The commit started passing a nullptr port into GDBRemoteCommunication::StartDebugserverProcess.
The function was mostly handling the null value correctly, but it one case it did not check it's
value before assigning to it. Fix that.
llvm-svn: 278662
This change opens a socket pair and passes the second socket pair file descriptor down to the debugserver binary using a new option: "--fd=N" where N is the file descriptor. This file descriptor gets passed via posix_spawn() so that there is no need to do any bind/listen or bind/accept calls and eliminates the hanshake unix socket that is used to pass the result of the actual port that ends up being used so it can save time on launch as well as being faster.
This is currently only enabled on __APPLE__ builds. Other OSs should try modifying the #define from ProcessGDBRemote.cpp but the first person will need to port the --fd option over to lldb-server. Any OSs that enable USE_SOCKETPAIR_FOR_LOCAL_CONNECTION in their native builds can use the socket pair stuff. The #define is Apple only right now, but looks like:
#if defined (__APPLE__)
#define USE_SOCKETPAIR_FOR_LOCAL_CONNECTION 1
#endif
<rdar://problem/27814880>
llvm-svn: 278524
It's always hard to remember when to include this file, and
when you do include it it's hard to remember what preprocessor
check it needs to be behind, and then you further have to remember
whether it's windows.h or win32.h which you need to include.
This patch changes the name to PosixApi.h, which is more appropriately
named, and makes it independent of any preprocessor setting.
There's still the issue of people not knowing when to include this,
because there's not a well-defined set of things it exposes other
than "whatever is missing on Windows", but at least this should
make it less painful to fix when problems arise.
This patch depends on LLVM revision r278170.
llvm-svn: 278177
Resumbitting the commit after fixing the following problems:
- broken unit tests on windows: incorrect gtest usage on my part (TEST vs. TEST_F)
- the new code did not correctly handle the case where we went to interrupt the process, but it
stopped due to a different reason - the interrupt request would remain queued and would
interfere with the following "continue". I also added a unit test for this case.
This reapplies r277156 and r277139.
llvm-svn: 278118
It only contained a reimplementation of std::to_string, which I have replaced with usages of
pre-existing llvm::to_string (also, injecting members into the std namespace is evil).
llvm-svn: 278000
This reverts commit r277139, because:
- broken unittest on windows (likely typo on my part)
- seems to break TestCallThatRestart (needs investigation)
llvm-svn: 277154
SendContinuePacketAndWaitForResponse was huge function with very complex interactions with
several other functions (SendAsyncSignal, SendInterrupt, SendPacket). This meant that making any
changes to how packet sending functions and threads interact was very difficult and error-prone.
This change does not add any functionality yet, it merely paves the way for future changes. In a
follow-up, I plan to add the ability to have multiple query packets in flight (i.e.,
request,request,response,response instead of the usual request,response sequences) and use that
to speed up qModuleInfo packet processing.
Here, I introduce two special kinds of locks: ContinueLock, which is used by the continue thread,
and Lock, which is used by everyone else. ContinueLock (atomically) sends a continue packet, and
blocks any other async threads from accessing the connection. Other threads create an instance of
the Lock object when they want to access the connection. This object, while in scope prevents the
continue from being send. Optionally, it can also interrupt the process to gain access to the
connection for async processing.
Most of the syncrhonization logic is encapsulated within these two classes. Some of it still
had to bleed over into the SendContinuePacketAndWaitForResponse, but the function is still much
more manageable than before -- partly because of most of the work is done in the ContinueLock
class, and partly because I have factored out a lot of the packet processing code separate
functions (this also makes the functionality more easily testable). Most importantly, there is
none of syncrhonization code in the async thread users -- as far as they are concerned, they just
need to declare a Lock object, and they are good to go (SendPacketAndWaitForResponse is now a
very thin wrapper around the NoLock version of the function, whereas previously it had over 100
lines of synchronization code). This will make my follow up changes there easy.
I have written a number of unit tests for the new code and I have ran the test suite on linux and
osx with no regressions.
Subscribers: tberghammer
Differential Revision: https://reviews.llvm.org/D22629
llvm-svn: 277139
This finally removes the use of the Mutex and Condition classes. This is an
intricate patch as the Mutex and Condition classes were tied together.
Furthermore, many places had slightly differing uses of time values. Convert
timeout values to relative everywhere to permit the use of
std::chrono::duration, which is required for the use of
std::condition_variable's timeout. Adjust all Condition and related Mutex
classes over to std::{,recursive_}mutex and std::condition_variable.
This change primarily comes at the cost of breaking the TracingMutex which was
based around the Mutex class. It would be possible to write a wrapper to
provide similar functionality, but that is beyond the scope of this change.
llvm-svn: 277011
This change implements dumping the executable, triple,
args and environment when using ProcessInfo::Dump().
It also tweaks the way Args::Dump() works so that it prints
a configurable label rather than argv[{index}]={value}. By
default it behaves the same, but if the Dump() method with
the additional arg is provided, it can be overridden. The
environment variables dumped as part of ProcessInfo::Dump()
make use of that.
lldb-server has been modified to dump the gdb-remote stub's
ProcessInfo before launching if the "gdb-remote process" channel
is logged.
llvm-svn: 271312
This is a pretty straightforward first pass over removing a number of uses of
Mutex in favor of std::mutex or std::recursive_mutex. The problem is that there
are interfaces which take Mutex::Locker & to lock internal locks. This patch
cleans up most of the easy cases. The only non-trivial change is in
CommandObjectTarget.cpp where a Mutex::Locker was split into two.
llvm-svn: 269877
the xcode project file to catch switch statements that have a
case that falls through unintentionally.
Define LLVM_FALLTHROUGH to indicate instances where a case has code
and intends to fall through. This should be in llvm/Support/Compiler.h;
Peter Collingbourne originally checked in there (r237766), then
reverted (r237941) because he didn't have time to mark up all the
'case' statements that were intended to fall through. I put together
a patch to get this back in llvm http://reviews.llvm.org/D17063 but
it hasn't been approved in the past week. I added a new
lldb-private-defines.h to hold the definition for now.
Every place in lldb where there is a comment that the fall-through
is intentional, I added LLVM_FALLTHROUGH to silence the warning.
I haven't tried to identify whether the fallthrough is a bug or
not in the other places.
