Directly use raw pointers in `AtomicPtr` store/load #77611

oli-obk · 2020-10-06T10:05:25Z

I was unable to find any reason for this limitation in the latest source of LLVM or in the documentation here.

fixes rust-lang/miri#1574

rust-highfive · 2020-10-06T10:05:30Z

r? @sfackler

(rust_highfive has picked a reviewer for you, use r? to override)

RalfJung · 2020-10-06T11:31:24Z

With this you can probably remove the use of miri_static_root in library/std/src/sys/windows/thread_local_key.rs.

oli-obk · 2020-10-06T12:25:34Z

Removed that and miri still tests successfully

oli-obk · 2020-10-28T16:38:27Z

r? @nagisa (I don't know who else to ping for llvm stuff)

RalfJung · 2020-10-28T16:39:30Z

AFAIK @cuviper is also knowledgeable for matters of LLVM

nagisa · 2020-10-28T16:52:37Z

There's an entire workgroup of us! (rust-lang/wg-llvm)

This seems just fine to me, @bors r+ rollup=iffy

bors · 2020-10-28T16:52:39Z

📌 Commit 31c1dfbe12870cea1144b5d0c77785acc1b4ce70 has been approved by nagisa

bors · 2020-10-28T17:20:19Z

⌛ Testing commit 31c1dfbe12870cea1144b5d0c77785acc1b4ce70 with merge 3610b9b4bb45c31e2e7af4814c95fe2106738100...

bors · 2020-10-28T17:47:31Z

💔 Test failed - checks-actions

oli-obk · 2020-10-28T17:53:51Z

😆 nevermind, we found a platform where this doesn't work

Intrinsic has incorrect argument type!
void ({}*)* @llvm.ppc.cfence.p0sl_s
in function rust_oom
LLVM ERROR: Broken function found, compilation aborted!
[RUSTC-TIMING] std test:false 20.161
error: could not compile `std`

I'll adjust the llvm backend to do the appropriate transformation instead of having that in libcore

kennytm · 2020-11-12T06:05:19Z

@bors r-

RalfJung · 2020-11-29T14:09:28Z

Oh, this looks great, I feared it would be much more complicated. :)

If I understood correctly, since this happens in the general SSA backend, the Cranelift backend will not need to do anything itself? Cc @bjorn3

bjorn3 · 2020-11-29T14:14:41Z

I don't use cg_ssa yet. Cranelift doesn't have a separate pointer type, so no changes to the actual codegen are necessary. There is a validation for the used type though. The fix for cg_clif should be a matter of adding ty::RawPtr to the match arm at

rust/compiler/rustc_codegen_cranelift/src/intrinsics/mod.rs

Line 149 in 2e57231

ty::Uint(_) | ty::Int(_) => {}

bjorn3

The cg_clif change LGTM

bjorn3 · 2020-11-29T15:05:22Z

library/core/src/sync/atomic.rs

+        // SAFETY: This intrinsic is unsafe because it operates on a raw pointer
+        // but we know for sure that the pointer is valid (we just got it from
+        // an `UnsafeCell` that we have by reference) and the atomic operation
+        // itself allows us to safely mutate the `UnsafeCell` contents.


This doesn't match the safety comment for the #[cfg(bootstrap)] case.

I know, but I felt like it wasn't worth updating that, too, since it'll go away with the next beta anyway

nagisa · 2020-12-09T14:03:01Z

@bors r+ rollup=never

bors · 2020-12-09T14:03:03Z

📌 Commit 41f988e has been approved by nagisa

bors · 2020-12-09T16:36:16Z

⌛ Testing commit 41f988e with merge 0bddb8da0ff78a761169b9ac1b18cde962fb88f6...

bors · 2020-12-09T17:27:29Z

💔 Test failed - checks-actions

bjorn3 · 2020-12-09T17:41:55Z

It says "This check failed" without any logs. I guess it is spurious

@bors retry

bors · 2020-12-09T19:53:29Z

⌛ Testing commit 41f988e with merge f0f6877...

bors · 2020-12-09T22:21:48Z

☀️ Test successful - checks-actions
Approved by: nagisa
Pushing f0f6877 to master...

