GDB Remote Protocol Extensions#
LLDB has added new GDB server packets to better support multi-threaded and
remote debugging. These extend the
protocol defined by GDB (and this page for vFile
packets).
If a packet is restated here it is because LLDB’s version has some behaviour difference to GDB’s version, or it provides some context for a following LLDB extension packet.
Why did we add these? The most common reason is flexibility. Normally you need to start the correct GDB and the correct GDB server when debugging. If you have mismatch, then things go wrong very quickly. LLDB makes extensive use of the GDB remote protocol and we wanted to make sure that the experience was a bit more dynamic where we can discover information about a remote target without having to know anything up front.
We also ran into performance issues with the existing GDB remote protocol that can be overcome when using a reliable communications layer.
Some packets improve performance, others allow for remote process launching (if you have an OS), and others allow us to dynamically figure out what registers a thread might have. Again with GDB, both sides pre-agree on how the registers will look (how many, their register number,name and offsets).
We prefer to be able to dynamically determine what kind of architecture, OS and vendor we are debugging, as well as how things are laid out when it comes to the thread register contexts.
_M<size>,<permissions>#
Allocate memory on the remote target with the specified size and permissions.
The allocate memory packet starts with _M<size>,<permissions>
. It returns a
raw big endian address value, or an empty response for unimplemented, or EXX
for an error
code. The packet is formatted as:
char packet[256];
int packet_len;
packet_len = ::snprintf (
packet,
sizeof(packet),
"_M%zx,%s%s%s",
(size_t)size,
permissions & lldb::ePermissionsReadable ? "r" : "",
permissions & lldb::ePermissionsWritable ? "w" : "",
permissions & lldb::ePermissionsExecutable ? "x" : "");
You request a size and give the permissions. This packet does NOT need to be implemented if you don’t want to support running JITed code. The return value is just the address of the newly allocated memory as raw big endian hex bytes.
Priority To Implement: High if you want LLDB to be able to JIT code and run that code. JIT code also needs data which is also allocated and tracked. Low if you don’t support running JIT’ed code.
_m<addr>#
Deallocate memory that was previously allocated using an allocate memory pack.
The deallocate memory packet is _m<addr>
where you pass in the address you
got back from a previous call to the allocate memory packet. It returns OK
if the memory was successfully deallocated, or EXX
” for an error, or an
empty response if not supported.
Priority To Implement: High if you want LLDB to be able to JIT code and run that code. JIT code also needs data which is also allocated and tracked. Low if you don’t support running JIT’ed code.
“A” - launch args packet#
Launch a program using the supplied arguments
We have added support for the “set program arguments” packet where we can start a connection to a remote server and then later supply the path to the executable and the arguments to use when executing:
GDB remote docs for this:
set program arguments(reserved) Aarglen,argnum,arg,...
Where A is followed by the length in bytes of the hex encoded argument, followed by an argument integer, and followed by the ASCII characters converted into hex bytes for each arg:
send packet: $A98,0,2f566f6c756d65732f776f726b2f67636c6179746f6e2f446f63756d656e74732f7372632f6174746163682f612e6f7574#00
read packet: $OK#00
The above packet helps when you have remote debugging abilities where you could launch a process on a remote host, this isn’t needed for bare board debugging.
Priority To Implement: Low. Only needed if the remote target wants to launch a target after making a connection to a GDB server that isn’t already connected to an inferior process.
“D” - Detach and stay stopped#
We extended the “D” packet to specify that the monitor should keep the target suspended on detach. The normal behavior is to resume execution on detach. We will send:
qSupportsDetachAndStayStopped:
to query whether the monitor supports the extended detach, and if it does, when we want the monitor to detach but not resume the target, we will send:
D1
In any case, if we want the normal detach behavior we will just send:
D
jGetDyldProcessState#
This packet fetches the process launch state, as reported by libdyld on Darwin systems, most importantly to indicate when the system libraries have initialized sufficiently to safely call utility functions.
LLDB SENDS: jGetDyldProcessState
STUB REPLIES: {"process_state_value":48,"process_state string":"dyld_process_state_libSystem_initialized"}
Priority To Implement: Low. This packet is needed to prevent lldb’s utility functions for scanning the Objective-C class list from running very early in process startup.
jGetLoadedDynamicLibrariesInfos#
This packet asks the remote debug stub to send the details about libraries being added/removed from the process as a performance optimization.
There are two ways this packet can be used. Both return a dictionary of binary images formatted the same way.
One requests information on all shared libraries:
jGetLoadedDynamicLibrariesInfos:{"fetch_all_solibs":true}
with an optional "report_load_commands":false
which can be added, asking
that only the dyld SPI information (load addresses, filenames) be returned.
The default behavior is that debugserver scans the mach-o header and load
commands of each binary, and returns it in the JSON reply.
And the second requests information about a list of shared libraries, given their load addresses:
jGetLoadedDynamicLibrariesInfos:{"solib_addresses":[8382824135,3258302053,830202858503]}
The second call is both a performance optimization (instead of having lldb read the mach-o header/load commands out of memory with generic read packets) but also adds additional information in the form of the filename of the shared libraries (which is not available in the mach-o header/load commands.)
An example using the OS X 10.11 style call:
LLDB SENDS: jGetLoadedDynamicLibrariesInfos:{"image_count":1,"image_list_address":140734800075128}
STUB REPLIES: ${"images":[{"load_address":4294967296,"mod_date":0,"pathname":"/tmp/a.out","uuid":"02CF262C-ED6F-3965-9E14-63538B465CFF","mach_header":{"magic":4277009103,"cputype":16777223,"cpusubtype":18446744071562067971,"filetype":2},"segments":{"name":"__PAGEZERO","vmaddr":0,"vmsize":4294967296,"fileoff":0,"filesize":0,"maxprot":0},{"name":"__TEXT","vmaddr":4294967296,"vmsize":4096,"fileoff":0,"filesize":4096,"maxprot":7},{"name":"__LINKEDIT","vmaddr":4294971392,"vmsize":4096,"fileoff":4096,"filesize":152,"maxprot":7}}]}#00
Or pretty-printed:
STUB REPLIES: ${"images":
[
{"load_address":4294967296,
"mod_date":0,
"pathname":"/tmp/a.out",
"uuid":"02CF262C-ED6F-3965-9E14-63538B465CFF",
"mach_header":
{"magic":4277009103,
"cputype":16777223,
"cpusubtype":18446744071562067971,
"filetype":2
},
"segments":
[
{"name":"__PAGEZERO",
"vmaddr":0,
"vmsize":4294967296,
"fileoff":0,
"filesize":0,
"maxprot":0
},
{"name":"__TEXT",
"vmaddr":4294967296,
"vmsize":4096,
"fileoff":0,
"filesize":4096,
"maxprot":7
},
{"name":"__LINKEDIT",
"vmaddr":4294971392,
"vmsize":4096,
"fileoff":4096,
"filesize":152,
"maxprot":7
}
]
}
]
}
This is similar to the qXfer:libraries:read
packet, and it could
be argued that it should be merged into that packet. A separate
packet was created primarily because lldb needs to specify the
number of images to be read and the address from which the initial
information is read. Also the XML DTD would need to be extended
quite a bit to provide all the information that the DynamicLoaderMacOSX
would need to work correctly on this platform.
Priority To Implement:
On OS X 10.11, iOS 9, tvOS 9, watchOS 2 and older: Low. If this packet is absent, lldb will read the Mach-O headers/load commands out of memory. On macOS 10.12, iOS 10, tvOS 10, watchOS 3 and newer: High. If this packet is absent, lldb will not know anything about shared libraries in the inferior, or where the main executable loaded.
jModulesInfo:[{“file”:”…”,triple:”…”}, …]#
Get information for a list of modules by given module path and architecture.
The response is a JSON array of dictionaries containing the following keys:
uuid
triple
file_path
file_offset
file_size
The meaning of the fields is the same as in the qModuleInfo
packet. The server
signals the failure to retrieve the module info for a file by ommiting the
corresponding array entry from the response. The server may also
include entries the client did not ask for, if it has reason to
the modules will be interesting to the client.
Priority To Implement: Optional. If not implemented, qModuleInfo
packet
will be used, which may be slower if the target contains a large number of modules
and the communication link has a non-negligible latency.
jLLDBTraceGetBinaryData#
Get binary data given a trace technology and a data identifier. The input is specified as a JSON object and the response has the same format as the “binary memory read” (aka “x”) packet. In case of failures, an error message is returned.
send packet: jLLDBTraceGetBinaryData:{"type":<type>,"kind":<query>,"tid":<tid>,"offset":<offset>,"size":<size>}]
read packet: <binary data>/E<error code>;AAAAAAAAA
Schema#
The schema for the input is:
{
"type": <string>,
Tracing technology name, e.g. intel-pt, arm-etm.
"kind": <string>,
Identifier for the data.
"cpuId": <Optional decimal>,
Core id in decimal if the data belongs to a CPU core.
"tid"?: <Optional decimal>,
Tid in decimal if the data belongs to a thread.
}
jLLDBTraceGetState#
Get the current state of the process and its threads being traced by a given trace technology. The response is a JSON object with custom information depending on the trace technology. In case of errors, an error message is returned.
send packet: jLLDBTraceGetState:{"type":<type>}]
read packet: {...object}/E<error code>;AAAAAAAAA
Input Schema#
{
"type": <string>
Tracing technology name, e.g. intel-pt, arm-etm.
}
Output Schema#
{
"tracedThreads": [{
"tid": <decimal integer>,
"binaryData": [
{
"kind": <string>,
Identifier for some binary data related to this thread to
fetch with the jLLDBTraceGetBinaryData packet.
"size": <decimal integer>,
Size in bytes of this thread data.
},
]
}],
"processBinaryData": [
{
"kind": <string>,
Identifier for some binary data related to this process to
fetch with the jLLDBTraceGetBinaryData packet.
"size": <decimal integer>,
Size in bytes of this thread data.
},
],
"cpus"?: [
"id": <decimal integer>,
Identifier for this CPU logical core.
"binaryData": [
{
"kind": <string>,
Identifier for some binary data related to this thread to
fetch with the jLLDBTraceGetBinaryData packet.
"size": <decimal integer>,
Size in bytes of this cpu core data.
},
]
],
"warnings"?: [<string>],
Non-fatal messages useful for troubleshooting.
