d731cd6df39c4f
Sreeram Ramachandran [Tue, 24 Jun 2014 23:09:21 +0000 (16:09 -0700)]
Unrevert the 'revert the "talk to netlink directly" change.'
http://ag/486277 was reverted in http://ag/491263 and is being unreverted in
this change. The bug in the original CL was a typo ("interface" instead of the
desired "ifindex"), which is now fixed.
Bug:
15840054
Change-Id: If66987c74cc86e9ba4f7a35d36f0a39afe939a68
Sreeram Ramachandran [Tue, 24 Jun 2014 22:03:52 +0000 (15:03 -0700)]
Temporarily revert the "talk to netlink directly" change.
The change being reverted is http://ag/486277. Instead, use /sbin/ip again. The
code to talk to netlink fails on volantis. I.e., instead of this:
$ ip route show table 1006
default via 100.110.191.254 dev wlan0
100.110.128.0/18 dev wlan0 scope link
we end up with this:
$ ip route show table 1006
default dev wlan0 proto static
100.110.128.0/18 dev wlan0 proto static
Notice the lack of the nexthop and the addition of "proto static". I think the
netlink message is somehow not properly aligned on volantis, causing the kernel
to misinterpret it.
Bug:
15840054
Change-Id: Ief60473e337410f7cb35890de0a5a74a21723a41
Sasha Levitskiy [Mon, 23 Jun 2014 17:51:05 +0000 (10:51 -0700)]
Netd: Interface controller: Demote a log message severity from warning to verbose.
Change-Id: Ia8b2e6a0b039c2bda418a5f7138c0aac6b6c643b
Signed-off-by: Sasha Levitskiy <sanek@google.com>
Lorenzo Colitti [Mon, 23 Jun 2014 16:41:13 +0000 (16:41 +0000)]
Merge "Pass route add/delete errors back to CommandListener."
Sreeram Ramachandran [Sun, 22 Jun 2014 18:02:57 +0000 (11:02 -0700)]
Return negative errno (instead of positive) on failure.
http://ag/489245 changed some return values from bools to errno values. However,
in forthcoming CLs, @lorenzo uses the convention of negative errno to indicate
failure. So, be consistent with that style.
Change-Id: I3eac8f142c36a2e779cda289c07ee374c49e2f6b
Lorenzo Colitti [Tue, 17 Jun 2014 15:41:58 +0000 (00:41 +0900)]
Pass route add/delete errors back to CommandListener.
Change-Id: Id1d6d578963080e141f71bc1303801fc53bce40a
Lorenzo Colitti [Tue, 17 Jun 2014 15:22:17 +0000 (00:22 +0900)]
Use native netlink code instead of /sbin/ip to manipulate routes
Shelling out to /sbin/ip is slow, and more importantly it does
not preserve the error messages returned by the kernel when
adding or deleting a route fails. Instead, use netlink directly.
This change does not yet pass the errors back to CommandListener;
that is done in the next change in the series.
Change-Id: I5ad3c8583580857be6386a620ff5c4f3872d685b
Lorenzo Colitti [Thu, 12 Jun 2014 04:51:05 +0000 (13:51 +0900)]
Notify network observers of route changes.
This reverts commit
20d3f5e856a38ef22851e32d7f9b6a7cd02eb459.
Bug:
9180552
Change-Id: I004448e2f0c731f6ca2c6d3535dc4dd51a622293
Sreeram Ramachandran [Fri, 20 Jun 2014 18:59:40 +0000 (11:59 -0700)]
Add getNetworkForSocket().
This doesn't require a roundtrip to netd, since anybody can read the SO_MARK.
Change-Id: I51dd17725c4534cb0d5dbc8e93e844e6a7847959
Sreeram Ramachandran [Fri, 20 Jun 2014 18:51:48 +0000 (11:51 -0700)]
Return errors explicitly instead of using errno.
Change-Id: Ia29f500e747a8c72d13a8f38c3b08c319c8c029a
Sreeram Ramachandran [Fri, 20 Jun 2014 17:23:46 +0000 (17:23 +0000)]
Merge "Add a client API to protect a socket from VPNs."
