package/dbusbroker: new package
PATCH v2 1/1 package/dbusbroker: new package
Add dbus-broker, which is a drop-in replacement
for the dbus-daemon.
Its possible to use this package standalone (without the dbus package - if buildroot's systemd would not depend on dbus). This is sufficient to provide systemd's (d)bus functionality. To allow standalone usage, the necessary config files are copied and adopted over from dbus.
bases on Yanns changes, and
- add an own config entry for dbus-broker-launch
enabled by default if systemd init is used
- undo BR2_COREUTILS_HOST_DEPENDENCY - undo adding dbus user - never used by this package - add condtional audit dependency - cleanup conditional logic a bit
--- DEVELOPERS | 1 + package/Config.in | 1 + package/dbus-broker/Config.in | 35 ++++++++ package/dbus-broker/dbus-broker.hash | 3 + package/dbus-broker/dbus-broker.mk | 72 ++++++++++++++++ package/dbus-broker/dbus.socket | 5 ++ package/dbus-broker/session.conf | 65 +++++++++++++++ package/dbus-broker/system.conf | 120 +++++++++++++++++++++++++++ 8 files changed, 302 insertions(+) create mode 100644 package/dbus-broker/Config.in create mode 100644 package/dbus-broker/dbus-broker.hash create mode 100644 package/dbus-broker/dbus-broker.mk create mode 100644 package/dbus-broker/dbus.socket create mode 100644 package/dbus-broker/session.conf create mode 100644 package/dbus-broker/system.conf
DEVELOPERS +1 -0
@@ -1884,6 +1884,7 @@ F: package/tpm-tools/ F: package/trousers/ N: Norbert Lange <[email protected]> +F: package/dbusbroker/ F: package/tcf-agent/ N: Nylon Chen <[email protected]>
package/Config.in +1 -0
@@ -432,6 +432,7 @@ endmenu source "package/dahdi-linux/Config.in" source "package/dahdi-tools/Config.in" source "package/dbus/Config.in" + source "package/dbus-broker/Config.in" source "package/dbus-cpp/Config.in" source "package/dbus-glib/Config.in" source "package/dbus-python/Config.in"
package/dbus-broker/Config.in new file +35 -0
@@ -0,0 +1,35 @@ +config BR2_PACKAGE_DBUS_BROKER + bool "dbus-broker" + depends on BR2_USE_MMU + depends on BR2_TOOLCHAIN_HAS_THREADS + help + Linux D-Bus Message Broker. + + The dbus-broker project is an implementation of a message bus + as defined by the D-Bus specification. Its aim is to provide + high performance and reliability, while keeping compatibility + to the D-Bus reference implementation. + + It is exclusively written for Linux systems, and makes use of + many modern features provided by recent linux kernel releases. + + https://github.com/bus1/dbus-broker/wiki + +if BR2_PACKAGE_DBUS_BROKER +config BR2_PACKAGE_DBUS_BROKER_LAUNCH + bool "dbus-broker-launch" + default y + depends on BR2_INIT_SYSTEMD + select BR2_PACKAGE_EXPAT + help + Launcher for D-Bus Message Brokers. + + dbus-broker-launch is a launcher for dbus-broker, spawning and + managing a D-Bus Message Bus. The launcher aims to be fully + compatible to the D-Bus reference implementation dbus-daemon, + supporting the same configuration syntax and runtime environment. +endif + +comment "dbus-broker needs a toolchain w/ threads" + depends on BR2_USE_MMU + depends on !BR2_TOOLCHAIN_HAS_THREADS
package/dbus-broker/dbus-broker.hash new file +3 -0
@@ -0,0 +1,3 @@ +# Locally calculated +sha256 95adfde56bce898c3b69eee0524732365e802348dd8189a35d5d00c30990dc81 dbus-broker-23.tar.xz +sha256 3cda3630283eda0eab825abe5ac84d191248c6b3fe1c232a118124959b96c6a4 LICENSE
package/dbus-broker/dbus-broker.mk new file +72 -0
72 lines not on this page · raw message
package/dbus-broker/dbus.socket new file +5 -0
@@ -0,0 +1,5 @@ +[Unit] +Description=D-Bus System Message Bus Socket + +[Socket] +ListenStream=/run/dbus/system_bus_socket
package/dbus-broker/session.conf new file +65 -0
65 lines not on this page · raw message
package/dbus-broker/system.conf new file +120 -0
120 lines not on this page · raw message
Signature · 1 line
2.26.2
PATCH v2 1/1 package/dbusbroker: new package
Norbert, All,
Thanks for this new improved version. :-)
I still have a few concerns about it, see below...
On 2020-06-11 01:24 +0200, Norbert Lange spake thusly:
Quoted 2 lines
Add dbus-broker, which is a drop-in replacement
for the dbus-daemon.
Sorry, but this commit log is far from enough. See below for all the pieces I find are missing.
A commit log is not here to describe what is being done, but why it is being done. It is here so that the others can understand it. The more details you can add (up to a certain externt, of course!), the easier the patch can be reviewed, especially when there are misundersstanding like what I provided in my own submission.
Quoted 6 lines
Its possible to use this package standalone (without the dbus package - if buildroot's systemd would not depend on dbus).
This is sufficient to provide systemd's (d)bus functionality. To allow standalone usage, the necessary config files are copied and adopted over from dbus.
As I explained previously, if you want to make systemd use dbus-broker, you change the systemd Config.in, as I did in my series. And if this will be done in a followup patch, you can write (instead of the two paragraphs above):
dbus-broker is sufficient to provide a dbus-daemon that can fullfil the requirements for systemd's dbus functionality. This will be done in a followup change.
Also, I find it lacking the part that describes how dbus-broker is split between a launcher and a bus daemon, and that the launcher can only be used with systemd, because the launcher makes heavy use of systemd functionalities.
The commit log also misses the explanations about the licensing information. This is *very* important to have, because it is not obvious hy the terms are as they are, and wh we have two files for each sub-projects. You could have simply duplicated what I wrote in my own submission.