... other attributes specific to the given tracing technology
}
Note: tracedThreads
includes all threads traced by both “process tracing”
and “thread tracing”.
Intel Pt#
If per-cpu process tracing is enabled, “tracedThreads” will contain all the threads of the process without any trace buffers. Besides that, the “cpus” field will also be returned with per cpu core trace buffers. A side effect of per-cpu tracing is that all the threads of unrelated processes will also be traced, thus polluting the tracing data.
Binary data kinds:
iptTrace: trace buffer for a thread or a cpu.
perfContextSwitchTrace: context switch trace for a cpu generated by perf_event_open.
procfsCpuInfo: contents of the /proc/cpuinfo file.
Additional attributes:
tscPerfZeroConversion
This field allows converting Intel processor’s TSC values to nanoseconds. It is available through the Linux perf_event API when cap_user_time and cap_user_time_zero are set. See the documentation of time_zero in https://man7.org/linux/man-pages/man2/perf_event_open.2.html for more information about the calculation and the meaning of the values in the schema below.
Schema for this field:
"tscPerfZeroConversion": { "timeMult": <decimal integer>, "timeShift": <decimal integer>, "timeZero": <decimal integer>, }
jLLDBTraceStart#
Start tracing a process or its threads using a provided tracing technology. The input and output are specified as JSON objects. In case of success, an OK response is returned, or an error otherwise.
Process Tracing#
This traces existing and future threads of the current process. An error is returned if the process is already being traced.
send packet: jLLDBTraceStart:{"type":<type>,...other params}]
read packet: OK/E<error code>;AAAAAAAAA
Thread Tracing#
This traces specific threads.
send packet: jLLDBTraceStart:{"type":<type>,"tids":<tids>,...other params}]
read packet: OK/E<error code>;AAAAAAAAA
Input Schema#
{
"type": <string>,
Tracing technology name, e.g. intel-pt, arm-etm.
/* thread tracing only */
"tids"?: [<decimal integer>],
Individual threads to trace.
... other parameters specific to the provided tracing type
}
Notes:
If “tids” is not provided, then the operation is “process tracing”, otherwise it’s “thread tracing”.
Each tracing technology can have different levels of support for “thread tracing” and “process tracing”.
Intel-Pt#
intel-pt supports both “thread tracing” and “process tracing”.
“Process tracing” is implemented in two different ways. If the “perCpuTracing” option is false, then each thread is traced individually but managed by the same “process trace” instance. This means that the amount of trace buffers used is proportional to the number of running threads. This is the recommended option unless the number of threads is huge. If “perCpuTracing” is true, then each cpu core is traced invidually instead of each thread, which uses a fixed number of trace buffers, but might result in less data available for less frequent threads. See “perCpuTracing” below for more information.
Each actual intel pt trace buffer, either from “process tracing” or “thread tracing”, is stored in an in-memory circular buffer, which keeps the most recent data.
Additional params in the input schema:
{
"iptTraceSize": <decimal integer>,
Size in bytes used by each individual per-thread or per-cpu trace
buffer. It must be a power of 2 greater than or equal to 4096 (2^12)
bytes.
"enableTsc": <boolean>,
Whether to enable TSC timestamps or not. This is supported on
all devices that support intel-pt. A TSC timestamp is generated along
with PSB (synchronization) packets, whose frequency can be configured
with the "psbPeriod" parameter.
"psbPeriod"?: <Optional decimal integer>,
This value defines the period in which PSB packets will be generated.
A PSB packet is a synchronization packet that contains a TSC
timestamp and the current absolute instruction pointer.
This parameter can only be used if
/sys/bus/event_source/devices/intel_pt/caps/psb_cyc
is 1. Otherwise, the PSB period will be defined by the processor.
If supported, valid values for this period can be found in
/sys/bus/event_source/devices/intel_pt/caps/psb_periods
which contains a hexadecimal number, whose bits represent valid
values e.g. if bit 2 is set, then value 2 is valid.
The psb_period value is converted to the approximate number of
raw trace bytes between PSB packets as:
2 ^ (value + 11)
e.g. value 3 means 16KiB between PSB packets. Defaults to
0 if supported.
/* process tracing only */
"perCpuTracing": <boolean>
Instead of having an individual trace buffer per thread, this option
triggers the collection on a per cpu core basis. This effectively
traces the entire activity on all cores. At decoding time, in order
to correctly associate a decoded instruction with a thread, the
context switch trace of each core is needed, as well as a record per
cpu indicating which thread was running on each core when tracing
started. These secondary traces are correlated with the intel-pt
trace by comparing TSC timestamps.
This option forces the capture of TSC timestamps (see "enableTsc").
Note: This option can't be used simulatenously with any other trace
sessions because of its system-wide nature.
/* process tracing only */
"processBufferSizeLimit": <decimal integer>,
Maximum total buffer size per process in bytes.
This limit applies to the sum of the sizes of all thread or cpu core
buffers for the current process, excluding the ones started with
"thread tracing".
If "perCpuTracing" is false, whenever a thread is attempted to be
traced due to "process tracing" and the limit would be reached, the
process is stopped with a "tracing" reason along with a meaningful
description, so that the user can retrace the process if needed.
If "perCpuTracing" is true, then starting the system-wide trace
session fails if all the individual per-cpu trace buffers require
in total more memory that the limit impossed by this parameter.
}
Notes:
Modifying the parameters of an existing trace is not supported. The user needs to stop the trace and start a new one.
If “process tracing” is attempted and there are individual threads already being traced with “thread tracing”, these traces are left unaffected and the threads not traced twice.
If “thread tracing” is attempted on a thread already being traced with either “thread tracing” or “process tracing”, it fails.
jLLDBTraceStop#
Stop tracing a process or its threads using a provided tracing technology. The input and output are specified as JSON objects. In case of success, an OK response is returned, or an error otherwise.
Process Trace Stopping#
Stopping a process trace stops the active traces initiated with “thread tracing”.
send packet: jLLDBTraceStop:{"type":<type>}]
read packet: OK/E<error code>;AAAAAAAAA
Thread Trace Stopping#
This is a best effort request, which tries to stop as many traces as possible.
send packet: jLLDBTraceStop:{"type":<type>,"tids":<tids>}]
read packet: OK/E<error code>;AAAAAAAAA
Input Schema#
The schema for the input is
{
"type": <string>
Tracing technology name, e.g. intel-pt, arm-etm.
/* thread trace stopping only */
"tids": [<decimal integer>]
Individual thread traces to stop.
}
Note: If tids
is not provided, then the operation is “process trace stopping”.
Intel Pt#
Stopping a specific thread trace started with “process tracing” is allowed.
jLLDBTraceSupported#
Get the processor tracing type supported by the gdb-server for the current inferior. Responses might be different depending on the architecture and capabilities of the underlying OS.
send packet: jLLDBTraceSupported
read packet: {"name":<name>, "description":<description>}/E<error code>;AAAAAAAAA
Output Schema#
{
"name": <string>,
Tracing technology name, e.g. intel-pt, arm-etm.
"description": <string>,
Description for this technology.
}
If no tracing technology is supported for the inferior, or no process is running, then an error message is returned.
Note: This packet is used by Trace plug-ins (see lldb_private::Trace.h
) to
do live tracing. Specifically, the name of the plug-in should match the name
of the tracing technology returned by this packet.
jThreadExtendedInfo#
This packet, which takes its arguments as JSON and sends its reply as JSON, allows the gdb remote stub to provide additional information about a given thread.
This packet takes its arguments in JSON. At a minimum, a thread must be specified, for example:
jThreadExtendedInfo:{"thread":612910}
Because this is a JSON string, the thread number is provided in base 10. Additional key-value pairs may be provided by lldb to the gdb remote stub. For instance, on some versions of macOS, lldb can read offset information out of the system libraries. Using those offsets, debugserver is able to find the Thread Specific Address (TSD) for a thread and include that in the return information. So lldb will send these additional fields like so:
jThreadExtendedInfo:{"plo_pthread_tsd_base_address_offset":0,"plo_pthread_tsd_base_offset":224,"plo_pthread_tsd_entry_size":8,"thread":612910}
There are no requirements for what is included in the response. A simple reply on a OS X Yosemite / iOS 8 may include the pthread_t value, the Thread Specific Data (TSD) address, the dispatch_queue_t value if the thread is associated with a GCD queue, and the requested Quality of Service (QoS) information about that thread. For instance, a reply may look like:
{"tsd_address":4371349728,"requested_qos":{"enum_value":33,"constant_name":"QOS_CLASS_USER_INTERACTIVE","printable_name":"User Interactive"},"pthread_t":4371349504,"dispatch_queue_t":140735087127872}
tsd_address
, pthread_t
, and dispatch_queue_t
are all simple key-value pairs.
The JSON standard requires that numbers be expressed in base 10 - so all of
these are. requested_qos
is a dictionary with three key-value pairs in it -
so the UI layer may choose the form most appropriate for displaying to the user.
Sending JSON over gdb-remote protocol introduces some problems. We may be
sending strings with arbitrary contents in them, including the #
, $
, and *
characters that have special meaning in gdb-remote protocol and cannot occur
in the middle of the string. The standard solution for this would be to require
ascii-hex encoding of all strings, or ascii-hex encode the entire JSON payload.
Instead, the binary escaping convention is used for JSON data. This convention
(e.g. used for the X
packet) says that if #
, $
, *
, or }
are to occur in
the payload, the character }
(0x7d
) is emitted, then the metacharacter is emitted
xor’ed by 0x20
. The }
character occurs in every JSON payload at least once, and
} ^ 0x20
happens to be ]
so the raw packet characters for a request will look
like:
jThreadExtendedInfo:{"thread":612910}]
Priority To Implement: Low. This packet is only needed if the gdb remote stub wants to provide interesting additional information about a thread for the user.
jThreadsInfo#
Ask for the server for thread stop information of all threads.
