Commit c918cbc added this reference to ensure
sun.misc.Unsafe.getLongVolatile could be implemented efficiently on
32-bit platforms. However, I neglected to ensure the reference was
updated to point to the final class instance instead of the temporary
one used in parseClass. This led to extra memory usage and
inconsistent locking behavior, plus broken bootimage builds.
If we don't clear these references, we risk finalizing objects which
can still be reached by one of the special reference types.
It's a bit of a chicken-and-egg problem. We need to visit finalizable
objects before visiting weak references, since some of the weak
references and/or their targets may become reachable once the
finalizable objects are visited. However, that ordering means we have
no efficient way of distinguishing between objects which are reachable
from one or more normal GC roots and those which are only reachable
via the finalization queue. The solution is to clear all weak
references to finalizable objects before visiting them.
The original stub implementation just echoed back its argument, but
that confused URLClassLoader when dealing with sealed JARs --
returning a non-null value for a non-system class from
JVM_GetSystemPackage made URLClassloader think it had already loaded a
class from a package which was supposed to be sealed, resulting in
SecurityExceptions which ultimately triggered NoClassDefFoundErrors.
The solution is to only return non-null values for actual system
classes.
We weren't wrapping exceptions thrown by invoked methods in
InvocationTargetExceptions in JVM_InvokeMethod or
JVM_NewInstanceFromConstructor. Also, JVM_GetCallerClass is supposed
to ignore Method.invoke frames when walking the stack.
My earlier fix (f8e8609) was almost -- but not quite -- sufficient.
It asked the heap to mark the dead fixies too early, so some of them
were marked dead even though they ultimately survived, causing us to
clear weak JNI references when we shouldn't.
The existing code did not handle static field lookups for
synchronization on 32-bit systems, which is necessary because such
systems generally don't support atomic operations on 64-bit values.
resolveClass was correctly respecting throw_ == false if the requested
class was not found, but it still threw an exception if e.g. the
superclass was missing. Now we catch such exceptions and return null
as appropriate.
This led to fixed-position objects being considered unreachable when
they were actually still reachable, causing global weak JNI references
to be cleared prematurely, most notably leading to crashes in AWT
buffered image code.
This commit also fixes a field offset calculation mismatch in
bootimage.cpp relative to machine.cpp.
We were assuming the array element size was always the native word
size, which is not correct in general for primitive arrays, and this
led to wasted space at best and memory corruption at worst.
The first problem was that, on x86, we failed to properly keep track
of whether to expect the return address to be on the stack or not when
unwinding through a frame. We were relying on a "stackLimit" pointer
to tell us whether we were looking at the most recently-called frame
by comparing it with the stack pointer for that frame. That was
inaccurate in the case of a thread executing at the beginning of a
method before a new frame is allocated, in which case the most recent
two frames share a stack pointer, confusing the unwinder. The
solution involves keeping track of how many frames we've looked at
while walking the stack.
The other problem was that compareIpToMethodBounds assumed every
method was followed by at least one byte of padding before the next
method started. That assumption was usually valid because we were
storing the size following method code prior to the code itself.
However, the last method of an AOT-compiled code image is not followed
by any such method header and may instead be followed directly by
native code with no intervening padding. In that case, we risk
interpreting that native code as part of the preceding method, with
potentially bizarre results.
The reason for the compareIpToMethodBounds assumption was that methods
which throw exceptions as their last instruction generate a
non-returning call, which nonetheless push a return address on the
stack which points past the end of the method, and the unwinder needs
to know that return address belongs to that method. A better solution
is to add an extra trap instruction to the end of such methods, which
is what this patch does.
It seems that GCC 4.6.1 gets confused at LTO time when we take the
address of inline functions, so I'm switching them to non-inline
linkage to make it happy.
It seems that GCC 4.6.1 gets confused at LTO time when we take the
address of inline functions, so I'm switching them to non-inline
linkage to make it happy.
Apple's linker tends to remove functions which are never called, which
is not what we want for e.g. vmPrintTrace, since that function is only
intended to be called interactively from within GDB.
My previous attempt wasn't quite sufficient, since it was too late to
call join on a thread which had already exited given the code was
written to aggressively dispose of system handles as soon as the
thread exited. The solution is to delay disposing these handles until
after we're able to join the thread.