I haven't tried to add this to the cmake option build flags.
This warning will only work for clang.
This build cleanly (with some new warnings) on macosx with clang
under xcodebuild, but if this causes problems for people on other
configurations, I'll back it out.
llvm-svn: 260930
The standard remote debugging workflow with gdb is to start the
application on the remote host under gdbserver (e.g.: gdbserver :5039
a.out) and then connect to it with gdb.
The same workflow is supported by debugserver/lldb-gdbserver with a very
similar syntax but to access all features of lldb we need to be
connected also to an lldb-platform instance running on the target.
Before this change this had to be done manually with starting a separate
lldb-platform on the target machine and then connecting to it with lldb
before connecting to the process.
This change modifies the behavior of "platform connect" with
automatically connecting to the process instance if it was started by
the remote platform. With this command replacing gdbserver in a gdb
based worflow is usually as simple as replacing the command to execute
gdbserver with executing lldb-platform.
Differential revision: http://reviews.llvm.org/D14952
llvm-svn: 255016
This allows open source MacOSX clients to not have to build debugserver and the current LLDB can find debugserver inside the selected Xcode.app on your system.
<rdar://problem/23167253>
llvm-svn: 250735
on iOS devices; fallout from Vince's cleanups made
in r237218 back in May. iOS native lldbs will call
StartDebugserverProcess() with a random port #
(see ProcessGDBRemote::LaunchAndConnectToDebugserver)
and neither side of this conditional expression should
be followed in that case.
I added an "if (in_port == 0) { ..." check around the
entire if/then/else and indented the block of code so
the diff looks larger than it really is.
<rdar://problem/21712643>
llvm-svn: 248343
When lldb receives a gdb-remote protocol packet that has
nonprintable characters, it will print the packet in
gdb-remote logging with binary-hex encoding so we don't
dump random 8-bit characters into the packet log.
I'm changing this check to allow whitespace characters
(newlines, linefeeds, tabs) to be printed if those are
the only non-printable characters in the packet.
This is primarily to get the response to the
qXfer:features:read:target.xml packet to show up in the
packet logs in human readable form. Right now we just
get a dozen kilobytes of hex-ascii and it's hard to
figure out what register number scheme is being used.
llvm-svn: 247120
working with (the Communication m_bytes ivar) contained a single packet.
Instead, it may contain multitudes. Find the boundaries of the first packet
in the buffer and replace that with the decompressed version leaving the
rest of the buffer unmodified.
<rdar://problem/21841377>
llvm-svn: 243846
Summary:
If we used unnamed pipes instead of named pipes, we can avoid having the
file system littered with debugserver-named-pipes if lldb-server happens to
crash for whatever reason. Also, on some buggy systems, it's possible to be
able to create but not to delete a fifo. Ideally, support for unnamed pipes
should be added to debugserver as well, so we can avoid the `#ifdef` here.
Reviewers: clayborg, vharron, chying
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D11609
llvm-svn: 243667
For some communication channels, sending large packets can be very
slow. In those cases, it may be faster to compress the contents of
the packet on the target device and decompress it on the debug host
system. For instance, communicating with a device using something
like Bluetooth may be an environment where this tradeoff is a good one.
This patch adds a new field to the response to the "qSupported" packet
(which returns a "qXfer:features:" response) -- SupportedCompressions
and DefaultCompressionMinSize. These tell you what the remote
stub can support.
lldb, if it wants to enable compression and can handle one of those
algorithms, it can send a QEnableCompression packet specifying the
algorithm and optionally the minimum packet size to use compression
on. lldb may have better knowledge about the best tradeoff for
a given communication channel.
I added support to debugserver an lldb to use the zlib APIs
(if -DHAVE_LIBZ=1 is in CFLAGS and -lz is in LDFLAGS) and the
libcompression APIs on Mac OS X 10.11 and later
(if -DHAVE_LIBCOMPRESSION=1). libz "zlib-deflate" compression.
libcompression can support deflate, lz4, lzma, and a proprietary
lzfse algorithm. libcompression has been hand-tuned for Apple
hardware so it should be preferred if available.
debugserver currently only adds the SupportedCompressions when
it is being run on an Apple watch (TARGET_OS_WATCH). Comment
that #if out from RNBRemote.cpp if you want to enable it to
see how it works. I haven't tested this on a native system
configuration but surely it will be slower to compress & decompress
the packets in a same-system debug session.
I haven't had a chance to add support for this to
GDBRemoteCommunciationServer.cpp yet.
<rdar://problem/21090180>
llvm-svn: 240066
In order to support asynchronous notifications for non-stop mode this patch adds a packet read thread. This is done by implementing AppendBytesToCache() from the communications class, which continually reads packets into a packet queue. To initialize this thread StartReadThread() must be called by the client, so since llgs and platform tools use the GBDRemoteCommunicatos code they must also call this function as well as ProcessGDBRemote.
When the read thread detects an async notify packet it broadcasts this event, where the matching listener will be added in the next non-stop patch.
Packets are now accessed by calling ReadPacket() which pops a packet from the queue, instead of using WaitForPacketWithTimeoutMicroSecondsNoLock()
Reviewers: vharron, clayborg
Subscribers: lldb-commits, labath, ted, domipheus, deepak2427
Differential Revision: http://reviews.llvm.org/D10085
llvm-svn: 239824
qEcho:%s
where '%s' is any valid string. The response to this packet is the exact packet itself with no changes, just reply with what you received!
This will help us to recover from packets timing out much more gracefully. Currently if a packet times out, LLDB quickly will hose up the debug session. For example, if we send a "abc" packet and we expect "ABC" back in response, but the "abc" command takes longer than the current timeout value this will happen:
--> "abc"
<-- <<<error: timeout>>>
Now we want to send "def" and get "DEF" back:
--> "def"
<-- "ABC"
We got the wrong response for the "def" packet because we didn't sync up with the server to clear any current responses from previously issues commands.
The fix is to modify GDBRemoteCommunication::WaitForPacketWithTimeoutMicroSecondsNoLock() so that when it gets a timeout, it syncs itself up with the client by sending a "qEcho:%u" where %u is an increasing integer, one for each time we timeout. We then wait for 3 timeout periods to sync back up. So the above "abc" session would look like:
--> "abc"
<-- <<<error: timeout>>> 1 second
--> "qEcho:1"
<-- <<<error: timeout>>> 1 second
<-- <<<error: timeout>>> 1 second
<-- "abc"
<-- "qEcho:1"
The first timeout is from trying to get the response, then we know we timed out and we send the "qEcho:1" packet and wait for 3 timeout periods to get back in sync knowing that we might actually get the response for the "abc" packet in the mean time...
