When the app_process is shutting down the main thread will close the
binder fd while pool threads are executing an ioctl (in
IPCThreadState::stopProcess called by AppRuntime::onStarted in
app_main.c).
The binder driver will then return all pending calls in ioctl
without any error and with a command. One of the threads gets a
BR_SPAWN_LOOPER which will create a new thread (the other thread
gets a BR_NOOP). This new thread then calls
vm->AttachCurrentThread. Usually this results in a log entry with
"AndroidRuntime: NOTE: attach of thread 'Binder Thread #3' failed",
but sometimes it also causes a SIGSEGV. This depends on the timing
between the new thread an the main thread that calls DestroyJavaVM
(in AndroidRuntime::start).
If IPCThreadState.cpp is compiled with "#define LOG_NDEBUG 0" the
pool thread will loop and hit the
ALOG_ASSERT(mProcess->mDriverFD >= 0) in
IPCThreadState::talkWithDriver.
Crashes like this has been seen when running the am command and
other commands that use the app_process.
This fix makes sure that any command that is received when the driver
fd is closed are ignored and IPCThreadState::talkWithDriver instead
returns an error which will cause the pool thread to exit and detach
itself from the vm. A check to avoid calling ioctl to a fd with -1
was also added in IPCThreadState::threadDestructor.
Another solution might be to change the binder driver so that it
returns an error when the fd is closed (or atleast not a
BR_SPAWN_LOOPER command). It might also be possible to call exit(0)
which is done when System.exit(0) is called from java.
Change-Id: I3d1f0ff64896c44be2a5994b3a90f7a06d27f429
The Java implementation of writing the RPC response header
calculates the length of the header including the 4 bytes
specifying the header length but the native implementation
excludes the 4 bytes specifying the length from the header
length.
The native implementation has been aligned to the Java impl.
Change-Id: I325bf272a63152d8fded4cf4e51a906b5a9bfe19
Every IBinder object can accept a new transaction to tell it that
it might want to reload system properties, and in the process
anyone can register a callback to be executed when this happens.
Use this to reload the trace property.
This is very much ONLY for debugging.
Change-Id: I55c67c46f8f3fa9073bef0dfaab4577ed1d47eb4
Remove C++ APIs androidSetThreadSchedulingGroup and
androidGetThreadSchedulingGroup, and the ANDROID_TGROUP_* constants.
Former callers of these should now use the C APIs set_sched_policy and
get_sched_policy, and the SP_* constants.
Note: debug.sys.noschedgroups is not supported by the C APIs,
this needs to be discussed.
Change-Id: I32bbfc539ef4090faf9ef0320380e8cca9eae07c
As part of scheduling policy cleanup, remove or isolate
all references to the scheduling policy APIs.
Change-Id: Ia1ea2fe711a399039f25217309e061267744b856
prctl(PR_SET_NAME) limits to 15 characters. Before we had names like
"Binder Thread #" and the counter was cut off :-( Also remove redundant
"thread" at end of name; it's always a thread.
Change-Id: I1f99c2730ba0787ed9b59c15914356cddf698e2f
MemoryHeapPmem is not used any longer. PMEM is not a supported
type of memory by the system anymore. a particular device might
use PMEM and need something like MemoryHeapPmem, in this case this
should be implemented in device specific code (HAL).
This will most likely break older no longer supported targets.
Change-Id: I434e4291219950018de8b793b0403bb2d92dd5cc
Services now must explicitly opt in to being accessed by isolated
processes. Currently only the activity manager and surface flinger
allow this. Activity manager is needed so that we can actually
bring up the process; SurfaceFlinger is needed to be able to get the
display information for creating the Configuration. The SurfaceFlinger
should be safe because the app doesn't have access to the window
manager so can't actually get a surface to do anything with.
The activity manager now protects most of its entry points against
isolated processes.
Change-Id: I0dad8cb2c873575c4c7659c3c2a7eda8e98f46b0
Switching activity stacks
Cache ContentProvider per user
Long-press power to switch users (on phone)
Added ServiceMap for separating services by user
Launch PendingIntents on the correct user's uid
Fix task switching from Recents list
AppWidgetService is mostly working.
Commands added to pm and am to allow creating and switching profiles.
Change-Id: I15810e8cfbe50a04bd3323a7ef5a8ff4230870ed
Currently, madvise(MADV_REMOVE) is called after deallocation.
Another thread might allocate (and even write) the same region between
deallocation and madvise(), in which case the new thread will fail to read
what it have written. So, call deallocate() after madvise(MADV_REMOVE).