The data in this packet is very similar to the stop reply packets, but is packaged in JSON and uses JSON arrays where applicable. The JSON output looks like:
[
{ "tid":1580681,
"metype":6,
"medata":[2,0],
"reason":"exception",
"qaddr":140735118423168,
"registers": {
"0":"8000000000000000",
"1":"0000000000000000",
"2":"20fabf5fff7f0000",
"3":"e8f8bf5fff7f0000",
"4":"0100000000000000",
"5":"d8f8bf5fff7f0000",
"6":"b0f8bf5fff7f0000",
"7":"20f4bf5fff7f0000",
"8":"8000000000000000",
"9":"61a8db78a61500db",
"10":"3200000000000000",
"11":"4602000000000000",
"12":"0000000000000000",
"13":"0000000000000000",
"14":"0000000000000000",
"15":"0000000000000000",
"16":"960b000001000000",
"17":"0202000000000000",
"18":"2b00000000000000",
"19":"0000000000000000",
"20":"0000000000000000"
},
"memory":[
{"address":140734799804592,"bytes":"c8f8bf5fff7f0000c9a59e8cff7f0000"},
{"address":140734799804616,"bytes":"00000000000000000100000000000000"}
]
}
]
It contains an array of dictionaries with all of the key value pairs that are
normally in the stop reply packet, including the expedited registers. The registers are
passed as hex-encoded JSON string in debuggee-endian byte order. Note that the register
numbers are decimal numbers, unlike the stop-reply packet, where they are written in
hex. The packet also contains expedited memory in the memory
key. This allows the
server to expedite memory that the client is likely to use (e.g., areas around the
stack pointer, which are needed for computing backtraces) and it reduces the packet
count.
On macOS with debugserver, we expedite the frame pointer backchain for a thread (up to 256 entries) by reading 2 pointers worth of bytes at the frame pointer (for the previous FP and PC), and follow the backchain. Most backtraces on macOS and iOS now don’t require us to read any memory!
Priority To Implement: Low
This is a performance optimization, which speeds up debugging by avoiding
multiple round-trips for retrieving thread information. The information from this
packet can be retrieved using a combination of qThreadStopInfo
and m
packets.
QEnvironment:NAME=VALUE#
Setup the environment up for a new child process that will soon be launched using the “A” packet.
NB: key/value pairs are sent as-is so gdb-remote protocol meta characters
(e.g. #
or $
) are not acceptable. If any non-printable or
metacharacters are present in the strings, QEnvironmentHexEncoded
should be used instead if it is available. If you don’t want to
scan the environment strings before sending, prefer
the QEnvironmentHexEncoded
packet over QEnvironment
, if it is
available.
Both GDB and LLDB support passing down environment variables. Is it ok to
respond with a $#00
(unimplemented):
send packet: $QEnvironment:ACK_COLOR_FILENAME=bold yellow#00
read packet: $OK#00
This packet can be sent one or more times prior to sending a “A” packet.
Priority To Implement: Low. Only needed if the remote target wants to launch a target after making a connection to a GDB server that isn’t already connected to an inferior process.
QEnvironmentHexEncoded:HEX-ENCODING(NAME=VALUE)#
Setup the environment up for a new child process that will soon be launched using the “A” packet.
The only difference between this packet and QEnvironment
is that the
environment key-value pair is ascii hex encoded for transmission.
This allows values with gdb-remote metacharacters like #
to be sent.
Both GDB and LLDB support passing down environment variables. Is it ok to
respond with a $#00
(unimplemented):
send packet: $QEnvironment:41434b5f434f4c4f525f46494c454e414d453d626f6c642379656c6c6f77#00
read packet: $OK#00
This packet can be sent one or more times prior to sending a “A” packet.
Priority To Implement: Low. Only needed if the remote target wants to launch a target after making a connection to a GDB server that isn’t already connected to an inferior process.
QEnableCompression#
This packet enables compression of the packets that the debug stub sends to lldb. If the debug stub can support compression, it indictes this in the reply of the “qSupported” packet. For example:
LLDB SENDS: qSupported:xmlRegisters=i386,arm,mips
STUB REPLIES: qXfer:features:read+;SupportedCompressions=lzfse,zlib-deflate,lz4,lzma;
If lldb knows how to use any of these compression algorithms, it can ask that this compression mode be enabled.
QEnableCompression:type:zlib-deflate;
The debug stub should reply with an uncompressed OK
packet to indicate that the
request was accepted. All further packets the stub sends will use this compression.
Packets are compressed as the last step before they are sent from the stub, and decompressed as the first step after they are received. The packet format in compressed mode becomes one of two:
$N<uncompressed payload>#00
$C<size of uncompressed payload in base 10>:<compressed payload>#00
Where #00
is the actual checksum value if noack mode is not enabled. The checksum
value is for the N<uncompressed payload>
or
C<size of uncompressed payload in base 10>:<compressed payload>
bytes in the packet.
The size of the uncompressed payload in base 10 is provided because it will simplify decompression if the final buffer size needed is known ahead of time.
Compression on low-latency connections is unlikely to be an improvement. Particularly when the debug stub and lldb are running on the same host. It should only be used for slow connections, and likely only for larger packets.
Example compression algorithms that may be used include:
zlib-deflate
- The raw DEFLATE format as described in IETF RFC 1951. With the ZLIB library, you can compress to this format with an initialization like deflateInit2 (&stream, 5, Z_DEFLATED, -15, 8, Z_DEFAULT_STRATEGY) and you can decompress with an initialization like inflateInit2 (&stream, -15).lz4
- https://en.wikipedia.org/wiki/LZ4_(compression_algorithm) https://github.com/Cyan4973/lz4 The libcompression APIs on darwin systems call thisCOMPRESSION_LZ4_RAW
.lzfse
- Compression algorithm added in macOS 10.11, with open source C reference implementation on github. https://en.wikipedia.org/wiki/LZFSE https://github.com/lzfse/lzfselzma
- libcompression implements “LZMA level 6”, the default compression for the open source LZMA implementation.
QEnableErrorStrings#
This packet enables reporting of Error strings in remote packet replies from the server to client. If the server supports this feature, it should send an OK response.
send packet: $QEnableErrorStrings
read packet: $OK#00
The client can expect the following error replies if this feature is enabled in the server:
EXX;AAAAAAAAA
where AAAAAAAAA
will be a hex encoded ASCII string.
`XX`` is hex encoded byte number.
It must be noted that even if the client has enabled reporting strings in error replies, it must not expect error strings to all error replies.
Priority To Implement: Low. Only needed if the remote target wants to provide strings that are human readable along with an error code.
QLaunchArch#
Set the architecture to use when launching a process for hosts that can run multiple architecture slices that are contained in a single universal program file.
send packet: $QLaunchArch:<architecture>
The response is OK
if the value in <architecture>
was recognised as valid
and will be used for the next launch request. E63
if not.
Priority To Implement: Only required for hosts that support program files that contain code for multiple architectures.
QListThreadsInStopReply#
Enable the threads:
and thread-pcs:
data in the question-mark packet
(“T packet”) responses when the stub reports that a program has
stopped executing.
send packet: QListThreadsInStopReply
read packet: OK
Priority To Implement: Performance. This is a performance benefit to lldb if the thread id’s and thread pc values are provided to lldb in the T stop packet – if they are not provided to lldb, lldb will likely need to send one to two packets per thread to fetch the data at every private stop.
QRestoreRegisterState:<save_id> / QRestoreRegisterState:<save_id>;thread:XXXX;#
The QRestoreRegisterState
packet tells the remote debugserver to
restore all registers using the save_id
which is an unsigned
integer that was returned from a previous call to
QSaveRegisterState
. The restoration process can only be done once
as the data backing the register state will be freed upon the
completion of the QRestoreRegisterState
command.
If thread suffixes are enabled the second form of this packet is used, otherwise the first form is used.
The response is either:
OK
- if all registers were successfully restoredEXX
- for any errors
Priority To Implement: Low, this is mostly a convenience packet to avoid
having to send all registers with a g
packet. It should only be implemented
if support for the QSaveRegisterState
is added.
QSaveRegisterState / QSaveRegisterState;thread:XXXX;#
The QSaveRegisterState
packet tells the remote debugserver to save
all registers and return a non-zero unique integer ID that
represents these save registers. If thread suffixes are enabled the
second form of this packet is used, otherwise the first form is
used. This packet is called prior to executing an expression, so
the remote GDB server should do anything it needs to in order to
ensure the registers that are saved are correct. On macOS this
involves calling thread_abort_safely(mach_port_t thread)
to
ensure we get the correct registers for a thread in case it is
currently having code run on its behalf in the kernel.
The response is either:
<unsigned int>
- The save_id result is a non-zero unsigned integer value that can be passed back to the GDB server using aQRestoreRegisterState
packet to restore the registers one time.EXX
- or an error code in the form ofEXX
whereXX
is a hex error code.
Priority To Implement: Low, this is mostly a convenience packet to avoid
having to send all registers with a g
packet. It should only be implemented if
support for the QRestoreRegisterState
is added.
QSetDetachOnError#
Sets what the server should do when the communication channel with LLDB
goes down. Either kill the inferior process (0
) or remove breakpoints and
detach (1
).
The data in this packet is a single a character, which should be 0
if the
inferior process should be killed, or 1
if the server should remove all
breakpoints and detach from the inferior.
Priority To Implement: Low. Only required if the target wants to keep the inferior process alive when the communication channel goes down.
QSetDisableASLR:<bool>#
Enable or disable ASLR on the next “A” packet.
Or control if ASLR is enabled/disabled:
send packet: QSetDisableASLR:1
read packet: OK
send packet: QSetDisableASLR:0
read packet: OK
This packet must be sent prior to sending a “A” packet.
Priority To Implement: Low. Only needed if the remote target wants to launch a target after making a connection to a GDB server that isn’t already connected to an inferior process and if the target supports disabling ASLR (Address space layout randomization).
QSetSTDIN:<ascii-hex-path> / QSetSTDOUT:<ascii-hex-path> / QSetSTDERR:<ascii-hex-path>#
Setup where STDIN, STDOUT, and STDERR go prior to sending an “A” packet.
When launching a program through the GDB remote protocol with the “A” packet, you might also want to specify where stdin/out/err go:
QSetSTDIN:<ascii-hex-path>
QSetSTDOUT:<ascii-hex-path>
QSetSTDERR:<ascii-hex-path>
These packets must be sent prior to sending a “A” packet.
Priority To Implement: Low. Only needed if the remote target wants to launch a target after making a connection to a GDB server that isn’t already connected to an inferior process.
QSetWorkingDir:<ascii-hex-path>#
Set the working directory prior to sending an “A” packet.
Or specify the working directory:
QSetWorkingDir:<ascii-hex-path>
This packet must be sent prior to sending a “A” packet.
