Modern network agents are supposed to have the following
lifecycle :
• Construct the agent
• Declare itself to connectivity by calling #register()
• Possibly do some setup
• Declare connected by calling #markConnected()
• Declare disconnected by calling #unregister()
Internally ConnectivityService still manages the state
with deprecated NetworkInfo, which has a lot of possible
detailed states, most of which are now unused. Internally,
the steps above translate to the states below :
• When the agent calls #register() it starts in state
CONNECTING
• When it calls #markConnected() it goes to CONNECTED
• When it calls #unregister() it goes to DISCONNECTED which
triggers the flow that disconnects it immediately
• ConnectivityService also synthesizes the SUSPENDED state
under some situations, but that's not relevant here.
The translation above is done by the NetworkAgent class,
which translates #register(), #markConnected() and
The only valid lifecyle for new agents is thus :
CONNECTING → CONNECTED → disconnect
↕ ↑
SUSPENDED ───┘
Note that in the above scheme, there is no space for the
UNKNOWN state, which is only used by legacy network agents.
However, the constructor for NetworkInfo still starts with
UNKNOWN. Ideally the constructor for NetworkInfo would set
the info to CONNECTING instead, but this would be a very
visible change in behavior affecting many apps (especially
their tests), so for backward compatibility we can't do
that.
New network agents however don't expose their NetworkInfo
to the client code, so they can just be made to start in
CONNECTING the way it was meant.
Note that this didn't really have a concrete impact beyond
printing some logs as by and large the ConnectivityService
code handles UNKNOWN and CONNECTING the same way, or more
exactly it handles transitions to CONNECTED from UNKNOWN
and CONNECTING in the same way. The exception to this is
VPNs, which are not using the new agent API yet, and the
upcoming conversion is why this fix is actually needed now.
Bug: 170904131
Test: FrameworksNetTests NetworkStackTests
Test: new test in aosp/1460705
Change-Id: Ib7d906a7b29a0b9f000b90bc78558a7402852fa3
Adding network capability to be used with networks to be only used by
OEM apps.
Bug: 171505922
Test: atest CtsNetTestCasesLatestSdk:NetworkCapabilitiesTest
Change-Id: Ib55ebec0e178b6317d92f17e1aff060dacdfb2b1
Create a new target service-connectivity to split
ConnectivityService from services.core.
Add ConnectivityServiceInitializer for initializing
ConnectivityService and add systemReady() in
ConnectivityManager so that SystemServer can call systemReady()
through ConnectivityManager which won't change current behavior.
Bug: 158268939
Test: make target-java, make host-java
atest FrameworksNetIntegrationTests
atest FrameworksNetTests
make, device can boot,
atest CtsStrictJavaPackagesTestCases
wifi and mobile data work.
Change-Id: I99401772ba9c1c34adca20040da3c7c72d86ddd9
Merged-In: Ie732bfaf381404af0bb599ca2f421a96e7aa4257
LinkProperties.compare{Dnses,ValidatedPrivateDnses,AllRoutes,
AllInterfaceNames} actually have no users.
Test: LinkPropertiesTest
Change-Id: Ic54ab3c5520fbbc2be1309aeaf1e5a7857dbd194
The utility classes are useful for mainline modules and their tests, for
example IpClientIntegrationTest.
Bug: 168868607
Test: m && atest NetworkStaticLibTests
Change-Id: If2b1613aa18a7990391e2d31cc2951ca93f1cf3c
Currently, the callbacks of stopping were fired when stop procedure
is started, because the upper layer apps only care about the reason
of stopping instead of stopping result. Thus, there is no need to
wait for the result comes back. However, this behavior generates
races if apps want to re-start keepalive immediately since the
resources are not released yet.
Fix: 134891441
Fix: 140305589
Test: atest com.android.server.ConnectivityServiceTest#testPacketKeepalives \
--rerun-until-failure 1000
Change-Id: I987776a9211a50e964c4675b747bc10e203750f1
This patch is still needed and should go in now that the
error is fixed.
The patch was submitted into rvc-dev and is already in
rvc-dev-plus-aosp (patch in in ag/11923559, revert
skipped in ag/12072199). A follow-up will remove the
unused services.net-module-wifi target.