Bug: 5654596
Change-Id: I26f36cd6013de499090768a0ddc68206a4a68219
Was mistakenly assuming that Parcel::writeFileDescriptor took
ownership of the fd that was passed in. It does not!
Added some comments and a default parameter to allow the caller
to specify whether it wishes the Parcel to take ownership.
Bug: 5563374
Change-Id: I5a12f51d582bf246ce90133cce7690bb9bca93f6
On user-debug and eng builds, you can set the
"db.log.slow_query_threshold" system property to queries that
take longer than the specified number of milliseconds.
Set it to 0 to log all queries.
This property has been around for a while but it was implemented
poorly. In particular, it *changed* the behavior of the query
by calling getCount() while holding the Db connection.
In normal operation, the query will not actually run until later.
By putting the timing logic into fillWindow() instead, we ensure
that we only measure queries that actually ran. We also capture
cases where the cursor window gets filled multiple times.
Bug: 5520301
Change-Id: I174f5e1ea15831a1d22a36e9a804d7755f230b38
Bug: 5520301
When an application requests a row from a SQLiteCursor that
is not in the window, instead of filling from the requested
row position onwards, fill from a little bit ahead of the
requested row position.
This fixes a problem with applications that seek backwards
in large cursor windows. Previously the application could
end up refilling the window every time it moved back
one position.
We try to fill about 1/3 before the requested position and
2/3 after which substantially improves scrolling responsiveness
when the list is bound to a data set that does not fit
entirely within one cursor window.
Change-Id: I168ff1d3aed1a41ac96267be34a026c108590e52
There is no difference and has never really been a difference
between local-only and remotable CursorWindows. By removing the
distinction officially in the API, we will make it easier to
implement CrossProcessCursor correctly. CrossProcessCursor
is problematic currently because it's not clear whether a call
to getWindow() will return a local-only window or a remotable window.
As a result, the bulk cursor adaptor has special case handling
for AbstractWindowedCursors vs. ordinary CrossProcessCursors
so that it can set a remotable window before the cursor fills it.
All these problems go away if we just forget about local-only
windows being special in any way.
Change-Id: Ie59f517968e33d0ecb239c3c4f60206495e8f376
Bug: 5332296
The memory dealer introduces additional delays for reclaiming
the memory owned by CursorWindows because the Binder object must
be finalized. Using ashmem instead gives CursorWindow more
direct control over the lifetime of the shared memory region.
The provider now allocates the CursorWindows and returns them
to clients with a read-only protection bit set on the ashmem
region.
Improved the encapsulation of CursorWindow. Callers shouldn't
need to care about details like how string fields are allocated.
Removed the compile-time configuration of string and numeric
storage modes to remove some dead weight.
Change-Id: I07c2bc2a9c573d7e435dcaecd269d25ea9807acd
Bug: 5332296
The code is functionally equivalent, but a little more efficient
and much easier to maintain.
Change-Id: I90670a13799df05831843a5137ab234929281b7c
This reverts commit 56c58f66b97d22fe7e7de1f7d9548bcbe1973029
This CL was causing the browser to crash when adding bookmarks, visiting the bookmarks page, and sharing pages (see bug http://b/issue?id=5369231
Updated the command name lists, and masked off the additional bits in
the command word when doing the name lookup.
Made descriptor values easier to grep for and consistent with kernel
output (i.e. decimal rather than hex). Attempt to show transaction
descriptors as such (they're in a union with a pointer).
Also, the writeLines() function in Static was using a no-op
logging call to write an iovec. It looks like all callers are using
N=1, so I just added a log for the first string.
Bug 5155269
Change-Id: I417b8d77da3eb6ee1d2069ba94047210f75738bc
This is intended to absorb the cost of the IPC
to the permission controller.
Cached permission checks cost about 3us, while
full blown ones are two orders of magnitude slower.
CAVEAT: PermissionCache can only handle system
permissions safely for now, because the cache is
not purged upon global permission changes.
Change-Id: I8b8a5e71e191e3c01e8f792f253c379190eee62e
The offset that is used in the creation of the MemoryHeapBase must be saved, so
that it can be used to recreate the Heap when an IMemory object is passed
across process boundary through the binder.
Change-Id: Ie618fb5c0718e6711f55ed9235616fd801e648dc
Signed-off-by: Anu Sundararajan <sanuradha@ti.com>
Signed-off-by: Iliyan Malchev <malchev@google.com>
We now write battery history directly into a buffer, instead of
creating objects. This allows for more efficient storage; later
it can be even better because we can only write deltas.