Priority To Implement: Low. Only needed if the remote target wants to launch a target after making a connection to a GDB server that isn’t already connected to an inferior process.
QStartNoAckMode#
Try to enable no ACK mode to skip sending ACKs and NACKs.
Having to send an ACK/NACK after every packet slows things down a bit, so we have a way to disable ACK packets to minimize the traffic for reliable communication interfaces (like sockets). Below GDB or LLDB will send this packet to try and disable ACKs. All lines that start with “send packet: “ are from GDB/LLDB, and all lines that start with “read packet: “ are from the GDB remote server:
send packet: $QStartNoAckMode#b0
read packet: +
read packet: $OK#9a
send packet: +
Priority To Implement: High. Any GDB remote server that can implement this should if the connection is reliable. This improves packet throughput and increases the performance of the connection.
QSupported#
Query the GDB remote server for features it supports
QSupported is a standard GDB Remote Serial Protocol packet, but there are several additions to the response that lldb can parse. They are not all listed here.
An example exchange:
send packet: qSupported:xmlRegisters=i386,arm,mips,arc;multiprocess+;fork-events+;vfork-events+
read packet: qXfer:features:read+;PacketSize=20000;qEcho+;native-signals+;SupportedCompressions=lzfse,zlib-deflate,lz4,lzma;SupportedWatchpointTypes=aarch64-mask,aarch64-bas;
In the example above, three lldb extensions are shown:
PacketSize=20000
The base 16 maximum packet size that the stub can handle.
SupportedCompressions=<item,item,...>
A list of compression types that the stub can use to compress packets when the QEnableCompression packet is used to request one of them.
SupportedWatchpointTypes=<item,item,...>
A list of watchpoint types that this stub can manage. Currently defined names are:
x86_64
- 64-bit x86-64 watchpoints (1, 2, 4, 8 byte watchpoints aligned to those amounts)aarch64-bas
AArch64 Byte Address Select watchpoints (any number of contiguous bytes within a doubleword)aarch64-mask
AArch64 MASK watchpoints (any power-of-2 region of memory from 8 to 2GB, aligned)
If nothing is specified, lldb will default to sending power-of-2 watchpoints, up to a pointer size,
sizeof(void*)
, a reasonable baseline assumption.
Priority To Implement: Optional
QThreadSuffixSupported#
Try to enable thread suffix support for the g
, G
, p
, and P
packets.
When reading thread registers, you currently need to set the current
thread, then read the registers. This is kind of cumbersome, so we added the
ability to query if the remote GDB server supports adding a thread:<tid>;
suffix to all packets that request information for a thread. To test if the
remote GDB server supports this feature:
send packet: $QThreadSuffixSupported#00
read packet: OK
If OK
is returned, then the g
, G
, p
and P
packets can accept a
thread suffix. So to send a g
packet (read all register values):
send packet: $g;thread:<tid>;#00
read packet: ....
send packet: $G;thread:<tid>;#00
read packet: ....
send packet: $p1a;thread:<tid>;#00
read packet: ....
send packet: $P1a=1234abcd;thread:<tid>;#00
read packet: ....
otherwise, without this you would need to always send two packets:
send packet: $Hg<tid>#00
read packet: ....
send packet: $g#00
read packet: ....
We also added support for allocating and deallocating memory. We use this to allocate memory so we can run JITed code.
Priority To Implement: High
Adding a thread suffix allows us to read and write registers
more efficiently and stops us from having to select a thread with
one packet and then read registers with a second packet. It also
makes sure that no errors can occur where the debugger thinks it
already has a thread selected (see the Hg
packet from the standard
GDB remote protocol documentation) yet the remote GDB server actually
has another thread selected.
qAttachOrWaitSupported#
This is a binary “is it supported” query. Return OK if you support
vAttachOrWait
.
Priority To Implement: Low. This is required if you support vAttachOrWait
,
otherwise no support is needed since the standard “I don’t recognize this packet”
response will do the right thing.
qFileLoadAddress:<file_path>#
Get the load address of a memory mapped file. The load address is defined as the address of the first memory region what contains data mapped from the specified file.
The response is either:
<unsigned-hex64>
- Load address of the file in big endian encodingE01
- the requested file isn’t loadedEXX
- for any other errors
Priority To Implement: Low, required if dynamic linker don’t fill in the load address of some object file in the rendezvous data structure.
qfProcessInfo / qsProcessInfo (Platform Extension)#
Get the first process info (qfProcessInfo
) or subsequent process
info (qsProcessInfo
) for one or more processes on the remote
platform. The first call gets the first match and subsequent calls
to qsProcessInfo
gets the subsequent matches. Return an error EXX
,
where XX
are two hex digits, when no more matches are available.
The qfProcessInfo
packet can be followed by a :
and
some key value pairs. The key value pairs in the command are:
name
-ascii-hex
- An ASCII hex string that contains the name of the process that will be matched.name_match
-enum
- One of:equals
starts_with
ends_with
contains
regex
pid
-integer
- A string value containing the decimal process IDparent_pid
-integer
- A string value containing the decimal parent process IDuid
-integer
- A string value containing the decimal user IDgid
-integer
- A string value containing the decimal group IDeuid
-integer
- A string value containing the decimal effective user IDegid
-integer
- A string value containing the decimal effective group IDall_users
-bool
- A boolean value that specifies if processes should be listed for all users, not just the user that the platform is running astriple
-string
- An ASCII triple string (x86_64
,x86_64-apple-macosx
,armv7-apple-ios
)args
-string
- A string value containing the process arguments separated by the character-
, where each argument is hex-encoded. It includesargv[0]
.
The response consists of key/value pairs where the key is separated from the
values with colons and each pair is terminated with a semi colon. For a list
of the key/value pairs in the response see the qProcessInfoPID
packet
documentation.
Sample packet/response:
send packet: $qfProcessInfo#00
read packet: $pid:60001;ppid:59948;uid:7746;gid:11;euid:7746;egid:11;name:6c6c6462;triple:x86_64-apple-macosx;#00
send packet: $qsProcessInfo#00
read packet: $pid:59992;ppid:192;uid:7746;gid:11;euid:7746;egid:11;name:6d64776f726b6572;triple:x86_64-apple-macosx;#00
send packet: $qsProcessInfo#00
read packet: $E04#00
Priority To Implement: Required
qGDBServerVersion#
Get version information about this implementation of the gdb-remote protocol.
The goal of this packet is to provide enough information about an implementation of the gdb-remote-protocol server that lldb can work around implementation problems that are discovered after the version has been released/deployed. The name and version number should be sufficiently unique that lldb can unambiguously identify the origin of the program (for instance, debugserver from lldb) and the version/submission number/patch level of the program - whatever is appropriate for your server implementation.
The packet follows the key-value pair model, semicolon separated.
send packet: $qGDBServerVersion#00
read packet: $name:debugserver;version:310.2;#00
Other clients may find other key-value pairs to be useful for identifying a gdb stub. Patch level, release name, build number may all be keys that better describe your implementation’s version.
Suggested key names:
name
: the name of your remote server - “debugserver” is the lldb standard implementationversion
: identifies the version number of this serverpatch_level
: the patch level of this serverrelease_name
: the name of this release, if your project uses namesbuild_number
: if you use a build system with increasing build numbers, this may be the right key name for your servermajor_version
: major version numberminor_version
: minor version number
Priority To Implement: High. This packet is usually very easy to implement and can help LLDB to work around bugs in a server’s implementation when they are found.
qGetWorkingDir#
Get the current working directory of the platform stub in ASCII hex encoding.
receive: qGetWorkingDir
send: 2f4170706c65496e7465726e616c2f6c6c64622f73657474696e67732f342f5465737453657474696e67732e746573745f646973617373656d626c65725f73657474696e6773
qHostInfo#
Get information about the host we are remotely connected to.
LLDB supports a host info call that gets all sorts of details of the system that is being debugged:
send packet: $qHostInfo#00
read packet: $cputype:16777223;cpusubtype:3;ostype:darwin;vendor:apple;endian:little;ptrsize:8;#00
Key value pairs are one of:
cputype
: is a number that is the mach-o CPU type that is being debugged (base 10)cpusubtype
: is a number that is the mach-o CPU subtype type that is being debugged (base 10)triple
: a string for the target triple (x86_64-apple-macosx) that can be used to specify arch + vendor + os in one entryvendor
: a string for the vendor (apple), not needed if “triple” is specifiedostype
: a string for the OS being debugged (macosx, linux, freebsd, ios, watchos), not needed if “triple” is specifiedendian
: is one of “little”, “big”, or “pdp”ptrsize
: an unsigned number that represents how big pointers are in bytes on the debug targethostname
: the hostname of the host that is running the GDB server if availableos_build
: a string for the OS build for the remote host as a string valueos_kernel
: a string describing the kernel versionos_version
: a version string that represents the current OS version (10.8.2)watchpoint_exceptions_received
: one of “before” or “after” to specify if a watchpoint is triggered before or after the pc when it stopsdefault_packet_timeout
: an unsigned number that specifies the default timeout in secondsdistribution_id
: optional. For linux, specifies distribution id (e.g. ubuntu, fedora, etc.)osmajor
: optional, specifies the major version number of the OS (e.g. for macOS 10.12.2, it would be 10)osminor
: optional, specifies the minor version number of the OS (e.g. for macOS 10.12.2, it would be 12)ospatch
: optional, specifies the patch level number of the OS (e.g. for macOS 10.12.2, it would be 2)vm-page-size
: optional, specifies the target system VM page size, base 10. Needed for the “dirty-pages:” list in the qMemoryRegionInfo packet, where a list of dirty pages is sent from the remote stub. This page size tells lldb how large each dirty page is.addressing_bits
: optional, specifies how many bits in addresses are significant for addressing, base 10. If bits 38..0 in a 64-bit pointer are significant for addressing, then the value is 39. This is needed on e.g. AArch64 v8.3 ABIs that use pointer authentication, so lldb knows which bits to clear/set to get the actual addresses.low_mem_addressing_bits
: optional, specifies how many bits in addresses in low memory are significant for addressing, base 10. AArch64 can have different page table setups for low and high memory, and therefore a different number of bits used for addressing.high_mem_addressing_bits
: optional, specifies how many bits in addresses in high memory are significant for addressing, base 10. AArch64 can have different page table setups for low and high memory, and therefore a different number of bits used for addressing.