Lorenzo Colitti [Fri, 20 Jun 2014 07:03:19 +0000 (07:03 +0000)]
Merge "Revert "Notify network observers of route changes.""
Lorenzo Colitti [Fri, 20 Jun 2014 07:02:36 +0000 (07:02 +0000)]
Revert "Notify network observers of route changes."
This reverts commit
d2f23fb6285000831f14cacc16881d96f6046c33.
Change-Id: Ifabbc6848ab33842461465fccb9886863a08aec6
Lorenzo Colitti [Thu, 12 Jun 2014 04:51:05 +0000 (13:51 +0900)]
Notify network observers of route changes.
Bug:
9180552
Change-Id: Ic23025e4d7cb4e58ae59017777267ac45f00b434
Lorenzo Colitti [Fri, 20 Jun 2014 06:02:48 +0000 (06:02 +0000)]
Merge "Flush both IPv4 and IPv6 routes."
Lorenzo Colitti [Tue, 17 Jun 2014 07:14:17 +0000 (16:14 +0900)]
Flush both IPv4 and IPv6 routes.
The current code unintentionally flushes only IPv4 routes
because it reuses a function that is normally used to add and
remove routes (where the IP version is implicitly specified by
the route to add or remove).
Instead of fixing the runIpRouteCommand function, add a new
flushIpRoutes function because runIpRouteCommand will be
replaced by a netlink implementation in an upcoming CL.
Change-Id: Ie96ae4124baca3edb8e0d0841e7abadb6b3ee9ab
Lorenzo Colitti [Fri, 20 Jun 2014 05:52:15 +0000 (05:52 +0000)]
Merge "Notify network observers of route changes."
JP Abgrall [Fri, 20 Jun 2014 01:35:24 +0000 (18:35 -0700)]
server: check interface names in RPC arguments for validity
This patch introduces a method isIfaceName that checks interface
names from various RPCs for validity before e.g. using them as
part of iptables arguments or in filenames.
All of these RPC calls can only be called from applications
with at least the CONNECTIVITY_INTERNAL permission in recent
Android versions, so the impact of the missing checks luckily
isn't very high.
Orig-Author: Jann Horn <jann@thejh.net>
Change-Id: I80df8d745a3de99ad02d6649f0d10562c81f6b98
Signed-off-by: JP Abgrall <jpa@google.com>
Sreeram Ramachandran [Thu, 19 Jun 2014 17:03:07 +0000 (10:03 -0700)]
Add a client API to protect a socket from VPNs.
The server side hasn't been implemented yet (see FwmarkServer.cpp).
A UID can only be in a single VPN at any time, so there's no need
to specify a netId or vpnId.
Change-Id: Ie9c4590a9900e1ebf28418c4b9c4760cc0a5501a
Paul Jensen [Thu, 12 Jun 2014 20:46:37 +0000 (16:46 -0400)]
Cache interface indices in case interfaces go away.
Without caching them netd will fail to remove rules and routes,
for example, when the Bluetooth reverse-tether interface ("bt-pan")
goes away.
bug:
15407087
Change-Id: I99fcf00f9645a0b029455516a705b70110f62ff6
Lorenzo Colitti [Fri, 13 Jun 2014 02:35:29 +0000 (02:35 +0000)]
Merge "Pass a socket mark to clatd so it can bind to its network."
Lorenzo Colitti [Mon, 9 Jun 2014 05:09:20 +0000 (14:09 +0900)]
Pass a socket mark to clatd so it can bind to its network.
Bug:
15340961
Change-Id: If15e90cbd5526f6c8fd839d4d009846d64d9e77a
Paul Jensen [Thu, 12 Jun 2014 19:56:50 +0000 (19:56 +0000)]
am
b72f6d71: (-s ours) DO NOT MERGE: Fix use-after-free of interface name during network destruction.
* commit '
b72f6d71840afeb1923f64f8c44417f4f208c282':
DO NOT MERGE: Fix use-after-free of interface name during network destruction.