The old code is still there temporarily for validation.
Change-Id: I9707d4d8ff30855be8ebdc93bc078911040d8e0b
HAVE_ANDROID_OS was defined as "1" for targets, but never defined as "0"
for non-targets. Changing them to #ifdef should be safe and matches
all the other uses of HAVE_ANDROID_OS throughout the system.
Change-Id: I82257325a8ae5e4e4371ddfc4dbf51cea8ea0abb
If writeString8 is called with the following sequence:
writeString8(String8(""));
writeString8(String8("TempString"));
Then in the readString8, the 2nd String i.e. "TempString" is not read,
instead an empty string is read.
The bug comes because of the write call for String8("") where there are
no String bytes present. In the write Statement, an extra ‘\0’ is
written. During the Marshalling, Following bytes are written:
1 2 3 4 5 ...
0x0 0x0 0xB ‘T’ ‘e’ ...
The readString8 function has a check that, if String length is 0, don’t
read anything. So the first byte is read as the length for the first
string. The second byte i.e. ‘\0’ is read as the length for the second
string and hence the second string becomes empty too.
Change-Id: Id7acc0c80ae16e77be4331f1ddf69ea87e758420
grief from people who think this message is bad news.
but in reality, this message is really just an informational message
to aid in debugging
Change-Id: I1a2ab1666a27adb7d3fd210528b2c5218640d53d
- New feature to "am monitor" to have it automatically launch
gdbserv for you when a crash/ANR happens, and tell you how to
run the client.
- Update dumpstate to match new location of binder debug logs
- Various commented out logs that are being used to track down
issues.
Change-Id: Ia5dd0cd2df983a1fc6be697642a4590aa02a26a5
This was causing stack stitching problems where a one-way call with
violations followed by a two-way call without violations was getting
the previous one-way call's violation stack stitched on to the second
caller's stack.
The solution is a little more indirect than I would've liked
(preserving the binder's onTransact flags until enforceInterface) but
was seemingly necessary to work without changing the AIDL compiler.
It should also be sufficiently cheap, since no new calls to
thread-local IPCThreadState lookups were required. The additional
work is just same-thread getter/setters on the existing
IPCThreadState.
Change-Id: I4b6db1d445c56e868e6d0d7be3ba6849f4ef23ae
Now, when Thread A has a strict mode policy in effect and does a
Binder call to Thread B (most likely in another process), the strict
mode policy is passed along, but with the GATHER penalty bit set which
overrides other policies and instead gathers all offending stack
traces to a threadlocal which are then written back in the Parcel's
reply header.
Change-Id: I7d4497032a0609b37b1a2a15855f5c929ba0584d
Add native Parcel methods analogous to the Java versions.
Currently, these don't do much, but upcoming StrictMode work changes
the RPC calling conventions in some cases, so it's important that
everybody uses these consistently, rather than having a lot of code
trying to parse RPC responses out of Parcels themselves.
As a summary, the current convention that Java Binder services use is
to prepend the reply Parcel with an int32 signaling the exception
status:
0: no exception
-1: Security exception
-2: Bad Parcelable
-3: ...
-4: ...
-5: ...
... followed by Parceled String if the exception code is non-zero.
With an upcoming change, it'll be the case that a response Parcel can,
non-exceptionally return rich data in the header, and also return data
to the caller. The important thing to note in this new case is that
the first int32 in the reply parcel *will not be zero*, so anybody
manually checking for it with reply.readInt32() will get false
negative failures.
Short summary: If you're calling into a Java service and manually
checking the exception status with reply.readInt32(), change it to
reply.readExceptionCode().
Change-Id: I23f9a0e53a8cfbbd9759242cfde16723641afe04
Merge commit 'efcf68aa1fd7fcfd52cf3d2837ed8db8e797194b'
* commit 'efcf68aa1fd7fcfd52cf3d2837ed8db8e797194b':
Start of work on passing around StrictMode policy over Binder calls.
This is (intendend to be) a no-op change.
At this stage, Binder RPCs just have an additional uint32 passed around
in the header, right before the interface name. But nothing is actually
done with them yet. That value should right now always be 0.
This now boots and seems to work.
Change-Id: I135b7c84f07575e6b9717fef2424d301a450df7b
To allow use of the native CursorWindow class outside of the core framework jni
Change-Id: I72e8dcb91a2c691130c33cdfd9a25d343da1c592
Signed-off-by: Mike Lockwood <lockwood@android.com>
Fix some small static-initialization-order issues (and a static-
initializers-missing issue) that result from doing so. The static
libraries don't actually get used for anything real at the moment --
they're used for perf tests of bug 2660235.
