| Date | Commit message (Collapse) |
|---|
|
This fixes compile errors on platforms we can't explicitly
support from pure Perl due to the lack of syscall stability
guarantees by the OS developers.
Reported-by: Gaelan Steele <gbs@canishe.com>
Tested-by: Gaelan Steele <gbs@canishe.com>
|
|
While syscall symbols (e.g. SYS_*) have changed on us in FreeBSD
during the history of Sys::Syscall and this project and did bite
us in some cases; the actual numbers don't get recycled for new
syscalls. We're also fortunate that sendmsg and recvmsg syscalls
and associated msghdr and cmsg structs predate the BSD forks and
are compatible across all the BSDs I've tried.
OpenBSD routes Perl `syscall' through libc; while NetBSD + FreeBSD
document procedures for maintaining backwards compatibility.
It looks like Dragonfly follows FreeBSD, here.
Tested on i386 OpenBSD, and amd64 {Free,Net,Open,Dragonfly}BSD
This enables *BSD users to use lei, -cindex and future SCM_RIGHTS-only
features without needing Inline::C.
[1] https://cvsweb.openbsd.org/src/gnu/usr.bin/perl/gen_syscall_emulator.pl
[2] https://www.netbsd.org/docs/internals/en/chap-processes.html#syscall_versioning
[3] https://wiki.freebsd.org/AddingSyscalls#Backward_compatibily
|
|
Sys::Syscall needs separate patches anyways (if it ever gets
updated), and having a mix of indentation styles in our codebase
gets confusing. We'll also update cfarm-related comments for
the current URL.
|
|
This is a step towards improving the out-of-the-box experience
in achieving notifications without XS, extra downloads, and .so
loading + runtime mmap overhead.
This also fixes loongarch support of all Linux syscalls due to
a bad regexp :x
All the reachable Linux architectures listed at
<https://portal.cfarm.net/machines/list/> should be supported.
At the moment, there appears to be no reachable sparc* Linux
machines available to cfarm users.
Fixes: b0e5093aa3572a86 (syscall: add support for riscv64, 2022-08-11)
|
|
We use this in various places to minimize or maximize pipe
size on Linux. So keep it all in one place.
|
|
This ensures script/lei $send_cmd usage is EINTR-safe (since
I prefer to avoid loading PublicInbox::IPC for startup time).
Overall, it saves us some code, too.
|
|
We need to allocate CMSG_SPACE for the `struct cmsghdr', not the
smaller CMSG_LEN. AFAIK this isn't a real world problem since
the Linux kernel doesn't care about the uninitialized space as
long as memory region belongs to the user, but valgrind complains.
|
|
Instead of converting to bytes to bits and asking `vec' to
operate on single bits, we can just have `vec' work on 8 bits
at-a-time.
This also fixes an overallocation in pure Perl Linux recv_cmd4.
Adding an extra byte ourselves for "\0" isn't necessary: Perl
already does it internally everywhere when creating/resizing
scalars.
|
|
Handling this should be done at the lowest levels possible;
so away from higher-level lei code.
|
|
This allows us to cut down on imports and reduce code.
This also makes it easier (in the next commit) to provide an option
to disable epoll/kqueue when saving an FD is valued over scalability.
|
|
The awaitpid API turns out to be quite handy for managing
long-lived worker processes. This allows us to ensure all our
uses of signalfd (and kevent emulation) are non-blocking.
|
|
We use it to dump SIGWINCH and _SC_NPROCESSORS_ONLN, so
"sysdefs" is a more appropriate list for *BSD users.
|
|
It's an informative message that's harmless, so hopefully
the `#' prefix puts the users mind at ease.
(I saw it on an `lei import' against an IMAP source)
|
|
Both __ILP32__ and __x86_64__ need to be defined for a system to
be considered x32. Without this, my 32-bit Debian VM on a
64-bit kernel would fail after upgrading to Perl 5.32.1 on
Debian 11 (bullseye).
|
|
h2ph-generated *.ph files are often wrong or incomplete and IME
they cause more problems than they solve. Furthermore, we need
knowledge of struct layouts which h2ph-generated files can't get
us. So trim down some bloat and leave a note for porters.
|
|
We can just check defined() on the `our' var itself and
save the process several kilobytes of memory.
|
|
SIGWINCH is actually different for these architectures on Linux
according to the signal(7) man page.
