virtiofsd: get/set log level via DBus
cover virtiofsd: get/set log level via DBus
It is likely that virtiofsd will need to support "management commands" for reconfiguring it at runtime. The first use case was proposed by Eryu Guan for getting/setting the current log level.
I promised to try out DBus as the management interface because it has a rich feature set and is accessible from most programming languages. It should be able to support all the use cases we come up with.
This patch series is a prototype that implements the get-log-level and set-log-level management commands via DBus. Use the new virtiofsctl tool to talk to a running virtiofsd process:
# dbus-run-session ./virtiofsd ... ... Using dbus address unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # export DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # ./virtiofsctl set-log-level err
Most of the work is done by gdbus-codegen(1). It generates code for the org.qemu.Virtiofsd.xml interface definition. Our code can use the simple virtiofsd_get/set_log_level() APIs and it will make corresponding DBus calls.
I'm pretty happy with this approach because the code is straightforward. It hasn't even triggered seccomp failures yet :).
Error handling is a little problematic. I noticed that virtiofsctl silently returns success even if it cannot talk to virtiofsd. This is due to the code generated by gdbus-codegen(1) which has no error reporting :(. This can be solved by writing more low-level GDBus code instead of using the high-level generated bindings.
What do you think about this approach?
Stefan Hajnoczi (3):
virtiofsd: add org.qemu.Virtiofsd interface
virtiofsd: add DBus server to handle log level changes
virtiofsd: add virtiofsctl command-line management tool
configure | 7 + Makefile | 16 +++ Makefile.objs | 1 + contrib/virtiofsd/Makefile.objs | 10 +- contrib/virtiofsd/dbus.h | 9 ++ contrib/virtiofsd/dbus.c | 162 +++++++++++++++++++++++ contrib/virtiofsd/passthrough_ll.c | 8 +- contrib/virtiofsd/virtiofsctl.c | 55 ++++++++ .gitignore | 1 + contrib/virtiofsd/org.qemu.Virtiofsd.xml | 7 + 10 files changed, 274 insertions(+), 2 deletions(-) create mode 100644 contrib/virtiofsd/dbus.h create mode 100644 contrib/virtiofsd/dbus.c create mode 100644 contrib/virtiofsd/virtiofsctl.c create mode 100644 contrib/virtiofsd/org.qemu.Virtiofsd.xml
Signature · 1 line
2.21.0
RFC 1/3 virtiofsd: add org.qemu.Virtiofsd interface
Define a DBus interface for virtiofsd management. It only allows querying and changing the log level at the moment.
In the future more methods and properties could be added.
--- configure | 7 +++++++ Makefile | 13 +++++++++++++ contrib/virtiofsd/Makefile.objs | 6 +++++- .gitignore | 1 + contrib/virtiofsd/org.qemu.Virtiofsd.xml | 7 +++++++ 5 files changed, 33 insertions(+), 1 deletion(-) create mode 100644 contrib/virtiofsd/org.qemu.Virtiofsd.xml
configure +7 -0
@@ -3674,10 +3674,16 @@ if $pkg_config --atleast-version=$glib_req_ver gio-2.0; then gio=yes gio_cflags=$($pkg_config --cflags gio-2.0) gio_libs=$($pkg_config --libs gio-2.0) + gdbus_codegen=$($pkg_config --variable=gdbus_codegen gio-2.0) else gio=no fi +if $pkg_config --atleast-version=$glib_req_ver gio-unix-2.0; then + gio_cflags="$gio_cflags $($pkg_config --cflags gio-unix-2.0)" + gio_libs="$gio_libs $($pkg_config --libs gio-unix-2.0)" +fi + # Sanity check that the current size_t matches the # size that glib thinks it should be. This catches # problems on multi-arch where people try to build
@@ -6856,6 +6862,7 @@ if test "$gio" = "yes" ; then echo "CONFIG_GIO=y" >> $config_host_mak echo "GIO_CFLAGS=$gio_cflags" >> $config_host_mak echo "GIO_LIBS=$gio_libs" >> $config_host_mak + echo "GDBUS_CODEGEN=$gdbus_codegen" >> $config_host_mak fi echo "CONFIG_TLS_PRIORITY=\"$tls_priority\"" >> $config_host_mak if test "$gnutls" = "yes" ; then
Makefile +13 -0
@@ -124,6 +124,11 @@ GENERATED_QAPI_FILES += qapi/qapi-doc.texi generated-files-y += $(GENERATED_QAPI_FILES) +ifdef CONFIG_LINUX +generated-files-y += contrib/virtiofsd/gdbus_generated.h +generated-files-y += contrib/virtiofsd/gdbus_generated.c +endif + generated-files-y += trace/generated-tcg-tracers.h generated-files-y += trace/generated-helpers-wrappers.h
@@ -646,6 +651,14 @@ rdmacm-mux$(EXESUF): $(rdmacm-mux-obj-y) $(COMMON_LDADDS) $(call LINK, $^) ifdef CONFIG_LINUX # relies on Linux-specific syscalls +contrib/virtiofsd/gdbus_generated.h contrib/virtiofsd/gdbus_generated.c: contrib/virtiofsd/gdbus_generated.c-timestamp ; +contrib/virtiofsd/gdbus_generated.c-timestamp: $(SRC_PATH)/contrib/virtiofsd/org.qemu.Virtiofsd.xml + $(call quiet-command,$(GDBUS_CODEGEN) $< \ + --interface-prefix org.qemu \ + --generate-c-code contrib/virtiofsd/gdbus_generated, \ + "GEN","$(@:%-timestamp=%)") + @>$@ + virtiofsd$(EXESUF): $(virtiofsd-obj-y) libvhost-user.a $(COMMON_LDADDS) $(call LINK, $^) endif
contrib/virtiofsd/Makefile.objs +5 -1
@@ -7,9 +7,13 @@ virtiofsd-obj-y = buffer.o \ fuse_virtio.o \ helper.o \ passthrough_ll.o \ - seccomp.o + seccomp.o \ + gdbus_generated.o seccomp.o-cflags := $(SECCOMP_CFLAGS) seccomp.o-libs := $(SECCOMP_LIBS) +gdbus_generated.o-cflags := $(GIO_CFLAGS) +gdbus_generated.o-libs := $(GIO_LIBS) + passthrough_ll.o-libs += -lcap
.gitignore +1 -0
@@ -6,6 +6,7 @@ /config-target.* /config.status /config-temp +/contrib/virtiofsd/gdbus_generated.* /elf2dmp /trace-events-all /trace/generated-events.h
contrib/virtiofsd/org.qemu.Virtiofsd.xml new file +7 -0
@@ -0,0 +1,7 @@ +<!DOCTYPE node PUBLIC "-//freedesktop//DTD D-BUS Object Introspection 1.0//EN" + "http://www.freedesktop.org/standards/dbus/1.0/introspect.dtd"> +<node> + <interface name="org.qemu.Virtiofsd"> + <property name="LogLevel" type="s" access="readwrite"/> + </interface> +</node>
Signature · 1 line
2.21.0
RFC 2/3 virtiofsd: add DBus server to handle log level changes
In reply to Stefan HajnocziIntroduce a DBus server thread that runs alongside the other virtiofsd threads. It processes changes to the /org/qemu/virtiofsd object which can be accessed at the org.qemu.virtiofsd location on the bus.