In this case we would actually succeed in getting the response for "abc". But lets say the remote GDB server is deadlocked and will never response, it would look like:
--> "abc"
<-- <<<error: timeout>>> 1 second
--> "qEcho:1"
<-- <<<error: timeout>>> 1 second
<-- <<<error: timeout>>> 1 second
<-- <<<error: timeout>>> 1 second
We then disconnect and say we lost connection.
We might also have a bad GDB server that just dropped the "abc" packet on the floor. We can still recover in this case and it would look like:
--> "abc"
<-- <<<error: timeout>>> 1 second
--> "qEcho:1"
<-- "qEcho:1"
Then we know our remote GDB server is still alive and well, and it just dropped the "abc" response on the floor and we can continue to debug.
<rdar://problem/21082939>
llvm-svn: 238530
In ProcessGDBRemote we currently have a single packet, m_last_stop_packet, used to set the thread stop info.
However in non-stop mode we can receive several stop reply packets in a sequence for different threads. As a result we need to use a container to hold them before they are processed.
This patch also changes the return type of CheckPacket() so we can detect async notification packets.
Reviewers: clayborg
Subscribers: labath, ted, deepak2427, lldb-commits
Differential Revision: http://reviews.llvm.org/D9853
llvm-svn: 238323
Summary:
This patch is the beginnings of support for Non-stop mode in the remote protocol. Letting a user examine stopped threads, while other threads execute freely.
Non-stop mode is enabled using the setting target.non-stop-mode, which sends a QNonStop packet when establishing the remote connection.
Changes are also made to treat the '?' stop reply packet differently in non-stop mode, according to spec https://sourceware.org/gdb/current/onlinedocs/gdb/Remote-Non_002dStop.html#Remote-Non_002dStop.
A setting for querying the remote for default thread on setup is also included.
Handling of '%' async notification packets will be added next.
Reviewers: clayborg
Subscribers: lldb-commits, ADodds, ted, deepak2427
Differential Revision: http://reviews.llvm.org/D9656
llvm-svn: 237239
Removed some unused variables, added some consts, changed some casts
to const_cast. I don't think any of these changes are very
controversial.
Differential Revision: http://reviews.llvm.org/D9674
llvm-svn: 237218
Summary:
New dotest options that allow arbitrary log channels and
categories to be enabled. Also enables logging for locally run
debug servers.
Log messages are separated into separate files per test case.
(this makes it possible to log in dosep runs)
These new log files are stored side-by-side with trace files in the
session directory.
These files are deleted by default if the test run is successful.
If --log-success is specified, even successful logs are retained.
--log-success is useful for creating reference log files.
Test Plan:
add '--channel "lldb all" --channel "gdb-remote packets" --log-success'
to your dotest options
Tested on OSX and Linux
Differential Revision: http://reviews.llvm.org/D9594
llvm-svn: 236956
Summary:
Currently, launching lldb-gdbserver from platform on Android requires root for
mkfifo() and an explicit TMPDIR variable. This should remove both requirements.
Test Plan: Successfully launched lldb-gdbserver on a non-rooted Android device.
Reviewers: tberghammer, vharron, clayborg
Reviewed By: clayborg
Subscribers: tberghammer, lldb-commits
Differential Revision: http://reviews.llvm.org/D9307
llvm-svn: 235940
This new class makes it easier to change the timeout of a
GDBRemoteCommunication instance for a short time and then restore it to
its original value.
Differential revision: http://reviews.llvm.org/D7826
llvm-svn: 230319
This commit merges lldb-platform and lldb-gdbserver into a single binary
of the same size as each of the previous individual binaries. Execution
mode is controlled by the first argument being either platform or
gdbserver.
Patch from: flackr <flackr@google.com>
Differential revision: http://reviews.llvm.org/D7545
llvm-svn: 229683
The refactor was motivated by some comments that Greg made
http://reviews.llvm.org/D6918
and also to break a dependency cascade that caused functions linking
in string->int conversion functions to pull in most of lldb
llvm-svn: 226199
This reverts commit 4a5ad2c077166cc3d6e7ab4cc6e3dcbbe922af86.
Windows doesn't support select() for pipe objects, and this also fails
to compile on Windows. Reverting this until we can get it sorted out
to keep the windows build working.
llvm-svn: 223392
The details are: large packets (like large memory reads (m packets) or large binary memory reads (x packet)) can get responses that come in across multiple read() calls. The while loop that was added meant that if only a partial packet came in (like only "$abc" coming for a response) GDBRemoteCommunication::CheckForPacket() was called, it would deadlock in the while loop because no more data is going to come in as this function needs to be called again with more data from another read. So the original fix will need to be corrected and resubmitted.
<rdar://problem/18853744>
llvm-svn: 221181
As part of getting ConnectionFileDescriptor working on Windows,
there is going to be alot of platform specific work to be done.
As a result, the implementation is moving into Host. This patch
performs the code move and fixes up call-sites appropriately.
Reviewed by: Greg Clayton
Differential Revision: http://reviews.llvm.org/D5548
llvm-svn: 219143
This patch moves creates a thread abstraction that represents a
thread running inside the LLDB process. This is a replacement for
otherwise using lldb::thread_t, and provides a platform agnostic
interface to managing these threads.
Differential Revision: http://reviews.llvm.org/D5198
Reviewed by: Jim Ingham
llvm-svn: 217460
This patch accepts environment variables of the form:
LLDB_DEBUGSERVER_EXTRA_ARG_n
where n starts with 1, and may continue nearly indefinitely (up through std::numeric_limits<uint32_t>::max()).
The code loops around, starting with 1, until it doesn't find one of the environment variables. For each one it does find defined, it appends the environment variable's contents to the end of the debugserver/llgs startup command line issued when the stub is started for local debugging.