Finally, the folowing, from here [*]:
Quoted 3 lines
bases on Yanns changes, and
- add an own config entry for dbus-broker-launch
enabled by default if systemd init is used
We usually do not enable options by default. But see below...
Quoted 1 line
- undo
BR2_COREUTILS_HOST_DEPENDENCY
So, I see you don;t like it, but BR2_COREUTILS_HOST_DEPENDENCY is already a dependency of systemd, so adding it to dbus-broker is not in fact adding any new build-time overhead. And if your build machine has a recent-enough, BR2_COREUTILS_HOST_DEPENDENCY will be empty already.
Quoted 1 line
- undo adding dbus user - never used by this package
So, how does the does the message bus daemon runs as non-root? In the original dbus pakcage, we define a user, that is used to switch the mesage bus to run as non-root. Pleas explain why the user is nopt needed.
Review
Quoted 2 lines
(quoted diff) +0
-2- add condtional audit dependency - cleanup conditional logic a bit
[*] ... to here, should have been after the --- line.
Quoted 2 lines
Signed-off-by Norbert Lange <[email protected]>---
[--SNIP--]
Quoted 18 lines
package/dbus-broker/Config.in new file +12
-0@@ -0,0 +1,35 @@ +config BR2_PACKAGE_DBUS_BROKER + bool "dbus-broker" + depends on BR2_USE_MMU + depends on BR2_TOOLCHAIN_HAS_THREADS + help + Linux D-Bus Message Broker. + + The dbus-broker project is an implementation of a message bus + as defined by the D-Bus specification. Its aim is to provide + high performance and reliability, while keeping compatibility + to the D-Bus reference implementation. +
package/dbus-broker/Config.in line 22
Quoted 10 lines
package/dbus-broker/Config.in new file +10
-0+ It is exclusively written for Linux systems, and makes use of + many modern features provided by recent linux kernel releases. + + https://github.com/bus1/dbus-broker/wiki + +if BR2_PACKAGE_DBUS_BROKER +config BR2_PACKAGE_DBUS_BROKER_LAUNCH + bool "dbus-broker-launch" + default y + depends on BR2_INIT_SYSTEMD
Do not depend on the init config option, but on the package (it is the package that provides libsystemd et al., not the init feature):
depends on BR2_PACKAGE_SYSTEMD
But see below...
package/dbus-broker/Config.in line 31
Quoted 9 lines
package/dbus-broker/Config.in new file +9
-0+ select BR2_PACKAGE_EXPAT + help + Launcher for D-Bus Message Brokers. + + dbus-broker-launch is a launcher for dbus-broker, spawning and + managing a D-Bus Message Bus. The launcher aims to be fully + compatible to the D-Bus reference implementation dbus-daemon, + supporting the same configuration syntax and runtime environment. +endif
Why do you add an option to enable the launcher? Just do it unconditionally when systemd is enabled. And select expat as I did in my own patch, in the main symbol:
config BR2_PACKAGE_DBUS_BROKER
[...]
select BR2_PACKAGE_EXPAT if BR2_PACKAGE_SYSTEMD
Note: if the option were to stay, which I doubt is interesting), then it would miss a comment that explains why the launcher is not availble:
comment "dbus-broker-launch needs systemd"
deepnds on !BR2_INIT_SYSTEMD
package/dbus-broker/Config.in line 35
Quoted 3 lines
package/dbus-broker/Config.in new file +3
-0+comment "dbus-broker needs a toolchain w/ threads" + depends on BR2_USE_MMU + depends on !BR2_TOOLCHAIN_HAS_THREADS
Yes, good. :-)
Quoted 6 lines
package/dbus-broker/dbus-broker.mk new file +0
-0
package/dbus-broker/dbus-broker.mk line 9
Quoted 9 lines
package/dbus-broker/dbus-broker.mk new file +9
-0@@ -0,0 +1,72 @@ +################################################################################ +# +# dbus-broker +# +# Launching services is delegated to systemd so there is very little else +# needed. No separate user is necessary and no helper for launching. +# +# Service + Config files were copied over from dbus, +# uneeded / unecessary entries removed for clarity.
Do not add any comment in this section. If you have help to provide, write in the help entry of the Config.in option, or in the commit log.
Quoted 9 lines
package/dbus-broker/dbus-broker.mk new file +9
-0+################################################################################ + +DBUS_BROKER_VERSION = 23 +DBUS_BROKER_SOURCE = dbus-broker-$(DBUS_BROKER_VERSION).tar.xz +DBUS_BROKER_SITE = https://github.com/bus1/dbus-broker/releases/download/v$(DBUS_BROKER_VERSION) +DBUS_BROKER_LICENSE = \ + Apache-2.0, \ + Apache-2.0 and/or LGPL-2.1+ (c-dvar, c-ini, c-list, c-rbtree, c-shquote, c-stdaux, c-utf8) +DBUS_BROKER_LICENSE_FILES = \
package/dbus-broker/dbus-broker.mk line 29
Quoted 10 lines
package/dbus-broker/dbus-broker.mk new file +10
-0+ LICENSE \ + subprojects/c-dvar/AUTHORS subprojects/c-dvar/README.md \ + subprojects/c-ini/AUTHORS subprojects/c-ini/README.md \ + subprojects/c-list/AUTHORS subprojects/c-list/README.md \ + subprojects/c-rbtree/AUTHORS subprojects/c-rbtree/README.md \ + subprojects/c-shquote/AUTHORS subprojects/c-shquote/README.md \ + subprojects/c-stdaux/AUTHORS subprojects/c-stdaux/README.md \ + subprojects/c-utf8/AUTHORS subprojects/c-utf8/README.md + +ifeq ($(BR2_PACKAGE_DBUS_BROKER_LAUNCH),y)
ifeq ($(BR2_PACKAGE_SYSTEMD),y)
package/dbus-broker/dbus-broker.mk line 34
Quoted 5 lines
package/dbus-broker/dbus-broker.mk new file +5
-0+DBUS_BROKER_DEPENDENCIES += expat systemd +DBUS_BROKER_CONF_OPTS += -Dlauncher=true +else +DBUS_BROKER_CONF_OPTS += -Dlauncher=false +endif
[--SNIP--]
package/dbus-broker/dbus-broker.mk line 58
Quoted 3 lines
package/dbus-broker/dbus-broker.mk new file +3
-0+# Only install config and service files if dbus is not available +ifeq ($(BR2_PACKAGE_DBUS)X$(BR2_PACKAGE_DBUS_BROKER_LAUNCH),Xy) +define DBUS_BROKER_INSTALL_CONFIG_FILES
This is not entirey wrong, but I think still incorrect. All those files are only releveant when systemd is used as an init system. As such, the DBUS_BROKER_INSTALL_INIT_SYSTEMD hook should be used.
