Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20191205174635.18758-13-vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
build_guest_fsinfo_for_virtual_device() dereferences @errp when
build_guest_fsinfo_for_device() fails. That's wrong; see the big
comment in error.h. Introduced in commit 46d4c5723e "qga: Add
guest-get-fsinfo command".
No caller actually passes null.
Fix anyway: splice in a local Error *err, and error_propagate().
Cc: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20191204093625.14836-10-armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
The Posix implementation of guest-set-time invokes hwclock to
set/retrieve the time to/from the hardware clock. If hwclock
is not available, the user is currently informed that "hwclock
failed to set hardware clock to system time", which is quite
misleading. This may happen e.g. on s390x, which has a different
timekeeping concept anyway.
Let's check for the availability of the hwclock command and
return QERR_UNSUPPORTED for guest-set-time if it is not available.
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Message-Id: <20191205115350.18713-1-cohuck@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Memory block commands are only supported for linux with sysfs,
"guest-get-memory-block-info" was not in blacklist for other
cases.
Reported on:
https://bugzilla.redhat.com/show_bug.cgi?id=1751431
Signed-off-by: Basil Salman <bsalman@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Network interface name is fetched as an encoded WCHAR array, (wide
character), then it is decoded using the guest's CP_ACP Windows code
page, which is the default code page as configure in the guest's
Windows, then it is returned as a byte array, (char array).
As stated in the BZ#1733165, when renaming a network interface to a
Chinese name and invoking this command, the returned name field has
the (\ufffd) value for each Chinese character the name had, this
value is an indication that the code page does not have the decoding
information for the given character.
This bug is a result of using the CP_ACP code page for decoding which
is an interchangeable code page, instead CP_UTF8 code page should be
used for decoding the network interface's name.
https://bugzilla.redhat.com/show_bug.cgi?id=1733165
Signed-off-by: Bishara AbuHattoum <bishara@daynix.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Current parameter was always one. We continue with that value for now
in all callers.
Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
---
Moved trace to socket_listen
No header includes qemu-common.h after this commit, as prescribed by
qemu-common.h's file comment.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190523143508.25387-5-armbru@redhat.com>
[Rebased with conflicts resolved automatically, except for
include/hw/arm/xlnx-zynqmp.h hw/arm/nrf51_soc.c hw/arm/msf2-soc.c
block/qcow2-refcount.c block/qcow2-cluster.c block/qcow2-cache.c
target/arm/cpu.h target/lm32/cpu.h target/m68k/cpu.h target/mips/cpu.h
target/moxie/cpu.h target/nios2/cpu.h target/openrisc/cpu.h
target/riscv/cpu.h target/tilegx/cpu.h target/tricore/cpu.h
target/unicore32/cpu.h target/xtensa/cpu.h; bsd-user/main.c and
net/tap-bsd.c fixed up]
Leading underscores are ill-advised because such identifiers are
reserved. Trailing underscores are merely ugly. Strip both.
Our header guards commonly end in _H. Normalize the exceptions.
Done with scripts/clean-header-guards.pl.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190315145123.28030-7-armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
[Changes to slirp/ dropped, as we're about to spin it off]
The win2qemu[] is supposed to be the conversion table to convert between
STORAGE_BUS_TYPE in Windows SDK and GuestDiskBusType in qga.
But it was incorrectly written that it forces to set a GuestDiskBusType
value to STORAGE_BUS_TYPE, which generates an enum conversion warning in clang.
Suggested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Cao Jiaxi <driver1998@foxmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-id: 20190503003650.10137-1-driver1998@foxmail.com
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Simplify the code around qmp_dispatch():
- rely on qmp_dispatch/check_obj() for message checking
- have a single send_response() point
- constify send_response() argument
It changes a couple of error messages:
* When @req isn't a dictionary, from
Invalid JSON syntax
to
QMP input must be a JSON object
* When @req lacks member "execute", from
this feature or command is not currently supported
to
QMP input lacks member 'execute'
CC: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
The Windows QEMU guest agent erroneously tries to collect PCI information
directly from the physical drive. However, windows stores SCSI/IDE information
with the drive and PCI information with the underlying storage controller
This changes get_pci_info to use the physical drive's underlying storage
controller to get PCI information.
* Additionally Fixes incorrect size being passed to DeviceIoControl
when getting volume extents. Can occasionally crash the guest agent
Signed-off-by: Matt Hines <mhines@scalecomputing.com>
*fix up some checkpatch warnings
*fix domain reporting and add some sanity checks for debug
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Commit 7be41675f7 set -std=gnu99 for C code via QEMU_CFLAGS. Currently
we generate a "custom" QEMU_CXXFLAGS for VSS DLL C++ build by
filtering out some options from QEMU_CFLAGS and adding some others.