I am using this to add arbitrary startup commands to the llgs command line for turning on additional logging. For example:
export LLDB_DEBUGSERVER_EXTRA_ARG_1="-c"
export LLDB_DEBUGSERVER_EXTRA_ARG_2="log enable -f /tmp/llgs_packets.log gdb-remote packets"
export LLDB_DEBUGSERVER_EXTRA_ARG_3="-c"
export LLDB_DEBUGSERVER_EXTRA_ARG_4="log enable -f /tmp/llgs_process.log lldb process"
llvm-svn: 216745
This continues the effort to get Host code moved over to HostInfo,
and removes many more instances of preprocessor defines along the
way.
llvm-svn: 216195
More specifically, this change can be summarized as follows:
1) Makes an lldbHostPosix library which contains code common to
all posix platforms.
2) Creates Host/FileSystem.h which defines a common FileSystem
interface.
3) Implements FileSystem.h in Host/windows and Host/posix.
4) Creates Host/FileCache.h, implemented in Host/common, which
defines a class useful for storing handles to open files needed
by the debugger.
Differential Revision: http://reviews.llvm.org/D4889
llvm-svn: 215775
This patch moves the logic of many common socket operations into
its own class lldb_private::Socket. It then modifies the
ConnectionFileDescriptor class, and a few users of that class,
to use this new Socket class instead of hardcoding socket logic
directly.
Finally, this patch creates a common interface called IOObject for
any objects that support reading and writing, so that endpoints
such as sockets and files can be treated the same.
Differential Revision: http://reviews.llvm.org/D4641
Reviewed by: Todd Fiala, Greg Clayton
llvm-svn: 214984
This change enables lldb-platform for Linux. In addition, it does the following:
* fixes Host::GetLLDBPath() to work on Linux/*BSD for ePathTypeSupportExecutableDir-relative paths.
* adds more logging and comments around lldb-platform startup and remote lldb-platform usage.
* refactors lldb-platform remote-* support for Darwin and Linux into PlatformPOSIX. This, in theory, is the bulk of what is needed for *BSD to make remote connections to lldb-platform as well (although I haven't tested that yet). FreeBSD can make similar changes to their Platform* as was made here for PlatformLinux to pick up the rest of the bits.
* teaches GDBRemoteCommunication to use lldb-gdbserver for non-Apple hosts.
llvm-svn: 213707
- CTRL+C wasn't clearing the command in lldb
- CTRL+C doesn't work in python macros in lldb
- Ctrl+C no longer interrupts the running process that you attach to
<rdar://problem/15949205>
<rdar://problem/16778652>
<rdar://problem/16774411>
llvm-svn: 207816
right after the space for it is allocated on the stack, instead of trying
to initialize it in all the different places in this method. It's too easy
for another uninitialized code path to sneak in as it is written right now.
llvm-svn: 200066
<rdar://problem/15600045>
Due to other recent changes, all connections to GDB servers that didn't support the "QStartNoAckMode" packet would cause us to fail to attach to the remote GDB server.
The problem was that SendPacket* and WaitForResponse* packets would return a size_t indicating the number of bytes sent/received. The other issue was WaitForResponse* packets would strip the leading '$' and the trailing "#CC" (checksum) bytes, so the unimplemented response packet of "$#00" would get stripped and the WaitForResponse* packets would return 0.
These new error codes give us flexibility to to more intelligent things in response to what is returned.
llvm-svn: 196610
This gets rid of our hacky "get_random_port()" which would grab a random port and tell debugserver to open that port. Now LLDB creates, binds, listens and accepts a connection by binding to port zero and sending the correctly bound port down as the host:port to connect back to.
Fixed the "ConnectionFileDescriptor" to be able to correctly listen for connections from a specified host, localhost, or any host. Prior to this fix "listen://" only accepted the following format:
listen://<port>
But now it can accept:
listen://<port> // Listen for connection from localhost on port <port>
listen://<host>:<port> // Listen for connection from <host> and <port>
listen://*:<port> // Listen for connection from any host on port <port>
llvm-svn: 196547
This helps ensure that the launched debugserver is ready and listening for a connection. Prior to this we had a race condition.
Consolidate the launching of debugserver into a single place: a static function in GDBRemoteCommunication.
llvm-svn: 196401
Summary:
This merge brings in the improved 'platform' command that knows how to
interface with remote machines; that is, query OS/kernel information, push
and pull files, run shell commands, etc... and implementation for the new
communication packets that back that interface, at least on Darwin based
operating systems via the POSIXPlatform class. Linux support is coming soon.
Verified the test suite runs cleanly on Linux (x86_64), build OK on Mac OS
X Mountain Lion.
Additional improvements (not in the source SVN branch 'lldb-platform-work'):
- cmake build scripts for lldb-platform
- cleanup test suite
- documentation stub for qPlatform_RunCommand
- use log class instead of printf() directly
- reverted work-in-progress-looking changes from test/types/TestAbstract.py that work towards running the test suite remotely.
- add new logging category 'platform'
Reviewers: Matt Kopec, Greg Clayton
Review: http://llvm-reviews.chandlerc.com/D1493
llvm-svn: 189295
LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down.
All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down.
llvm-svn: 178191
than being given the pthread_mutex_t from the Mutex and locks that. That allows us to
track ownership of the Mutex better.
Used this to switch the LLDB_CONFIGURATION_DEBUG enabled assert when we can't get the
gdb-remote sequence mutex to assert when the thread that had the mutex releases it. This
is generally more useful information than saying just who failed to get it (since the
code that had it locked often had released it by the time the assert fired.)
llvm-svn: 158240
Fixed a case where multiple threads can be asking to send a packet to the GDB server and one of three things will happen:
1 - everything works
2 - one thread will fail to send the packet due to not being able to get the sequence mutex
3 - one thread will try and interrupt the other packet sending and fail and not send the packet
Now the flow is a bit different. Prior to this fix we did:
if (try_get_sequence_mutex()) {
send_packet()
return success;
} else {
if (async_ok) {
interrupt()
send_packet()
resume()
return success;
}
}
return fail
The issue is that the call to "try_get_sequence_mutex()" could fail if another thread was sending a packet and could cause us to just not send the packet and an error would be returned.
