Add a hash generation helper for NMEtherAddr struct. This can be used
for HashTables containing pointers to NMEtherAddr structs.
(cherry picked from commit a63eec924c)
(cherry picked from commit 6371802087)
(cherry picked from commit 73aa5b47fa)
GCC 14 complans with:
src/libnm-glib-aux/nm-uuid.c: In function 'nm_uuid_generate_from_strings_strv':
src/libnm-glib-aux/nm-uuid.c:492:12: error: '_1' may be used uninitialized [-Werror=maybe-uninitialized]
492 | return nm_uuid_generate_from_string_str(s, slen, uuid_type, type_args);
| ^
src/libnm-glib-aux/nm-uuid.c:392:1: note: by argument 1 of type 'const char *' to 'nm_uuid_generate_from_string_str' declared here
392 | nm_uuid_generate_from_string_str(const char *s,
| ^
"-Wmaybe-uninitialized" diagnoses passing pointers or references to
uninitialized memory to functions taking const-qualified arguments.
In this case, nm_uuid_generate_from_string_str()'s first argument is a
"const char *" and so the compiler expects that the string is always
initialized. However, it is not initialized when len is zero.
A non-null zero-length array can be specified in two ways: by setting
len to zero, or by setting len to -1 and having NULL as first
element. Handle both cases in the same way.
(cherry picked from commit 2386c0f52d)
It's simply not valid to read the ref-count without an atomic.
The compiler might optimize out the assignment to "r" and read the
_ref_count field multiple times. Thereby, we might at first appear
to be larger than > 1, and later pass 1 to compare-and-exchange.
We need an atomic get here.
Fixes: 19d4027824 ('refstr: inline nm_ref_string_{ref,unref}()')
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/1847
(cherry picked from commit 5f7a027f59)
There is g_variant_equal(), which can handle all variant types (however
that is not a compare function).
There is g_variant_compare(), which is a compare function but only works for
basic types.
Add nm_g_variant_cmp() which works with all variant types.
This is based on nm_property_compare(), with some differences:
- nm_property_compare() tries (wrongly) to accept string dictionaries in
any order. That functionality seems wrong, and nm_g_variant_cmp()
doesn't do that.
- nm_property_compare() does possibly not support all variant types.
This can be a problem, if we call the function on untrusted data
(and it can be hard to validate first, whether the function can
be called with a particular variant). Instead, nm_g_variant_cmp()
should work with all variants.
The unit tests are copied from "src/libnm-core-impl/tests/test-compare.c"
with some adjustments (because nm_property_compare() is not the same as
nm_g_variant_cmp()).
Note that the code is actually unused. It was written as replacement for
nm_property_compare(), but turns out not to be used there. For now,
leave it, because it might still be useful to have in the toolbox and it
exists (including tests).
Rename _nm_gobject_notify_together_impl##suffix() to
_nm_gobject_notify_together_full_v##suffix(). This name makes a bit more
sense. The "_v" suffix indicates that this takes an array of properties.
Also, commonly, when we have an array and a length parameter, the array comes
first. Reorder the arguments.
A SSID of zero length, really looks "empty". Let
nm_utils_is_empty_ssid() indicate so too.
This affects some places, that try to guess what a hidden SSID looks
like. In general, it seems that treating a length of zero as empty, is
suitable also then.
We have nm_gobject_notify_together(), which accepts a list of
_PropertyEnums arguments. Add nm_gobject_notify_together_by_pspec(),
which can use a param spec.
"nm-glib" h contains compat wrappers for older glib versions. This file
used to be copied over to VPN plugins, to use the same compat code. It
was thus interesting to also have compat code for glib versions, that
were no longer supported by NetworkManager itself.
This was fine. But glib 2.42 is more than 8 years old. At this point,
there really is no need to support that, even if you copy the file out
of NetworkManager source tree.
Drop those compat wrappers.
This patch add support to HSR/PRP interface. Please notice that PRP
driver is represented as HSR too. They are different drivers but on
kernel they are integrated together.
HSR/PRP is a network protocol standard for Ethernet that provides
seamless failover against failure of any network component. It intends
to be transparent to the application. These protocols are useful for
applications that request high availability and short switchover time
e.g electrical substation or high power inverters.
https://gitlab.freedesktop.org/NetworkManager/NetworkManager/-/merge_requests/1791
Unfortunately, there are several possibilities how to handle NULL and
empty arrays. Therefore we have different variants.
Clean this up, and add a way to preserve whether the array is empty
(previous variants could not distinguish that).
Functions are also renamed, so that if you backport a user of the new
API, you'll get a compiler error if this patch is missing.
Also, nm_strvarray_get_strv_notnull() no longer takes a pointer to a
"GArray*". Previously, it used that to fake an empty strv array. Now
this returns NM_STRV_EMPTY_CC().
The check "sizeof(const char *const *) ==
g_array_get_element_size((GArray *) strv)" is wrong, but probably
harmless, because most likely on our supported architectures all pointer
sizes are the same size.
Also, just use `sizeof(char *)` instead of `sizeof(const char *)`. Not
that it matters, but the GArray holds pointers of `char *`.
Also, consistently place the "sizeof()" on the left side of the
comparison.
NM variants:
- evaluate arguments only once
- have a static assertion that the signedness of the argument agrees.