Note: AFAICS there's no parisc machine in the GCC Farm[1],
so it remains untested. I've only tested mips64 for mips,
but I expect them to both work.
OpenBSD (on gcc231) octeon defines SIGWINCH as the common `28',
so it appears Linux is the only one with arch-dependent signal
numbers (ditto with syscalls).
[1] https://cfarm.tetaneutral.net/machines/list/
|
|
For common x86-64 systems, we can avoid a needless
string comparison on `mips64' by restructuring the
branches for architecture detection.
|
|
This is needed for older Perls (tested perl 5.16.3 on CentOS 7).
|
|
Tested on gcc92.fsffrance.org from cfarm.
|
|
Socket.pm still loads strict.pm, unfortunately, which hurts
startup time; but we'll save some LoC this way.
|
|
While `vec' is useful for user-supplied buffers to avoid excess
memory traffic, but provides no benefit when we need to allocate
our own buffers as we do in nodatacow_fh, since Perl can't elide
memset(ptr, 0, len). So just use the idiomatic `"\0" x $LEN' here.
|
|
Since we know the space required under Linux, we can use the
same initialization as the Inline::C version instead of
hard-coding 256 as we do for Socket::MsgHdr.
|
|
aarch64, ppc64le, sparc64, loongarch64, and mips (32-bit userspace)
are all tested via machines from the GCC Farm Project
<https://cfarm.tetaneutral.net/>
Remaining syscall numbers are from musl <https://musl.libc.org/>
|
|
Socket::MsgHdr is only packaged for Debian and derivatives at
the moment, and Inline::C pulling in gcc/clang is a huge amount
of disk space and bandwidth for some users.
This enables disk space and/or bandwidth-limited users to use lei.
Only Linux guarantees a stable ABI and syscall numbers, but
that's the majority of our userbase. FreeBSD users will still
have to use Inline::C (or get Socket::MsgHdr packaged).
x86, x32, and x86-64 are all currently supported, more to be added.
|
|
We've never used it, actually.
|
|
It turns out these Linux ioctls are unfortunately
architecture-dependent, and not endian-dependent.
Fixup some warning messages while we're at it, too.
Fixes: 14fa0abdcc7b6513 ("rewrite Linux nodatacow use in pure Perl w/o system")
Link: https://public-inbox.org/meta/YfdYqLhDVQRQ9NGT@codewreck.org/
Noticed-by: Dominique Martinet <asmadeus@codewreck.org>
|
|
ZFS appears to incorrectly return EINVAL on renameat2 when the operation is not
supported:
renameat2(AT_FDCWD, "...", AT_FDCWD, "...", RENAME_NOREPLACE) = -1 EINVAL
Fall back to the racy rename in this case as well:
|
|
btrfs is Linux-only at the moment (and likely to remain that way
for practical purposes). So rely on Linux ABI stability and use
the `syscall' and `ioctl' perlops rather than relying on Inline::C.
Inline::C (and gcc||clang) are monstrous dependencies which we
can't expect users to have.
This makes supporting new architectures more difficult, but new
architectures come along rarely and this reduces the burden for
the majority of Linux users on popular architectures (while
still avoiding the distribution of pre-built binaries).
Link: https://public-inbox.org/meta/YbCPWGaJEkV6eWfo@codewreck.org/
|
|
One syscall is better than two for atomicity in Maildirs. This
means there's no window where another process can see both the
old and new file at the same time (link && unlink), nor a window
where we might inadvertantly clobber an existing file if we were
to do `stat && rename'.
|
|
Since signalfd is often combined with our event loop, give it a
convenient API and reduce the code duplication required to use it.
EventLoop is replaced with ::event_loop to allow consistent
parameter passing and avoid needlessly passing the package name
on stack.
We also avoid exporting SFD_NONBLOCK since it's the only flag we
support. There's no sense in having the memory overhead of a
constant function when it's in cold code.
|
|
We'll be supporting inotify directly as we do with epoll so so
Linux users won't have to deal with XS, extra DSOs or install
Linux::Inotify2 (and common::sense) modules.
|
|
FreeBSD (and other *BSDs) do not have stable syscall numbers, so
drop no-op checks for it and add a note to use Inline::C,
instead. Drop an implicit return for the syscall.ph loading
while we're at it, too.
On Linux, epoll_create(2) ignores the size arg since Linux
2.6.8, so just hard code it to some non-zero value.