What we really want is to try and get the sequence mutex, and if this succeeds, send the packet. Else check if we are running and if we are, do what we used to do. The big difference is when we aren't running, we wait for the sequence mutex so we don't drop packets. Pseudo code is:
if (try_get_sequence_mutex()) {
// Safe to send the packet right away
send_packet()
return success;
} else {
if (running) {
// We are running, interrupt and send async packet if ok to do so,
// else it is ok to fail
if (async_ok) {
interrupt()
send_packet()
resume()
return success;
}
}
else {
// Not running, wait for the sequence mutex so we don't drop packets
get_sequence_mutex()
send_packet()
return success;
}
}
return fail
llvm-svn: 157751
No one was using it and Locker(pthread_mutex_t *) immediately asserts for
pthread_mutex_t's that don't come from a Mutex anyway. Rather than try to make
that work, we should maintain the Mutex abstraction and not pass around the
platform implementation...
Make Mutex::Locker::Lock take a Mutex & or a Mutex *, and remove the constructor
taking a pthread_mutex_t *. You no longer need to call Mutex::GetMutex to pass
your mutex to a Locker (you can't in fact, since I made it private.)
llvm-svn: 156221
Enable logging the packet history when registers fail to read due to not getting the sequence mutex if "--verbose" is enabled on the log channel for the "gdb-remote" log category.
This will help us track down some issues.
llvm-svn: 154704
Cleaned up the Mutex::Locker and the ReadWriteLock classes a bit.
Also cleaned up the GDBRemoteCommunication class to not have so many packet functions. Used the "NoLock" versions of send/receive packet functions when possible for a bit of performance.
llvm-svn: 154458
The current ProcessGDBRemote function that updates the threads could end up with an empty list if any other thread had the sequence mutex. We now don't clear the thread list when we can't access it, and we also have changed how lldb_private::Process handles the return code from the:
virtual bool
Process::UpdateThreadList (lldb_private::ThreadList &old_thread_list,
lldb_private::ThreadList &new_thread_list) = 0;
A bool is now returned to indicate if the list was actually updated or not and the lldb_private::Process class will only update the stop ID of the validity of the thread list if "true" is returned.
The ProcessGDBRemote also got an extra assertion that will hopefully assert when running debug builds so we can find the source of this issue.
llvm-svn: 154365
lldb_private::Error objects the rules are:
- short strings that don't start with a capitol letter unless the name is a
class or anything else that is always capitolized
- no trailing newline character
- should be one line if possible
Implemented a first pass at adding "--gdb-format" support to anything that
accepts format with optional size/count.
llvm-svn: 142999
stdarg formats to use __attribute__ format so the compiler can flag
incorrect uses. Fix all incorrect uses. Most of these are innocuous,
a few were resulting in crashes.
llvm-svn: 140185
invoking the Read(...) method to read in bytes. This seems to fix the infinite looping
I was seeing on SnowLeopard while running the test suite.
llvm-svn: 135461
variables prior to running your binary. Zero filled sections now get
section data correctly filled with zeroes when Target::ReadMemory
reads from the object file section data.
Added new option groups and option values for file lists. I still need
to hook up all of the options to "target variable" to allow more complete
introspection by file and shlib.
Added the ability for ValueObjectVariable objects to be created with
only the target as the execution context. This allows them to be read
from the object files through Target::ReadMemory(...).
Added a "virtual Module * GetModule()" function to the ValueObject
class. By default it will look to the parent variable object and
return its module. The module is needed when we have global variables
that have file addresses (virtual addresses that are specific to
module object files) and in turn allows global variables to be displayed
prior to running.
Removed all of the unused proxy object support that bit rotted in
lldb_private::Value.
Replaced a lot of places that used "FileSpec::Compare (lhs, rhs) == 0" code
with the more efficient "FileSpec::Equal (lhs, rhs)".
Improved logging in GDB remote plug-in.
llvm-svn: 134579
a file or socket. We now make a getsockopt call to check if the fd is a socket.
Also, the previous logic in the GDB communication needs to watch for success
with an error so we can deal with EAGAIN and other normal "retry" error codes.
llvm-svn: 134359
_only_ in the resulting stream, not in the error objects (lldb_private::Error).
lldb_private::Error objects should always just have an error string with no
terminating newline characters or periods.
Fixed an issue with GDB remote packet detection that could end up deadlocking
if a full packet wasn't received in one chunk. Also modified the packet
checking function to properly toss one or more bytes when it detects bad
data.
llvm-svn: 134357
libraries and headers exist. This can be specified using the platform select
function:
platform select --sysroot /Volumes/remote-root remote-macosx
Each platform subclass is free to interpret the sysroot as needed.
Expose the new SDK root directory through the SBDebugger class.
Fixed an issue with the GDB remote protocol where unimplemented packets were
not being handled correctly.
llvm-svn: 133231
darwin (not sure about other platforms).
Modified the communication and connection classes to not require the
BytesAvailable function. Now the "Read(...)" function has a timeout in
microseconds.
Fixed a lot of assertions that were firing off in certain cases and replaced
them with error output and code that can deal with the assertion case.
llvm-svn: 133224
the "payload_length" argument for the "payload" packet data. This meant we
could end up sending random extra data with a packet depending on how the
packet was constructed.
Fixed GDBRemoteRegisterContext to properly save and restore all registers.
Previous fixes had been added to work around the "payload_length" issues fixed
above and aren't needed anymore.
Fix logging in GDBRemoteCommunication to make sure we log the correct packet
data being sent by using the packet length when dumping the packet contents.
Added register definitions for 'arm-lldb' in the "disasm-gdb-remote.pl" script
so if you have a register dump from the GDB remote that doesn't include the
qRegisterInfo packets, you can manually tell the script which registers are
which.
llvm-svn: 131715
the CommandInterpreter where it was always being used.
Make sure that Modules can track their object file offsets correctly to
allow opening of sub object files (like the "__commpage" on darwin).
Modified the Platforms to be able to launch processes. The first part of this
move is the platform soon will become the entity that launches your program
and when it does, it uses a new ProcessLaunchInfo class which encapsulates
all process launching settings. This simplifies the internal APIs needed for
launching. I want to slowly phase out process launching from the process
classes, so for now we can still launch just as we used to, but eventually
the platform is the object that should do the launching.
Modified the Host::LaunchProcess in the MacOSX Host.mm to correctly be able
to launch processes with all of the new eLaunchFlag settings. Modified any
code that was manually launching processes to use the Host::LaunchProcess
functions.