Since we don't filter out -std=gnu99 currently this breaks builds when
VSS support is enabled.
We could keep the existing approach, filter out -std=gnu99 from
QEMU_CFLAGS, and add -std=gnu++98, like configure currently does for
QEMU_CXXFLAGS, but as it turns out our resulting QEMU_CXXFLAGS would
be exactly what configure already generates, just with these filtered
out:
-fstack-protector-all -fstack-protector-strong
and these added:
-Wno-unknown-pragmas -Wno-delete-non-virtual-dtor
So fix the issue by re-using configure-generated QEMU_CXXFLAGS and
just handling these specific changes.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Commit 3ebee3b191 defined assert() as g_assert(), but when we build
the VSS DLL component of QGA (to handle fsfreeze) we do not include
glib, which results in breakage when building with VSS support enabled.
Fix this by including glib (along with the -lintl and -lws2_32
dependencies it brings).
Since the VSS DLL is built statically, this introduces an additional
dependency on static glib and supporting libs for the mingw environment
(possibly why we didn't include glib originally), but VSS support
already has very specific prerequisites so it shouldn't affect too many
build environments.
Since the VSS DLL code does use qemu/osdep.h, this should also help
avoid future breakages and possibly allow for some clean ups in current
VSS code.
Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
Cc: Daniel P. Berrangé <berrange@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Since Windows Server 2016, Microsoft stopped upgrading the major and minor
versions of their new Windows Server product, so, the current functionality
of checking major and minor version numbers to determine the Windows Server
version wont work as expected.
The implemented solution here is to use the build number in addition to the
major and minor version numbers of the product to determine the Windows
Server product version.
The final build number of Windows Server 2016 is 14939, and
the final build number of Windows Server 2019 is 17764, so any Windows
Server product that has the major version of 10 and minor version of 0
with a build number lower or equal to 14939 will resemble 2016 and if the
build number is lower or equal to 17763 will resemble 2019.
Reference:
https://techcommunity.microsoft.com/t5/Windows-Server-Insiders/Windows-Server-2019-version-info/m-p/293112/highlight/true#M859
Signed-off-by: Bishara AbuHattoum <bishara@daynix.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Building QGA for XP seems possible so far: the dependency on
libqemuutil.a implies building qemu-thread-win32.c, which requires
Vista API since commit 12f8def0 (v2.9). But qemu-thread isn't being
used in QGA, the resulting binary may still work on XP. XP is no
longer supported for the past 4.5y, it's time to drop support for it.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20181122110039.15972-5-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Both qemu & qga build with Vista API by default already, by defining
_WIN32_WINNT 0x0600. Set it globally in osdep.h instead.
This replaces WINVER by _WIN32_WINNT in osdep.h. WINVER doesn't seem
to be really useful these days.
(see also https://blogs.msdn.microsoft.com/oldnewthing/20070411-00/?p=27283)
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20181122110039.15972-4-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch updates the descriptions of 'guest-suspend-ram' and
'guest-suspend-hybrid' to mention that both commands relies now
on the proper support for wake up from suspend, retrieved by the
'wakeup-suspend-support' attribute of the 'query-current-machine'
QMP command.
Reported-by: Balamuruhan S <bala24@linux.vnet.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Message-Id: <20181205194701.17836-3-danielhb413@gmail.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
The guest-agent-core.h header was missing the usual guards
against multiple inclusion; add them.
(Spotted by lgtm.com's static analyzer.)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Introduced in commit b1ba8890e6, vol_h
handle should be closed, and "out" cleanup should be done after
DeviceIoControl() fails.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Currently whenever the qemu-ga's service doesn't find the virtio-serial
the run_agent() loops in a QGA_RETRY_INTERVAL (default 5 seconds)
intervals and try to restart the qemu-ga which causes a synchronous loop.
Changed to wait and listen for the serial events by registering for
notifications a proper serial event handler that deals with events:
DBT_DEVICEARRIVAL indicates that the device has been inserted and
is available
DBT_DEVICEREMOVECOMPLETE indicates that the devive has been removed
Which allow us to determine when the channel path is available for the
qemu-ga to restart.