On a side note, we can probably drop epoll_create(2) support
soon and just use epoll_create1(2) which appeared in 2.6.27+
(2008-10-09). Our userspace (Perl and git) requirements are
already further ahead.
|
|
Older Perls (tested 5.16.3) would warn on uninitialized scalars while
newer (tested 5.28.1) do not. Just initialize it to an empty string
since it'll be filled in by `vec'.
|
|
Using "make update-copyrights" after setting GNULIB_PATH in my
config.mak
|
|
Since Perl exposes O_NONBLOCK as a constant, we can safely make
SFD_NONBLOCK a constant, too. This is not the case for
SFD_CLOEXEC, since O_CLOEXEC is not exposed by Perl despite
being used internally in the interpreter.
|
|
Consistently returning the equivalent of pollfd.revents in a
portable manner was never worth the effort for us, as we use the
same ->event_step callback regardless of POLLIN/POLLOUT/POLLHUP.
Being a Perl, @events knows it size and we don't have to return
a maximum index for the caller to iterate on.
We can also avoid redundant integer coercion ("+0") since we
ensure everything is an IV in other places.
Finally, vec() is preferable to ("\0" x $size) for resizing
buffers because it only needs to write the extended portion
and not overwrite the entire buffer.
|
|
Thanks to the GCC compile farm project, we can wire up syscalls
for sparc64 and set system-specific SFD_* constants properly.
I've FINALLY figured out how to use POSIX::SigSet to generate
a usable buffer for the syscall perlfunc. This is required
for endian-neutral behavior and relevant to sparc64, at least.
There's no need for signalfd-related stuff to be constants,
either. signalfd initialization is never a hot path and a stub
subroutine for constants uses several KB of memory in the
interpreter.
We'll drop the needless SEEK_CUR import while we're importing
O_NONBLOCK, too.
|
|
The x32 ABI allows users to take advantage of the extra
registers on x86-64 without the bloat of 64-bit pointers and
longs.
This ought to be significant since Perl was designed when 32-bit
was prevalent; and the common structs for ops, hashes, scalars,
and arrays use longs (SSize_t/Size_t) for things which should
never need 64-bits when processing emails.
Debian's x32 port seems to work quite nicely under a chroot
on an amd64 Linux system. All tests pass under x32, now.
|
|
I didn't wait until September to do it, this year!
|
|
"use vars" was superseded by "our" in Perl 5.6, and we
can "use parent qw(Exporter)" in favor of manipulating
@ISA directly (or the bigger "use base ...");
While we're at it, avoid multiple invocations of constant->import
by passing a hashref as a "use" parameter.
|
|
Our attempt at using a self-pipe in signal handlers was
ineffective, since pure Perl code execution is deferred
and Perl doesn't use an internal self-pipe/eventfd. In
retrospect, I actually prefer the simplicity of Perl in
this regard...
We can use sigprocmask() from Perl, so we can introduce
signalfd(2) and EVFILT_SIGNAL support on Linux and *BSD-based
systems, respectively. These OS primitives allow us to avoid a
race where Perl checks for signals right before epoll_wait() or
kevent() puts the process to sleep.
The (few) systems nowadays without signalfd(2) or IO::KQueue
will now see wakeups every second to avoid missed signals.
|
|
I'm not sure they'll make a measurable difference or will
be worth the effort in the future given the prevalance
of HTTPS and giant socket buffers.
Using Inline::C for this may make more sense in the
future, too, especially if we want to be able to use
GnuTLS.
|
|
We don't need extra wakeups from the kernel when we know a
listener is already active.
|
|
We don't need to code multiple event loops or have branches in
watch() if we can easily make the IO::KQueue-based interface
look like our lower-level epoll_* API.
|
|
We don't need to keep track of that field since we always
know what events we're interested in when using one-shot
wakeups.
|
|
We don't need to keep information from uname(2) around outside
of startup.
|
|
EPOLLRDBAND is used for DECnet; and I'm pretty sure I won't be
updating any of our code to work with DECnet.
I've never found use for EPOLLHUP or EPOLLERR, either; so
disable those for now and add comments for things I might
actually use: EPOLLET and EPOLLONESHOT.
|
|
No backwards compatibility to worry about for us; and fadvise
is superior anyways.
|
|
Since our listen sockets are non-blocking and we may run
multiple httpd|nntpd processes; we need a way to avoid
thundering herds when there are multiple httpd|nntpd worker
processes.
EPOLLEXCLUSIVE was added just for that in Linux 4.5
|