Fixed an issue where lldb_private::Args had implicitly defined copy
constructors that could do the wrong thing. This has now been fixed by adding
an appropriate copy constructor and assignment operator.
Make sure we don't add empty ModuleSP entries to a module list.
Fixed the commpage module creation on MacOSX, but we still need to train
the MacOSX dynamic loader to not get rid of it when it doesn't have an entry
in the all image infos.
Abstracted many more calls from in ProcessGDBRemote down into the
GDBRemoteCommunicationClient subclass to make the classes cleaner and more
efficient.
Fixed the default iOS ARM register context to be correct and also added support
for targets that don't support the qThreadStopInfo packet by selecting the
current thread (only if needed) and then sending a stop reply packet.
Debugserver can now start up with a --unix-socket (-u for short) and can
then bind to port zero and send the port it bound to to a listening process
on the other end. This allows the GDB remote platform to spawn new GDB server
instances (debugserver) to allow platform debugging.
llvm-svn: 129351
class now implements the Host functionality for a lot of things that make
sense by default so that subclasses can check:
int
PlatformSubclass::Foo ()
{
if (IsHost())
return Platform::Foo (); // Let the platform base class do the host specific stuff
// Platform subclass specific code...
int result = ...
return result;
}
Added new functions to the platform:
virtual const char *Platform::GetUserName (uint32_t uid);
virtual const char *Platform::GetGroupName (uint32_t gid);
The user and group names are cached locally so that remote platforms can avoid
sending packets multiple times to resolve this information.
Added the parent process ID to the ProcessInfo class.
Added a new ProcessInfoMatch class which helps us to match processes up
and changed the Host layer over to using this new class. The new class allows
us to search for processs:
1 - by name (equal to, starts with, ends with, contains, and regex)
2 - by pid
3 - And further check for parent pid == value, uid == value, gid == value,
euid == value, egid == value, arch == value, parent == value.
This is all hookup up to the "platform process list" command which required
adding dumping routines to dump process information. If the Host class
implements the process lookup routines, you can now lists processes on
your local machine:
machine1.foo.com % lldb
(lldb) platform process list
PID PARENT USER GROUP EFF USER EFF GROUP TRIPLE NAME
====== ====== ========== ========== ========== ========== ======================== ============================
99538 1 username usergroup username usergroup x86_64-apple-darwin FileMerge
94943 1 username usergroup username usergroup x86_64-apple-darwin mdworker
94852 244 username usergroup username usergroup x86_64-apple-darwin Safari
94727 244 username usergroup username usergroup x86_64-apple-darwin Xcode
92742 92710 username usergroup username usergroup i386-apple-darwin debugserver
This of course also works remotely with the lldb-platform:
machine1.foo.com % lldb-platform --listen 1234
machine2.foo.com % lldb
(lldb) platform create remote-macosx
Platform: remote-macosx
Connected: no
(lldb) platform connect connect://localhost:1444
Platform: remote-macosx
Triple: x86_64-apple-darwin
OS Version: 10.6.7 (10J869)
Kernel: Darwin Kernel Version 10.7.0: Sat Jan 29 15:17:16 PST 2011; root:xnu-1504.9.37~1/RELEASE_I386
Hostname: machine1.foo.com
Connected: yes
(lldb) platform process list
PID PARENT USER GROUP EFF USER EFF GROUP TRIPLE NAME
====== ====== ========== ========== ========== ========== ======================== ============================
99556 244 username usergroup username usergroup x86_64-apple-darwin trustevaluation
99548 65539 username usergroup username usergroup x86_64-apple-darwin lldb
99538 1 username usergroup username usergroup x86_64-apple-darwin FileMerge
94943 1 username usergroup username usergroup x86_64-apple-darwin mdworker
94852 244 username usergroup username usergroup x86_64-apple-darwin Safari
The lldb-platform implements everything with the Host:: layer, so this should
"just work" for linux. I will probably be adding more stuff to the Host layer
for launching processes and attaching to processes so that this support should
eventually just work as well.
Modified the target to be able to be created with an architecture that differs
from the main executable. This is needed for iOS debugging since we can have
an "armv6" binary which can run on an "armv7" machine, so we want to be able
to do:
% lldb
(lldb) platform create remote-ios
(lldb) file --arch armv7 a.out
Where "a.out" is an armv6 executable. The platform then can correctly decide
to open all "armv7" images for all dependent shared libraries.
Modified the disassembly to show the current PC value. Example output:
(lldb) disassemble --frame
a.out`main:
0x1eb7: pushl %ebp
0x1eb8: movl %esp, %ebp
0x1eba: pushl %ebx
0x1ebb: subl $20, %esp
0x1ebe: calll 0x1ec3 ; main + 12 at test.c:18
0x1ec3: popl %ebx
-> 0x1ec4: calll 0x1f12 ; getpid
0x1ec9: movl %eax, 4(%esp)
0x1ecd: leal 199(%ebx), %eax
0x1ed3: movl %eax, (%esp)
0x1ed6: calll 0x1f18 ; printf
0x1edb: leal 213(%ebx), %eax
0x1ee1: movl %eax, (%esp)
0x1ee4: calll 0x1f1e ; puts
0x1ee9: calll 0x1f0c ; getchar
0x1eee: movl $20, (%esp)
0x1ef5: calll 0x1e6a ; sleep_loop at test.c:6
0x1efa: movl $12, %eax
0x1eff: addl $20, %esp
0x1f02: popl %ebx
0x1f03: leave
0x1f04: ret
This can be handy when dealing with the new --line options that was recently
added:
(lldb) disassemble --line
a.out`main + 13 at test.c:19
18 {
-> 19 printf("Process: %i\n\n", getpid());
20 puts("Press any key to continue..."); getchar();
-> 0x1ec4: calll 0x1f12 ; getpid
0x1ec9: movl %eax, 4(%esp)
0x1ecd: leal 199(%ebx), %eax
0x1ed3: movl %eax, (%esp)
0x1ed6: calll 0x1f18 ; printf
Modified the ModuleList to have a lookup based solely on a UUID. Since the
UUID is typically the MD5 checksum of a binary image, there is no need
to give the path and architecture when searching for a pre-existing
image in an image list.