Like MIN()/MAX(), NM_MIN()/NM_MAX() now also evaluate to a constant
expression, if the arguments are already constant. That means, the only
reason why MIN()/MAX() was preferable over NM_MIN()/NM_MAX() is no
longer relevant. Except there are a few places where NM_MIN()/NM_MAX()
cannot be used. In those places use NM_MIN_CONST()/NM_MAX_CONST().
glib's MIN()/MAX() will be replaced by NM_MIN()/NM_MAX().
There are however a few places where NM_MIN()/NM_MAX() cannot
be used.
Adjust those places to use NM_MIN_CONST()/NM_MAX_CONST() instead.
Glib's MIN()/MAX() should not be used, in favor of NM_MIN()/NM_MAX().
That's because the NM variants
- evaluate arguments only once
- have a static assertion that the signedness of the arguments matches
However, previously those macros never evaluated to a compile time
constant. Unlike the glib variants, which do so when the arguments are
compile time constants. That is sometimes important when using the
macros in a context that requires a constant.
Extend NM_MIN()/NM_MAX() to be a compile time constant, when possible.
Note that there are still a few places where NM_MIN()/NM_MAX() cannot be
used due to the expression statement. For those cases, there is
NM_MIN_CONST()/NM_MAX_CONST().
nm_hash_siphash42() uses a randomized seed like nm_hash*(). In this case,
we want to always generate the same fake timestamp, based on the host-id.
In practice, it doesn't really matter, because this is only the fallback
path for something gone horribly wrong already.
c_siphash_init() requires a 16 bytes array. That is cumbersome to use.
We have NM_HASH_SEED_16() macro for helping with that. It's still
cumbersome.
Most of the time, the caller just wants to pick an arbitrarily chosen,
fixed number. Add NM_HASH_SEED_16_U64() which takes a number and gives
a 16 seed array. The argument is in host endianness, but the resulting
seed array has it encoded in big endianness, to be architecture
independent.
Calling getpwuid_r() is cumbersome, because it has a separate passwd and
string buffer, and you shall retry, when the buffer is too small.
Extract nm_getpwuid() for that. This one always allocates a suitable
buffer, that the caller can free.
This will allow callers to get the full passwd struct. It will also
allow callers to avoid the additional strdup() of nm_utils_uid_to_name(),
when we don't need a clone of the string.
nm_g_array_index() performs additional nm_assert() checks for
correctness.
In this case, it's pretty clear that the assertion will hold and that
the code is correct. Note that most of the time when having assertions,
we expect that they hold. Since nm_assert() is disabled in release
build, arguing that an assertion holds is not a strong argument against
having the assert (they are always supposed to hold, quite obviously so!).
The reason to change is that we should use the wrappers that perform
additional checks. Especially when the additional checks are nm_assert()
or static-asserts, as they are not present in release builds. To find
how well we are doing in this regard we can check `git grep -w
g_array_index`. If that gives many uses of the unchecked function, then
we cannot manually check them all to be really obviously correct.
Instead, we should not use g_array_index() and trivially see that all
array accesses are guarded by assertions.
"checkpatch.pl" also recommends against g_array_index().
"skip_repeated" sounds as if the function would only drop duplicate
elements that follow each other (in which case, the operation would be
O(n)). But it does search the entire array to prevent duplicates (resulting
in O(n^2)). Rename the argument "skip_repeated" to "no_duplicates"
to make that clearer.
Also, rename "skip_{empty,duplicates}" to "no_{empty,duplicates}". The
function removes those elements from the list, so "skip" is a bit
misleading too.
nm_strv_find_first() is useful (and used) to find the first index (if
any). I can thus also used to check for membership.
However, we also have nm_strv_contains(), which seems better for
readability, when we check for membership. Use it.
nm_streq() is better for readability. Prefer it over strcmp(). Note that
nm_streq() will be inlined, so it should make no difference performance
wise.
While at it, drop wrong comment.
sysfs is deprecated and kernel will not add new bridge port options to
sysfs. Netlink is a stable API and therefore is the right method to
communicate with kernel in order to set the link options.
Show a new field called "Valid values" in those properties that only
accept a limited set of values, like enums, ints with a valid range of
values, etc.
As there is some complex logic behind getting this information, this
logic has been put in nm-meta-setting-desc and nm-enum-utils so they can
be re-used, avoiding duplicity and errors. Some refactor has been done
in nm-meta-setting-desc in this direction, too.
We have nm_strerror_native_r(), which is the wrapper around strerror_r() that
we want to use in glib components (it also will ensure that the string is valid
UTF-8). However, it's not usable from non-glib components.
Move the part that abstracts strerror_r() out to libnm-std-aux as _nm_strerror_r().
The purpose is that non-glib componenent can use the thread-safe wrapper around
strerror_r().
Kernel's dev_valid_name() calls isspace(), which also rejects '\v'
and '\240'.
As this tightens the check, the change can break code that partly worked
before. It surely didn't work to the point, where an interface with such
name could be created in kernel.
# ip link add name $'foo\240bar' type dummy
RTNETLINK answers: Invalid argument
This is the version shipped in Fedora 38. As Fedora 38 is now out, the
core developers switch to it. Our gitlab-ci will also use that as base
image for the check-{patch.tree} tests and to generate the pages. There
is a need that everybody agrees on which clang-format version to use,
and that version should be the one of the currently used Fedora release.
Also update the used Fedora image in "contrib/scripts/nm-code-format-container.sh"
script.
The gitlab-ci still needs update in the following commit. This change
in isolation will break the "check-tree" test.
