Replace RwLock by a futex based one on Linux
This replaces the pthread-based RwLock on Linux by a futex based one.
This implementation is similar to [the algorithm](https://gist.github.com/kprotty/3042436aa55620d8ebcddf2bf25668bc) suggested by `@kprotty,` but modified to prefer writers and spin before sleeping. It uses two futexes: One for the readers to wait on, and one for the writers to wait on. The readers futex contains the state of the RwLock: The number of readers, a bit indicating whether writers are waiting, and a bit indicating whether readers are waiting. The writers futex is used as a simple condition variable and its contents are meaningless; it just needs to be changed on every notification.
Using two futexes rather than one has the obvious advantage of allowing a separate queue for readers and writers, but it also means we avoid the problem a single-futex RwLock would have of making it hard for a writer to go to sleep while the number of readers is rapidly changing up and down, as the writers futex is only changed when we actually want to wake up a writer.
It always prefers writers, as we decided [here](https://github.com/rust-lang/rust/issues/93740#issuecomment-1070696128).
To be able to prefer writers, it relies on futex_wake to return the number of awoken threads to be able to handle write-unlocking while both the readers-waiting and writers-waiting bits are set. Instead of waking both and letting them race, it first wakes writers and only continues to wake the readers too if futex_wake reported there were no writers to wake up.
r? `@Amanieu`
The current implementation uses a `hlt` instruction, which is the most
direct way to notify a connected debugger but is not the most flexible
way. This commit changes it to a call to the `abort` libc function,
making it possible for a system designer to override its behavior as
they see fit.
make windows compat_fn (crudely) work on Miri
With https://github.com/rust-lang/rust/pull/95469, Windows `compat_fn!` now has to be supported by Miri to even make stdout work. Unfortunately, it relies on some outside-of-Rust linker hacks (`#[link_section = ".CRT$XCU"]`) that are rather hard to make work in Miri. So I came up with this crude hack to make this stuff work in Miri regardless. It should come at no cost for regular executions, so I hope this is okay.
Cc https://github.com/rust-lang/rust/issues/95627 `@ChrisDenton`
Update libc to 0.2.121
With the updated libc, UNIX stack overflow handling in libstd can now
use the common `si_addr` accessor function, rather than attempting to
use a field from that name in `siginfo_t`. This simplifies the
collection of the fault address, particularly on platforms where that
data resides within a union in `siginfo_t`.
Don't cast thread name to an integer for prctl
`libc::prctl` and the `prctl` definitions in glibc, musl, and the kernel headers are C variadic functions. Therefore, all the arguments (except for the first) are untyped. It is only the Linux man page which says that `prctl` takes 4 `unsigned long` arguments. I have no idea why it says this.
In any case, the upshot is that we don't need to cast the pointer to an integer and confuse Miri.
But in light of this... what are we doing with those three `0`s? We're passing 3 `i32`s to `prctl`, which doesn't fill me with confidence. The man page says `unsigned long` and all the constants in the linux kernel are macros for expressions of the form `1UL << N`. I'm mostly commenting on this because looks a whole lot like some UB that was found in SQLite a few years ago: <https://youtu.be/LbzbHWdLAI0?t=1925> that was related to accidentally passing a 32-bit value from a literal `0` instead of a pointer-sized value. This happens to work on x86 due to the size of pointers and happens to work on x86_64 due to the calling convention. But also, there is no good reason for an implementation to be looking at those arguments. Some other calls to `prctl` require that other arguments be zeroed, but not `PR_SET_NAME`... so why are we even passing them?
I would prefer to end such questions by either passing 3 `libc::c_ulong`, or not passing those at all, but I'm not sure which is better.
Fix unsound `File` methods
This is a draft attempt to fix#81357. *EDIT*: this PR now tackles `read()`, `write()`, `read_at()`, `write_at()` and `read_buf`. Still needs more testing though.
cc `@jstarks,` can you confirm the the Windows team is ok with the Rust stdlib using `NtReadFile` and `NtWriteFile`?
~Also, I'm provisionally using `CancelIo` in a last ditch attempt to recover but I'm not sure that this is actually a good idea. Especially as getting into this state would be a programmer error so aborting the process is justified in any case.~ *EDIT*: removed, see comments.
With the updated libc, UNIX stack overflow handling in libstd can now
use the common `si_addr` accessor function, rather than attempting to
use a field from that name in `siginfo_t`. This simplifies the
collection of the fault address, particularly on platforms where that
data resides within a union in `siginfo_t`.
Windows: Synchronize asynchronous pipe reads and writes
On Windows, the pipes used for spawned processes are opened for asynchronous access but `read` and `write` are done using the standard methods that assume synchronous access. This means that the buffer (and variables on the stack) may be read/written to after the function returns.
This PR ensures reads/writes complete before returning. Note that this only applies to pipes we create and does not affect the standard file read/write methods.
Fixes#95411
libc::prctl and the prctl definitions in glibc, musl, and the kernel
headers are C variadic functions. Therefore, all the arguments (except
for the first) are untyped. It is only the Linux man page which says
that prctl takes 4 unsigned long arguments. I have no idea why it says
this.
In any case, the upshot is that we don't need to cast the pointer to an
integer and confuse Miri.
