fallback hostname fixes #5366
fallback hostname fixes #5366
Conversation
|
Hmm, CI fialure appears unrelated... I wonder what's going on there... |
This is the result of #5298 This diff --git a/Makefile.am b/Makefile.am
index 10839e9..a70ebe6 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -2469,7 +2469,7 @@ test_journal_importer_SOURCES = \
test_journal_importer_LDADD = \
libsystemd-shared.la
-EXTRA_DIST += \
+TEST_DATA_FILES += \
test/journal-data/journal-1.txt \
test/journal-data/journal-2.txtfixes that test |
Fixes: ``` Found container virtualization none. Assertion 'imp.fd >= 0' failed at ../src/test/test-journal-importer.c:43, function test_basic_parsing(). Aborting. FAIL: test-journal-importer (code: 134) ``` See systemd#5366 (comment)
|
Right, this started failing as PR #4526 was merged with test failures (that introduced the new test journal test files). |
|
Defaulting to "localhost" and sending "linux" out to the local network seems really arbitrary and inconsistent to me. I see how resolving "localhost" over mdns/llmnr is actively dangerous, but if multiple machines in a network are all called "linux" that's not really much better. Forgive the naïve question, but is it an option to simply not send a host name at all in this case? If you don't set one, I don't think it's reasonable to expect being able to use and resolve a host name. (Tests are fixed, btw, I retried them.) |
well, it means the collision logic ensures that one of the systems will pick a different name. both llmnr and mdns have strategies to deal with that in place. In the case of mdns this is particular important, as dns-sd services refer to mdns hostnames, and while hostnames might conflict service names might not, but in order not to break them you need the hostnames. Now, we currently don't implement service registration in resolved, but the goal definitely is to, and we shouldn't really alter behaviour if we gain support for them eventually. So yeah, they might not be overly useful as long as humans want to identify machines in the network, but for stuff built on this we need them, it's the way mDNS/DNS-SD is defined... Hope that makes sense... |
…tname_malloc() Currently, if the hostname is not set gethostname_malloc() defaults to the "sysname", which is "linux" on Linux. Let's change that to also honour the compile-time fallback hostname as specified on the configure command line.
…ostname When /etc/hostname isn't set, default to the configured compile-time fallback hostname instead of "localhost" for the kernel hostname.
This changes resolved to use the compile-time fallback hostname the configured one is not set. Note that if the local hostname is set to "localhost" then we'll instead default to "linux" here, as for mDNS/LLMNR exposing "localhost" is actively dangerous.
poettering
force-pushed
the
default-hostname-fix
branch
from
f0c0900
to
a25b0dc
Compare
martinpitt
added
the
good-to-merge/waiting-for-ci 👍
|
ci passed, merging |
|
FTR, looks good. I booted without /etc/hostname now, and everything seems to behave as expected. Thanks for fixing this. |
A follow-up for #5175, actually enforcing the fallback hostname also for resolved, and for the hostname PID1 sets if none is configured.
Fixes: #5253.