This code does not use locking because we are the only writer to the int current_log_level variable. More advanced management commands would require locking to prevent race conditions with the other threads.
--- contrib/virtiofsd/Makefile.objs | 3 +- contrib/virtiofsd/dbus.h | 9 ++ contrib/virtiofsd/dbus.c | 162 +++++++++++++++++++++++++++++ contrib/virtiofsd/passthrough_ll.c | 8 +- 4 files changed, 180 insertions(+), 2 deletions(-) create mode 100644 contrib/virtiofsd/dbus.h create mode 100644 contrib/virtiofsd/dbus.c
contrib/virtiofsd/Makefile.objs +2 -1
@@ -8,7 +8,8 @@ virtiofsd-obj-y = buffer.o \ helper.o \ passthrough_ll.o \ seccomp.o \ - gdbus_generated.o + gdbus_generated.o \ + dbus.o seccomp.o-cflags := $(SECCOMP_CFLAGS) seccomp.o-libs := $(SECCOMP_LIBS)
contrib/virtiofsd/dbus.h new file +9 -0
@@ -0,0 +1,9 @@ +#ifndef DBUS_H +#define DBUS_H + +#include <stdbool.h> + +bool setup_dbus (void ); +void cleanup_dbus (void ); + +#endif /* DBUS_H */
contrib/virtiofsd/dbus.c new file +162 -0
162 lines not on this page · raw message
contrib/virtiofsd/passthrough_ll.c +7 -1
@@ -66,6 +66,7 @@#include <gmodule.h> #include "fuse_log.h" #include "seccomp.h" +#include "dbus.h" /* Keep track of inode posix locks for each owner. */ struct lo_inode_plock {
@@ -2989,6 +2990,9 @@ int main(int argc, char *argv[])if (fuse_session_mount (se) !=0 ) goto err_out3; +if (!setup_dbus ()) + goto err_out4; +fuse_daemonize (opts.foreground);if (lo.ireg_sock !=-1 ) {
@@ -2998,7 +3002,7 @@ int main(int argc, char *argv[])if (ret) {warnx ("pthread_create: %s" ,strerror (ret)); ret =1 ; - goto err_out4; + goto err_out5; }get_shared (&lo, &lo.root);
@@ -3014,6 +3018,8 @@ int main(int argc, char *argv[])/* Block until ctrl+c or fusermount -u */ ret =virtio_loop (se); +err_out5: +cleanup_dbus (); err_out4:fuse_session_unmount (se); err_out3:
Signature · 1 line
2.21.0
Re: [Qemu-devel] [RFC 2/3] virtiofsd: add DBus server to handle log level changes
Quoted 9 lines · * Stefan Hajnoczi ([email protected]) wrote:
Introduce a DBus server thread that runs alongside the other virtiofsd threads. It processes changes to the /org/qemu/virtiofsd object which can be accessed at the org.qemu.virtiofsd location on the bus.
This code does not use locking because we are the only writer to the int
current_log_levelvariable. More advanced management commands would require locking to prevent race conditions with the other threads.Signed-off-by Stefan Hajnoczi <[email protected]>
OK, that is less complex than I'd feared.
I guess there's something probably nice to do with name/integer mapping for warning levels that we could use from one of the libraries.
Dave
Quoted 249 lines
--- contrib/virtiofsd/Makefile.objs | 3 +- contrib/virtiofsd/dbus.h | 9 ++ contrib/virtiofsd/dbus.c | 162 +++++++++++++++++++++++++++++ contrib/virtiofsd/passthrough_ll.c | 8 +- 4 files changed, 180 insertions(+), 2 deletions(-) create mode 100644 contrib/virtiofsd/dbus.h create mode 100644 contrib/virtiofsd/dbus.ccontrib/virtiofsd/Makefile.objs +2
-1@@ -8,7 +8,8 @@ virtiofsd-obj-y = buffer.o \ helper.o \ passthrough_ll.o \ seccomp.o \ - gdbus_generated.o + gdbus_generated.o \ + dbus.o seccomp.o-cflags := $(SECCOMP_CFLAGS) seccomp.o-libs := $(SECCOMP_LIBS)contrib/virtiofsd/dbus.h new file +9
-0@@ -0,0 +1,9 @@ +#ifndef DBUS_H +#define DBUS_H + +#include <stdbool.h> + +bool setup_dbus (void ); +void cleanup_dbus (void ); + +#endif /* DBUS_H */ contrib/virtiofsd/dbus.c new file +162
-0@@ -0,0 +1,162 @@ +#include <assert.h> +#include <stdio.h> +#include <glib.h> +#include "fuse_log.h" +#include "dbus.h" +#include "gdbus_generated.h" + +static GThread *the_dbus_thread; +static GMainContext *the_dbus_context; +static GMainLoop *the_dbus_loop; + +/* Set the string property based on the current log level */ +static void refresh_log_level (Virtiofsd *virtiofsd) +{ +switch (current_log_level) { +case LOG_ERR : +virtiofsd_set_log_level (virtiofsd,"err" ); +break ; +case LOG_WARNING : +virtiofsd_set_log_level (virtiofsd,"warning" ); +break ; +case LOG_INFO : +virtiofsd_set_log_level (virtiofsd,"info" ); +break ; +case LOG_DEBUG : +virtiofsd_set_log_level (virtiofsd,"debug" ); +break ; + } +} + +/* Handle changes to Virtiofsd object properties */ +static void notify (GObject *object,GParamSpec *pspec) +{ +Virtiofsd *virtiofsd =VIRTIOFSD (object); + +fprintf (stderr,"%s %s\n" , __func__, pspec->name); + +if (strcmp (pspec->name,"log-level" ) ==0 ) { +const char *s =virtiofsd_get_log_level (virtiofsd); + +if (strcmp (s,"err" ) ==0 ) { + current_log_level =LOG_ERR ; + }else if (strcmp (s,"warning" ) ==0 ) { + current_log_level =LOG_WARNING ; + }else if (strcmp (s,"info" ) ==0 ) { + current_log_level =LOG_INFO ; + }else if (strcmp (s,"debug" ) ==0 ) { + current_log_level =LOG_DEBUG ; + }else { +/* Invalid, reset the log level property */ +refresh_log_level (virtiofsd); + } + } +} + +typedef struct { +Virtiofsd *virtiofsd; +pthread_barrier_t *started; +}GBusOwnNameData ; + +static void bus_acquired (GDBusConnection *connection,const gchar *name, +gpointer user_data) +{ +GBusOwnNameData *data = user_data; +GError *error =NULL ; + +if (!g_dbus_interface_skeleton_export ( +G_DBUS_INTERFACE_SKELETON (data->virtiofsd), + connection,"/org/qemu/virtiofsd" , &error)) { +fuse_err ("g_dbus_interface_skeleton_export: %s\n" , error->message); +g_error_free (error); +exit (EXIT_FAILURE ); + } +} + +static void name_acquired (GDBusConnection *connection,const gchar *name, +gpointer user_data) +{ +GBusOwnNameData *data = user_data; + +pthread_barrier_wait (data->started); +} + +static void name_lost (GDBusConnection *connection,const gchar *name, +gpointer user_data) +{ +if (connection) { +fuse_err ("unable to own dbus name\n" ); + }else { +fuse_err ("unable to connect to dbus\n" ); + } +exit (EXIT_FAILURE ); +} + +static gpointer dbus_thread (gpointer opaque) +{ +GBusOwnNameData data; +Virtiofsd *virtiofsd; +guint owner_id; + +g_main_context_push_thread_default (the_dbus_context); + + virtiofsd =virtiofsd_skeleton_new (); +refresh_log_level (virtiofsd); +g_signal_connect (virtiofsd,"notify" ,G_CALLBACK (notify),NULL ); + + data.virtiofsd = virtiofsd; + data.started = opaque; + + owner_id =g_bus_own_name (G_BUS_TYPE_SESSION ,"org.qemu.virtiofsd" , +G_BUS_NAME_OWNER_FLAGS_DO_NOT_QUEUE , bus_acquired, name_acquired, + name_lost, &data,NULL ); + +g_main_loop_run (the_dbus_loop); +g_bus_unown_name (owner_id); +g_object_unref (virtiofsd); + +g_main_context_pop_thread_default (the_dbus_context); +return NULL ; +} + +/** +* Start DBus server thread. +* +* Returns: true on success, false on failure +*/ +bool setup_dbus (void ) +{ +pthread_barrier_t started; + +assert (!the_dbus_thread); + +fuse_info ("Using dbus address %s\n" , +getenv ("DBUS_SESSION_BUS_ADDRESS" ) ?:"(null)" ); + +pthread_barrier_init (&started,NULL ,2 ); + + the_dbus_context =g_main_context_new (); + the_dbus_loop =g_main_loop_new (the_dbus_context, FALSE); + the_dbus_thread =g_thread_new ("dbus-thread" , dbus_thread, &started); + +pthread_barrier_wait (&started); +pthread_barrier_destroy (&started); + +return true; +} + +/** +* Stop DBus server thread. +*/ +void cleanup_dbus (void ) +{ +g_main_loop_quit (the_dbus_loop); +g_thread_join (the_dbus_thread); + the_dbus_thread =NULL ; + +g_main_loop_unref (the_dbus_loop); + the_dbus_loop =NULL ; + +g_main_context_unref (the_dbus_context); + the_dbus_context =NULL ; +}contrib/virtiofsd/passthrough_ll.c +7
-1@@ -66,6 +66,7 @@#include <gmodule.h> #include "fuse_log.h" #include "seccomp.h" +#include "dbus.h" /* Keep track of inode posix locks for each owner. */ struct lo_inode_plock {@@ -2989,6 +2990,9 @@ int main(int argc, char *argv[])if (fuse_session_mount (se) !=0 ) goto err_out3; +if (!setup_dbus ()) + goto err_out4; +fuse_daemonize (opts.foreground);if (lo.ireg_sock !=-1 ) {@@ -2998,7 +3002,7 @@ int main(int argc, char *argv[])if (ret) {warnx ("pthread_create: %s" ,strerror (ret)); ret =1 ; - goto err_out4; + goto err_out5; }get_shared (&lo, &lo.root);@@ -3014,6 +3018,8 @@ int main(int argc, char *argv[])/* Block until ctrl+c or fusermount -u */ ret =virtio_loop (se); +err_out5: +cleanup_dbus (); err_out4:fuse_session_unmount (se); err_out3:Signature · 1 line
2.21.0
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Re: [Qemu-devel] [RFC 2/3] virtiofsd: add DBus server to handle log level changes
Quoted 9 lines · Dr. David Alan Gilbert
Quoted 8 lines · * Stefan Hajnoczi ([email protected]) wrote:
Introduce a DBus server thread that runs alongside the other virtiofsd threads. It processes changes to the /org/qemu/virtiofsd object which can be accessed at the org.qemu.virtiofsd location on the bus.
This code does not use locking because we are the only writer to the int
current_log_levelvariable. More advanced management commands would require locking to prevent race conditions with the other threads.
Quoted 5 lines
Quoted 1 line
Signed-off-by Stefan Hajnoczi <[email protected]>OK, that is less complex than I'd feared.
I guess there's something probably nice to do with name/integer mapping for warning levels that we could use from one of the libraries.
I used a free-form string because it's what systemd's LogLevel property also does. But I can investigate the cleanest approach for limiting it to a set of string constants.
Stefan
Re: [Qemu-devel] [RFC 2/3] virtiofsd: add DBus server to handle log level changes
Quoted 12 lines · Stefan Hajnoczi
Quoted 11 lines · Dr. David Alan Gilbert
Quoted 9 lines · * Stefan Hajnoczi ([email protected]) wrote:
Introduce a DBus server thread that runs alongside the other virtiofsd threads. It processes changes to the /org/qemu/virtiofsd object which can be accessed at the org.qemu.virtiofsd location on the bus.
This code does not use locking because we are the only writer to the int
current_log_levelvariable. More advanced management commands would require locking to prevent race conditions with the other threads.Signed-off-by Stefan Hajnoczi <[email protected]>
Quoted 7 lines
Quoted 3 lines
OK, that is less complex than I'd feared.
I guess there's something probably nice to do with name/integer mapping for warning levels that we could use from one of the libraries.I used a free-form string because it's what systemd's LogLevel property also does. But I can investigate the cleanest approach for limiting it to a set of string constants.
There's no concept of "enums" at the DBus protocol level. Sending enums in string form is the normal practice - avoiding integer values means you are not vulnerable to enum values changing if someone inserts a new constant in the middlle of the enum. This same reason is why QAPI uses strings for enums instead of ints.
Regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [Qemu-devel] [RFC 2/3] virtiofsd: add DBus server to handle log level changes
Quoted 21 lines · * Daniel P. Berrangé ([email protected]) wrote:
Quoted 12 lines · Stefan Hajnoczi
Quoted 11 lines · Dr. David Alan Gilbert
Quoted 9 lines · * Stefan Hajnoczi ([email protected]) wrote:
Introduce a DBus server thread that runs alongside the other virtiofsd threads. It processes changes to the /org/qemu/virtiofsd object which can be accessed at the org.qemu.virtiofsd location on the bus.
This code does not use locking because we are the only writer to the int
current_log_levelvariable. More advanced management commands would require locking to prevent race conditions with the other threads.Signed-off-by Stefan Hajnoczi <[email protected]>Quoted 7 lines
Quoted 3 lines
OK, that is less complex than I'd feared.
I guess there's something probably nice to do with name/integer mapping for warning levels that we could use from one of the libraries.I used a free-form string because it's what systemd's LogLevel property also does. But I can investigate the cleanest approach for limiting it to a set of string constants.
Quoted 5 lines
There's no concept of "enums" at the DBus protocol level. Sending enums in string form is the normal practice - avoiding integer values means you are not vulnerable to enum values changing if someone inserts a new constant in the middlle of the enum. This same reason is why QAPI uses strings for enums instead of ints.
Oh, I wasn't talking aobut changing protocol; I just meant there was probably a neater way of doing the string look up than the opencoded way it was done.
Dave
Quoted 6 lines
Regards,
DanielSignature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Re: [Qemu-devel] [RFC 2/3] virtiofsd: add DBus server to handle log level changes
Quoted 28 lines · Dr. David Alan Gilbert
Quoted 21 lines · * Daniel P. Berrangé ([email protected]) wrote:
Quoted 12 lines · Stefan Hajnoczi
Quoted 11 lines · Dr. David Alan Gilbert
Quoted 9 lines · * Stefan Hajnoczi ([email protected]) wrote:
Introduce a DBus server thread that runs alongside the other virtiofsd threads. It processes changes to the /org/qemu/virtiofsd object which can be accessed at the org.qemu.virtiofsd location on the bus.
This code does not use locking because we are the only writer to the int
current_log_levelvariable. More advanced management commands would require locking to prevent race conditions with the other threads.Signed-off-by Stefan Hajnoczi <[email protected]>Quoted 7 lines
Quoted 3 lines
OK, that is less complex than I'd feared.
I guess there's something probably nice to do with name/integer mapping for warning levels that we could use from one of the libraries.I used a free-form string because it's what systemd's LogLevel property also does. But I can investigate the cleanest approach for limiting it to a set of string constants.
Quoted 5 lines
There's no concept of "enums" at the DBus protocol level. Sending enums in string form is the normal practice - avoiding integer values means you are not vulnerable to enum values changing if someone inserts a new constant in the middlle of the enum. This same reason is why QAPI uses strings for enums instead of ints.
Quoted 3 lines
Oh, I wasn't talking aobut changing protocol; I just meant there was probably a neater way of doing the string look up than the opencoded way it was done.
Oh sure, you can declare the enum in a header, and then run glib-mkenums with that header file and it will spit out the code neccessary to register a GEnum class. This gives you ability to do int/string conversions with a simple api
https://developer.gnome.org/gobject/stable/gobject-Enumeration-and-Flag-Types.html
https://developer.gnome.org/gobject/stable/glib-mkenums.html
It makes sense to use this, since virtiofsd is already using GObject via GDbus.
Regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
RFC 3/3 virtiofsd: add virtiofsctl command-line management tool
In reply to Stefan Hajnoczivirtiofsctl can control a running virtiofsd process:
usage: ./virtiofsctl COMMAND [args...]
Commands:
get-log-level - show current log level
set-log-level LEVEL - set current log level to one of
"err", "warning", "info", "debug"
Make sure it is running in the same DBus session as virtiofsd. This may require setting the DBUS_SESSION_BUS_ADDRESS environment variable to the same value as used by virtiofsd.
--- Makefile | 3 ++ Makefile.objs | 1 + contrib/virtiofsd/Makefile.objs | 3 ++ contrib/virtiofsd/virtiofsctl.c | 55 +++++++++++++++++++++++++++++++++ 4 files changed, 62 insertions(+) create mode 100644 contrib/virtiofsd/virtiofsctl.c
Makefile +3 -0
@@ -419,6 +419,7 @@ dummy := $(call unnest-vars,, \ ivshmem-client-obj-y \ ivshmem-server-obj-y \ virtiofsd-obj-y \ + virtiofsctl-obj-y \ rdmacm-mux-obj-y \ libvhost-user-obj-y \ vhost-user-scsi-obj-y \
@@ -661,6 +662,8 @@ contrib/virtiofsd/gdbus_generated.c-timestamp: $(SRC_PATH)/contrib/virtiofsd/org virtiofsd$(EXESUF): $(virtiofsd-obj-y) libvhost-user.a $(COMMON_LDADDS) $(call LINK, $^) +virtiofsctl$(EXESUF): $(virtiofsctl-obj-y) + $(call LINK, $^) endif vhost-user-gpu$(EXESUF): $(vhost-user-gpu-obj-y) $(libvhost-user-obj-y) libqemuutil.a libqemustub.a
Makefile.objs +1 -0
@@ -126,6 +126,7 @@ rdmacm-mux-obj-y = contrib/rdmacm-mux/ vhost-user-input-obj-y = contrib/vhost-user-input/ vhost-user-gpu-obj-y = contrib/vhost-user-gpu/ virtiofsd-obj-y = contrib/virtiofsd/ +virtiofsctl-obj-y = contrib/virtiofsd/ ###################################################################### trace-events-subdirs =
contrib/virtiofsd/Makefile.objs +3 -0
@@ -11,6 +11,9 @@ virtiofsd-obj-y = buffer.o \ gdbus_generated.o \ dbus.o +virtiofsctl-obj-y = virtiofsctl.o \ + gdbus_generated.o + seccomp.o-cflags := $(SECCOMP_CFLAGS) seccomp.o-libs := $(SECCOMP_LIBS)
contrib/virtiofsd/virtiofsctl.c new file +55 -0
55 lines not on this page · raw message
Signature · 1 line
2.21.0
Re: [Qemu-devel] [RFC 3/3] virtiofsd: add virtiofsctl command-line management tool
Quoted 110 lines · * Stefan Hajnoczi ([email protected]) wrote:
virtiofsctl can control a running virtiofsd process:
usage: ./virtiofsctl COMMAND [args...]