Quoted 4 lines
package/dbus-broker/dbus-broker.mk new file +4
-0+ $(INSTALL) -D -m644 $(DBUS_BROKER_PKGDIR)/dbus.socket \ + $(TARGET_DIR)/usr/lib/systemd/system/dbus.socket + $(INSTALL) -D -m644 $(DBUS_BROKER_PKGDIR)/session.conf \ + $(TARGET_DIR)/usr/share/dbus-1/session.conf
package/dbus-broker/dbus-broker.mk line 72
Quoted 10 lines
package/dbus-broker/dbus-broker.mk new file +10
-0+ $(INSTALL) -D -m644 $(DBUS_BROKER_PKGDIR)/system.conf \ + $(TARGET_DIR)/usr/share/dbus-1/system.conf + ln -sf ../dbus.socket \ + $(TARGET_DIR)/usr/lib/systemd/system/sockets.target.wants/dbus.socket +endef + +DBUS_BROKER_POST_INSTALL_TARGET_HOOKS += DBUS_BROKER_INSTALL_CONFIG_FILES +endif + +$(eval $(meson-package))
So, in addition to all the above, this patch is lacking two other things that I did provide:
- switching systemd to work with dbus-broker (rather than whining
about it in the commit log;
- a runtime test that demonstrates that systemd does run fine with dbus-broker, and that the original dbus still takes precendence when both dbus and dbus-broker are enabled.
Note that the runtime test is not only about demonstrating the feature; it is also and foremost a way to guarantee that any regression will be caught, since we automatically run the runtime tests weekly in gitlab-ci.
Regards,
Yann E. MORIN.
Signature · 6 lines
.-----------------.--------------------.------------------.--------------------. | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: | | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ | | +33 561 099 427 `------------.-------: X AGAINST | \e/ There is no | | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. | '------------------------------^-------^------------------^--------------------'
PATCH v2 1/1 package/dbusbroker: new package
Am Fr., 12. Juni 2020 um 07:32 Uhr schrieb Yann E. MORIN <[email protected]>:
Quoted 11 lines
Norbert, All,
Thanks for this new improved version. :-)
I still have a few concerns about it, see below...
On 2020-06-11 01:24 +0200, Norbert Lange spake thusly:
Quoted 2 lines
Add dbus-broker, which is a drop-in replacement
for the dbus-daemon.
Quoted 8 lines
Sorry, but this commit log is far from enough. See below for all the pieces I find are missing.
A commit log is not here to describe what is being done, but why it is being done. It is here so that the others can understand it. The more details you can add (up to a certain externt, of course!), the easier the patch can be reviewed, especially when there are misundersstanding like what I provided in my own submission.
Yes, I hoped you could merge this from your version, which I commented BTW. I have more a problem with the workflow of the ML, especially if someone "branches out" with another patch.
Quoted 8 lines
Quoted 7 lines
Its possible to use this package standalone (without the dbus package - if buildroot's systemd would not depend on dbus).
This is sufficient to provide systemd's (d)bus functionality. To allow standalone usage, the necessary config files are copied and adopted over from dbus.
Quoted 8 lines
As I explained previously, if you want to make systemd use dbus-broker, you change the systemd Config.in, as I did in my series. And if this will be done in a followup patch, you can write (instead of the two paragraphs above):
dbus-broker is sufficient to provide a dbus-daemon that can fullfil the requirements for systemd's dbus functionality. This will be done in a followup change.
Ok, that was my intention.
Quoted 11 lines
Also, I find it lacking the part that describes how dbus-broker is split between a launcher and a bus daemon, and that the launcher can only be used with systemd, because the launcher makes heavy use of systemd functionalities.
The commit log also misses the explanations about the licensing information. This is *very* important to have, because it is not obvious hy the terms are as they are, and wh we have two files for each sub-projects. You could have simply duplicated what I wrote in my own submission.
Yes, I am not up to speed about commit-log stuff, and I usually keep explanations where the code is.
Quoted 7 lines
Finally, the folowing, from here [*]:
Quoted 3 lines
bases on Yanns changes, and
- add an own config entry for dbus-broker-launch
enabled by default if systemd init is used
Quoted 8 lines
We usually do not enable options by default. But see below...
Quoted 1 line
- undo
BR2_COREUTILS_HOST_DEPENDENCYSo, I see you don;t like it, but
BR2_COREUTILS_HOST_DEPENDENCYis already a dependency of systemd, so adding it to dbus-broker is not in fact adding any new build-time overhead. And if your build machine has a recent-enough,BR2_COREUTILS_HOST_DEPENDENCYwill be empty already.
There is still the point of keeping things simple, and I dont get why ../dbus.socket
cant be used, instead of an gnu-specific option.
Quoted 7 lines
Quoted 2 lines
- undo adding dbus user - never used by this package
So, how does the does the message bus daemon runs as non-root? In the original dbus pakcage, we define a user, that is used to switch the mesage bus to run as non-root. Pleas explain why the user is nopt needed.
Ok, I take everything back.
I thought this was handled in the service files by adding isolation options (as systemd does the "launching"). Seems like it does drop to the uid, dunno what I tested months ago when first created that package
Quoted 45 lines
Review
Quoted 3 lines
(quoted diff) +0
-2- add condtional audit dependency - cleanup conditional logic a bit[*] ... to here, should have been after the --- line.