Priority To Implement: High. This packet is usually very easy to implement and can help LLDB select the correct plug-ins for the job based on the target triple information that is supplied.
qKillSpawnedProcess (Platform Extension)#
Kill a process running on the target system.
receive: qKillSpawnedProcess:1337
send: OK
The request packet has the process ID in base 10.
qLaunchGDBServer (Platform Extension)#
Have the remote platform launch a GDB server.
The qLaunchGDBServer
packet must be followed by a :
and
some key value pairs. The key value pairs in the command are:
port
-integer
- A string value containing the decimal port ID or zero if the port should be bound and returnedhost
-integer
- The host that connections should be limited to when the GDB server is connected to.
Sample packet/response:
send packet: $qLaunchGDBServer:port:0;host:lldb.apple.com;#00
read packet: $pid:60025;port:50776;#00
The pid
key/value pair is only specified if the remote platform launched
a separate process for the GDB remote server and can be omitted if no
process was separately launched.
The port
key/value pair in the response lets clients know what port number
to attach to in case zero was specified as the “port” in the sent command.
Priority To Implement: Required
qLaunchSuccess#
Check whether launching a process with the A
packet succeeded.
Returns the status of the last attempt to launch a process.
Either OK
if no error ocurred, or E
followed by a string
describing the error.
Priority To Implement: High, launching processes is a key part of LLDB’s platform mode.
qMemoryRegionInfo:<addr>#
Get information about the address range that contains <addr>
.
We added a way to get information for a memory region. The packet is:
qMemoryRegionInfo:<addr>
Where <addr>
is a big endian hex address. The response is returned in a series
of tuples like the data returned in a stop reply packet. The currently valid
tuples to return are:
start:<start-addr>;
-<start-addr>
is a big endian hex address that is the start address of the range that contains<addr>
size:<size>;
-<size>
is a big endian hex byte size of the address of the range that contains<addr>
permissions:<permissions>;
-<permissions>
is a string that contains one or more of the characters fromrwx
name:<name>;
-<name>
is a hex encoded string that contains the name of the memory region mapped at the given address. In case of regions backed by a file it have to be the absolute path of the file while for anonymous regions it have to be the name associated to the region if that is available.flags:<flags-string>;
- where<flags-string>
is a space separated string of flag names. Currently the only supported flag ismt
for AArch64 memory tagging. lldb will ignore any other flags in this field.type:[<type>][,<type>];
- memory types that apply to this region, e.g.stack
for stack memory.error:<ascii-byte-error-string>;
- where<ascii-byte-error-string>
is a hex encoded string value that contains an error stringdirty-pages:[<hexaddr>][,<hexaddr];
- A list of memory pages within this region that are “dirty” – they have been modified. Page addresses are in base 16. The size of a page can be found from theqHostInfo
’spage-size
key-value.If the stub supports identifying dirty pages within a memory region, this key should always be present for all
qMemoryRegionInfo
replies. This key with no pages listed (dirty-pages:;
) indicates no dirty pages in this memory region. The absence of this key means that this stub cannot determine dirty pages.
If the address requested is not in a mapped region (e.g. we’ve jumped through a NULL pointer and are at 0x0) currently lldb expects to get back the size of the unmapped region – that is, the distance to the next valid region. For instance, with a macOS process which has nothing mapped in the first 4GB of its address space, if we’re asking about address 0x2:
qMemoryRegionInfo:2
start:2;size:fffffffe;
The lack of permissions:
indicates that none of read/write/execute are valid
for this region.
The stub must include permissions:
key-value on all memory ranges
that are valid to access in the inferior process – the lack of
permissions:
means that this is an inaccessible (no page table
entries exist, in a system using VM) memory range. If a stub cannot
determine actual permissions, return rwx
.
Priority To Implement: Medium
This is nice to have, but it isn’t necessary. It helps LLDB do stack unwinding when we branch into memory that isn’t executable. If we can detect that the code we are stopped in isn’t executable, then we can recover registers for stack frames above the current frame. Otherwise we must assume we are in some JIT’ed code (not JIT code that LLDB has made) and assume that no registers are available in higher stack frames.
qModuleInfo:<module_path>;<arch triple>#
Get information for a module by given module path and architecture.
The response is either:
(uuid|md5):...;triple:...;file_offset:...;file_size...;
EXX
- for any errors
Priority To Implement: Optional, required if dynamic loader cannot fetch module’s information like UUID directly from inferior’s memory.
qPathComplete (Platform Extension)#
Get a list of matched disk files/directories by passing a boolean flag and a partial path.
receive: qPathComplete:0,6d61696e
send: M6d61696e2e637070
receive: qPathComplete:1,746573
send: M746573742f,74657374732f
If the first argument is zero, the result should contain all files (including directories) starting with the given path. If the argument is one, the result should contain only directories.
The result should be a comma-separated list of hex-encoded paths.
Paths denoting a directory should end with a directory separator (/
or \
.
qPlatform_mkdir#
Creates a new directory on the connected remote machine.
Request: qPlatform_mkdir:<hex-file-mode>,<ascii-hex-path>
The request packet has the fields:
mode bits in base 16
file path in ascii-hex encoding
Reply:
F<mkdir-return-code>
(mkdir called successfully and returned with the given return code)Exx
(An error occurred)
Priority To Implement: Low
qPlatform_shell#
Run a command in a shell on the connected remote machine.
The request consists of the command to be executed encoded in ASCII characters converted into hex bytes.
The response to this packet consists of the letter F followed by the return code, followed by the signal number (or 0 if no signal was delivered), and escaped bytes of captured program output.
Below is an example communication from a client sending an “ls -la” command:
send packet: $qPlatform_shell:6c73202d6c61,00000002#ec
read packet: $F,00000000,00000000,total 4736
drwxrwxr-x 16 username groupname 4096 Aug 15 21:36 .
drwxr-xr-x 17 username groupname 4096 Aug 10 16:39 ..
-rw-rw-r-- 1 username groupname 73875 Aug 12 16:46 notes.txt
drwxrwxr-x 5 username groupname 4096 Aug 15 21:36 source.cpp
-rw-r--r-- 1 username groupname 2792 Aug 12 16:46 a.out
-rw-r--r-- 1 username groupname 3190 Aug 12 16:46 Makefile
Priority To Implement: High
qProcessInfo#
Get information about the process we are currently debugging.
Priority To Implement: Medium
On systems which can launch multiple different architecture processes, the qHostInfo may not disambiguate sufficiently to know what kind of process is being debugged.
For example on a 64-bit x86 Mac system both 32-bit and 64-bit user processes are possible, and with Mach-O universal files, the executable file may contain both 32- and 64-bit slices so it may be impossible to know until you’re attached to a real process to know what you’re working with.
All numeric fields return base 16 numbers without any “0x” prefix.
An i386 process:
send packet: $qProcessInfo#00
read packet: $pid:42a8;parent-pid:42bf;real-uid:ecf;real-gid:b;effective-uid:ecf;effective-gid:b;cputype:7;cpusubtype:3;ostype:macosx;vendor:apple;endian:little;ptrsize:4;#00
An x86_64 process:
send packet: $qProcessInfo#00
read packet: $pid:d22c;parent-pid:d34d;real-uid:ecf;real-gid:b;effective-uid:ecf;effective-gid:b;cputype:1000007;cpusubtype:3;ostype:macosx;vendor:apple;endian:little;ptrsize:8;#00
Key value pairs include:
pid
: the process idparent-pid
: the process of the parent process (often debugserver will become the parent when attaching)real-uid
: the real user id of the processreal-gid
: the real group id of the processeffective-uid
: the effective user id of the processeffective-gid
: the effective group id of the processcputype
: the Mach-O CPU type of the process (base 16)cpusubtype
: the Mach-O CPU subtype of the process (base 16)ostype
: is a string the represents the OS being debugged (darwin, linux, freebsd)vendor
: is a string that represents the vendor (apple)endian
: is one of “little”, “big”, or “pdp”ptrsize
: is a number that represents how big pointers are in bytesmain-binary-uuid
: is the UUID of a firmware type binary that the gdb stub knows aboutmain-binary-address
: is the load address of the firmware type binarymain-binary-slide
: is the slide of the firmware type binary, if address isn’t knownbinary-addresses
: A comma-separated list of binary load addresses base 16. lldb will parse the binaries in memory to get UUIDs, then try to find the binaries & debug info by UUID. Intended for use with a small number of firmware type binaries where the search for binary/debug info may be expensive.
qProcessInfoPID:PID (Platform Extension)#
Have the remote platform get detailed information on a process by ID. PID is specified as a decimal integer.
The response consists of key/value pairs where the key is separated from the values with colons and each pair is terminated with a semi colon.
The key value pairs in the response are:
pid
-integer
- Process ID as a decimal integer stringppid
-integer
- Parent process ID as a decimal integer stringuid
-integer
- A string value containing the decimal user IDgid
-integer
- A string value containing the decimal group IDeuid
-integer
- A string value containing the decimal effective user IDegid
-integer
- A string value containing the decimal effective group IDname
-ascii-hex
- An ASCII hex string that contains the name of the processtriple
-string
- A target triple (x86_64-apple-macosx
,armv7-apple-ios
)
Sample packet/response:
send packet: $qProcessInfoPID:60050#00
read packet: $pid:60050;ppid:59948;uid:7746;gid:11;euid:7746;egid:11;name:6c6c6462;triple:x86_64-apple-macosx;#00
Priority To Implement: Optional
qQueryGDBServer#
Ask the platform for the list of gdbservers we have to connect
If the remote platform automatically started one or more gdbserver instance (without lldb asking it) then it have to return the list of port number or socket name for each of them what can be used by lldb to connect to those instances.
The data in this packet is a JSON array of JSON objects with the following keys:
port
:<the port number to connect>
(optional)socket_name
:<the name of the socket to connect>
(optional)
Example packet:
[
{ "port": 1234 },
{ "port": 5432 },
{ "socket_name": "foo" }
]
Priority To Implement: Low
The packet is required to support connecting to gdbserver started by the platform instance automatically.
qRegisterInfo<hex-reg-id>#
Discover register information from the remote GDB server.
