Refactor stack overflow handling

[joboet]

Currently, every platform must implement a Guard that protects a thread from stack overflow. However, UNIX is the only platform that actually does so. Windows has a different mechanism for detecting stack overflow, while the other platforms don't detect it at all. Also, the UNIX stack overflow handling is split between sys::pal::unix::stack_overflow, which implements the signal handler, and sys::pal::unix::thread, which detects/installs guard pages.

This PR cleans this by getting rid of Guard and unifying UNIX stack overflow handling inside stack_overflow (commit 1). Therefore we can get rid of sys_common::thread_info, which stores Guard and the current Thread handle and move the thread::current TLS variable into thread (commit 2).

The second commit is not strictly speaking necessary. To keep the implementation clean, I've included it here, but if it causes too much noise, I can split it out without any trouble.

[rustbot]

r? @ChrisDenton

rustbot has assigned @ChrisDenton.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[ChrisDenton]

I'm very much in favour of containing platform specific stuff within platform specific directories. I'm less opinionated on where exactly to put the guard page handling but, sure, stack_overflow.rs makes sense to me.

Anyways this looks good to me, just a small nit

[ChrisDenton]

@bors r+

[bors]

📌 Commit d7b55e4 has been approved by ChrisDenton

It is now in the queue for this repository.

[bors]

⌛ Testing commit d7b55e4 with merge c518e5a...

[bors]

☀️ Test successful - checks-actions
Approved by: ChrisDenton
Pushing c518e5a to master...

[rust-timer]

Finished benchmarking commit (c518e5a): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	7.8%	[7.8%, 7.8%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	7.8%	[7.8%, 7.8%]	1

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.3%	[2.3%, 2.3%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Binary size

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.4%	[0.0%, 1.2%]	13
Regressions ❌ (secondary)	1.1%	[0.2%, 1.4%]	39
Improvements ✅ (primary)	-0.0%	[-0.0%, -0.0%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.4%	[-0.0%, 1.2%]	14

Bootstrap: 667.497s -> 668.277s (0.12%)
Artifact size: 315.72 MiB -> 315.71 MiB (-0.00%)

[Kobzol]

The binary size regressions are for actual binaries, not just metadata, which is a bit unfortunate. Was this expected?

[joboet]

That is extremely weird, because this PR didn't really add any new code, it mainly just moved things around.

[Kobzol]

./target/release/collector binary_stats \
    +6bb6b816bfcf9e26fac5175e3e907dcefe5ecdbc \
    --rustc2 +c518e5aeecd3421a8ed5acb96946badb217dc64f \
    --include helloworld \
    --profile Opt \
    --backend Llvm
┌─────────────────────┬───────────────┬──────────────┬───────┬──────────┐
│ Section             │ Size (before) │ Size (after) │  Diff │ Diff (%) │
├─────────────────────┼───────────────┼──────────────┼───────┼──────────┤
│ .text               │    245.96 KiB │   246.67 KiB │  +720 │    +0.3% │
│ .strtab             │     47.13 KiB │    47.81 KiB │  +700 │    +1.5% │
│ .rela.dyn           │     17.04 KiB │    17.44 KiB │  +408 │    +2.3% │
│ .data.rel.ro        │      8.95 KiB │     9.23 KiB │  +280 │    +3.1% │
│ .eh_frame           │     21.14 KiB │    21.39 KiB │  +256 │    +1.2% │
│ .symtab             │     17.46 KiB │    17.70 KiB │  +240 │    +1.3% │
│ .rodata             │     24.20 KiB │    24.03 KiB │  -176 │    -0.7% │
│ .eh_frame_hdr       │      3.89 KiB │     3.95 KiB │   +56 │    +1.4% │
│ .dynstr             │      1.02 KiB │     1.04 KiB │   +26 │    +2.5% │
│ .dynsym             │      1.57 KiB │     1.59 KiB │   +24 │    +1.5% │
│ .got                │      1.67 KiB │     1.69 KiB │   +16 │    +0.9% │
│ .gcc_except_table   │      4.06 KiB │     4.07 KiB │   +12 │    +0.3% │
│ .bss                │         216 B │        208 B │    -8 │    -3.7% │
│ .tbss               │          80 B │         73 B │    -7 │    -8.8% │
│ .gnu.version        │         134 B │        136 B │    +2 │    +1.5% │
│ <17 unchanged rows> │      1.59 KiB │     1.59 KiB │     0 │     0.0% │
│─────────────────────│───────────────│──────────────│───────│──────────│
│ Total               │    396.12 KiB │   398.61 KiB │ +2549 │    +0.6% │
└─────────────────────┴───────────────┴──────────────┴───────┴──────────┘