Quoted 2 lines
Signed-off-by Norbert Lange <[email protected]>---
[--SNIP--]
Quoted 18 lines
package/dbus-broker/Config.in new file +12
-0@@ -0,0 +1,35 @@ +config BR2_PACKAGE_DBUS_BROKER + bool "dbus-broker" + depends on BR2_USE_MMU + depends on BR2_TOOLCHAIN_HAS_THREADS + help + Linux D-Bus Message Broker. + + The dbus-broker project is an implementation of a message bus + as defined by the D-Bus specification. Its aim is to provide + high performance and reliability, while keeping compatibility + to the D-Bus reference implementation. +package/dbus-broker/Config.in line 22
Quoted 10 lines
package/dbus-broker/Config.in new file +10
-0+ It is exclusively written for Linux systems, and makes use of + many modern features provided by recent linux kernel releases. + + https://github.com/bus1/dbus-broker/wiki + +if BR2_PACKAGE_DBUS_BROKER +config BR2_PACKAGE_DBUS_BROKER_LAUNCH + bool "dbus-broker-launch" + default y + depends on BR2_INIT_SYSTEMDDo not depend on the init config option, but on the package (it is the package that provides libsystemd et al., not the init feature):
depends on
BR2_PACKAGE_SYSTEMDBut see below...
Quoted 13 lines
package/dbus-broker/Config.in line 31
Quoted 9 lines
package/dbus-broker/Config.in new file +9
-0+ select BR2_PACKAGE_EXPAT + help + Launcher for D-Bus Message Brokers. + + dbus-broker-launch is a launcher for dbus-broker, spawning and + managing a D-Bus Message Bus. The launcher aims to be fully + compatible to the D-Bus reference implementation dbus-daemon, + supporting the same configuration syntax and runtime environment. +endifWhy do you add an option to enable the launcher? Just do it unconditionally when systemd is enabled. And select expat as I did in my own patch, in the main symbol:
Well, my first attempt was to only make dbus-broker(-launch) available with systemd,
given that there is probably no one using it differently yet. If you argue that it makes sense to provide the plain dbus-broker, then it makes sense doing so with systemd aswell.
(see below, I use system without dbus or dbus-broker-launch).
Quoted 90 lines
config
BR2_PACKAGE_DBUS_BROKER
[...]
selectBR2_PACKAGE_EXPATifBR2_PACKAGE_SYSTEMDNote: if the option were to stay, which I doubt is interesting), then it would miss a comment that explains why the launcher is not availble:
comment "dbus-broker-launch needs systemd"
deepnds on !BR2_INIT_SYSTEMDpackage/dbus-broker/Config.in line 35
Quoted 3 lines
package/dbus-broker/Config.in new file +3
-0+comment "dbus-broker needs a toolchain w/ threads" + depends on BR2_USE_MMU + depends on !BR2_TOOLCHAIN_HAS_THREADSYes, good. :-)
Quoted 6 lines
package/dbus-broker/dbus-broker.mk new file +0
-0package/dbus-broker/dbus-broker.mk line 9
Quoted 9 lines
package/dbus-broker/dbus-broker.mk new file +9
-0@@ -0,0 +1,72 @@ +################################################################################ +# +# dbus-broker +# +# Launching services is delegated to systemd so there is very little else +# needed. No separate user is necessary and no helper for launching. +# +# Service + Config files were copied over from dbus, +# uneeded / unecessary entries removed for clarity.Do not add any comment in this section. If you have help to provide, write in the help entry of the Config.in option, or in the commit log.
Quoted 9 lines
package/dbus-broker/dbus-broker.mk new file +9
-0+################################################################################ + +DBUS_BROKER_VERSION = 23 +DBUS_BROKER_SOURCE = dbus-broker-$(DBUS_BROKER_VERSION).tar.xz +DBUS_BROKER_SITE = https://github.com/bus1/dbus-broker/releases/download/v$(DBUS_BROKER_VERSION) +DBUS_BROKER_LICENSE = \ + Apache-2.0, \ + Apache-2.0 and/or LGPL-2.1+ (c-dvar, c-ini, c-list, c-rbtree, c-shquote, c-stdaux, c-utf8) +DBUS_BROKER_LICENSE_FILES = \package/dbus-broker/dbus-broker.mk line 29
Quoted 10 lines
package/dbus-broker/dbus-broker.mk new file +10
-0+ LICENSE \ + subprojects/c-dvar/AUTHORS subprojects/c-dvar/README.md \ + subprojects/c-ini/AUTHORS subprojects/c-ini/README.md \ + subprojects/c-list/AUTHORS subprojects/c-list/README.md \ + subprojects/c-rbtree/AUTHORS subprojects/c-rbtree/README.md \ + subprojects/c-shquote/AUTHORS subprojects/c-shquote/README.md \ + subprojects/c-stdaux/AUTHORS subprojects/c-stdaux/README.md \ + subprojects/c-utf8/AUTHORS subprojects/c-utf8/README.md + +ifeq ($(BR2_PACKAGE_DBUS_BROKER_LAUNCH),y)ifeq ($(BR2_PACKAGE_SYSTEMD),y)
package/dbus-broker/dbus-broker.mk line 34
Quoted 5 lines
package/dbus-broker/dbus-broker.mk new file +5
-0+DBUS_BROKER_DEPENDENCIES += expat systemd +DBUS_BROKER_CONF_OPTS += -Dlauncher=true +else +DBUS_BROKER_CONF_OPTS += -Dlauncher=false +endif[--SNIP--]
package/dbus-broker/dbus-broker.mk line 58
Quoted 3 lines
package/dbus-broker/dbus-broker.mk new file +3
-0+# Only install config and service files if dbus is not available +ifeq ($(BR2_PACKAGE_DBUS)X$(BR2_PACKAGE_DBUS_BROKER_LAUNCH),Xy) +define DBUS_BROKER_INSTALL_CONFIG_FILESThis is not entirey wrong, but I think still incorrect. All those files are only releveant when systemd is used as an init system. As such, the
DBUS_BROKER_INSTALL_INIT_SYSTEMDhook should be used.Quoted 4 lines
package/dbus-broker/dbus-broker.mk new file +4
-0+ $(INSTALL) -D -m644 $(DBUS_BROKER_PKGDIR)/dbus.socket \ + $(TARGET_DIR)/usr/lib/systemd/system/dbus.socket + $(INSTALL) -D -m644 $(DBUS_BROKER_PKGDIR)/session.conf \ + $(TARGET_DIR)/usr/share/dbus-1/session.confpackage/dbus-broker/dbus-broker.mk line 72
Quoted 10 lines
package/dbus-broker/dbus-broker.mk new file +10
-0+ $(INSTALL) -D -m644 $(DBUS_BROKER_PKGDIR)/system.conf \ + $(TARGET_DIR)/usr/share/dbus-1/system.conf + ln -sf ../dbus.socket \ + $(TARGET_DIR)/usr/lib/systemd/system/sockets.target.wants/dbus.socket +endef + +DBUS_BROKER_POST_INSTALL_TARGET_HOOKS += DBUS_BROKER_INSTALL_CONFIG_FILES +endif + +$(eval $(meson-package))So, in addition to all the above, this patch is lacking two other things that I did provide:
Quoted 2 lines
- switching systemd to work with dbus-broker (rather than whining
about it in the commit log;
I got a patch series for systemd, just a matter of finding the time (and retesting).