Now that we support remote debugging a bit better, our lldb_private::Module
needs to be able to track what the original path for file was as the platform
knows it, as well as where the file is locally. The module has the two
following functions to retrieve both paths:
const FileSpec &Module::GetFileSpec () const;
const FileSpec &Module::GetPlatformFileSpec () const;
llvm-svn: 128563
On Mac OS X we now have 3 platforms:
PlatformDarwin - must be subclassed to fill in the missing pure virtual funcs
but this implements all the common functionality between
remote-macosx and remote-ios. It also allows for another
platform to be used (remote-gdb-server for now) when doing
remote connections. Keeping this pluggable will allow for
flexibility.
PlatformMacOSX - Now implements both local and remote macosx desktop platforms.
PlatformRemoteiOS - Remote only iOS that knows how to locate SDK files in the
cached SDK locations on the host.
A new agnostic platform has been created:
PlatformRemoteGDBServer - this implements the platform using the GDB remote
protocol and uses the built in lldb_private::Host
static functions to implement many queries.
llvm-svn: 128193
GDBRemoteCommunication - The base GDB remote communication class
GDBRemoteCommunicationClient - designed to be used for clients the connect to
a remote GDB server
GDBRemoteCommunicationServer - designed to be used on the server side of a
GDB server implementation.
llvm-svn: 128070
member variable (m_packet_timeout which is a value in seconds). This value is
then used for all packets sent to/from the remote GDB server.
llvm-svn: 127392
Modifed lldb_private::Process to be able to handle connecting to a remote
target that isn't running a process. This leaves lldb_private::Process in the
eStateConnected state from which we can then do an attach or launch.
Modified ProcessGDBRemote to be able to set stdin, stdout, stderr, working
dir, disable ASLR and a few other settings down by using new GDB remote
packets. This allows us to keep all of our current launch flags and settings
intact and still be able to communicate them over to the remote GDB server.
Previously these were being sent as arguments to the debugserver binary that
we were spawning. Also modified ProcessGDBRemote to handle losing connection
to the remote GDB server and always exit immediately. We do this by watching
the lldb_private::Communication event bit for the read thread exiting in the
ProcessGDBRemote async thread.
Added support for many of the new 'Q' packets for setting stdin, stdout,
stderr, working dir and disable ASLR to the GDBRemoteCommunication class for
easy accesss.
Modified debugserver for all of the new 'Q' packets and also made it so that
debugserver always exists if it loses connection with the remote debugger.
llvm-svn: 126444
of Stephen Wilson's idea (thanks for the input Stephen!). What I ended up
doing was:
- Got rid of ArchSpec::CPU (which was a generic CPU enumeration that mimics
the contents of llvm::Triple::ArchType). We now rely upon the llvm::Triple
to give us the machine type from llvm::Triple::ArchType.
- There is a new ArchSpec::Core definition which further qualifies the CPU
core we are dealing with into a single enumeration. If you need support for
a new Core and want to debug it in LLDB, it must be added to this list. In
the future we can allow for dynamic core registration, but for now it is
hard coded.
- The ArchSpec can now be initialized with a llvm::Triple or with a C string
that represents the triple (it can just be an arch still like "i386").
- The ArchSpec can still initialize itself with a architecture type -- mach-o
with cpu type and subtype, or ELF with e_machine + e_flags -- and this will
then get translated into the internal llvm::Triple::ArchSpec + ArchSpec::Core.
The mach-o cpu type and subtype can be accessed using the getter functions:
uint32_t
ArchSpec::GetMachOCPUType () const;
uint32_t
ArchSpec::GetMachOCPUSubType () const;
But these functions are just converting out internal llvm::Triple::ArchSpec
+ ArchSpec::Core back into mach-o. Same goes for ELF.
All code has been updated to deal with the changes.
This should abstract us until later when the llvm::TargetSpec stuff gets
finalized and we can then adopt it.
llvm-svn: 126278
now, in addition to cpu type/subtype and architecture flavor, contains:
- byte order (big endian, little endian)
- address size in bytes
- llvm::Triple for true target triple support and for more powerful plug-in
selection.
llvm-svn: 125602
are supported by the remote GDB target. We can also now deal with the lack of
vCont support and send packets that the remote GDB stub can use. We also error
out of the continue if LLDB tries to do something too complex when vCont isn't
supported.
llvm-svn: 125433
(lldb) process connect <remote-url>
Currently when you specify a file with the file command it helps us to find
a process plug-in that is suitable for debugging. If you specify a file you
can rely upon this to find the correct debugger plug-in:
% lldb a.out
Current executable set to 'a.out' (x86_64).
(lldb) process connect connect://localhost:2345
...
If you don't specify a file, you will need to specify the plug-in name that
you wish to use:
% lldb
(lldb) process connect --plugin process.gdb-remote connect://localhost:2345
Other connection URL examples:
(lldb) process connect connect://localhost:2345
(lldb) process connect tcp://127.0.0.1
(lldb) process connect file:///dev/ttyS1
We are currently treating the "connect://host:port" as a way to do raw socket
connections. If there is a URL for this already, please let me know and we
will adopt it.
So now you can connect to a remote debug server with the ProcessGDBRemote
plug-in. After connection, it will ask for the pid info using the "qC" packet
and if it responds with a valid process ID, it will be equivalent to attaching.
If it response with an error or invalid process ID, the LLDB process will be
in a new state: eStateConnected. This allows us to then download a program or
specify the program to run (using the 'A' packet), or specify a process to
attach to (using the "vAttach" packets), or query info about the processes
that might be available.
llvm-svn: 124846
sessions: When continue packet has been sent and an interrupt packet was
quickly sent, it would get read at the same time:
$c#00\x03
There was an error where the packet end index was always being computed
incorrectly by debugserver, but it wouldn't matter if there weren't extra
bytes on the end (the hex \x03 interrupt byte in this case). The first
'$' last 3 bytes of the data in the packet buffer were being trimmed
(trying to trim the '#' + checksum (#XX)) which made:
c#
And this would then be passed to the handle routine for the 'c' packet which
would see an extra character at the end and assume it was going to be in the
form c[addr] where "[addr]" was a hex address to resume at and this would
result in a malformed packet response. This is now fixed and everything works
great.
Another issue was issuing async packets correctly by doing correct handshakes
between the thread that wants to send the async packet, and the thread that
is tracking the current run.