The bug here is that when a thread exits and becomes a "zombie", the
OS resources associated with it are not necessarily released until we
actually join and dispose of that thread. Since that only happens
during garbage collection, and collection normally only happens in
response to heap memory pressure, there's no guarantee that we'll GC
frequently enough to clean up zombies promptly and avoid running out
of resources.
The solution is to force a GC whenever we start a new thread and there
are at least N zombies waiting to be disposed, where N=16 for now.
There was a subtle race condition in the VM shutdown process such that
a System::Thread would be disposed after the System instance it was
created under has been disposed, in which case doing a virtual call to
System::free with that instance would potentially cause a crash. The
solution is to just call the C library version of free directly, since
that's all System::free does.
Until now, the bootimage build hasn't supported using the Java
invocation API to create a VM, destroy it, and create another in the
same process. Ideally, we would be able to create multiple VMs
simultaneously without any interference between them. In fact, Avian
is designed to support this for the most part, but there are a few
places we use global, mutable state which prevent this from working.
Most notably, the bootimage is modified in-place at runtime, so the
best we can do without extensive changes is to clean up the bootimage
when the VM is destroyed so it's ready for later instances. Hence
this commit.
Ultimately, we can move towards a fully reentrant VM by making the
bootimage immutable, but this will require some care to avoid
performance regressions. Another challenge is our Posix signal
handlers, which currently rely on a global handle to the VM, since you
can't, to my knowledge, pass a context pointer when registering a
signal handler. Thread local variables won't necessarily help, since
a thread might attatch to more than one VM at a time.
This avoids the requirement of putting the code image in a
section/segment which is both writable and executable, which is good
for security and avoids trouble with systems like iOS which disallow
such things.
The implementation relies on relative addressing such that the offset
of the desired address is fixed as a compile-time constant relative to
the start of the memory area of interest (e.g. the code image, heap
image, or thunk table). At runtime, the base pointer to the memory
area is retrieved from the thread structure and added to the offset to
compute the final address. Using the thread pointer allows us to
generate read-only, position-independent code while avoiding the use
of IP-relative addressing, which is not available on all
architectures.
This monster commit is the first step towards supporting
cross-architecture bootimage builds. The challenge is to build a heap
and code image for the target platform where the word size and
endianess may differ from those of the build architecture. That means
the memory layout of objects may differ due to alignment and size
differences, so we can't just copy objects into the heap image
unchanged; we must copy field by field, resizing values, reversing
endianess and shifting offsets as necessary.
This commit also removes POD (plain old data) type support from the
type generator because it added a lot of complication and little
value.
Internally, the VM augments the method tables for abstract classes
with any inherited abstract methods to make code simpler elsewhere,
but that means we can't use that table to construct the result of
Class.getDeclaredMethods since it would include methods not actually
declared in the class. This commit ensures that we preserve and use
the original, un-augmented table for that purpose.
Previously, we would abort the process if we encountered a truncated
multibyte character in parseUtf8NonAscii (called by the JNI method
NewStringUTF). Now we simply terminate the string at that point.
Also, assume any class which has an ancestor class which has a static
initializer needs initialization even if it doesn't have one itself,
per the Java Language Spec.
The result of Class.getInterfaces should not include interfaces
declared to be implemented/extended by superclasses/superinterfaces,
only those declared by the class itself. This is important because it
influences how java.io.ObjectStreamClass calculates serial version
IDs.
This includes a proper implementation of JVM_ActiveProcessorCount, as
well as JVM_SetLength and JVM_NewMultiArray. Also, we now accept up
to JNI_VERSION_1_6 in JVM_IsSupportedJNIVersion.
We must not allocate heap objects from doCollect, since it might
trigger a GC while one is already in progress, which can cause trouble
when we're still queuing up objects to finalize, among other things.
To avoid this, I've added extra fields to the finalizer and cleaner
types which we can use to link instances up during GC without
allocating new memory.
OpenJDK uses an alternative to Object.finalize for resource cleanup in
the form of sun.misc.Cleaner. Normally, OpenJDK's
java.lang.ref.Reference.ReferenceHandler thread handles this, calling
Cleaner.clean on any instances it finds in its "pending" queue.