But I would just simply *take out* the dependency to DBUS ( and UTIL_LINUX_BINARIES and UTIL_LINUX_NOLOGIN, getting a systemd rootfs below 20MB).
I have been running systemd without either for more than a year.
What would be your pick here? no dependency and a warning if neither is available?
adding some BR2_HAS_DBUS_DAEMON that is set by both, so systemd features (like logind) and packages depending on that (and potentially on PACKAGE_DBUS if they need the library or tools)?
(https://github.com/nolange/buildroot/commits/improve_systemd_nodbus)
Quoted 8 lines
- a runtime test that demonstrates that systemd does run fine with dbus-broker, and that the original dbus still takes precendence when both dbus and dbus-broker are enabled.
Note that the runtime test is not only about demonstrating the feature; it is also and foremost a way to guarantee that any regression will be caught, since we automatically run the runtime tests weekly in gitlab-ci.
Ok. Since you already have that, and I know little about your test framework, then could you please incorporate this version into your patch-set.
Norbert
PATCH v2 1/1 package/dbusbroker: new package
Norbert, All,
On 2020-06-12 09:02 +0200, Norbert Lange spake thusly:
Quoted 20 lines
Am Fr., 12. Juni 2020 um 07:32 Uhr schrieb Yann E. MORIN <[email protected]>:
Quoted 9 lines
Thanks for this new improved version. :-)
I still have a few concerns about it, see below...
On 2020-06-11 01:24 +0200, Norbert Lange spake thusly:
Quoted 2 lines
Add dbus-broker, which is a drop-in replacement
for the dbus-daemon.Quoted 8 lines
Sorry, but this commit log is far from enough. See below for all the pieces I find are missing.
A commit log is not here to describe what is being done, but why it is being done. It is here so that the others can understand it. The more details you can add (up to a certain externt, of course!), the easier the patch can be reviewed, especially when there are misundersstanding like what I provided in my own submission.
Quoted 3 lines
Yes, I hoped you could merge this from your version, which I commented BTW. I have more a problem with the workflow of the ML, especially if someone "branches out" with another patch.
When I am suggesting an alternate solution, but I have some doubts about it, I would usually take the original patch, massage it into my idea of what it should look like, test it, and resubmit to make my point.
Then the original submitter (and others!) have a way to see the point, and comment further by providing additional review and explanations on why the alternate proposal is not correct.
Eventually, the original submitter can further improve by keeping the good pieces, removing the bad ones, and so on... and resubmit a new iteration.
So you are very welcome to have re-spun this new iteration. :-)
[--SNIP--]
Quoted 7 lines
Quoted 5 lines
The commit log also misses the explanations about the licensing information. This is *very* important to have, because it is not obvious hy the terms are as they are, and wh we have two files for each sub-projects. You could have simply duplicated what I wrote in my own submission.
Yes, I am not up to speed about commit-log stuff, and I usually keep explanations where the code is.
Usually, we like the nitty-gritty details in the commit log, for posterity, while we keep the comments in the code to the bare minimal.
In this case, the licensing dirtiness should be kept in the commit log, and a comment could be added in the .mk, like:
# Inconsistency between AUTHORS and README, keep both, and # interpret as per the README
Quoted 10 lines
Quoted 10 lines
Finally, the folowing, from here [*]:
Quoted 3 lines
bases on Yanns changes, and
- add an own config entry for dbus-broker-launch
enabled by default if systemd init is usedWe usually do not enable options by default. But see below...
Quoted 1 line
- undo
BR2_COREUTILS_HOST_DEPENDENCY
Quoted 8 lines
Quoted 4 lines
So, I see you don;t like it, but
BR2_COREUTILS_HOST_DEPENDENCYis already a dependency of systemd, so adding it to dbus-broker is not in fact adding any new build-time overhead. And if your build machine has a recent-enough,BR2_COREUTILS_HOST_DEPENDENCYwill be empty already.There is still the point of keeping things simple, and I dont get why ../dbus.socket
cant be used, instead of an gnu-specific option.
I tend to like it better that we use it, because this makes it explicit.
Also, there is always the ambiguity about what a relative symlink means when created: is it relative to the current working directory, or relative to the symlink itself?
And since we already anyway are sure we do have it (either from the host or as a dependency of systemd itself), let's just use it...
And BTW, we do not explicitly support building on a non-GNU system anyway, so it being a GNU extension is moot (IMHO).