Signed-off-by: Bishara AbuHattoum <bishara@daynix.com>
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Useful in general, but especially now that errors might occur more
frequently with --retry-path set.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
It's nicer from a management perspective that the agent can survive
hotplug/unplug of the channel device, or be started prior to the
installation of the channel device's driver without and still be able
to resume normal function afterward. On linux there are alternatives
like systemd to support this, but on w32 --retry-path is the only
option so it makes sense to set it by default when installed as a
w32 service.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
This adds an option to instruct the agent to periodically attempt
re-opening the communication channel after a channel error has
occurred. The main use-case for this is providing an OS-independent
way of allowing the agent to survive situations like hotplug/unplug of
the communication channel, or initial guest set up where the agent may
be installed/started prior to the installation of the channel device's
driver.
There are nicer ways of implementing this functionality via things
like systemd services, but this option is useful for platforms like
*BSD/w32.
Currently a channel error will result in the GSource for that channel
being removed from the GMainLoop, but the main loop continuing to run.
That behavior results in a dead loop when --retry-path isn't set, and
prevents us from knowing when to attempt re-opening the channel when
it is set, so we also force the loop to exit as part of this patch.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Eventually we want a w32 service to be able to restart the qga main
loop from within service_main(). To allow for this we move service
handling out of run_agent() such that service_main() calls
run_agent() instead of the reverse.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Bishara AbuHattoum <bishara@daynix.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
For w32 services we rely on the global GAState to access resources
associated with the agent within service_main(). Currently this is
sufficient for starting the agent since we open the channel once prior
to calling service_main(), and simply start the GMainLoop to start the
agent from within service_main().
Eventually we want to be able to also [re-]open the communication
channel from within service_main(), which requires access to
config/socket_activation variables, so we hang them off GAState in
preparation for that.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
*dont move GAConfig struct, just the typedef
*fix build bisect for w32
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
This patch better separates the init/cleanup routines out into
separate functions to make the start-up procedure a bit easier to
follow. This will be useful when we eventually break out the actual
start/stop of the agent's main loop into separates routines that
can be called multiple times after the init phase.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Li Qiang <liq3ea@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
It was not obvious what exactly the cryptic string copying does to the
GUID. This change makes the intent clearer.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Report device UNC of the disk. It is reported as "\\.\PhysicalDriveX".
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Probe the volume for disk extents and return list of all disks.
Originally only first disk of composite volume was returned.
Note that the patch changes get_pci_info() from one state of brokenness
into a different state of brokenness. In other words it still does not do
what it's supposed to do (see comment in code). If anyone knows how to
fix it, please step in.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Refactor building of disk info into a function that builds the list and
a function that returns infor for single disk. This will be used in
future commit that will handle multi-disk volumes.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
*coding style fix-ups (declarations at beginning of block)
*improve readability for user-visible errors
*cover additional edge-cases with debug statements
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Refactor code that queries bus type to be more generic. The function
get_disk_bus_type() has been renamed to build_guest_disk_info().
Following commit(s) will extend this function.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The windows code generaly lacks debug information (compared to posix
code). This patch adds some related to HW info in guest-get-fsinfo
command.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The call to SetupDiGetDeviceRegistryProperty might fail because the
value doesn't exist in the registry, in this case we shouldn't exit from
the loop but instead continue to look for other available values in the
registry and set this value as unavailable (-1).
Signed-off-by: Sameeh Jubran <sjubran@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
*squash in fix for when get_pci_info() returns NULL pci_controller field
*fix handling for error_set() cases in get_pci_info(), not just NULL return
*force all -1 PCI addr fields if any single one of them isn't found
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The fsinfo command is currently implemented for Windows only and it's disk
parameter can be enabled by adding the define "CONFIG_QGA_NTDDSCSI" to the qga
code. When enabled and executed the qemu-ga crashed with the following message:
------------------------------------------------
File qapi/qapi-visit-core.c, Line 49
Expression: !(v->type & VISITOR_OUTPUT) || *obj)
------------------------------------------------
After some digging, turns out that the GuestPCIAddress is null and the
qapi visitor doesn't like that, so we can always allocate it instead and
initiate all it's members to -1.
Signed-off-by: Sameeh Jubran <sjubran@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Report device node of the disk on Linux (e.g. "/dev/sda2").
Requirs libudev.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Add reporting of disk serial number on Linux guests. The feature depends
on libudev.
Example:
{
"name": "dm-2",
"mountpoint": "/",
...
"disk": [
{
"serial": "SAMSUNG_MZ7LN512HCHP-000L1_S1ZKNXAG822493",
...
}
],
}
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
If VM has VCPUs plugged sparselly (for example a VM started with
3 VCPUs (cpu0, cpu1 and cpu2) and then cpu1 was hotunplugged so
only cpu0 and cpu2 are present), QGA will rise a error
error: internal error: unable to execute QEMU agent command 'guest-get-vcpus':
open("/sys/devices/system/cpu/cpu1/"): No such file or directory
when
virsh vcpucount FOO --guest
is executed.