With LLDB, for register information, remote GDB servers can add support for the “qRegisterInfoN” packet where “N” is a zero based base 16 register number that must start at zero and increase by one for each register that is supported. The response is done in typical GDB remote fashion where a series of “KEY:VALUE;” pairs are returned. An example for the x86_64 registers is included below:
send packet: $qRegisterInfo0#00
read packet: $name:rax;bitsize:64;offset:0;encoding:uint;format:hex;set:General Purpose Registers;gcc:0;dwarf:0;#00
send packet: $qRegisterInfo1#00
read packet: $name:rbx;bitsize:64;offset:8;encoding:uint;format:hex;set:General Purpose Registers;gcc:3;dwarf:3;#00
send packet: $qRegisterInfo2#00
read packet: $name:rcx;bitsize:64;offset:16;encoding:uint;format:hex;set:General Purpose Registers;gcc:2;dwarf:2;#00
send packet: $qRegisterInfo3#00
read packet: $name:rdx;bitsize:64;offset:24;encoding:uint;format:hex;set:General Purpose Registers;gcc:1;dwarf:1;#00
send packet: $qRegisterInfo4#00
read packet: $name:rdi;bitsize:64;offset:32;encoding:uint;format:hex;set:General Purpose Registers;gcc:5;dwarf:5;#00
send packet: $qRegisterInfo5#00
read packet: $name:rsi;bitsize:64;offset:40;encoding:uint;format:hex;set:General Purpose Registers;gcc:4;dwarf:4;#00
send packet: $qRegisterInfo6#00
read packet: $name:rbp;alt-name:fp;bitsize:64;offset:48;encoding:uint;format:hex;set:General Purpose Registers;gcc:6;dwarf:6;generic:fp;#00
send packet: $qRegisterInfo7#00
read packet: $name:rsp;alt-name:sp;bitsize:64;offset:56;encoding:uint;format:hex;set:General Purpose Registers;gcc:7;dwarf:7;generic:sp;#00
send packet: $qRegisterInfo8#00
read packet: $name:r8;bitsize:64;offset:64;encoding:uint;format:hex;set:General Purpose Registers;gcc:8;dwarf:8;#00
send packet: $qRegisterInfo9#00
read packet: $name:r9;bitsize:64;offset:72;encoding:uint;format:hex;set:General Purpose Registers;gcc:9;dwarf:9;#00
send packet: $qRegisterInfoa#00
read packet: $name:r10;bitsize:64;offset:80;encoding:uint;format:hex;set:General Purpose Registers;gcc:10;dwarf:10;#00
send packet: $qRegisterInfob#00
read packet: $name:r11;bitsize:64;offset:88;encoding:uint;format:hex;set:General Purpose Registers;gcc:11;dwarf:11;#00
send packet: $qRegisterInfoc#00
read packet: $name:r12;bitsize:64;offset:96;encoding:uint;format:hex;set:General Purpose Registers;gcc:12;dwarf:12;#00
send packet: $qRegisterInfod#00
read packet: $name:r13;bitsize:64;offset:104;encoding:uint;format:hex;set:General Purpose Registers;gcc:13;dwarf:13;#00
send packet: $qRegisterInfoe#00
read packet: $name:r14;bitsize:64;offset:112;encoding:uint;format:hex;set:General Purpose Registers;gcc:14;dwarf:14;#00
send packet: $qRegisterInfof#00
read packet: $name:r15;bitsize:64;offset:120;encoding:uint;format:hex;set:General Purpose Registers;gcc:15;dwarf:15;#00
send packet: $qRegisterInfo10#00
read packet: $name:rip;alt-name:pc;bitsize:64;offset:128;encoding:uint;format:hex;set:General Purpose Registers;gcc:16;dwarf:16;generic:pc;#00
send packet: $qRegisterInfo11#00
read packet: $name:rflags;alt-name:flags;bitsize:64;offset:136;encoding:uint;format:hex;set:General Purpose Registers;#00
send packet: $qRegisterInfo12#00
read packet: $name:cs;bitsize:64;offset:144;encoding:uint;format:hex;set:General Purpose Registers;#00
send packet: $qRegisterInfo13#00
read packet: $name:fs;bitsize:64;offset:152;encoding:uint;format:hex;set:General Purpose Registers;#00
send packet: $qRegisterInfo14#00
read packet: $name:gs;bitsize:64;offset:160;encoding:uint;format:hex;set:General Purpose Registers;#00
send packet: $qRegisterInfo15#00
read packet: $name:fctrl;bitsize:16;offset:176;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo16#00
read packet: $name:fstat;bitsize:16;offset:178;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo17#00
read packet: $name:ftag;bitsize:8;offset:180;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo18#00
read packet: $name:fop;bitsize:16;offset:182;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo19#00
read packet: $name:fioff;bitsize:32;offset:184;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo1a#00
read packet: $name:fiseg;bitsize:16;offset:188;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo1b#00
read packet: $name:fooff;bitsize:32;offset:192;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo1c#00
read packet: $name:foseg;bitsize:16;offset:196;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo1d#00
read packet: $name:mxcsr;bitsize:32;offset:200;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo1e#00
read packet: $name:mxcsrmask;bitsize:32;offset:204;encoding:uint;format:hex;set:Floating Point Registers;#00
send packet: $qRegisterInfo1f#00
read packet: $name:stmm0;bitsize:80;offset:208;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:33;dwarf:33;#00
send packet: $qRegisterInfo20#00
read packet: $name:stmm1;bitsize:80;offset:224;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:34;dwarf:34;#00
send packet: $qRegisterInfo21#00
read packet: $name:stmm2;bitsize:80;offset:240;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:35;dwarf:35;#00
send packet: $qRegisterInfo22#00
read packet: $name:stmm3;bitsize:80;offset:256;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:36;dwarf:36;#00
send packet: $qRegisterInfo23#00
read packet: $name:stmm4;bitsize:80;offset:272;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:37;dwarf:37;#00
send packet: $qRegisterInfo24#00
read packet: $name:stmm5;bitsize:80;offset:288;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:38;dwarf:38;#00
send packet: $qRegisterInfo25#00
read packet: $name:stmm6;bitsize:80;offset:304;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:39;dwarf:39;#00
send packet: $qRegisterInfo26#00
read packet: $name:stmm7;bitsize:80;offset:320;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:40;dwarf:40;#00
send packet: $qRegisterInfo27#00
read packet: $name:xmm0;bitsize:128;offset:336;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:17;dwarf:17;#00
send packet: $qRegisterInfo28#00
read packet: $name:xmm1;bitsize:128;offset:352;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:18;dwarf:18;#00
send packet: $qRegisterInfo29#00
read packet: $name:xmm2;bitsize:128;offset:368;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:19;dwarf:19;#00
send packet: $qRegisterInfo2a#00
read packet: $name:xmm3;bitsize:128;offset:384;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:20;dwarf:20;#00
send packet: $qRegisterInfo2b#00
read packet: $name:xmm4;bitsize:128;offset:400;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:21;dwarf:21;#00
send packet: $qRegisterInfo2c#00
read packet: $name:xmm5;bitsize:128;offset:416;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:22;dwarf:22;#00
send packet: $qRegisterInfo2d#00
read packet: $name:xmm6;bitsize:128;offset:432;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:23;dwarf:23;#00
send packet: $qRegisterInfo2e#00
read packet: $name:xmm7;bitsize:128;offset:448;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:24;dwarf:24;#00
send packet: $qRegisterInfo2f#00
read packet: $name:xmm8;bitsize:128;offset:464;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:25;dwarf:25;#00
send packet: $qRegisterInfo30#00
read packet: $name:xmm9;bitsize:128;offset:480;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:26;dwarf:26;#00
send packet: $qRegisterInfo31#00
read packet: $name:xmm10;bitsize:128;offset:496;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:27;dwarf:27;#00
send packet: $qRegisterInfo32#00
read packet: $name:xmm11;bitsize:128;offset:512;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:28;dwarf:28;#00
send packet: $qRegisterInfo33#00
read packet: $name:xmm12;bitsize:128;offset:528;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:29;dwarf:29;#00
send packet: $qRegisterInfo34#00
read packet: $name:xmm13;bitsize:128;offset:544;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:30;dwarf:30;#00
send packet: $qRegisterInfo35#00
read packet: $name:xmm14;bitsize:128;offset:560;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:31;dwarf:31;#00
send packet: $qRegisterInfo36#00
read packet: $name:xmm15;bitsize:128;offset:576;encoding:vector;format:vector-uint8;set:Floating Point Registers;gcc:32;dwarf:32;#00
send packet: $qRegisterInfo37#00
read packet: $name:trapno;bitsize:32;offset:696;encoding:uint;format:hex;set:Exception State Registers;#00
send packet: $qRegisterInfo38#00
read packet: $name:err;bitsize:32;offset:700;encoding:uint;format:hex;set:Exception State Registers;#00
send packet: $qRegisterInfo39#00
read packet: $name:faultvaddr;bitsize:64;offset:704;encoding:uint;format:hex;set:Exception State Registers;#00
send packet: $qRegisterInfo3a#00
read packet: $E45#00
As we see above we keep making subsequent calls to the remote server to
discover all registers by increasing the number appended to qRegisterInfo
and
we get a response back that is a series of key=value;
strings.
The offset:
fields should not leave a gap anywhere in the g/G packet – the
register values should be appended one after another. For instance, if the
register context for a thread looks like:
struct rctx {
uint32_t gpr1; // offset 0
uint32_t gpr2; // offset 4
uint32_t gpr3; // offset 8
uint64_t fp1; // offset 16
};
You may end up with a 4-byte gap between gpr3 and fp1 on architectures that align values like this. The correct offset: value for fp1 is 12 - in the g/G packet fp1 will immediately follow gpr3, even though the in-memory thread structure has an empty 4 bytes for alignment between these two registers.