Quoted 11 lines
Quoted 6 lines
Quoted 1 line
- undo adding dbus user - never used by this package
So, how does the does the message bus daemon runs as non-root? In the original dbus pakcage, we define a user, that is used to switch the mesage bus to run as non-root. Pleas explain why the user is nopt needed.
Ok, I take everything back.
I thought this was handled in the service files by adding isolation options (as systemd does the "launching"). Seems like it does drop to the uid, dunno what I tested months ago when first created that package
OK, I knew that would have been systemd doing the uid-drop, by I'd still like the info appear somewhere (so that we do not later question this in light of the original dbus package which defines one).
[--SNIP--]
Quoted 9 lines
Quoted 3 lines
Why do you add an option to enable the launcher? Just do it unconditionally when systemd is enabled. And select expat as I did in my own patch, in the main symbol:
Well, my first attempt was to only make dbus-broker(-launch) available with systemd,
given that there is probably no one using it differently yet. If you argue that it makes sense to provide the plain dbus-broker, then it makes sense doing so with systemd aswell.
(see below, I use system without dbus or dbus-broker-launch).
I guess you meant 'systemd' in that last sentence.
But OK, I would be fine to introduce this package as a systemd-only option for now, until someone actually provides a non-systemd launcher.
So, we would still not need an extra option, but just:
config BR2_PACKAGE_DBUS_BROKER
bool "dbus-broker"
depends on BR2_USE_MMU
depends on BR2_TOOLCHAIN_HAS_THREADS
depends on BR2_PACKAGE_SYSTEMD
select BR2_PACKAGE_EXPAT
And explain in the commit log that the launcher needs systemd, and that without the launcher there is no point in dbus-broker.
[--SNIP--]
Quoted 6 lines
Quoted 5 lines
So, in addition to all the above, this patch is lacking two other things that I did provide:
- switching systemd to work with dbus-broker (rather than whining
about it in the commit log;
Quoted 8 lines
I got a patch series for systemd, just a matter of finding the time (and retesting).
But I would just simply *take out* the dependency to DBUS ( andUTIL_LINUX_BINARIESandUTIL_LINUX_NOLOGIN, getting a systemd rootfs below 20MB).
I have been running systemd without either for more than a year. What would be your pick here? no dependency and a warning if neither is available?
What would be the reason for not wanting dbus on a systemd-based system (honest question) ?
Should we move that select to a (existing?) sub-option of systemd?
Quoted 4 lines
adding some
BR2_HAS_DBUS_DAEMONthat is set by both, so systemd features (like logind) and packages depending on that (and potentially onPACKAGE_DBUSif they need the library or tools)?
(https://github.com/nolange/buildroot/commits/improve_systemd_nodbus)
In the current case, we would have the systemd's sub-option(s) select what they require, and move the select down to logind (and other sub-options, maybe?)
config BR2_PACKAGE_SYSTEMD_LOGIND
bool "logind"
depends on BR2_USE_MMU # dbus && dbus-broker depends on BR2_TOOLCHAIN_HAS_THREADS # dbus && dbus-broker select BR2_PACKAGE_DBUS if !BR2_PACKAGE_DBUS_BROKER
comment "logind needs a toolchain w/ threads"
depends on BR2_USE_MMU
depends on !BR2_TOOLCHAIN_HAS_THREADS
... and keep the systemd.mk as it is now.
Note that is there are three or more sub-option that require dbus, we could go with an intermediate option:
config BR2_PACKAGE_SYSTEMD_NEEDS_DBUS
bool
depends on BR2_USE_MMU # dbus && dbus-broker depends on BR2_TOOLCHAIN_HAS_THREADS # dbus && dbus-broker select BR2_PACKAGE_DBUS if !BR2_PACKAGE_DBUS_BROKER
config BR2_PACKAGE_SYSTEMD_LOGIND
bool "logind"
depends on BR2_USE_MMU # needs_dbus depends on BR2_TOOLCHAIN_HAS_THREADS # needs_dbus select BR2_PACKAGE_SYSTEMD_NEEDS_DBUS
comment "logind needs a toolchain w/ threads"
depends on BR2_USE_MMU
depends on !BR2_TOOLCHAIN_HAS_THREADS
Quoted 9 lines
Quoted 7 lines
- a runtime test that demonstrates that systemd does run fine with dbus-broker, and that the original dbus still takes precendence when both dbus and dbus-broker are enabled.
Note that the runtime test is not only about demonstrating the feature; it is also and foremost a way to guarantee that any regression will be caught, since we automatically run the runtime tests weekly in gitlab-ci.
Ok. Since you already have that, and I know little about your test framework, then could you please incorporate this version into your patch-set.
If you agree, then I'll try to merge your new submission with your comments into my series and re-spin it soon-ish, if that's OK with you.
Regards,
Yann E. MORIN.
Signature · 6 lines
.-----------------.--------------------.------------------.--------------------. | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: | | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ | | +33 561 099 427 `------------.-------: X AGAINST | \e/ There is no | | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. | '------------------------------^-------^------------------^--------------------'
PATCH v2 1/1 package/dbusbroker: new package
Am So., 14. Juni 2020 um 15:10 Uhr schrieb Yann E. MORIN <[email protected]>:
Quoted 36 lines
Norbert, All,
On 2020-06-12 09:02 +0200, Norbert Lange spake thusly:
Quoted 20 lines
Am Fr., 12. Juni 2020 um 07:32 Uhr schrieb Yann E. MORIN <[email protected]>:
Quoted 9 lines
Thanks for this new improved version. :-)
I still have a few concerns about it, see below...
On 2020-06-11 01:24 +0200, Norbert Lange spake thusly:
Quoted 2 lines
Add dbus-broker, which is a drop-in replacement
for the dbus-daemon.Quoted 8 lines
Sorry, but this commit log is far from enough. See below for all the pieces I find are missing.
A commit log is not here to describe what is being done, but why it is being done. It is here so that the others can understand it. The more details you can add (up to a certain externt, of course!), the easier the patch can be reviewed, especially when there are misundersstanding like what I provided in my own submission.