Test: originally tested in aosp/1324109
Test: m; manual: flashed, wifi and telephony working
Test: atest NetworkStackCoverageTests
Change-Id: I1074eedb0b0f156a1135e11210ec102de15ea674
Merged-In: Icd141a992c46290c74929785e261a1cd57bc001b
For non-telephony networks, this was always set to 0 before R.
In R, it is currently set to the same value as the network type.
This is incorrect because the two have different namespaces.
or example, currently, any network of type WIFI (==1) will have
a subtype of NETWORK_TYPE_GPRS (==1). Similarly, all ETHERNET
networks will have subtype NETWORK_TYPE_1XRTT, all VPN networks
will have a subtype of NETWORK_TYPE_TD_SCDMA, etd.
Bug: 161653721
Test: builds, boots
Change-Id: I07e111c1762e0021c931cefc27f193f78578748b
(cherry picked from commit a40ff70cad00a6c298e07c2c38e95d58074a7157)
Lorenzo Colittiauthored and android-build-team Robotcommited 2020-07-29 04:04:42 +00:00
Avoid recording negative data by doing integer-based
math as much as possible, but switch to double-based
math if we detect that we'd end up causing an overflow.
Test :
- Builds, Boots
- High data throughput scenarios
- NetworkStatsHistoryTest, NetworkStatsCollectionTest,
NetworkStatsTest
Bug: 119527458
Change-Id: I55d4aa649bea3ae201ad513074ff2b71cb918dad
Update all callers to use ProxyInfo(Uri) and remove ProxyInfo(String).
Bug: 160557668
Test: build pass
atest FrameworksNetTests
Change-Id: I4c980764c8839c48acf03208bb68cd53eb910121
This reverts commit c98e38195d.
Reason for revert: Breaks the build b/160390899
Merged-In: Icd141a992c46290c74929785e261a1cd57bc001b
Change-Id: Ie10e66f61393602f17fbb61bf17230b176bf1f44
Chalard Jeanauthored and Remi NGUYEN VANcommited 2020-07-02 23:47:35 +00:00
The classes should not be picked up from frameworks/base, as they are
part of several mainline modules.
Also refine comments in DhcpResults following feedback in previous
change.
(cherry-pick from internal branch skipping wifi classes; no other
conflict)
Bug: 151052811
Test: m; manual: flashed, wifi and telephony working
Test: atest NetworkStackCoverageTests
Merged-In: I7074651c6a2a7a6b11bcf13cc4bb03833d7d655f
Change-Id: I7074651c6a2a7a6b11bcf13cc4bb03833d7d655f
This is a Client-only solution.
- Add to NetdClient a per-process std::atomic_boolean
similar to netIdForProcess and netIdForResolv.
- The boolean says whether the process should be
allowed Internet connectivity.
- Add an @hide method to NetUtils.java to set the boolean;
call it from the initialization code of the new
process just after forking from zygote.
- Make netdClientSocket and dnsOpenProxy check the
boolean. If the boolean is false, return EPERM from
socket calls.
Bug: 150028556
Test: atest NetworkUtilsTest
Test: atest CtsAppSecurityHostTestCases:UseProcessTest
(cherry-pick from internal branch, ag/11881939)
Merged-In: If002280fbad493dfc2db3d9d505c0257d49a9056
Change-Id: If002280fbad493dfc2db3d9d505c0257d49a9056
This class might be used by some mainline modules.
Bug: 151052811
Test: atest DnsPacketTest
Test: atest DnsResolverTest
(clean cherry-pick from internal branch)
Merged-In: I8841d91456952ded5efbf8ea221289aecc7746ad
Change-Id: I8841d91456952ded5efbf8ea221289aecc7746ad
This class might be used by some mainline modules.
Bug: 151052811
Test: atest DnsPacketTest
Test: atest DnsResolverTest
Change-Id: I8841d91456952ded5efbf8ea221289aecc7746ad
This is a Client-only solution.
- Add to NetdClient a per-process std::atomic_boolean
similar to netIdForProcess and netIdForResolv.