However, Avian handles reference queuing internally, so it never
actually adds anything to that queue, so the VM must call
Cleaner.clean itself.
The main changes here are:
* fixes for runtime annotation support
* proper support for runtime generic type introspection
* throw NoClassDefFoundErrors instead of ClassNotFoundExceptions
where appropriate
This commit ensures that we use the proper memory barriers or locking
necessary to preserve volatile semantics for such fields when accessed
or updated via JNI.
Unlike the interpreter, the JIT compiler tries to resolve all the
symbols referenced by a method when compiling that method. However,
this can backfire if a symbol cannot be resolved: we end up throwing
an e.g. NoClassDefFoundError for code which may never be executed.
This is particularly troublesome for code which supports multiple
APIs, choosing one at runtime.
The solution is to defer to stub code for symbols which can't be
resolved at JIT compile time. Such a stub will try again at runtime
to resolve the needed symbol and throw an appropriate error if it
still can't be found.
It is possible to create an Exception with no stack trace by
overriding Throwable.fillInStackTrace, so we can't assume any given
instance will have one.
There was a race between these two functions such that one thread A
would run dispose on thread B just before thread B finishes exit, with
the result that Thread::lock and/or Thread::systemThread would be
disposed twice, resulting in a crash.
It seems that older versions of GCC (4.0 and older, at least) generate
assembly files with duplicate symbols for function templates which
differ only by the attributes of the templated types. Newer versions
have no such problem, but we need to support both, hence the
workaround in this commit of using a dedicated, non-template "alias"
function where we previously used "cast<alias_t>".
We use a template function called "cast" to get raw access to fields
in in the VM. In particular, we use this function in util.cpp to
treat reference fields as intptr_t fields so we can use the least
significant bit as the red/black flag in red/black tree nodes.
Unfortunately, this runs afoul of the type aliasing rules in C/C++,
and the compiler is permitted to optimize in a way that assumes such
aliasing cannot occur. Such optimization caused all the nodes in the
tree to be black, leading to extremely unbalanced trees and thus slow
performance.
The fix in this case is to use the __may_alias__ attribute to tell the
compiler we're doing something devious. I've also used this technique
to avoid other potential aliasing problems. There may be others
lurking, so a complete audit of the VM might be a good idea.
When loading a class which extends another class that contained a
field of primitive array type using defineClass in a bootimage=true
build, the VM was unable to find the primitive array class, and
makeArrayClass refused to create one since it should already have
existed.
The problem was that the bootimage=true build uses an empty
Machine::BootstrapClassMap, and resolveArrayClass expected to find the
primitive array classes there. The fix is to check the
Machine::BootLoader map if we can't find it in
Machine::BootstrapClassMap.
Previously, we unwound the stack by following the chain of frame
pointers for normal returns, stack trace creation, and exception
unwinding. On x86, this required reserving EBP/RBP for frame pointer
duties, making it unavailable for general computation and requiring
that it be explicitly saved and restored on entry and exit,
respectively.
On PowerPC, we use an ABI that makes the stack pointer double as a
frame pointer, so it doesn't cost us anything. We've been using the
same convention on ARM, but it doesn't match the native calling
convention, which makes it unusable when we want to call native code
from Java and pass arguments on the stack.
So far, the ARM calling convention mismatch hasn't been an issue
because we've never passed more arguments from Java to native code
than would fit in registers. However, we must now pass an extra
argument (the thread pointer) to e.g. divideLong so it can throw an
exception on divide by zero, which means the last argument must be
passed on the stack. This will clobber the linkage area we've been
using to hold the frame pointer, so we need to stop using it.
One solution would be to use the same convention on ARM as we do on
x86, but this would introduce the same overhead of making a register
unavailable for general use and extra code at method entry and exit.
Instead, this commit removes the need for a frame pointer. Unwinding
involves consulting a map of instruction offsets to frame sizes which
is generated at compile time. This is necessary because stack trace
creation can happen at any time due to Thread.getStackTrace being
called by another thread, and the frame size varies during the
execution of a method.
So far, only x86(_64) is working, and continuations and tail call
optimization are probably broken. More to come.