Commands:
get-log-level - show current log level
set-log-level LEVEL - set current log level to one of
"err", "warning", "info", "debug"Make sure it is running in the same DBus session as virtiofsd. This may require setting the
DBUS_SESSION_BUS_ADDRESSenvironment variable to the same value as used by virtiofsd.Signed-off-by Stefan Hajnoczi <[email protected]>--- Makefile | 3 ++ Makefile.objs | 1 + contrib/virtiofsd/Makefile.objs | 3 ++ contrib/virtiofsd/virtiofsctl.c | 55 +++++++++++++++++++++++++++++++++ 4 files changed, 62 insertions(+) create mode 100644 contrib/virtiofsd/virtiofsctl.cMakefile +3
-0@@ -419,6 +419,7 @@ dummy := $(call unnest-vars,, \ ivshmem-client-obj-y \ ivshmem-server-obj-y \ virtiofsd-obj-y \ + virtiofsctl-obj-y \ rdmacm-mux-obj-y \ libvhost-user-obj-y \ vhost-user-scsi-obj-y \@@ -661,6 +662,8 @@ contrib/virtiofsd/gdbus_generated.c-timestamp: $(SRC_PATH)/contrib/virtiofsd/org virtiofsd$(EXESUF): $(virtiofsd-obj-y) libvhost-user.a $(COMMON_LDADDS) $(call LINK, $^) +virtiofsctl$(EXESUF): $(virtiofsctl-obj-y) + $(call LINK, $^) endif vhost-user-gpu$(EXESUF): $(vhost-user-gpu-obj-y) $(libvhost-user-obj-y) libqemuutil.a libqemustub.aMakefile.objs +1
-0@@ -126,6 +126,7 @@ rdmacm-mux-obj-y = contrib/rdmacm-mux/ vhost-user-input-obj-y = contrib/vhost-user-input/ vhost-user-gpu-obj-y = contrib/vhost-user-gpu/ virtiofsd-obj-y = contrib/virtiofsd/ +virtiofsctl-obj-y = contrib/virtiofsd/ ###################################################################### trace-events-subdirs =contrib/virtiofsd/Makefile.objs +3
-0@@ -11,6 +11,9 @@ virtiofsd-obj-y = buffer.o \ gdbus_generated.o \ dbus.o +virtiofsctl-obj-y = virtiofsctl.o \ + gdbus_generated.o + seccomp.o-cflags := $(SECCOMP_CFLAGS) seccomp.o-libs := $(SECCOMP_LIBS)contrib/virtiofsd/virtiofsctl.c new file +35
-0@@ -0,0 +1,55 @@ +#include <stdio.h> +#include "gdbus_generated.h" + +static void get_log_level (Virtiofsd *virtiofsd) +{ +const char *value =virtiofsd_get_log_level (virtiofsd); + +printf ("%s\n" , value ? value :"(null)" ); +} + +static void set_log_level (Virtiofsd *virtiofsd,const char *value) +{ +virtiofsd_set_log_level (virtiofsd, value); +} + +static void usage (const char *progname) +{ +printf ("usage: %s COMMAND [args...]\n" , progname); +printf ("\n" ); +printf ("Commands:\n" ); +printf (" get-log-level - show current log level\n" ); +printf (" set-log-level LEVEL - set current log level to one of\n" ); +printf (" \"err\", \"warning\", \"info\", \"debug\"\n" ); +exit (0 ); +} + +int main (int argc,char **argv) +{ +Virtiofsd *virtiofsd; +GError *error =NULL ; + +if (argc <2 ) { +usage (argv[0 ]); + } +
contrib/virtiofsd/virtiofsctl.c line 45
Quoted 10 lines
contrib/virtiofsd/virtiofsctl.c new file +10
-0+ virtiofsd =virtiofsd_proxy_new_for_bus_sync (G_BUS_TYPE_SESSION , +G_DBUS_PROXY_FLAGS_NONE ,"org.qemu.virtiofsd" , +"/org/qemu/virtiofsd" ,NULL , &error); +if (error) { +fprintf (stderr,"%s\n" , error->message); +g_error_free (error); +return 1 ; + } + +if (strcmp (argv[0 ],"get-log-level" ) ==0 ) {
This and the one below works a lot better with argv[1] !
(I wonder if a little python script would be better for these type of wrappers).
Dave
Quoted 14 lines
contrib/virtiofsd/virtiofsctl.c new file +10
-0+get_log_level (virtiofsd); + }else if (strcmp (argv[0 ],"set-log-level" ) ==0 ) { +if (argc !=3 ) { +usage (argv[0 ]); + } +set_log_level (virtiofsd, argv[2 ]); + } +g_object_unref (virtiofsd); +return 0 ; +}Signature · 1 line
2.21.0
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Re: [Qemu-devel] [RFC 3/3] virtiofsd: add virtiofsctl command-line management tool
Hi
Quoted 125 lines · Dr. David Alan Gilbert
Quoted 110 lines · * Stefan Hajnoczi ([email protected]) wrote:
virtiofsctl can control a running virtiofsd process:
usage: ./virtiofsctl COMMAND [args...]