The keys and values are detailed below:
name
- The primary register name as a string (“rbp” for example)alt-name
- An alternate name for a register as a string (“fp” for example for the above “rbp”)bitsize
- Size in bits of a register (32, 64, etc). Base 10.offset
- The offset within the “g” and “G” packet of the register data for this register. This is the byte offset once the data has been transformed into binary, not the character offset into the g/G packet. Base 10.encoding
- The encoding type of the register which must be one of:uint
(unsigned integer)sint
(signed integer)ieee754
(IEEE 754 float)vector
(vector register)
format - The preferred format for display of this register. The value must be one of:
binary
decimal
hex
float
vector-sint8
vector-uint8
vector-sint16
vector-uint16
vector-sint32
vector-uint32
vector-float32
vector-uint128
set
- The register set name as a string that this register belongs to.gcc
- The GCC compiler registers number for this register (used for EH frame and other compiler information that is encoded in the executable files). The supplied number will be decoded like a string passed to strtoul() with a base of zero, so the number can be decimal, or hex if it is prefixed with “0x”.Note: If the compiler doesn’t have a register number for this register, this key/value pair should be omitted.
dwarf
- The DWARF register number for this register that is used for this register in the debug information. The supplied number will be decoded like a string passed to strtoul() with a base of zero, so the number can be decimal, or hex if it is prefixed with “0x”.Note: If the compiler doesn’t have a register number for this register, this key/value pair should be omitted.
generic
- If the register is a generic register that most CPUs have, classify it correctly so the debugger knows. Valid values are one of:pc
(a program counter register. for examplename=eip;
(i386),name=rip;
(x86_64),name=r15;
(32 bit arm) would include ageneric=pc;
key value pair)sp
(a stack pointer register. for examplename=esp;
(i386),name=rsp;
(x86_64),name=r13;
(32 bit arm) would include ageneric=sp;
key value pair)fp
(a frame pointer register. for examplename=ebp;
(i386),name=rbp;
(x86_64),name=r7;
(32 bit arm with macosx ABI) would include ageneric=fp;
key value pair)ra
(a return address register. for examplename=lr;
(32 bit ARM) would include ageneric=ra;
key value pair)flags
(a CPU flags register. for examplename=eflags;
(i386),name=rflags;
(x86_64),name=cpsr;
(32 bit ARM) would include ageneric=flags;
key value pair)arg1
-arg8
(specified for registers that contain function arguments when the argument fits into a register)
container-regs
- The value for this key is a comma separated list of raw hex (optional leading “0x”) register numbers.This specifies that this register is contained in other concrete register values. For example “eax” is in the lower 32 bits of the “rax” register value for x86_64, so “eax” could specify that it is contained in “rax” by specifying the register number for “rax” (whose register number is 0x00):
container-regs:00;
If a register is comprised of one or more registers, like “d0” is ARM which is a 64 bit register, it might be made up of “s0” and “s1”. If the register number for “s0” is 0x20, and the register number of “s1” is “0x21”, the “container-regs” key/value pair would be:
container-regs:20,21;
This is handy for defining what GDB used to call “pseudo” registers. These registers are never requested by LLDB via the register read or write packets, the container registers will be requested on behalf of this register.
invalidate-regs
- The value for this key is a comma separated list of raw hex (optional leading “0x”) register numbers.This specifies which register values should be invalidated when this register is modified. For example if modifying “eax” would cause “rax”, “eax”, “ax”, “ah”, and “al” to be modified where rax is 0x0, eax is 0x15, ax is 0x25, ah is 0x35, and al is 0x39, the “invalidate-regs” key/value pair would be:
invalidate-regs:0,15,25,35,39;
If there is a single register that gets invalidated, then omit the comma and just list a single register:
invalidate-regs:0;
This is handy when modifying a specific register can cause other register values to change. For example, when debugging an ARM target, modifying the CPSR register can cause the r8 - r14 and cpsr value to change depending on if the mode has changed.
Priority To Implement: High. Any target that can self describe its registers, should do so. This means if new registers are ever added to a remote target, they will get picked up automatically, and allows registers to change depending on the actual CPU type that is used.
Note: qRegisterInfo
is deprecated in favor of the standard gdb remote
serial protocol register description method, qXfer:features:read:target.xml
.
If qXfer:features:read:target.xml
is supported, qRegisterInfo
does
not need to be implemented. The target.xml format is used by most
gdb RSP stubs whereas qRegisterInfo
was an lldb-only design.
qRegisterInfo
requires one packet per register and can have undesirable
performance costs at the start of a debug session, whereas target.xml
may be able to describe all registers in a single packet.
qShlibInfoAddr#
Get an address where the dynamic linker stores information about where shared libraries are loaded.
LLDB and GDB both support the qShlibInfoAddr
packet which is a hint to each
debugger as to where to find the dynamic loader information. For darwin
binaries that run in user land this is the address of the all_image_infos
structure in the /usr/lib/dyld
executable, or the result of a TASK_DYLD_INFO
call. The result is returned as big endian hex bytes that are the address
value:
send packet: $qShlibInfoAddr#00
read packet: $7fff5fc40040#00
Priority To Implement: High
If you have a dynamic loader plug-in in LLDB for your target triple (see the “qHostInfo” packet) that can use this information. Many times address load randomization can make it hard to detect where the dynamic loader binary and data structures are located and some platforms know, or can find out where this information is.
Low if you have a debug target where all object and symbol files contain static load addresses.
qSpeedTest#
Test the maximum speed at which packets can be sent and received.
send packet: qSpeedTest:response_size:<response size>;
read packet: data:<response data>
<response size>
is a hex encoded unsigned number up to 64 bits in size.
The remote will respond with data:
followed by a block of a
characters
whose size should match <response size>
, if the connection is stable.
If there is an error parsing the packet, the response is E79
.
This packet is used by LLDB to discover how reliable the connection is by
varying the amount of data requested by <response size>
and checking whether
the expected amount and values were received.
Priority to Implemment: Not required for debugging on the same host, otherwise low unless you know your connection quality is variable.
qSymbol#
Notify the remote that LLDB is ready to do symbol lookups on behalf of the
debug server. The response is the symbol name the debug server wants to know the
value of, or OK
if the debug server does not need to know any more symbol values.
The exchange always begins with:
send packet: qSymbol::
The ::
are delimiters for fields that may be filled in future responses. These
delimiters must be included even in the first packet sent.
The debug server can reply one of two ways. If it doesn’t need any symbol values:
read packet: OK
If it does need a symbol value, it includes the ASCII hex encoded name of the symbol:
read packet: qSymbol:6578616D706C65
This should be looked up by LLDB then sent back to the server. Include the name again, with the vaue as a hex number:
read packet: qSymbol:6578616D706C65:CAFEF00D
If LLDB cannot find the value, it should respond with only the name. Note that
the second :
is not included here, whereas it is in the initial packet.
read packet: qSymbol:6578616D706C65
If LLDB is asked for any symbols that it cannot find, it should send the
initial qSymbol::
again at any point where new libraries are loaded. In case
the symbol can now be resolved.
If the debug server has requested all the symbols it wants, the final response
will be OK
(whether they were all found or not).
If LLDB did find all the symbols and recieves an OK
it does not need to send
qSymbol::
again during the debug session.
Priority To Implement: Low, this is rarely used.
qThreadStopInfo<tid>#
Get information about why a thread, whose ID is <tid>
, is stopped.
LLDB tries to use the qThreadStopInfo
packet which is formatted as
qThreadStopInfo%x
where %x
is the hex thread ID. This requests information
about why a thread is stopped. The response is the same as the stop reply
packets and tells us what happened to the other threads. The standard GDB
remote packets love to think that there is only one reason that one thread
stops at a time. This allows us to see why all threads stopped and allows us
to implement better multi-threaded debugging support.
Priority To Implement: High
If you need to support multi-threaded or multi-core debugging. Many times one thread will hit a breakpoint and while the debugger is in the process of suspending the other threads, other threads will also hit a breakpoint. This packet allows LLDB to know why all threads (live system debug) / cores (JTAG) in your program have stopped and allows LLDB to display and control your program correctly.
Stop reply packet extensions#
This section describes some of the additional information you can specify in stop reply packets that help LLDB to know more detailed information about your threads.
Standard GDB remote stop reply packets are reply packets sent in response to a packet that made the program run. They come in the following forms:
SAA
-S
means signal andAA
is a hex signal number that describes why the thread or stopped. It doesn’t specify which thread, so theT
packet is recommended to use instead of theS
packet.TAAkey1:value1;key2:value2;...
-T
means a thread stopped due to a unix signal whereAA
is a hex signal number that describes why the program stopped. This is followed by a series of key/value pairs:If key is a hex number, it is a register number and value is the hex value of the register in debuggee endian byte order.
If key == “thread”, then the value is the big endian hex thread-id of the stopped thread.
If key == “core”, then value is a hex number of the core on which the stop was detected.
If key == “watch” or key == “rwatch” or key == “awatch”, then value is the data address in big endian hex
If key == “library”, then value is ignore and “qXfer:libraries:read” packets should be used to detect any newly loaded shared libraries
WAA
-W
means the process exited andAA
is the exit status.XAA
-X
means the process exited andAA
is signal that caused the program to exit.O<ascii-hex-string>
-O
meansSTDOUT
has data that was written to its console and is being delivered to the debugger. This packet happens asynchronously and the debugger is expected to continue to wait for another stop reply packet.
Lldb Extensions#
We have extended the T
packet to be able to also understand the
following keys and values:
metype
-unsigned
- mach exception type (the value of theEXC_XXX
enumerations) as an unsigned integer. For targets with mach kernels only.mecount
-unsigned
- mach exception data count as an unsigned integer For targets with mach kernels only.medata
-unsigned
- There should bemecount
of these and it is the data that goes along with a mach exception (as an unsigned integer). For targets with mach kernels only.name
-string
- The name of the thread as a plain string. The string must not contain an special packet characters or contain a:
or a;
. Usehexname
if the thread name has special characters.hexname
-ascii-hex
- An ASCII hex string that contains the name of the threadqaddr
-hex
- Big endian hex value that contains thelibdispatch
queue address for the queue of the thread.reason
-enum
- The enumeration must be one of:trace
- the program stopped after a single instruction was executed on a core. Usually done when single stepping past a breakpointbreakpoint
- a breakpoint set using az
packet was hit.trap
- stopped due to user interruptionsignal
- stopped due to an actual unix signal, not just the debugger using a unix signal to keep the GDB remote client happy.watchpoint
- Can be used with of thewatch
/rwatch
/awatch
key value pairs. Or can be used instead of those keys, with the specially formatteddescription
field.exception
- an exception stop reason. Use with thedescription
key/value pair to describe the exceptional event the user should see as the stop reason.description
- An ASCII hex string that contains a more descriptive reason that the thread stopped. This is only needed if none of the key/value pairs are enough to describe why something stopped.For
reason:watchpoint
,description
is an ascii-hex encoded string with between one and three base 10 numbers, space separated. The three numbers are:Watchpoint address. This address should always be within a memory region lldb has a watchpoint on. On architectures where the actual reported hit address may be outside the watchpoint that was triggered, the remote stub should determine which watchpoint was triggered and report an address from within its range.