This rather large commit modifies the VM to use non-local returns to
throw exceptions instead of simply setting Thread::exception and
returning frame-by-frame as it used to. This has several benefits:
* Functions no longer need to check Thread::exception after each call
which might throw an exception (which would be especially tedious
and error-prone now that any function which allocates objects
directly or indirectly might throw an OutOfMemoryError)
* There's no need to audit the code for calls to functions which
previously did not throw exceptions but later do
* Performance should be improved slightly due to both the reduced
need for conditionals and because undwinding now occurs in a single
jump instead of a series of returns
The main disadvantages are:
* Slightly higher overhead for entering and leaving the VM via the
JNI and JDK methods
* Non-local returns can make the code harder to read
* We must be careful to register destructors for stack-allocated
resources with the Thread so they can be called prior to a
non-local return
The non-local return implementation is similar to setjmp/longjmp,
except it uses continuation-passing style to avoid the need for
cooperation from the C/C++ compiler. Native C++ exceptions would have
also been an option, but that would introduce a dependence on
libstdc++, which we're trying to avoid for portability reasons.
Finally, this commit ensures that the VM throws an OutOfMemoryError
instead of aborting when it reaches its memory ceiling. Currently, we
treat the ceiling as a soft limit and temporarily exceed it as
necessary to allow garbage collection and certain internal allocations
to succeed, but refuse to allocate any Java objects until the heap
size drops back below the ceiling.
There is a delay between when we tell the OS to start a thread and
when it actually starts, and during that time a thread might
mistakenly think it was the last to exit, try to shut down the VM, and
then block in joinAll when it finds it wasn't the last one after all.
The solution is to increment Machine::liveCount and add the new thread
to the process tree before starting it -- all while holding
Machine::stateLock for atomicity. This helps guarantee that when
liveCount is one, we can be sure there's really only one thread
running or staged to run.
If we don't do this, the VM will crash when it tries to create a stack
trace for the error because makeObjectArray will return null
immediately when it sees there is a pending exception.
When trying to create an array class, we try to resolve
java.lang.Object so we can use its vtable in the array class.
However, if Object is missing, we'll try to create and throw a
ClassNotFoundException, which requires creating an array to store the
stack trace, which requires creating an array class, which requires
resolving Object, etc.. This commit short-circuits this process by
telling resolveClass not to create and throw an exception if it can't
find Object.
While doing the above work, I noticed that the implementations of
Classpath::makeThrowable in classpath-avian.cpp and
classpath-openjdk.cpp were identical, so I made makeThrowable a
top-level function.
Finally, I discovered that Thread.setDaemon can only be called before
the target thread has been started, which allowed me to simplify the
code to track daemon threads in the VM.
* add libnet.so and libnio.so to built-in libraries for openjdk-src build
* implement sun.misc.Unsafe.park/unpark
* implement JVM_SetClassSigners/JVM_GetClassSigners
* etc.
The main change here is to use a lazily-populated vector to associate
runtime data with classes instead of referencing them directly from
the class which requires updating immutable references in the heap
image. The other changes employ other strategies to avoid trying to
update immutable references.
If the VM runs out of heap space and the "avian.heap.dump" system
property was specified at startup, the VM will write a heap dump to
the filename indicated by that property. This dump may be analyzed
using e.g. DumpStats.java.
My recent commit to ensure that OS resources are released immediately
upon thread exit introduced a race condition where interrupting or
joining a thread as it exited could lead to attempts to use
already-released resources. This commit adds locking to avoid the
race.
This makes heap dumps more useful since these classes are now refered
to by name instead of number.
This commit also adds a couple of utilities for parsing heap dumps:
PrintDump and DumpStats.
Previously, we waited until the next GC to do this, but that can be
too long for workloads which create a lot of short-lived threads but
don't do much allocation.
This allows OpenJDK to access time zone data which is normally found
under java.home, but which we must embed in the executable itself to
create a self-contained build. The VM intercepts various file
operations, looking for paths which start with a prefix specified by
the avian.embed.prefix property and redirecting those operations to an
embedded JAR.