Paul Jensen [Wed, 11 Jun 2014 14:58:16 +0000 (10:58 -0400)]
DO NOT MERGE: Fix use-after-free of interface name during network destruction.
bug:
15560990
Change-Id: I899827c4f50847a3a60b6359f829bee5d6eb4f00
(cherry picked from commit
6ff16785ae1f67f44a73ad00d6c86690af6772d2)
Paul Jensen [Wed, 11 Jun 2014 14:58:16 +0000 (10:58 -0400)]
Fix use-after-free of interface name during network destruction.
bug:
15560990
Change-Id: I899827c4f50847a3a60b6359f829bee5d6eb4f00
Sreeram Ramachandran [Fri, 6 Jun 2014 15:50:13 +0000 (15:50 +0000)]
am
41417388: (-s ours) Use AF_INET sockets when checking netIds in setNetworkFor{Process,Resolv}(). DO NOT MERGE
* commit '
414173882a87ded2fdd4f5181cd7dff3842a8917':
Use AF_INET sockets when checking netIds in setNetworkFor{Process,Resolv}(). DO NOT MERGE
Dan Albert [Thu, 5 Jun 2014 20:33:08 +0000 (20:33 +0000)]
am
d42cdf50: am
437eb9c3: Merge "Changes uses of index() to strchr()"
* commit '
d42cdf50ab883acb74b17d63646f0228d881df3c':
Changes uses of index() to strchr()
Dan Albert [Thu, 5 Jun 2014 20:29:33 +0000 (20:29 +0000)]
am
437eb9c3: Merge "Changes uses of index() to strchr()"
* commit '
437eb9c3092c96c1828e326edb6574090fe37aaa':
Changes uses of index() to strchr()
Dan Albert [Thu, 5 Jun 2014 19:55:52 +0000 (19:55 +0000)]
Merge "Changes uses of index() to strchr()"
Dan Albert [Thu, 5 Jun 2014 18:35:41 +0000 (11:35 -0700)]
Changes uses of index() to strchr()
Bionic is geting rid of index() because it was removed from POSIX 2008.
Change-Id: I182523734f372d1c6ed052b1373422332587f8ff
Sreeram Ramachandran [Tue, 3 Jun 2014 19:51:08 +0000 (12:51 -0700)]
Use std::atomic<> instead of volatile sig_atomic_t.
std::atomic<> comes with true load/store barriers, so it's preferable.
Change-Id: Ife47b0e404f41aa74aeb168befd1ff4a6603f3ed
Sreeram Ramachandran [Tue, 3 Jun 2014 00:47:36 +0000 (17:47 -0700)]
Remove unnecessary includes.
Neither of these actually does anything.
Change-Id: I82ee09a4565a8e8e4dc5f54f278bf83e002d7387
Paul Jensen [Tue, 3 Jun 2014 17:24:46 +0000 (17:24 +0000)]
Merge "Pass NetID to clatd as a command line argument."
Paul Jensen [Fri, 30 May 2014 17:29:41 +0000 (13:29 -0400)]
Pass NetID to clatd as a command line argument.
Change-Id: I77a47b24e68c7786f790974f05787a40a4934af5
Sreeram Ramachandran [Sat, 31 May 2014 02:59:51 +0000 (19:59 -0700)]
Use AF_INET sockets when checking netIds in setNetworkFor{Process,Resolv}(). DO NOT MERGE
AF_UNIX sockets don't need to be marked, so we don't give netd the permission to
operate on them (cf: netd.te). I.e., netd doesn't expect to receive them.
Make sure that the creation of the AF_INET socket doesn't trigger another
wasteful check with netd by calling the libc version directly.
Bug:
13885501
Change-Id: I6b549232e57cacd47501edcefa4c0b4b79df9da0
(cherry picked from commit
2756045bebaac342f7cb70dad11519f896d44833)
Sreeram Ramachandran [Sat, 31 May 2014 02:59:51 +0000 (19:59 -0700)]
Use AF_INET sockets when checking netIds in setNetworkFor{Process,Resolv}().