Quoted 3 lines
Yes, I hoped you could merge this from your version, which I commented BTW. I have more a problem with the workflow of the ML, especially if someone "branches out" with another patch.
When I am suggesting an alternate solution, but I have some doubts about it, I would usually take the original patch, massage it into my idea of what it should look like, test it, and resubmit to make my point.
Then the original submitter (and others!) have a way to see the point, and comment further by providing additional review and explanations on why the alternate proposal is not correct.
Quoted 5 lines
Eventually, the original submitter can further improve by keeping the good pieces, removing the bad ones, and so on... and resubmit a new iteration.
So you are very welcome to have re-spun this new iteration. :-)
As said, I have trouble with the ML flow, seems like the commit should then include the whole discussion as otherwise no one is able to follow?
(Might be just me)
Quoted 7 lines
[--SNIP--]
Quoted 5 lines
Quoted 5 lines
The commit log also misses the explanations about the licensing information. This is *very* important to have, because it is not obvious hy the terms are as they are, and wh we have two files for each sub-projects. You could have simply duplicated what I wrote in my own submission.
Quoted 8 lines
Quoted 2 lines
Yes, I am not up to speed about commit-log stuff, and I usually keep explanations where the code is.
Usually, we like the nitty-gritty details in the commit log, for posterity, while we keep the comments in the code to the bare minimal.
In this case, the licensing dirtiness should be kept in the commit log, and a comment could be added in the .mk, like:
I did not adjust the commit log, should've mentioned that ;) Consider it a continued discussion.
Quoted 23 lines
# Inconsistency between AUTHORS and README, keep both, and # interpret as per the README
Quoted 10 lines
Quoted 10 lines
Finally, the folowing, from here [*]:
Quoted 3 lines
bases on Yanns changes, and
- add an own config entry for dbus-broker-launch
enabled by default if systemd init is usedWe usually do not enable options by default. But see below...
Quoted 1 line
- undo
BR2_COREUTILS_HOST_DEPENDENCYQuoted 8 lines
Quoted 4 lines
So, I see you don;t like it, but
BR2_COREUTILS_HOST_DEPENDENCYis already a dependency of systemd, so adding it to dbus-broker is not in fact adding any new build-time overhead. And if your build machine has a recent-enough,BR2_COREUTILS_HOST_DEPENDENCYwill be empty already.There is still the point of keeping things simple, and I dont get why ../dbus.socket
cant be used, instead of an gnu-specific option.
Quoted 5 lines
I tend to like it better that we use it, because this makes it explicit.
Also, there is always the ambiguity about what a relative symlink means when created: is it relative to the current working directory, or relative to the symlink itself?
Seems clearer to me than some non-trivial path duplicated, and identical to what you see with a 'ln -l'.
But I am not really invested into keeping it one way or another.
Quoted 14 lines
And since we already anyway are sure we do have it (either from the host or as a dependency of systemd itself), let's just use it...
And BTW, we do not explicitly support building on a non-GNU system anyway, so it being a GNU extension is moot (IMHO).
Quoted 7 lines
Quoted 6 lines
Quoted 1 line
- undo adding dbus user - never used by this package
So, how does the does the message bus daemon runs as non-root? In the original dbus pakcage, we define a user, that is used to switch the mesage bus to run as non-root. Pleas explain why the user is nopt needed.
Quoted 8 lines
Quoted 4 lines
Ok, I take everything back.
I thought this was handled in the service files by adding isolation options (as systemd does the "launching"). Seems like it does drop to the uid, dunno what I tested months ago when first created that packageOK, I knew that would have been systemd doing the uid-drop, by I'd still like the info appear somewhere (so that we do not later question this in light of the original dbus package which defines one).
Not sure if we got some communication issues. The setting does something (drop privileges),
while I originally considered the launcher ignoring the Xml key. I guess you could tweak the systemd service to use a Dynamic user, but I won't touch that for now.
Quoted 26 lines
[--SNIP--]
Quoted 9 lines
Quoted 3 lines
Why do you add an option to enable the launcher? Just do it unconditionally when systemd is enabled. And select expat as I did in my own patch, in the main symbol:
Well, my first attempt was to only make dbus-broker(-launch) available with systemd,
given that there is probably no one using it differently yet. If you argue that it makes sense to provide the plain dbus-broker, then it makes sense doing so with systemd aswell.
(see below, I use system without dbus or dbus-broker-launch).I guess you meant 'systemd' in that last sentence.
But OK, I would be fine to introduce this package as a systemd-only option for now, until someone actually provides a non-systemd launcher.
So, we would still not need an extra option, but just:
config
BR2_PACKAGE_DBUS_BROKER
bool "dbus-broker"
depends onBR2_USE_MMU
depends onBR2_TOOLCHAIN_HAS_THREADS
depends onBR2_PACKAGE_SYSTEMD
selectBR2_PACKAGE_EXPAT
Quoted 2 lines
And explain in the commit log that the launcher needs systemd, and that without the launcher there is no point in dbus-broker.
Seems to be the best option for now IMHO. Revisit the launcher-less option once you have something to test.
Quoted 17 lines
[--SNIP--]
Quoted 6 lines
Quoted 5 lines
So, in addition to all the above, this patch is lacking two other things that I did provide:
- switching systemd to work with dbus-broker (rather than whining
about it in the commit log;Quoted 8 lines
I got a patch series for systemd, just a matter of finding the time (and retesting).
But I would just simply *take out* the dependency to DBUS ( andUTIL_LINUX_BINARIESandUTIL_LINUX_NOLOGIN, getting a systemd rootfs below 20MB).
I have been running systemd without either for more than a year. What would be your pick here? no dependency and a warning if neither is available?
Quoted 2 lines
What would be the reason for not wanting dbus on a systemd-based system (honest question) ?
Minimalism, both in RAM + storage aswell as in maintenance and attack points. You can have alot of "Dbus-like" just with unix sockets, and the core system and service manager portion is fully functional without it. systemd moved some early init stuff and core stuff to "varlink" the last couple years [1],
I can run my system without realizing dbus is missing.