For example, if avian.embed.prefix is "/avian-embedded", and code
calls File.exists() with a path of
"/avian-embedded/javahomeJar/foo.txt", the VM looks for a function
named javahomeJar via dlsym, calls the function to find the memory
region containing the embeded JAR, and finally consults the JAR to see
if the file "foo.txt" exists.
As described in readme.txt, a standalone OpenJDK build embeds all
libraries, classes, and other files needed at runtime in the resulting
binary, eliminating dependencies on external resources.
We now consult the JAVA_HOME environment variable to determine where
to find the system library JARs and SOs. Ultimately, we'll want to
support self-contained build, but this allows Avian to behave like a
conventional libjvm.so.
The main changes in this commit ensure that we don't hold the global
class lock when doing class resolution using application-defined
classloaders. Such classloaders may do their own locking (in fact,
it's almost certain), making deadlock likely when mixed with VM-level
locking in various orders.
Other changes include a fix to avoid overflow when waiting for
extremely long intervals and a GC root stack mapping bug.
The biggest change in this commit is to split the system classloader
into two: one for boot classes (e.g. java.lang.*) and another for
application classes. This is necessary to make OpenJDK's security
checks happy.
The rest of the changes include bugfixes and additional JVM method
implementations in classpath-openjdk.cpp.
Whereas the GNU Classpath port used the strategy of patching Classpath
with core classes from Avian so as to minimize changes to the VM, this
port uses the opposite strategy: abstract and isolate
classpath-specific features in the VM similar to how we abstract away
platform-specific features in system.h. This allows us to use an
unmodified copy of OpenJDK's class library, including its core classes
and augmented by a few VM-specific classes in the "avian" package.
In order to facilitate making the VM compatible with multiple class
libraries, it's useful to separate the VM-specific representation of
these classes from the library implementations. This commit
introduces VMClass, VMField, and VMMethod for that purpose.
A long time ago, I refactored the class initialization code in the VM,
but did not notice until today that it had caused the
process=interpret build to break on certain recursive initializations.
In particular, we were not always detecting when a thread recursively
tried to initialize a class it was already in the process of
initializing, leading to the mistaken assumption that another thread
was initializing it and that we should wait until it was done, in
which case we would wait forever.
This commit ensures that we always detect recursive initialization and
short-circuit it.
It's not safe to use malloc from a signal handler, so we can't
allocate new memory when handling segfaults or Thread.getStackTrace
signals. Instead, we allocate a fixed-size backup heap for each
thread ahead of time and use it if there's no space left in the normal
heap pool. In the rare case that the backup heap isn't large enough,
we fall back to using a preallocated exception without a stack trace
as a last resort.
See commit 8120bee4dc for the original
problem description and solution. That commit and a couple of related
ones had to be reverted when we found they had introduced GC-safety
regressions leading to crashes.
This commit restores the reverted code and fixes the regressions.
We're seeing race conditions which occasionally lead to assertion
failures and thus crashes, so I'm reverting these changes for now:
29309fb414e92674cb738120bee4dc
Due to SWT's nasty habit of creating a new object monitor for every
task added to Display.asyncExec, we've found that, on Windows at
least, we tend to run out of OS handles due to the large number of
mutexes we create between garbage collections.
One way to address this might be to trigger a GC when either the
number of monitors created since the last GC exceeds a certain number
or when the total number of monitors in the VM reaches a certain
number. Both of these risk hurting performance, especially if they
force major collections which would otherwise be infrequent. Also,
it's hard to know what the values of such thresholds should be on a
given system.
Instead, we reimplement Java monitors using atomic compare-and-swap
(CAS) and thread-specific native locks for blocking in the case of
contention. This way, we can create an arbitrary number of monitors
without creating any new native locks. The total number of native
locks needed by the VM is bounded instead by the number of live
threads plus a small constant.
Note that if we ever add support for an architecture which does not
support CAS, we'll need to provide a fallback monitor implementation.
If another thread succeeds in entering the "exclusive" state while we
use the fast path to transition the current thread to "active", we
must switch back to "idle" temporarily to allow the exclusive thread a
chance to continue, and then retry the transition to "active" via the
slow path.
These paths reduce contention among threads by using atomic operations
and memory barriers instead of mutexes where possible. This is
especially important for JNI calls, since each such call involves two
state transitions: from "active" to "idle" and back.