Commands:
get-log-level - show current log level
set-log-level LEVEL - set current log level to one of
"err", "warning", "info", "debug"Make sure it is running in the same DBus session as virtiofsd. This may require setting the
DBUS_SESSION_BUS_ADDRESSenvironment variable to the same value as used by virtiofsd.Signed-off-by Stefan Hajnoczi <[email protected]>--- Makefile | 3 ++ Makefile.objs | 1 + contrib/virtiofsd/Makefile.objs | 3 ++ contrib/virtiofsd/virtiofsctl.c | 55 +++++++++++++++++++++++++++++++++ 4 files changed, 62 insertions(+) create mode 100644 contrib/virtiofsd/virtiofsctl.cMakefile +3
-0@@ -419,6 +419,7 @@ dummy := $(call unnest-vars,, \ ivshmem-client-obj-y \ ivshmem-server-obj-y \ virtiofsd-obj-y \ + virtiofsctl-obj-y \ rdmacm-mux-obj-y \ libvhost-user-obj-y \ vhost-user-scsi-obj-y \@@ -661,6 +662,8 @@ contrib/virtiofsd/gdbus_generated.c-timestamp: $(SRC_PATH)/contrib/virtiofsd/org virtiofsd$(EXESUF): $(virtiofsd-obj-y) libvhost-user.a $(COMMON_LDADDS) $(call LINK, $^) +virtiofsctl$(EXESUF): $(virtiofsctl-obj-y) + $(call LINK, $^) endif vhost-user-gpu$(EXESUF): $(vhost-user-gpu-obj-y) $(libvhost-user-obj-y) libqemuutil.a libqemustub.aMakefile.objs +1
-0@@ -126,6 +126,7 @@ rdmacm-mux-obj-y = contrib/rdmacm-mux/ vhost-user-input-obj-y = contrib/vhost-user-input/ vhost-user-gpu-obj-y = contrib/vhost-user-gpu/ virtiofsd-obj-y = contrib/virtiofsd/ +virtiofsctl-obj-y = contrib/virtiofsd/ ###################################################################### trace-events-subdirs =contrib/virtiofsd/Makefile.objs +3
-0@@ -11,6 +11,9 @@ virtiofsd-obj-y = buffer.o \ gdbus_generated.o \ dbus.o +virtiofsctl-obj-y = virtiofsctl.o \ + gdbus_generated.o + seccomp.o-cflags := $(SECCOMP_CFLAGS) seccomp.o-libs := $(SECCOMP_LIBS)contrib/virtiofsd/virtiofsctl.c new file +35
-0@@ -0,0 +1,55 @@ +#include <stdio.h> +#include "gdbus_generated.h" + +static void get_log_level (Virtiofsd *virtiofsd) +{ +const char *value =virtiofsd_get_log_level (virtiofsd); + +printf ("%s\n" , value ? value :"(null)" ); +} + +static void set_log_level (Virtiofsd *virtiofsd,const char *value) +{ +virtiofsd_set_log_level (virtiofsd, value); +} + +static void usage (const char *progname) +{ +printf ("usage: %s COMMAND [args...]\n" , progname); +printf ("\n" ); +printf ("Commands:\n" ); +printf (" get-log-level - show current log level\n" ); +printf (" set-log-level LEVEL - set current log level to one of\n" ); +printf (" \"err\", \"warning\", \"info\", \"debug\"\n" ); +exit (0 ); +} + +int main (int argc,char **argv) +{ +Virtiofsd *virtiofsd; +GError *error =NULL ; + +if (argc <2 ) { +usage (argv[0 ]); + } +contrib/virtiofsd/virtiofsctl.c line 45
Quoted 10 lines
contrib/virtiofsd/virtiofsctl.c new file +10
-0+ virtiofsd =virtiofsd_proxy_new_for_bus_sync (G_BUS_TYPE_SESSION , +G_DBUS_PROXY_FLAGS_NONE ,"org.qemu.virtiofsd" , +"/org/qemu/virtiofsd" ,NULL , &error); +if (error) { +fprintf (stderr,"%s\n" , error->message); +g_error_free (error); +return 1 ; + } + +if (strcmp (argv[0 ],"get-log-level" ) ==0 ) {This and the one below works a lot better with argv[1] !
Quoted 2 lines
(I wonder if a little python script would be better for these type of wrappers).
Or just plain gdbus/busctl calls (which already has bash completion fwiw), and/or eventually a shell script.
Quoted 19 lines
Dave
Quoted 12 lines
contrib/virtiofsd/virtiofsctl.c new file +10
-1+get_log_level (virtiofsd); + }else if (strcmp (argv[0 ],"set-log-level" ) ==0 ) { +if (argc !=3 ) { +usage (argv[0 ]); + } +set_log_level (virtiofsd, argv[2 ]); + } +g_object_unref (virtiofsd); +return 0 ; +} --Quoted 2 lines
2.21.0
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Re: [Qemu-devel] [RFC 3/3] virtiofsd: add virtiofsctl command-line management tool
Quoted 129 lines · Marc-André Lureau
Hi
Quoted 125 lines · Dr. David Alan Gilbert
Quoted 110 lines · * Stefan Hajnoczi ([email protected]) wrote:
virtiofsctl can control a running virtiofsd process:
usage: ./virtiofsctl COMMAND [args...]
Commands:
get-log-level - show current log level
set-log-level LEVEL - set current log level to one of
"err", "warning", "info", "debug"Make sure it is running in the same DBus session as virtiofsd. This may require setting the
DBUS_SESSION_BUS_ADDRESSenvironment variable to the same value as used by virtiofsd.Signed-off-by Stefan Hajnoczi <[email protected]>--- Makefile | 3 ++ Makefile.objs | 1 + contrib/virtiofsd/Makefile.objs | 3 ++ contrib/virtiofsd/virtiofsctl.c | 55 +++++++++++++++++++++++++++++++++ 4 files changed, 62 insertions(+) create mode 100644 contrib/virtiofsd/virtiofsctl.cMakefile +3
-0@@ -419,6 +419,7 @@ dummy := $(call unnest-vars,, \ ivshmem-client-obj-y \ ivshmem-server-obj-y \ virtiofsd-obj-y \ + virtiofsctl-obj-y \ rdmacm-mux-obj-y \ libvhost-user-obj-y \ vhost-user-scsi-obj-y \@@ -661,6 +662,8 @@ contrib/virtiofsd/gdbus_generated.c-timestamp: $(SRC_PATH)/contrib/virtiofsd/org virtiofsd$(EXESUF): $(virtiofsd-obj-y) libvhost-user.a $(COMMON_LDADDS) $(call LINK, $^) +virtiofsctl$(EXESUF): $(virtiofsctl-obj-y) + $(call LINK, $^) endif vhost-user-gpu$(EXESUF): $(vhost-user-gpu-obj-y) $(libvhost-user-obj-y) libqemuutil.a libqemustub.aMakefile.objs +1
-0@@ -126,6 +126,7 @@ rdmacm-mux-obj-y = contrib/rdmacm-mux/ vhost-user-input-obj-y = contrib/vhost-user-input/ vhost-user-gpu-obj-y = contrib/vhost-user-gpu/ virtiofsd-obj-y = contrib/virtiofsd/ +virtiofsctl-obj-y = contrib/virtiofsd/ ###################################################################### trace-events-subdirs =contrib/virtiofsd/Makefile.objs +3
-0@@ -11,6 +11,9 @@ virtiofsd-obj-y = buffer.o \ gdbus_generated.o \ dbus.o +virtiofsctl-obj-y = virtiofsctl.o \ + gdbus_generated.o + seccomp.o-cflags := $(SECCOMP_CFLAGS) seccomp.o-libs := $(SECCOMP_LIBS)contrib/virtiofsd/virtiofsctl.c new file +35
-0@@ -0,0 +1,55 @@ +#include <stdio.h> +#include "gdbus_generated.h" + +static void get_log_level (Virtiofsd *virtiofsd) +{ +const char *value =virtiofsd_get_log_level (virtiofsd); + +printf ("%s\n" , value ? value :"(null)" ); +} + +static void set_log_level (Virtiofsd *virtiofsd,const char *value) +{ +virtiofsd_set_log_level (virtiofsd, value); +} + +static void usage (const char *progname) +{ +printf ("usage: %s COMMAND [args...]\n" , progname); +printf ("\n" ); +printf ("Commands:\n" ); +printf (" get-log-level - show current log level\n" ); +printf (" set-log-level LEVEL - set current log level to one of\n" ); +printf (" \"err\", \"warning\", \"info\", \"debug\"\n" ); +exit (0 ); +} + +int main (int argc,char **argv) +{ +Virtiofsd *virtiofsd; +GError *error =NULL ; + +if (argc <2 ) { +usage (argv[0 ]); + } +contrib/virtiofsd/virtiofsctl.c line 45
Quoted 10 lines
contrib/virtiofsd/virtiofsctl.c new file +10
-0+ virtiofsd =virtiofsd_proxy_new_for_bus_sync (G_BUS_TYPE_SESSION , +G_DBUS_PROXY_FLAGS_NONE ,"org.qemu.virtiofsd" , +"/org/qemu/virtiofsd" ,NULL , &error); +if (error) { +fprintf (stderr,"%s\n" , error->message); +g_error_free (error); +return 1 ; + } + +if (strcmp (argv[0 ],"get-log-level" ) ==0 ) {This and the one below works a lot better with argv[1] !