Added a write lock to the communication class as well to make sure you never
get two threads trying to write data at the same time. This wasn't happening,
but it is a good idea to make sure it doesn't.
llvm-svn: 124369
checking the validity of the shared pointer prior to using it.
Fixed the GDB remote plug-in to once again watch for a reply from the "k"
packet, and fixed the logic to make sure the thread requesting the kill
and the async thread play nice (and very quickly) by synchronizing the
packet sending and reply. I also tweaked some of the shut down packet
("k" kill, "D" detach, and the halt packet) to make sure they do the right
thing.
Fixed "StateType Process::WaitForProcessStopPrivate (...)" to correctly pass
the timeout along to WaitForStateChangedEventsPrivate() and made the function
behave correctly with respect to timing out.
Added separate STDIN, STDOUT, and STDERR support to debugserver. Also added
the start of being able to set the working directory for the inferior process.
llvm-svn: 124049
threads that we spawn let us know when they are going away and that we
don't timeout waiting for a message from threads that have gone away.
We also now don't expect the "k" packet (kill) to send a response. This
greatly speeds up debugger shutdown performance. The test suite now runs
quite a bit faster.
Added a fix to the variable display code that fixes the display of
base classes. We were assuming the virtual or normal base class offsets
were being given in bit sizes, but they were being given as character
sizes, so we needed to multiply the offset by 8. This wasn't affecting
the expression parser, but it was affecting the correct display of C++
class base classes and all of their children.
llvm-svn: 124024
while the inferior is running) to be fast. The previous code would always
cause the sender to timeout, yet still return success due to the way we
were waiting for a value (incorrect value) to change. Now the ProcessGDBRemote
plug-in has a public and private "is running" predicate. This allows things
that need to send async packets to interrupt and wait for the private "is running"
state to be flipped to false, and then resume quickly with no timeout.
llvm-svn: 123903
I added support for asking if the GDB remote server supports thread suffixes
for packets that should be thread specific (register read/write packets) because
the way the GDB remote protocol does it right now is to have a notion of a
current thread for register and memory reads/writes (set via the "$Hg%x" packet)
and a current thread for running ("$Hc%x"). Now we ask the remote GDB server
if it supports adding the thread ID to the register packets and we enable
that feature in LLDB if supported. This stops us from having to send a bunch
of packets that update the current thread ID to some value which is prone to
error, or extra packets.
llvm-svn: 123762
by LLDB. Instead of being materialized into the input structure
passed to the expression, variables are left in place and pointers
to them are materialzied into the structure. Variables not resident
in memory (notably, registers) get temporary memory regions allocated
for them.
Persistent variables are the most complex part of this, because they
are made in various ways and there are different expectations about
their lifetime. Persistent variables now have flags indicating their
status and what the expectations for longevity are. They can be
marked as residing in target memory permanently -- this is the
default for result variables from expressions entered on the command
line and for explicitly declared persistent variables (but more on
that below). Other result variables have their memory freed.
Some major improvements resulting from this include being able to
properly take the address of variables, better and cleaner support
for functions that return references, and cleaner C++ support in
general. One problem that remains is the problem of explicitly
declared persistent variables; I have not yet implemented the code
that makes references to them into indirect references, so currently
materialization and dematerialization of these variables is broken.
llvm-svn: 123371
new "hexname" key for the "key:value;" duple that is part of the packet. This
allows for thread names to contain special characters such as $ # : ; + -
Debugserver now detects if the thread name contains special characters and
sends the chars in hex format if needed.
llvm-svn: 123053
do. Closing on EOF is an option that can be set on the
lldb_private::Communication or the lldb::SBCommunication objects after they
are created. Of course the EOF support isn't hooked up, so they don't do
anything at the moment, but they are left in so when the code is fixed, it
will be easy to get working again.
llvm-svn: 120885
an error saying the resume timed out. Previously the thread that was trying
to resume the process would eventually call ProcessGDBRemote::DoResume() which
would broadcast an event over to the async GDB remote thread which would sent the
continue packet to the remote gdb server. Right after this was sent, it would
set a predicate boolean value (protected by a mutex and condition) and then the
thread that issued the ProcessGDBRemote::DoResume() would then wait for that
condition variable to be set. If the async gdb thread was too quick though, the
predicate boolean value could have been set to true and back to false by the
time the thread that issued the ProcessGDBRemote::DoResume() checks the boolean
value. So we can't use the predicate value as a handshake. I have changed the code
over to using a Event by having the GDB remote communication object post an
event:
GDBRemoteCommunication::eBroadcastBitRunPacketSent
This allows reliable handshaking between the two threads and avoids the erroneous
ProcessGDBRemote::DoResume() errors.
Added a host backtrace service to allow in process backtraces when trying to track
down tricky issues. I need to see if LLVM has any backtracing abilities abstracted
in it already, and if so, use that, but I needed something ASAP for the current issue
I was working on. The static function is:
void
Host::Backtrace (Stream &strm, uint32_t max_frames);
And it will backtrace at most "max_frames" frames for the current thread and can be
used with any of the Stream subclasses for logging.
llvm-svn: 120793
Add bool member to Communication class indicating whether the
Connection should be closed on receiving an EOF or not. Update the
Connection read to return an EOF status when appropriate. Modify the
Communication class to pass the EOF along or not, and to close the
Connection or not, as appropriate.
llvm-svn: 120723
with the Interrupted bit set. Process::HandlePrivateEvent ignores Interrupted events.
DoHalt is changed to ensure that the stop even is processed, and an event with
the Interrupted event is posted. Finally ClangFunction is rationalized to use this
facility so the that Halt is handled more deterministically.
llvm-svn: 119453
don't crash if we disable logging when some code already has a copy of the
logger. Prior to this fix, logs were handed out as pointers and if they were
held onto while a log got disabled, then it could cause a crash. Now all logs
are handed out as shared pointers so this problem shouldn't happen anymore.
We are also using our new shared pointers that put the shared pointer count
and the object into the same allocation for a tad better performance.
llvm-svn: 118319
than just the entire log channel.
Add checks, where appropriate, to make sure a log channel/category has
not been disabled before attempting to write to it.
llvm-svn: 117715
enabled LLVM make style building and made this compile LLDB on Mac OS X. We
can now iterate on this to make the build work on both linux and macosx.
llvm-svn: 108009
Pavel Labath