- The boolean says whether the process should be
allowed Internet connectivity.
- Add an @hide method to NetUtils.java to set the boolean;
call it from the initialization code of the new
process just after forking from zygote.
- Make netdClientSocket and dnsOpenProxy check the
boolean. If the boolean is false, return EPERM from
socket calls.
Bug: 150028556
Test: atest NetworkUtilsTest
Test: atest CtsAppSecurityHostTestCases:UseProcessTest
Change-Id: If002280fbad493dfc2db3d9d505c0257d49a9056
Exempt-From-Owner-Approval: OWNERS already approved identical patchset 5
On Android different interfaces usually use different routing tables.
As a result, a change in interface should not be treated as route
update, but rather a remove and an add.
This change fixes a bug in VPN seamless handover where routes
failed to be updated when a new tunnel interface replaces the existing
one within the same network.
Bug: 158696878
Test: atest com.android.cts.net.HostsideVpnTests
Test: atest NetworkStackTests
Test: atest CtsNetTestCases
Test: atest FrameworksNetTests
Original-Change: https://android-review.googlesource.com/1331916
Merged-In: I57987233d42a0253eaee2e1ca5f28728c2354620
Change-Id: I57987233d42a0253eaee2e1ca5f28728c2354620
Treehugger Robotauthored and Lorenzo Colitticommited 2020-06-17 13:34:11 +00:00
On Android different interfaces usually use different routing tables.
As a result, a change in interface should not be treated as route
update, but rather a remove and an add.
This change fixes a bug in VPN seamless handover where routes
failed to be updated when a new tunnel interface replaces the existing
one within the same network.
Bug: 158696878
Test: atest com.android.cts.net.HostsideVpnTests
Test: atest NetworkStackTests
Test: atest CtsNetTestCases
Test: atest FrameworksNetTests
Change-Id: I57987233d42a0253eaee2e1ca5f28728c2354620
This commit changes agentConnect to set the owner UID as the mOwnerUid
field instead of the Binder.getCallingUid().
Binder.getCallingUid() can return incorrect results for platform VPNs,
as agentConnect() is called under a clean calling UID.
Additionally, this relaxes the ownerUid sanitization check to allow a
VPN network's owner to see it's own ownership information.
Vpn.mOwnerUid is guaranteed to be correct, as all VPNs MUST have called
prepareInternal() at some previous point, which sets mOwnerUid as the
package's UID (or SYSTEM_UID if this is legacy VPN).
Bug: 150135470
Test: CTS tests showing ownership information
Merged-In: Ic979dad73983d722365849fbfb0becfd432b894c
Change-Id: Ic979dad73983d722365849fbfb0becfd432b894c
(cherry picked from commit e29bf99a7fc1067c546d7fa6cbcb9001fb110d16)
The classes should not be picked up from frameworks/base, as they are
part of several mainline modules.
Also refine comments in DhcpResults following feedback in previous
change.
Bug: 151052811
Test: m; manual: flashed, wifi and telephony working
Test: atest NetworkStackCoverageTests
Change-Id: I7074651c6a2a7a6b11bcf13cc4bb03833d7d655f
Some developers have been surprised by this limitation and had trouble
figuring out what the issue was. Add documentation to address this.
This also includes a drive-by removal of a duplicate check.
Bug: 149867479
Test: doc-only change
Original-Change: https://android-review.googlesource.com/1313813
Merged-In: I5911d01984695550b6c9afe7a8eb535bf5e320a1
Change-Id: I5911d01984695550b6c9afe7a8eb535bf5e320a1
James Mattis
commited 
Chalard Jean
commited 
Anton Hansson
authored and
Gerrit Code Review
commited 
Aaron Huang
authored and
Remi NGUYEN VAN
commited 
Jeff Sharkey
commited 
Xin Li
commited 
Junyu Lai
authored and
Lorenzo Colitti
authored and
android-build-team Robot
commited 
Chiachang Wang
authored and
Pavan Kumar M
commited 
Automerger Merge Worker
commited 
Luke Huang
authored and
Android (Google) Code Review
commited 
TreeHugger Robot
authored and
Rubin Xu
commited 
Benedict Wong
commited