Fix it by ignoring non present CPUs when fetching CPUs status from sysfs.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
This patch add support for freeze specified fs.
The valid mountpoints list member are [1]:
The path of a mounted folder, for example, Y:\MountX\
A drive letter, for example, D:\
A volume GUID path of the form \\?\Volume{GUID}\,
where GUID identifies the volume
A UNC path that specifies a remote file share,
for example, \\Clusterx\Share1\
[1] https://docs.microsoft.com/en-us/windows/desktop/api/vsbackup/nf-vsbackup-ivssbackupcomponents-addtosnapshotset
Cc: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Chen Hanxiao <chenhanxiao@gmail.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Currently, the win32 port of QEMU Guest Agent does not properly handle Unicode
paths. The JSON decoder produces a valid UTF-8 path string, but this is passed
directly to CreateFileA, which is expecting an ANSI string and not UTF-8. This
leads to mangled filenames.
This patch follows the example of qmp_guest_set_user_password() and uses
g_utf8_to_utf16() to convert the string to UTF-16 and calls CreateFileW()
explicitly.
Signed-off-by: Jonathon Reinhart <jreinhart@cc-sw.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The JSON parser has three public headers, json-lexer.h, json-parser.h,
json-streamer.h. They all contain stuff that is of no interest
outside qobject/json-*.c.
Collect the public interface in include/qapi/qmp/json-parser.h, and
everything else in qobject/json-parser-int.h.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20180823164025.12553-54-armbru@redhat.com>
The callback to consume JSON values takes QObject *json, Error *err.
If both are null, the callback is supposed to make up an error by
itself. This sucks.
qjson.c's consume_json() neglects to do so, which makes
qobject_from_json() null instead of failing. I consider that a bug.
The culprit is json_message_process_token(): it passes two null
pointers when it runs into a lexical error or a limit violation. Fix
it to pass a proper Error object then. Update the callbacks:
* monitor.c's handle_qmp_command(): the code to make up an error is
now dead, drop it.
* qga/main.c's process_event(): lumps the "both null" case together
with the "not a JSON object" case. The former is now gone. The
error message "Invalid JSON syntax" is misleading for the latter.
Improve it to "Input must be a JSON object".
* qobject/qjson.c's consume_json(): no update; check-qjson
demonstrates qobject_from_json() now sets an error on lexical
errors, but still doesn't on some other errors.
* tests/libqtest.c's qmp_response(): the Error object is now reliable,
so use it to improve the error message.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20180823164025.12553-40-armbru@redhat.com>
The classical way to structure parser and lexer is to have the client
call the parser to get an abstract syntax tree, the parser call the
lexer to get the next token, and the lexer call some function to get
input characters.
Another way to structure them would be to have the client feed
characters to the lexer, the lexer feed tokens to the parser, and the
parser feed abstract syntax trees to some callback provided by the
client. This way is more easily integrated into an event loop that
dispatches input characters as they arrive.
Our JSON parser is kind of between the two. The lexer feeds tokens to
a "streamer" instead of a real parser. The streamer accumulates
tokens until it got the sequence of tokens that comprise a single JSON
value (it counts curly braces and square brackets to decide). It
feeds those token sequences to a callback provided by the client. The
callback passes each token sequence to the parser, and gets back an
abstract syntax tree.
I figure it was done that way to make a straightforward recursive
descent parser possible. "Get next token" becomes "pop the first
token off the token sequence". Drawback: we need to store a complete
token sequence. Each token eats 13 + input characters + malloc
overhead bytes.
Observations:
1. This is not the only way to use recursive descent. If we replaced
"get next token" by a coroutine yield, we could do without a
streamer.
2. The lexer reports errors by passing a JSON_ERROR token to the
streamer. This communicates the offending input characters and
their location, but no more.
3. The streamer reports errors by passing a null token sequence to the
callback. The (already poor) lexical error information is thrown
away.
4. Having the callback receive a token sequence duplicates the code to
convert token sequence to abstract syntax tree in every callback.
5. Known bug: the streamer silently drops incomplete token sequences.
This commit rectifies 4. by lifting the call of the parser from the
callbacks into the streamer. Later commits will address 3. and 5.
The lifting removes a bug from qjson.c's parse_json(): it passed a
pointer to a non-null Error * in certain cases, as demonstrated by
check-qjson.c.
json_parser_parse() is now unused. It's a stupid wrapper around
json_parser_parse_err(). Drop it, and rename json_parser_parse_err()
to json_parser_parse().
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20180823164025.12553-35-armbru@redhat.com>