Quoted 5 lines
Quoted 2 lines
(I wonder if a little python script would be better for these type of wrappers).
Or just plain gdbus/busctl calls (which already has bash completion fwiw), and/or eventually a shell script.
A virtiofsctl program is useful because the sub-commands are easily discoverable and no DBus knowledge is required. Writing it in Python 3 would be nice.
Stefan
Quoted 18 lines · * Stefan Hajnoczi ([email protected]) wrote:
It is likely that virtiofsd will need to support "management commands" for reconfiguring it at runtime. The first use case was proposed by Eryu Guan for getting/setting the current log level.
I promised to try out DBus as the management interface because it has a rich feature set and is accessible from most programming languages. It should be able to support all the use cases we come up with.
This patch series is a prototype that implements the get-log-level and set-log-level management commands via DBus. Use the new virtiofsctl tool to talk to a running virtiofsd process:
# dbus-run-session ./virtiofsd ... ... Using dbus address unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # export DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # ./virtiofsctl set-log-level err
Quoted 6 lines
Most of the work is done by gdbus-codegen(1). It generates code for the org.qemu.Virtiofsd.xml interface definition. Our code can use the simple virtiofsd_get/set_log_level() APIs and it will make corresponding DBus calls.
I'm pretty happy with this approach because the code is straightforward. It hasn't even triggered seccomp failures yet :).
Yes it's less complex than I'd worried.
Now, I do think we've got to think about how qemu in general is going to use dbus as people were discussing it, because then we have to think what the security aspects are - do we need to look at some calls only available to some clients etc.
Dave
Quoted 13 lines
Error handling is a little problematic. I noticed that virtiofsctl silently returns success even if it cannot talk to virtiofsd. This is due to the code generated by gdbus-codegen(1) which has no error reporting :(. This can be solved by writing more low-level GDBus code instead of using the high-level generated bindings.
What do you think about this approach?
Stefan Hajnoczi (3):
virtiofsd: add org.qemu.Virtiofsd interface
virtiofsd: add DBus server to handle log level changes
virtiofsd: add virtiofsctl command-line management tool
Quoted 19 lines
configure | 7 + Makefile | 16 +++ Makefile.objs | 1 + contrib/virtiofsd/Makefile.objs | 10 +- contrib/virtiofsd/dbus.h | 9 ++ contrib/virtiofsd/dbus.c | 162 +++++++++++++++++++++++ contrib/virtiofsd/passthrough_ll.c | 8 +- contrib/virtiofsd/virtiofsctl.c | 55 ++++++++ .gitignore | 1 + contrib/virtiofsd/org.qemu.Virtiofsd.xml | 7 + 10 files changed, 274 insertions(+), 2 deletions(-) create mode 100644 contrib/virtiofsd/dbus.h create mode 100644 contrib/virtiofsd/dbus.c create mode 100644 contrib/virtiofsd/virtiofsctl.c create mode 100644 contrib/virtiofsd/org.qemu.Virtiofsd.xmlSignature · 1 line
2.21.0
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Quoted 23 lines · Dr. David Alan Gilbert
Quoted 22 lines · * Stefan Hajnoczi ([email protected]) wrote:
It is likely that virtiofsd will need to support "management commands" for reconfiguring it at runtime. The first use case was proposed by Eryu Guan for getting/setting the current log level.
I promised to try out DBus as the management interface because it has a rich feature set and is accessible from most programming languages. It should be able to support all the use cases we come up with.
This patch series is a prototype that implements the get-log-level and set-log-level management commands via DBus. Use the new virtiofsctl tool to talk to a running virtiofsd process:
# dbus-run-session ./virtiofsd ... ... Using dbus address unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # export DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # ./virtiofsctl set-log-level errMost of the work is done by gdbus-codegen(1). It generates code for the org.qemu.Virtiofsd.xml interface definition. Our code can use the simple virtiofsd_get/set_log_level() APIs and it will make corresponding DBus calls.
Quoted 8 lines
Quoted 2 lines
I'm pretty happy with this approach because the code is straightforward. It hasn't even triggered seccomp failures yet :).
Yes it's less complex than I'd worried.
Now, I do think we've got to think about how qemu in general is going to use dbus as people were discussing it, because then we have to think what the security aspects are - do we need to look at some calls only available to some clients etc.
The approach I took in this patch series is to launch a session bus just for this virtiofsd. The abstract socket unix(7) namespace used by GDBus by default does not offer any security. I think any process on the host can connect to it, regardless of uid/gid.
A path like unix:path=/tmp/foo would allow us to use UNIX Domain Socket permissions as the main security mechanism. I'm not enthusiastic about using SELinux or some kind of DBus-specific policy language if we can avoid it because it's complex and obscure.
Stefan
Quoted 33 lines · * Stefan Hajnoczi ([email protected]) wrote:
Quoted 23 lines · Dr. David Alan Gilbert
Quoted 22 lines · * Stefan Hajnoczi ([email protected]) wrote:
It is likely that virtiofsd will need to support "management commands" for reconfiguring it at runtime. The first use case was proposed by Eryu Guan for getting/setting the current log level.
I promised to try out DBus as the management interface because it has a rich feature set and is accessible from most programming languages. It should be able to support all the use cases we come up with.
This patch series is a prototype that implements the get-log-level and set-log-level management commands via DBus. Use the new virtiofsctl tool to talk to a running virtiofsd process:
# dbus-run-session ./virtiofsd ... ... Using dbus address unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # export DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # ./virtiofsctl set-log-level errMost of the work is done by gdbus-codegen(1). It generates code for the org.qemu.Virtiofsd.xml interface definition. Our code can use the simple virtiofsd_get/set_log_level() APIs and it will make corresponding DBus calls.