Bug: 2660235
Change-Id: Iee2f38f79cc93b395e8d0a5a144ed92461f5ada0
get rid off the MAP_ONCE flag is MemoryHeapBase (as well as it's functionality),
this feature should not be used anymore.
the software renderer was incorrectly using the default ctor which set MAP_ONCE,
causing the leak. the software renderer itself is incorrectly used while coming
back from sleep.
Change-Id: I123621f8d140550b864f352bbcd8a5729db12b57
On binder incalls, the handler thread is given the caller's priority by the
driver, but not the caller's cgroup. We have explicit code that sets the
handler's cgroup to match the caller's, *except* that the system process
explicitly disables this behavior. This led to a siuation in which we were
running binder incalls to the system process at nice=10 but cgroup=fg.
That's fine as far as it goes, except that if a GC happened in the handler
thread, it would be promoted to foreground priority and cgroup both, to avoid
having the GC take forever. Then, when GC finished, the original priority
is reset, and the cgroup set *based on that priority*. This would push the
handler thread into nice=10 cgroup=bg_non_interactive -- which matches the
caller, but is supposed to be impossible in the system process.
The end result of this was that we could be running "lengthy" operations in
the system process in the background. Unfortunately, some of the operations
that wound up like this would hold important global system locks for up to
twenty seconds as a result, making the entire device unresponsive to input
for that period.
This CL fixes the binder incall setup to ensure that within the system process,
a binder incall is always begun from the normal foreground priority as well
as cgroup. In practice now the device still becomes laggy/sluggish when the
offending lock-holding time-consuming incall occurs, but since it still runs
as a foreground task it is able to proceed to completion within a short time
rather than taking 20 seconds.
Fixes bug #2403717
Change-Id: Id046aeabd0e80c48eef94accc37842835eab308d
Add a Flattenable interface to libutils which can be used to flatten
an object into bytestream + filedescriptor stream.
Parcel is modified to handle Flattenable. And GraphicBuffer implements
Flattenable.
Except for the overlay classes libui is now independent of libbinder.
At some point the implementation became complicated because of
SurfaceFlinger's special needs, since we are now relying on gralloc
we can go back to much simpler MemoryDealer.
Removed HeapInterface and AllocatorInterface, since those don't need
to be paramterized anymore. Merged SimpleMemory and Allocation.
Made SimplisticAllocator non virtual.
Removed MemoryDealer flags (READ_ONLY, PAGE_ALIGNED)
Removed a lot of unneeded code.
This is a very simply implementation: upon receiving an IPC, if the handling
thread is at a background priority (the driver will have taken care of
propagating this from the calling thread), then stick it in to the background
scheduling group. Plus an API to turn this off for the process, which is
used by the system process.
This also pulls some of the code for managing scheduling classes out of
the Process JNI wrappers and in to some convenience methods in thread.h.
in the kernel requires a guard page, so 1M allocations fragment memory very
badly. Subtracting a couple of pages so that they fit in a power of
two allows the kernel to make more efficient use of its virtual address space.
Signed-off-by: Rebecca Schultz Zavin <rebecca@android.com>
To prevent buggy command implementations from poisoning binder threads'
scheduling class & priority for future command execution, we now reset the
cgroup and thread priority to foreground/normal when a binder service thread
finishes executing the designated command.
Change-Id: Ibc0ab2485751453f6dc96fdb4eb877fd02796e3f
- make sure that all binder Bn classes define a ctor and dtor in their respective library.
This avoids duplication of the ctor/dtor in libraries where these objects are instantiated.
This is also cleaner, should we want these ctor/dtor to do something one day.
- same change as above for some Bp classes and various other non-binder classes
- moved the definition of CHECK_INTERFACE() in IInterface.h instead of having it everywhere.
- improved the CHECK_INTERFACE() macro so it calls a single method in Parcel, instead of inlining its code everywhere
- IBinder::getInterfaceDescriptor() now returns a "const String16&" instead of String16, which saves calls to String16 and ~String16
- implemented a cache for BpBinder::getInterfaceDescriptor(), since this does an IPC. HOWEVER, this method never seems to be called.
The cache makes BpBinder bigger, so we need to figure out if we need this method at all.
Merge commit 'd50a458bb291801ab9fdc119301bc7b84b42a6e3'
* commit 'd50a458bb291801ab9fdc119301bc7b84b42a6e3':
Fix a major bug in Bundle when unparcelling from AIDL.