Wwatchpoint hardware register index number.
Actual watchpoint trap address, which may be outside the range of any watched region of memory. On MIPS, an addr outside a watched range means lldb should disable the wp, step, re-enable the wp and continue silently.
On MIPS, the low 3 bits are masked so if a watchpoint is on 0x1004, a 2-byte write to 0x1000 will trigger the watchpoint (a false positive hit), and lldb needs to disable the watchpoint at 0x1004, inst-step, then re-enable the watchpoint and not make this a user visible event. The description here would be “0x1004 0 0x1000”. lldb needs a known watchpoint address in the first field, so it can disable it and step.
On AArch64 we have a related issue, where you watch 4 bytes at 0x1004, an instruction does an 8-byte write starting at 0x1000 (a true watchpoint hit) and the hardware may report the trap address as 0x1000 - before the watched memory region - with the write extending into the watched region. This can be reported as “0x1004 0 0x1000”. lldb will use 0x1004 to identify which Watchpoint was triggered, and can report 0x1000 to the user. The behavior of silently stepping over the watchpoint, with an 3rd field addr outside the range, is restricted to MIPS.
There may be false-positive watchpoint hits on AArch64 as well, in the SVE Streaming Mode, but that is less common (see ESR register flag “WPF”, “Watchpoint might be False-Positive”) and not currently handled by lldb.
threads
-comma-sep-base16
- A list of thread ids for all threads (including the thread that we’re reporting as stopped) that are live in the process right now. lldb may request that this be included in the T packet via the QListThreadsInStopReply packet earlier in the debug session.Example:
threads:63387,633b2,63424,63462,63486;
thread-pcs
-comma-sep-base16
- A list of pc values for all threads that currently exist in the process, including the thread that thisT
packet is reporting as stopped. This key-value pair will only be emitted when thethreads
key is already included in theT
packet. The pc values correspond to the threads reported in thethreads
list. The number of pcs in thethread-pcs
list will be the same as the number of threads in thethreads
list. lldb may request that this be included in theT
packet via theQListThreadsInStopReply
packet earlier in the debug session.Example:
thread-pcs:dec14,2cf872b0,2cf8681c,2d02d68c,2cf716a8;
addressing_bits
-unsigned
(optional) - Specifies how many bits in addresses are significant for addressing, base 10. If bits 38..0 in a 64-bit pointer are significant for addressing, then the value is 39. This is needed on e.g. AArch64 v8.3 ABIs that use pointer authentication in the high bits. This value is normally sent in theqHostInfo
packet response, and if the value cannot change during the process lifetime, it does not need to be duplicated here in the stop packet. For a firmware environment with early start code that may be changing the page table setup, a dynamically set value may be needed.low_mem_addressing_bits
-unsigned
(optional) - Specifies how many bits in addresses in low memory are significant for addressing, base 10. AArch64 can have different page table setups for low and high memory, and therefore a different number of bits used for addressing.high_mem_addressing_bits
-unsigned
(optional) - Specifies how many bits in addresses in high memory are significant for addressing, base 10. AArch64 can have different page table setups for low and high memory, and therefore a different number of bits used for addressing.
Best Practices#
Since register values can be supplied with this packet, it is often useful to return the PC, SP, FP, LR (if any), and FLAGS registers so that separate packets don’t need to be sent to read each of these registers from each thread.
If a thread is stopped for no reason (like just because another thread
stopped, or because when one core stops all cores should stop), use a
T
packet with 00
as the signal number and fill in as many key values
and registers as possible.
LLDB likes to know why a thread stopped since many thread control operations like stepping over a source line, actually are implemented by running the process multiple times. If a breakpoint is hit while trying to step over a source line and LLDB finds out that a breakpoint is hit in the “reason”, we will know to stop trying to do the step over because something happened that should stop us from trying to do the step. If we are at a breakpoint and we disable the breakpoint at the current PC and do an instruction single step, knowing that we stopped due to a “trace” helps us know that we can continue running versus stopping due to a “breakpoint” (if we have two breakpoint instruction on consecutive instructions). So the more info we can get about the reason a thread stops, the better job LLDB can do when controlling your process. A typical GDB server behavior is to send a SIGTRAP for breakpoints and also when instruction single stepping, in this case the debugger doesn’t really know why we stopped and it can make it hard for the debugger to control your program correctly. What if a real SIGTRAP was delivered to a thread while we were trying to single step? We wouldn’t know the difference with a standard GDB remote server and we could do the wrong thing.
Priority To Implement: High. Having the extra information in your stop reply packets makes your debug session more reliable and informative.
vAttachName#
Same as vAttach
, except instead of a pid
you send a process name.
Priority To Implement: Low. Only needed for process attach -n
. If the
packet isn’t supported then process attach -n
will fail gracefully. So you need
only to support it if attaching to a process by name makes sense for your environment.
vAttachOrWait#
Same as vAttachWait
, except that the stub will attach to a process
by name if it exists, and if it does not, it will wait for a process
of that name to appear and attach to it.
Priority To Implement: Low
Only needed to implement process attach -w -i false -n
. If
you don’t implement it but do implement -n
AND lldb can somehow get
a process list from your device, it will fall back on scanning the
process list, and sending vAttach
or vAttachWait
depending on
whether the requested process exists already. This is racy,
however, so if you want to support this behavior it is better to
support this packet.
vAttachWait#
Same as vAttachName
, except that the stub should wait for the next instance
of a process by that name to be launched and attach to that.
Priority To Implement: Low. Only needed to support process attach -w -n
which will fail gracefully if the packet is not supported.
vFile Packets#
Though some of these may match the ones described in GDB’s protocol documentation, we include our own expectations here in case of mismatches or extensions.
vFile:chmod / qPlatform_chmod#
Change the permissions of a file on the connected remote machine.
Request: qPlatform_chmod:<hex-file-mode>,<ascii-hex-path>
Reply:
F<chmod-return-code>
(chmod called successfully and returned with the given return code)Exx
(An error occurred)
vFile:close#
Close a previously opened file descriptor.
receive: vFile:close:7
send: F0
File descriptor is in base 16. F-1,errno
with the errno if an error occurs,
errno is base 16.
vFile:exists#
Check whether the file at the given path exists.
receive: vFile:exists:2f746d702f61
send (exists): F,1
send (does not exist): F,0
Request packet contains the ASCII hex encoded filename.
The response is a return code where 1 means the file exists and 0 means it does not.
Priority To Implement: Low
vFile:MD5#
Generate an MD5 hash of the file at the given path.
receive: vFile:MD5:2f746d702f61
send (success): F,00000000000000001111111111111111
send (failure): F,x
Request packet contains the ASCII hex encoded filename.
If the hash succeeded, the response is F,
followed by the low 64
bits of the result, and finally the high 64 bits of the result. Both are in
hex format without a prefix.
The response is F,
, followed by x
if the file did not exist
or failed to hash.
vFile:mode#
Get the mode bits of a file on the target system, filename in ASCII hex.
receive: vFile:mode:2f746d702f61
send: F1ed
response is F
followed by the mode bits in base 16, this 0x1ed
would
correspond to 0755
in octal.
F-1,errno
with the errno if an error occurs, base 16.
vFile:open#
Open a file on the remote system and return the file descriptor of it.
receive: vFile:open:2f746d702f61,00000001,00000180
send: F8
request packet has the fields:
ASCII hex encoded filename
Flags passed to the open call, base 16. Note that these are not the
oflags
thatopen(2)
takes, but are the constant values inenum OpenOptions
from LLDB’sFile.h
.Mode bits, base 16
response is F
followed by the opened file descriptor in base 16.
F-1,errno
with the errno if an error occurs, base 16.
vFile:pread#
Read data from an opened file descriptor.
receive: vFile:pread:7,1024,0
send: F4;a'b\00
Request packet has the fields:
File descriptor, base 16
Number of bytes to be read, base 16
Offset into file to start from, base 16
Response is F
, followed by the number of bytes read (base 16), a
semicolon, followed by the data in the binary-escaped-data encoding.
vFile:pwrite#
Write data to a previously opened file descriptor.
receive: vFile:pwrite:8,0,\cf\fa\ed\fe\0c\00\00
send: F1024
Request packet has the fields:
File descriptor, base 16
Offset into file to start from, base 16
binary-escaped-data to be written
Response is F
, followed by the number of bytes written (base 16).
vFile:size#
Get the size of a file on the target system, filename in ASCII hex.
receive: vFile:size:2f746d702f61
send: Fc008
response is F
followed by the file size in base 16.
F-1,errno
with the errno if an error occurs, base 16.
vFile:symlink#
Create a symbolic link (symlink, soft-link) on the target system.
receive: vFile:symlink:<SRC-FILE>,<DST-NAME>
send: F0,0
Argument file paths are in ascii-hex encoding.
Response is F
plus the return value of symlink()
, base 16 encoding,
optionally followed by the value of errno if it failed, also base 16.
vFile:unlink#
Remove a file on the target system.
receive: vFile:unlink:2f746d702f61
send: F0
Argument is a file path in ascii-hex encoding.
Response is F
plus the return value of unlink()
, base 16 encoding.
Return value may optionally be followed by a comma and the base16
value of errno if unlink failed.
“x” - Binary memory read#
Like the m
(read) and M
(write) packets, this is a partner to the
X
(write binary data) packet, x
.
It is called like
xADDRESS,LENGTH
where both ADDRESS
and LENGTH
are big-endian base 16 values.
To test if this packet is available, send a addr/len of 0:
x0,0
You will get an OK
response if it is supported.
The reply will be the data requested in 8-bit binary data format.
The standard quoting is applied to the payload. Characters } # $ *
will all be escaped with }
(0x7d
) character and then XOR’ed with 0x20
.
A typical use to read 512 bytes at 0x1000 would look like:
x0x1000,0x200
The 0x
prefixes are optional - like most of the gdb-remote packets,
omitting them will work fine; these numbers are always base 16.
The length of the payload is not provided. A reliable, 8-bit clean, transport layer is assumed.