AF_UNIX sockets don't need to be marked, so we don't give netd the permission to
operate on them (cf: netd.te). I.e., netd doesn't expect to receive them.
Make sure that the creation of the AF_INET socket doesn't trigger another
wasteful check with netd by calling the libc version directly.
Bug:
13885501
Change-Id: I6b549232e57cacd47501edcefa4c0b4b79df9da0
Sreeram Ramachandran [Fri, 30 May 2014 22:11:54 +0000 (15:11 -0700)]
Fix build.
Change-Id: If170e46ff92c6a972bc1c2d838b1ac0eea6e23de
Sreeram Ramachandran [Fri, 30 May 2014 21:33:07 +0000 (21:33 +0000)]
am
82eab785: Support legacy routes added by apps via ensureRouteToHost().
* commit '
82eab785bd5cb2eff0a263f5b0dcde13e9139588':
Support legacy routes added by apps via ensureRouteToHost().
Sreeram Ramachandran [Thu, 22 May 2014 21:21:49 +0000 (14:21 -0700)]
Support legacy routes added by apps via ensureRouteToHost().
This adds the routes to two fixed tables:
+ LEGACY, which has higher priority than other non-explicit lookup tables
(per-network and default network).
+ PRIVILEGED_LEGACY, available only to system apps and has higher priority than
VPNs (system apps are those with the CONNECTIVITY_INTERNAL permission).
This will be changed to per-UID tables once the kernel supports UID-based
routing, so that these legacy routes are scoped to each app and not global.
Also, fix a TODO: The framework (as of http://ag/471599) will not set the
gateway argument if it's actually a direct-connected route.
Change-Id: I0ee1ca89fdc859d75a89021ca8c1902811b1e4a9
(cherry picked from commit
38b7af1f2cb9579895465fabc37865f5dadcac25)
Lorenzo Colitti [Fri, 30 May 2014 15:06:46 +0000 (15:06 +0000)]
am
ce8f583f: Set errno in NetworkController commands.
* commit '
ce8f583ff6620a19602d3c4604557e1b1501cafa':
Set errno in NetworkController commands.
The Android Automerger [Fri, 30 May 2014 15:06:46 +0000 (15:06 +0000)]
am
a5d68758: merge in master-release history after reset to
ec00884cac216d1cb79556ca23b21ce55e35af3e
* commit '
a5d6875887a63797711093a169e7ef19c5a1ab3b':
Fix for DNS resolutions when there is no default network set yet. We need to determine the actual netId to use prior to comparing it against the default netId. Also initialize the default network variable.
Sreeram Ramachandran [Wed, 28 May 2014 22:07:00 +0000 (15:07 -0700)]
Turn on C++11 and make all warnings into errors.
As a consequence:
+ Comment out the names of all unused parameters.
+ Remove all unused variables and functions.
In server/Android.mk, there are a couple of non-trivial changes:
+ Use libcxx instead of stlport. This is needed to fix a bunch of errors due to
specifying -std=c++11.
+ LOCAL_SHARED_LIBRARIES is sorted. Technically, the order in which libraries
are listed has an effect on linking, but nobody should be doing such brittle
things anyway.
Change-Id: I0aff5b745e04609da23144d0e8be4c5694321b8b
Sreeram Ramachandran [Thu, 22 May 2014 21:21:49 +0000 (14:21 -0700)]
Support legacy routes added by apps via ensureRouteToHost().
This adds the routes to two fixed tables:
+ LEGACY, which has higher priority than other non-explicit lookup tables
(per-network and default network).
+ PRIVILEGED_LEGACY, available only to system apps and has higher priority than
VPNs (system apps are those with the CONNECTIVITY_INTERNAL permission).
This will be changed to per-UID tables once the kernel supports UID-based
routing, so that these legacy routes are scoped to each app and not global.
Also, fix a TODO: The framework (as of http://ag/471599) will not set the
gateway argument if it's actually a direct-connected route.
Change-Id: I0ee1ca89fdc859d75a89021ca8c1902811b1e4a9
Lorenzo Colitti [Thu, 29 May 2014 03:20:55 +0000 (12:20 +0900)]
Set errno in NetworkController commands.
