Now that the necessary functions from log.h have been moved to macro.h,
we can stop including log.h in macro.h. This requires modifying source
files all over the tree to include log.h instead.
On musl, O_ACCMODE is defined as (03|O_SEARCH), unlike glibc which
defines it as (O_RDONLY|O_WRONLY|O_RDWR). Additionally, O_SEARCH is
simply defined as O_PATH.
This causes problems for systemd on musl, as it changes the
behaviour of open_mkdir_at_full() to return -EINVAL if O_PATH is
included in flags due to the fact that O_ACCMODE includes O_SEARCH
(i.e. O_PATH). Consequently, this makes the test-fs-util test fail.
Upstream musl seems content with this behaviour and doesn't seem
interested in matching glibc's behaviour due to that defining it this
way allows for O_SEARCH to match POSIX better by allowing it to open
directories where read permission is missing. Apparently musl does some
emulation in other places to make this work more consistently as well.
Initially I took the approach of working around this by redefining
O_SEARCH as O_RDONLY if O_SEARCH == O_PATH. This fixes the test and is
the approach taken by both XZ[1] and Gzip[2][3], but was not taken as
redefining system headers potentially could be problematic.
Instead, introduce O_ACCMODE_STRICT which just is a copy of glibc's
O_ACCMODE and use it everywhere. This way we don't have to deal with
unusual definitions of O_ACCMODE from C standard libraries other than
glibc.
[1]: https://git.tukaani.org/?p=xz.git;a=blob;f=src/xz/file_io.c;h=8c83269b13fa31284f7ea5f3627a1dfbce7d6e14;hb=HEAD#l72
[2]: https://git.savannah.gnu.org/cgit/gnulib.git/tree/lib/fcntl.in.h
(lines 380 and 396, commit d7f551b30f3f2a0fa57c1b10c12f4eea41a9b89e)
[3]: https://lists.gnu.org/archive/html/bug-gzip/2025-01/msg00000.html
To resolve conflict with sys/mount.h and linux/mount.h or linux/fs.h.
The conflict between sys/mount.h and linux/mount.h is resolved in
glibc-2.37 (774058d72942249f71d74e7f2b639f77184160a6), but our baseline
is still glibc-2.31. Also, even with the version or newer, still
sys/mount.h conflicts with linux/fs.h, which is included by
linux/btrfs.h.
This introduces our own implementation of sys/mount.h, that can be
simultaneously included with linux/mount.h and linux/fs.h. This also
imports linux/fs.h, linux/mount.h, and several other dependent headers.
The introduced sys/mount.h header itself may not be enough simple, but
by using the header, we can drop most of workarounds in other source files.
struct statx in glibc header was introduced in glibc-2.28
(fd70af45528d59a00eb3190ef6706cb299488fcd), but at that time,
sys/stat.h conflicts with linux/stat.h. Since glibc-2.30
(5dad6ffbb2b76215cfcd38c3001778536ada8e8a), sys/stat.h includes
linux/stat.h if exists.
Since now our baseline of glibc is 2.31. Hence, we can drop workarounds
for struct statx by importing linux/stat.h from newer kernel (v6.14-rc4).
Catch up with the nice little toys the kernel fs developers have added
for us. Preferably, let's make use of the new F_DUPFD_QUERY fcntl() call
that checks whether two fds are just duplicates of each other
(duplicates as in dup(), not as in open() of the same inode, i.e.
whether they share a single file offset and so on).
This API is much nicer, since it is a core kernel feature, unlike the
kcmp() call we so far used, which is part of the (optional)
checkpoint/restore stuff.
F_DUPFD_QUERY is available since kernel 6.10.
fd_cloexec_many promised to report if work was done, but that code was
not effective, because it always reported true if any fds were open.
But no callers care about the return value, so let's just drop this.
Currently, if proc_mounted() != 0, some functions
propagate -ENOENT while others return -EBADF.
Let's make things consistent, by introducing
a static inline helper responsible for finding out
the appropriate errno.
We may want to propagate O_APPEND, or (try to) keep the current file position,
even if we use fd_reopen() to re-initialize (and "unshare") other file
description status.
For now, used only with --pty to keep/propagate O_APPEND (and/or) position
if set on stdin/stdout.
If we re-open stdout and "drop" the O_APPEND,
we get rather "unexpected" behavior,
for example with repeated "systemd-run --pty >> some-log".
If someone carefully pre-positioned the passed in original file descriptors,
we avoid surprises if we do not reset file postition to zero.
fcntl F_GETFL first, and propagate O_APPEND if present in the existing flags.
Then use lseek to propagate the file position.
Let's make fd_verify_safe_flags() even more useful:
1. let's return the cleaned up flags (i.e. just the access mode) after
validation, hiding all the noise, such as O_NOFOLLOW, O_LARGEFILE and
similar.
2. let's add a "full" version of the call that allows passing additional
flags that are OK to be set.
This is useful for situations where an array of FDs is to be passed into
a child process (i.e. by passing it through safe_fork). This function
can be called in the child (before calling exec) to pack the FDs to all
be next to each-other starting from SD_LISTEN_FDS_START (i.e. 3)
So glibc exposes a close_range() syscall wrapper now, but they decided
to use "unsigned" as type for the fds. Which is a bit weird, because fds
are universally understood to be "int". The kernel internally uses
"unsigned", both for close() and for close_range(), but weirdly,
userspace didn't fix that for close_range() unlike what they did for
close()... Weird.
But anyway, let's follow suit, and make our wrapper match glibc's.
Fixes#31270
O_CREAT doesn't make sense for fd_reopen, since we're
working on an already opened fd. Also, in fd_reopen
we don't handle the mode parameter of open(2), which
means we may get runtime error like #29938.
This is just like FORMAT_PROC_FD_PATH() but goes via the PID number
rather than the "self" symlink.
This is useful whenever we want to generate a path that is useful
outside of our local scope.
We use it for more than just pipe() arrays. For example also for
socketpair(). Hence let's give it a generic name.
Also add EBADF_TRIPLET to mirror this for things like
stdin/stdout/stderr arrays, which we use a bunch of times.
This is supposed to be a help for compilers to apply optimizations on
functions where they can't determine whether they are const/pure on
their own. For static, local functions the compiler can do this on its
own easily however, hence the decoration with pure/const is just noise.
Let's drop it, and let the compiler to its thing better.
(Use it for exported functions, since compilers can't 'reach-over' into
other modules to determine if they are pure, except if LTO is used)
Previously, in path_is_root_at(), if statx() does not provide mount ID,
path_get_mnt_id_at() was called, but it also calls statx(). Let's avoid
the second trial.
Now, dir_fd_is_root() is heavily used in chaseat(), which is used at
various places. If the kernel is too old and /proc is not mounted, then
there is no way to get the mount ID of a directory. In that case, let's
silently skip the mount ID check.
Fixes https://github.com/systemd/systemd/pull/27299#issuecomment-1511403680.
The /proc/self/fd/ interface cannot be used to follow symlinks pinned
via O_PATH. Add a comment + test for that. Moreover, using fd_reopen()
with O_NOFOLLOW cannot work. Add an explicit check and test for that, to
make behaviour uniform.
Daan De Meyer
Yu Watanabe
Stefan Hansson
Mike Yuan
Lennart Poettering
Zbigniew Jędrzejewski-Szmek
Lars Ellenberg
Adrian Vovk
Luca Boccassi
David Tardon