After reading https://simonbyrne.github.io/notes/fastmath/ I think we
should drop -ffast-math. The JSON code actually looks for NaN, so the
fact it becomes unreliable kinda sucks.
Moreover, we don't do any number crunching. We use floating point fields
only sporadical for trivial math. Hence the optimization is entirely
unnecessary.
homework: rework how we disassemble a home dir in home_deactivate()
Let's first move the home dir to a new mount point that is only visible
in our own private namespace. Then, do FITRIM and stuff there, so that
we know the regular userspace can't interfere with that, and we know
that the home fs is not used anymore.
(This will become even more important once we add auto-grow/auto-shrink
for home dirs)
by moving the read permissions to the top level and
granting additional permissions to the specific jobs.
It should help to prevent new jobs that could be added
there eventually from having write access to resources they
most likely would never need.
Judging by https://docs.github.com/en/actions/security-guides/automatic-token-authentication#permissions-for-the-github_token
it should be enough to grant the "read contents" permission to
most of our actions. The "read metadata" permission is set impliciclty
somewhere and can't be set via the "permissions" setting:
```
The workflow is not valid. .github/workflows/linter.yml (Line: 14, Col: 3): Unexpected value 'metadata'
```
With glibc-2.34.9000-17.fc36.x86_64, dynamically programs newly fail in early
init with a restrictive syscall filter that does not include @system-service.
I think this is caused by 2dd87703d4386f2776c5b5f375a494c91d7f9fe4:
nptl: Move changing of stack permissions into ld.so
All the stack lists are now in _rtld_global, so it is possible
to change stack permissions directly from there, instead of
calling into libpthread to do the change.
It seems that this call will now be very widely used, so let's just move it to
default to avoid too many failures.
shared: split out UID allocation range stuff from user-record.h
user-record.[ch] are about the UserRecord JSON stuff, and the UID
allocation range stuff (i.e. login.defs handling) is a very different
thing, and complex enough on its own, let's give it its own c/h files.
homed: include actual fs type + access mode as part of "status" section of user record
So far we have two properties for the intended fstype + access mode of
home dirs, but they might differ from what is actually used (because the
user record changed from the home dir, after it was created, or vice
versa). Let's hence add these props also to the "status" section of user
record, which report the status quo. That way we can always show the
correct, current settings.
homed: allow querying disk free status separetely from generating JSON from it
We later want to query per-home free status for implementing automatic
grow/shrink of home directories, hence let's separate the JSON
generation from the disk free status determination.
This adds to new helpers: keyring_read() for reading a key data from a
keyring entry, and TAKE_KEY_SERIAL which is what TAKE_FD is for fds, but
for key_serial_t.
The former is immediately used by ask-password-api.c
homed: add env var for overriding default mount options
This adds an esay way to override the default mount options to use for
LUKS home dirs via the env vars SYSTEMD_HOME_MOUNT_OPTIONS_EXT4,
SYSTEMD_HOME_MOUNT_OPTIONS_BTRFS, SYSTEMD_HOME_MOUNT_OPTIONS_XFS.
This follows what Fedora did with 34: enables compression by default,
lowering IO bandwidth and reducing disk space use, at the price of
slightly higher CPU use.
In delete_rule(), we already checked that the rule name is a valid file name
(i.e. no slashes), so we can just trivially append.
Also, let's always reject rules that we would later fail to delete. It's
probably better to avoid such confusion.
And print the operations we do with file name and line number. I hope this
helps with cases like https://github.com/systemd/systemd/pull/21178. At least
we'll know what rule failed.
$ sudo SYSTEMD_LOG_LEVEL=debug build/systemd-binfmt
Flushed all binfmt_misc rules.
Applying /etc/binfmt.d/kshcomp.conf…
/etc/binfmt.d/kshcomp.conf:1: binary format 'kshcomp' registered.
namespace: make tmp dir handling code independent of umask too
Let's make all code in namespace.c robust towards weird umask. This
doesn't matter too much given that the parent dirs we deal here almost
certainly exist anyway, but let's clean this up anyway and make it fully
clean.
namespace: make whole namespace_setup() work regardless of configured umask
Let's reset the umask during the whole namespace_setup() logic, so that
all our mkdir() + mknod() are not subjected to whatever umask might
currently be set.
This mostly moves the umask save/restore logic out of
mount_private_dev() and into the stack frame of namespace_setup() that
is further out.
Jonas Witschel [Thu, 11 Nov 2021 21:25:40 +0000 (22:25 +0100)]
test: add regression test for systemd-run --scope [--user]
systemd-run --scope --user failed to run in system 249.6, cf. #21297. Add tests
for systemd-run --scope and systemd-run --scope --user to make sure this does
not regress again.
Apparently version updates aren't always disabled on old forks,
which leads to new PRs opened there. To somewhat mitigate the
issue let's limit the number of PRs Dependabot can create.
It was reported in https://github.com/yuwata/systemd/pull/2#issuecomment-967737195
I think we should stick to the rule that stuff defined in
types-fundamental.h either:
1. adds a prefixed concept "sd_xyz" that maps differently in the two
environments
2. adds a non-prefixed concept "xyz" that adds a type otherwise missing
in one of the two environments but with the same definition as in the
other.
i.e. if have have some concept that might differ the way its set up in
the two environments it really should be prefixed by "sd_" to make clear
it has semantics we defined. Only drop the prefix if it really means the
exact same thin in all environments.
Now, sd_bool is defined prefixed, because its either mapped to "BOOLEAN"
(which is an integer) in UEFI or "bool" (which is C99 _Bool) in
userspace. size_t is not defined prefixed, because it's mapped to the
same thing ultimately (on the UEFI its mapped to UINTN, but that in turn
is defined as being the type for the size of memory objects, thus it's
really the same as userspace size_t).
So far "true" and "false" where defined unprefixed even though they map
to values of different types. typeof(true) in userspace would reveal
_Bool, but typeof(false) in UEFI would reveal BOOLEAN. The distinction
actually does matter in comparisons (i.e. (_Bool) 1 == (_Bool) 2 holds
while (BOOLEAN) 1 == (BOOLEAN) 2 does not hold).
Hence, let's add sd_true and sd_false, thus indicating we defined our
own concept here, and it has similar but different semantics in UEFI and
in userspace.
"type.h" is a very generic name, but this header is very specific to
making the "fundaemtnal" stuff work, it maps genric types in two
distinct ways. Hence let's make clear in the header name already what
this is about.
Unfortunately they forgot the "const" decoration on the MetaiMatch()
prototype, but let that omission not leak into our code, let's hide it
away in the innermost use.
projects / thirdparty / systemd.git / log