(Back when I wasn't sure about that, I added dbus-broker for a cleaner dbus implementations)
Quoted 2 lines
Should we move that select to a (existing?) sub-option of systemd?
Dunno, add a warning like?
+ comment "systemd recommends enabling a dbus daemon"
+ depends on !BR2_PACKAGE_DBUS
+ depends on !BR2_PACKAGE_DBUS_BROKER
Quoted 37 lines
Quoted 5 lines
adding some
BR2_HAS_DBUS_DAEMONthat is set by both, so systemd features (like logind) and packages depending on that (and potentially onPACKAGE_DBUSif they need the library or tools)?
(https://github.com/nolange/buildroot/commits/improve_systemd_nodbus)In the current case, we would have the systemd's sub-option(s) select what they require, and move the select down to logind (and other sub-options, maybe?)
config
BR2_PACKAGE_SYSTEMD_LOGIND
bool "logind"
depends onBR2_USE_MMU# dbus && dbus-broker depends onBR2_TOOLCHAIN_HAS_THREADS# dbus && dbus-broker selectBR2_PACKAGE_DBUSif !BR2_PACKAGE_DBUS_BROKERcomment "logind needs a toolchain w/ threads"
depends onBR2_USE_MMU
depends on !BR2_TOOLCHAIN_HAS_THREADS... and keep the systemd.mk as it is now.
Note that is there are three or more sub-option that require dbus, we could go with an intermediate option:
config
BR2_PACKAGE_SYSTEMD_NEEDS_DBUS
bool
depends onBR2_USE_MMU# dbus && dbus-broker depends onBR2_TOOLCHAIN_HAS_THREADS# dbus && dbus-broker selectBR2_PACKAGE_DBUSif !BR2_PACKAGE_DBUS_BROKERconfig
BR2_PACKAGE_SYSTEMD_LOGIND
bool "logind"
depends onBR2_USE_MMU#needs_dbusdepends onBR2_TOOLCHAIN_HAS_THREADS#needs_dbusselectBR2_PACKAGE_SYSTEMD_NEEDS_DBUS
Quoted 4 lines
comment "logind needs a toolchain w/ threads"
depends onBR2_USE_MMU
depends on !BR2_TOOLCHAIN_HAS_THREADS
Problem is, that thi is not as clear-cut and might change btw version, see [1]. For the time being, a global "here be dragons" warning if no dbus-daemon is available would be best.
For example networkd is fine handling the system setup, but networkctl cant display
the state.
dbus/broker should be enabled by default with systemd, but with some way that the user still can disable those. Dunno what's the correct to do this in buildroot,
add a line to dbus 'default y if BR2_PACKAGE_SYSTEMD'?
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
- a runtime test that demonstrates that systemd does run fine with dbus-broker, and that the original dbus still takes precendence when both dbus and dbus-broker are enabled.
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Note that the runtime test is not only about demonstrating the feature; it is also and foremost a way to guarantee that any regression will be caught, since we automatically run the runtime tests weekly in gitlab-ci.
Ok. Since you already have that, and I know little about your test framework, then could you please incorporate this version into your patch-set.
If you agree, then I'll try to merge your new submission with your comments into my series and re-spin it soon-ish, if that's OK with you.
You doing the work, and I get to nag? Deal ;)
Norbert
PATCH v2 1/1 package/dbusbroker: new package
Norbert, All,
On 2020-06-14 23:30 +0200, Norbert Lange spake thusly:
Quoted 2 lines
Am So., 14. Juni 2020 um 15:10 Uhr schrieb Yann E. MORIN <[email protected]>:
[--SNIP--]
Quoted 3 lines
Quoted 1 line
So you are very welcome to have re-spun this new iteration. :-)
As said, I have trouble with the ML flow, seems like the commit should then include the whole discussion as otherwise no one is able to follow?
I guess that's what the post-commit history (all that is after the --- line) is for: keep a track of the changes between the various iteration of a patch (if I understood your isue correctly).
[--SNIP--]
Quoted 8 lines
Quoted 3 lines
OK, I knew that would have been systemd doing the uid-drop, by I'd still like the info appear somewhere (so that we do not later question this in light of the original dbus package which defines one).
Not sure if we got some communication issues. The setting does something (drop privileges),
while I originally considered the launcher ignoring the Xml key. I guess you could tweak the systemd service to use a Dynamic user, but I won't touch that for now.
So I did some testing:
- without the <user> entry in system.conf, the launcher and the message bus are running as root.
- with a <user> entry in system.conf (and the correct user in /etc/passwd and /etc/group), the launcher and the message bus are running as non-root with the specified user, and systemd is working happily.
So, I would like we do use a non-root user for dbus-broker, like we do for the original dbus.
I'll add that when I re-submit later tonight.
[--SNIP--]
Quoted 4 lines
Quoted 2 lines
And explain in the commit log that the launcher needs systemd, and that without the launcher there is no point in dbus-broker.
Seems to be the best option for now IMHO. Revisit the launcher-less option once you have something to test.
Agreed.
[--SNIP--]
Quoted 4 lines
dbus/broker should be enabled by default with systemd, but with some way that the user still can disable those. Dunno what's the correct to do this in buildroot,
add a line to dbus 'default y ifBR2_PACKAGE_SYSTEMD'?
We can revisit that in a later patch (I've seen and started to apply your otehr series, which conclusion is to indeed drop the dbus requirement; let's see that happen there, not here).
[--SNIP--]
Quoted 3 lines
Quoted 2 lines
If you agree, then I'll try to merge your new submission with your comments into my series and re-spin it soon-ish, if that's OK with you.
You doing the work, and I get to nag? Deal ;)
My pleasure! ;-)
Thanks for the discusion and the explanations. New iteration pending by the end of the day (hopefully).
Regards,
Yann E. MORIN.
Signature · 6 lines
.-----------------.--------------------.------------------.--------------------. | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: | | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ | | +33 561 099 427 `------------.-------: X AGAINST | \e/ There is no | | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. | '------------------------------^-------^------------------^--------------------'