Set errno in the NetworkController functions that are called by
CommandListener and where failure causes an operationError. This
is because operationError always appends errno to its messages,
and
"400 481 addInterfaceToNetwork() failed (Device or resource busy)"
is more useful than:
"400 481 addInterfaceToNetwork() failed (Success)"
We already call ALOGE to put descriptive messages in the system
log, but the bugreport already conveniently aggregates all netd
commands in one place in the network_management service dump, and
it would help if the errno in those messages were accurate.
Unfortunately many of the route and iptables commands shell out
to /sbin/ip and we can't return meaningful values, but we'll
fix that when (if?) we replace that with a proper netlink
implementation.
Bug:
15316764
Change-Id: Ia47f451029d611491aa72bca602de77333f2c6a0
The Android Automerger [Thu, 22 May 2014 14:22:41 +0000 (07:22 -0700)]
merge in master-release history after reset to
ec00884cac216d1cb79556ca23b21ce55e35af3e
The Android Automerger [Thu, 22 May 2014 12:51:13 +0000 (05:51 -0700)]
merge in master-release history after reset to
ec00884cac216d1cb79556ca23b21ce55e35af3e
Sreeram Ramachandran [Wed, 21 May 2014 21:01:16 +0000 (14:01 -0700)]
Implement the "select network" fwmark command.
Change-Id: Id4a49eb288b18022d53014d1ae2211ed7d1099a6
Sreeram Ramachandran [Wed, 21 May 2014 20:19:43 +0000 (13:19 -0700)]
Rework the determination of a "valid network".
+ isNetIdValid() doesn't make much sense. What we want is whether the netId has
actually been created (via createNetwork()).
+ It isn't an error to call deleteNetwork() or setDefaultNetwork() even when
there are no interfaces assigned to the network.
+ Secure all accesses to the maps in PermissionsController with locks; they are
called from many threads (CommandListener, DnsProxyListener and FwmarkServer).
+ Remove the redundant mIfaceNetidMap.
+ Minor cosmetic changes to things such as #includes and log messages.
Change-Id: Ieb154589b24f00ba8067eaaec4def3534aec4923
Sreeram Ramachandran [Wed, 21 May 2014 20:08:34 +0000 (13:08 -0700)]
Mark sockets on creation (socket()) and accept4().
Continued from: https://android-review.git.corp.google.com/#/c/95094/
Change-Id: Ib0b8f5d7c5013b91eae6bbc3847852eb355c7714
Sreeram Ramachandran [Wed, 21 May 2014 18:41:39 +0000 (11:41 -0700)]
New network selection APIs.
Continued from: https://android-review.git.corp.google.com/#/c/94977/
Change-Id: Ie0576888f50a8ce91bbb0a4794708b406eb0aa35
Sreeram Ramachandran [Wed, 21 May 2014 15:54:07 +0000 (08:54 -0700)]
Move netd_client into netd.
Change-Id: Ie4b6b303225c93f2448a503d6ea9cebb552cbad5
The Android Automerger [Wed, 21 May 2014 12:51:08 +0000 (05:51 -0700)]
merge in master-release history after reset to
5ff58d494a5ddf7e544adc77590300f04daf8e31
The Android Automerger [Tue, 20 May 2014 12:51:24 +0000 (05:51 -0700)]
merge in master-release history after reset to
5ff58d494a5ddf7e544adc77590300f04daf8e31
The Android Automerger [Mon, 19 May 2014 12:50:25 +0000 (05:50 -0700)]
merge in master-release history after reset to
5ff58d494a5ddf7e544adc77590300f04daf8e31
The Android Automerger [Sun, 18 May 2014 12:50:12 +0000 (05:50 -0700)]
merge in master-release history after reset to
5ff58d494a5ddf7e544adc77590300f04daf8e31
The Android Automerger [Sat, 17 May 2014 12:51:51 +0000 (05:51 -0700)]
merge in master-release history after reset to
5ff58d494a5ddf7e544adc77590300f04daf8e31