Quoted 8 lines
Quoted 2 lines
I'm pretty happy with this approach because the code is straightforward. It hasn't even triggered seccomp failures yet :).
Yes it's less complex than I'd worried.
Now, I do think we've got to think about how qemu in general is going to use dbus as people were discussing it, because then we have to think what the security aspects are - do we need to look at some calls only available to some clients etc.
Quoted 7 lines
The approach I took in this patch series is to launch a session bus just for this virtiofsd. The abstract socket unix(7) namespace used by GDBus by default does not offer any security. I think any process on the host can connect to it, regardless of uid/gid.
A path like unix:path=/tmp/foo would allow us to use UNIX Domain Socket permissions as the main security mechanism.
Yes, that I'm fine with; my worry is that there was talk of much more complex dbus setups coming out of the qemu world with many things being connected; and then we have to think about security upfront.
Quoted 3 lines
I'm not enthusiastic about
using SELinux or some kind of DBus-specific policy language if we can avoid it because it's complex and obscure.
Right.
Dave
Quoted 1 line
Stefan
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
Quoted 33 lines · Stefan Hajnoczi
Quoted 23 lines · Dr. David Alan Gilbert
Quoted 22 lines · * Stefan Hajnoczi ([email protected]) wrote:
It is likely that virtiofsd will need to support "management commands" for reconfiguring it at runtime. The first use case was proposed by Eryu Guan for getting/setting the current log level.
I promised to try out DBus as the management interface because it has a rich feature set and is accessible from most programming languages. It should be able to support all the use cases we come up with.
This patch series is a prototype that implements the get-log-level and set-log-level management commands via DBus. Use the new virtiofsctl tool to talk to a running virtiofsd process:
# dbus-run-session ./virtiofsd ... ... Using dbus address unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # export DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # ./virtiofsctl set-log-level errMost of the work is done by gdbus-codegen(1). It generates code for the org.qemu.Virtiofsd.xml interface definition. Our code can use the simple virtiofsd_get/set_log_level() APIs and it will make corresponding DBus calls.
Quoted 8 lines
Quoted 2 lines
I'm pretty happy with this approach because the code is straightforward. It hasn't even triggered seccomp failures yet :).
Yes it's less complex than I'd worried.
Now, I do think we've got to think about how qemu in general is going to use dbus as people were discussing it, because then we have to think what the security aspects are - do we need to look at some calls only available to some clients etc.
Quoted 4 lines
The approach I took in this patch series is to launch a session bus just for this virtiofsd. The abstract socket unix(7) namespace used by GDBus by default does not offer any security. I think any process on the host can connect to it, regardless of uid/gid.
Other users will be able to connect(), but you'll find that the dbus policy causes their connections to be dropped immediately - even if they are the root user in fact.
Quoted 4 lines
A path like unix:path=/tmp/foo would allow us to use UNIX Domain Socket permissions as the main security mechanism. I'm not enthusiastic about using SELinux or some kind of DBus-specific policy language if we can avoid it because it's complex and obscure.
Yep, that just needs you to supply a config file when launching to specify a desired filesystem path.
I don't think it is an either/or matter - I think we'll want all three in general - DAC controls on the socket, and DBus policy and SELinux policy. DAC controls on the socket alone are not sufficient if you want to separate each QEMU from each other and they're running the same UID which is common.
Regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Quoted 22 lines · Stefan Hajnoczi
It is likely that virtiofsd will need to support "management commands" for reconfiguring it at runtime. The first use case was proposed by Eryu Guan for getting/setting the current log level.
I promised to try out DBus as the management interface because it has a rich feature set and is accessible from most programming languages. It should be able to support all the use cases we come up with.
This patch series is a prototype that implements the get-log-level and set-log-level management commands via DBus. Use the new virtiofsctl tool to talk to a running virtiofsd process:
# dbus-run-session ./virtiofsd ... ... Using dbus address unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # export DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # ./virtiofsctl set-log-level errMost of the work is done by gdbus-codegen(1). It generates code for the org.qemu.Virtiofsd.xml interface definition. Our code can use the simple virtiofsd_get/set_log_level() APIs and it will make corresponding DBus calls.
Quoted 8 lines
I'm pretty happy with this approach because the code is straightforward. It hasn't even triggered seccomp failures yet :).
Error handling is a little problematic. I noticed that virtiofsctl silently returns success even if it cannot talk to virtiofsd. This is due to the code generated by gdbus-codegen(1) which has no error reporting :(. This can be solved by writing more low-level GDBus code instead of using the high-level generated bindings.
You declared a simple property for the log level, which gets mapped into GObject properties, and hence hit the error reporting limitations they have. DBus at the protocol level can report errors for properties.
For log level settings I'm not worried, but more generally we may well hit cases where error reporting is functionally critical. So we have a choice I think
- Enhance gdbus-codegen so that it can map DBus properties to a different impl, with explicit getters/setters, ignoring GObject's properties. This would allow for a GError ** parameter to handle/report errors
- Don't use properties at the DBus protocol level. Instead simply define explicit setter & getter methods in DBus. These woudl again bypass GObject's properties
Option 2 is probably easier to be honest, especially since in the code you end up calling what are setters & getters on the client side anyway.
Quoted 1 line
What do you think about this approach?
Regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Quoted 25 lines · Stefan Hajnoczi
It is likely that virtiofsd will need to support "management commands" for reconfiguring it at runtime. The first use case was proposed by Eryu Guan for getting/setting the current log level.
I promised to try out DBus as the management interface because it has a rich feature set and is accessible from most programming languages. It should be able to support all the use cases we come up with.
This patch series is a prototype that implements the get-log-level and set-log-level management commands via DBus. Use the new virtiofsctl tool to talk to a running virtiofsd process:
# dbus-run-session ./virtiofsd ... ... Using dbus address unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # export DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-H9WBbpjk3O,guid=0be16acefb868e6025a8737f5d7124d2 # ./virtiofsctl set-log-level errMost of the work is done by gdbus-codegen(1). It generates code for the org.qemu.Virtiofsd.xml interface definition. Our code can use the simple virtiofsd_get/set_log_level() APIs and it will make corresponding DBus calls.
I'm pretty happy with this approach because the code is straightforward. It hasn't even triggered seccomp failures yet :).
Quoted 7 lines
Error handling is a little problematic. I noticed that virtiofsctl silently returns success even if it cannot talk to virtiofsd. This is due to the code generated by gdbus-codegen(1) which has no error reporting :(. This can be solved by writing more low-level GDBus code instead of using the high-level generated bindings.
What do you think about this approach?
Hi Eryu Guan,
Have you had a chance to try these patches? Do they meet your requirements for being able to get/set the log level at runtime?
Stefan