Kernel manages each genl family by its name, e.g. "nlctrl" or WG_GENL_NAME,
and its ID (used for nlmsg_type) is determined dynamically when the
corresponding module is loaded.
This commit makes sd-netlink follow the same way; now, sd_genl_family_t
is dropped, and sd_genl_message_new() takes a genl family name. Each
genl family is resolved when it is used first time, and its information
is stored in GenericNetlinkFamily.
This makes the root type system for nfnl indexed by subsystem, and
itroduces a next level type system for each subsystem. The second
level type systems are indexed by message types correspond to each
subsystem.
All type systems of currently supported genl families do not depend on
commands. Hence, at least tentatively, let's drop the tables.
Note that type system for genl ethtool depends on commands. Let's
reintroduce a mechanism to support the deps when we support ethtool on
netlink.
We're only doing one thing in the child process which is connecting
to the D-Bus socket so let's mention that in the error message when
something goes wrong instead of having a generic error message.
Previously, when the connect() call in (sd-buscntr) failed, we returned
-EPROTO without ever reading the actual errno from the error pipe. To fix
the issue, delay checking the process exit status until after we've read
and processed any error from the error pipe.
Let's be paranoid and do something useful if we operate with empty
haystack/needle. This doesn't actually fix anything, as the places as
far as I can see check for non-emptyness already beforehand, but I will
sleep safer at night, if we don't even allow the trap to be fallen in,
ever, even if the code is changed sooner or later.
These have ignored the return value forever. Two are public APIs so
we can't really change what they return anyway, and the other one is
a cleanup path and the existing error code is more important.
CID#1461274
CID#1461275
CID#1461276
In general we almost never hit those asserts in production code, so users see
them very rarely, if ever. But either way, we just need something that users
can pass to the developers.
We have quite a few of those asserts, and some have fairly nice messages, but
many are like "WTF?" or "???" or "unexpected something". The error that is
printed includes the file location, and function name. In almost all functions
there's at most one assert, so the function name alone is enough to identify
the failure for a developer. So we don't get much extra from the message, and
we might just as well drop them.
Dropping them makes our code a tiny bit smaller, and most importantly, improves
development experience by making it easy to insert such an assert in the code
without thinking how to phrase the argument.
DISKSEQ is a monotonic per-use unique sequence number, incremented
on each block device create/attach. For example, the same loop device
will have different diskseq each time an object is attached to it.
Add sd/udev helpers to parse/expose it.
Library code should not call freeze(), this is something that should
only be done by "application code", so moving it into shared/ is appropriate.
The fallback to call _exit() is dropped: let's trust that the infinite loop
is infinite.
UIDs don't work well over ssh, but locally or with containers they are OK.
In particular, user@.service uses UIDs as identifiers, and it's nice to be
able to copy&paste that UID for interaction with the user's managers.
The arguments are where the interesting part is:
src/libsystemd/sd-bus/bus-socket.c:965: sd-bus: starting bus with systemd-run...
↓
src/libsystemd/sd-bus/bus-socket.c:972: sd-bus: starting bus with systemd-run -M.host -PGq --wait -pUser=1000 -pPAMName=login systemd-stdio-bridge "-punix:path=\${XDG_RUNTIME_DIR}/bus"
asprintf(3) says that the pointer is "undefined" after a failed call.
In the current glibc implementation it is just NULL. In principle the
call could return a valid pointer with bad contents or something.
We have two styles of error handling: in a majority of cases we would
check the return value, but sometimes we used (void) and relied on the
pointer not being set. In practice both styles should be equivalent,
but gcc doesn't like the second one with -Wunused-result. (Though only
sometimes. E.g. on my F34 box I don't get the same warnings as in CI,
even though the compiler version is very similar and the compilation
options are the same…). It's also nice to be consistent in our code base.
So let's always use the first style of error checking.
Yu Watanabe
Luca Boccassi
Lennart Poettering
Daan De Meyer
Zbigniew Jędrzejewski-Szmek