┌───────────────────────────────────────────────────────────────┬───────────────┬──────────────┬──────┬──────────┐
│ Symbol                                                        │ Size (before) │ Size (after) │ Diff │ Diff (%) │
├───────────────────────────────────────────────────────────────┼───────────────┼──────────────┼──────┼──────────┤
│ std::sys::pal::unix::stack_overflow::imp::make_handler        │         464 B │        903 B │ +439 │   +94.6% │
│ core::cell::once::OnceCell<T>::get_or_try_init::outlined_call │           0 B │        382 B │ +382 │  +100.0% │
│ std::sys::pal::unix::thread::Thread::get_name                 │         371 B │          0 B │ -371 │  -100.0% │
│ std::sys_common::thread_info::set                             │         323 B │          0 B │ -323 │  -100.0% │
│ std::sys::pal::unix::stack_overflow::imp::signal_handler      │         685 B │        420 B │ -265 │   -38.7% │
│ std::sys_common::thread_info::current_thread                  │         259 B │          0 B │ -259 │  -100.0% │
│ std::thread::try_current                                      │           0 B │        258 B │ +258 │  +100.0% │
│ std::thread::set_current                                      │           0 B │        249 B │ +249 │  +100.0% │
│ core::fmt::builders::DebugStruct::finish_non_exhaustive       │           0 B │        182 B │ +182 │  +100.0% │
│ <std::thread::Thread as core::fmt::Debug>::fmt                │           0 B │        158 B │ +158 │  +100.0% │
│ <&T as core::fmt::Debug>::fmt                                 │         465 B │        558 B │  +93 │   +20.0% │
│ <core::option::Option<T> as core::fmt::Debug>::fmt            │           0 B │         80 B │  +80 │  +100.0% │
│ std::sys_common::thread_info::THREAD_INFO::__getit::destroy   │          68 B │          0 B │  -68 │  -100.0% │
│ std::thread::CURRENT::__getit::destroy                        │           0 B │         63 B │  +63 │  +100.0% │
│ <std::thread::ThreadId as core::fmt::Debug>::fmt              │           0 B │         53 B │  +53 │  +100.0% │
│ std::sys_common::thread_info::THREAD_INFO::__getit::VAL       │          32 B │          0 B │  -32 │  -100.0% │
│ core::ptr::drop_in_place<std::thread::Thread>                 │           0 B │         21 B │  +21 │  +100.0% │
│ std::sys::pal::unix::stack_overflow::imp::GUARD::__getit::VAL │           0 B │         16 B │  +16 │  +100.0% │
│ std::rt::lang_start_internal                                  │      1.73 KiB │     1.72 KiB │  -14 │    -0.8% │
│ std::sys::pal::unix::thread::guard::PAGE_SIZE                 │           8 B │          0 B │   -8 │  -100.0% │
│ std::thread::CURRENT::__getit::VAL                            │           0 B │          8 B │   +8 │  +100.0% │
│ std::sys::pal::unix::stack_overflow::imp::PAGE_SIZE           │           0 B │          8 B │   +8 │  +100.0% │
│ std::sys::sync::once::futex::Once::call                       │      1.96 KiB │     1.95 KiB │   -7 │    -0.3% │
│ std::sys_common::thread_info::THREAD_INFO::__getit::STATE     │           1 B │          0 B │   -1 │  -100.0% │
│ std::thread::CURRENT::__getit::STATE                          │           0 B │          1 B │   +1 │  +100.0% │
│ <689 unchanged rows>                                          │    243.75 KiB │   243.75 KiB │    0 │     0.0% │
│───────────────────────────────────────────────────────────────│───────────────│──────────────│──────│──────────│
│ Total                                                         │    250.05 KiB │   250.70 KiB │ +663 │    +0.3% │
└───────────────────────────────────────────────────────────────┴───────────────┴──────────────┴──────┴──────────┘

[joboet]

Nearly all of that bloat is debug formatting. Filed #123356 to address what I believe to be the cause.