Directly use raw pointers in `AtomicPtr` store/load I was unable to find any reason for this limitation in the latest source of LLVM or in the documentation [here](http://llvm.org/docs/Atomics.html#libcalls-atomic). fixes rust-lang/miri#1574

…li-obk Only generate a ptrtoint in AtomicPtr codegen when absolutely necessary This special case was added in this PR: rust-lang#77611 in response to this error message: ``` Intrinsic has incorrect argument type! void ({}*)* `@llvm.ppc.cfence.p0sl_s` in function rust_oom LLVM ERROR: Broken function found, compilation aborted! [RUSTC-TIMING] std test:false 20.161 error: could not compile `std` ``` But when I tried searching for more information about that intrinsic I found this: llvm/llvm-project#55983 which is a report of someone hitting this same error and a fix was landed in LLVM, 2 years after the above Rust PR.

Only generate a ptrtoint in AtomicPtr codegen when absolutely necessary This special case was added in this PR: rust-lang/rust#77611 in response to this error message: ``` Intrinsic has incorrect argument type! void ({}*)* `@llvm.ppc.cfence.p0sl_s` in function rust_oom LLVM ERROR: Broken function found, compilation aborted! [RUSTC-TIMING] std test:false 20.161 error: could not compile `std` ``` But when I tried searching for more information about that intrinsic I found this: llvm/llvm-project#55983 which is a report of someone hitting this same error and a fix was landed in LLVM, 2 years after the above Rust PR.

rust-highfive assigned sfackler Oct 6, 2020

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 6, 2020

oli-obk force-pushed the atomic_miri_leakage branch from 9d1626f to 2da5960 Compare October 6, 2020 10:39

jyn514 added A-LLVM Area: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Oct 6, 2020

camelid added A-codegen Area: Code generation S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 23, 2020

rust-highfive assigned nagisa and unassigned sfackler Oct 28, 2020

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 28, 2020

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Oct 28, 2020

RalfJung mentioned this pull request Nov 2, 2020

Add current version of BTreeMap unit tests RalfJung/miri#1

Closed

ssomers mentioned this pull request Nov 4, 2020

BTreeMap: fix pointer provenance rules #78480

Merged

bors added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Nov 12, 2020

Allow cranelift to handle atomic pointers

41f988e

bjorn3 approved these changes Nov 29, 2020

View reviewed changes

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Dec 9, 2020

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Dec 9, 2020

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Dec 9, 2020

bors added the merged-by-bors This PR was explicitly merged by bors. label Dec 9, 2020

bors merged commit f0f6877 into rust-lang:master Dec 9, 2020

rustbot added this to the 1.50.0 milestone Dec 9, 2020

RalfJung mentioned this pull request Dec 20, 2020

Miri detects MPSC as having UB according to stacked borrows #80234

Closed

tmiasko mentioned this pull request Dec 20, 2020

Use pointer type in AtomicPtr::swap implementation #80236

Merged

oli-obk deleted the atomic_miri_leakage branch March 16, 2021 12:14

saethlin mentioned this pull request Mar 9, 2024

Only generate a ptrtoint in AtomicPtr codegen when absolutely necessary #122220

Merged

Directly use raw pointers in AtomicPtr store/load #77611

Directly use raw pointers in AtomicPtr store/load #77611

Uh oh!

Conversation

oli-obk commented Oct 6, 2020

Uh oh!

rust-highfive commented Oct 6, 2020

Uh oh!

RalfJung commented Oct 6, 2020

Uh oh!

oli-obk commented Oct 6, 2020

Uh oh!

oli-obk commented Oct 28, 2020

Uh oh!

RalfJung commented Oct 28, 2020

Uh oh!

nagisa commented Oct 28, 2020

Uh oh!

bors commented Oct 28, 2020

Uh oh!

bors commented Oct 28, 2020

Uh oh!

bors commented Oct 28, 2020

Uh oh!

oli-obk commented Oct 28, 2020

Uh oh!

kennytm commented Nov 12, 2020

Uh oh!

RalfJung commented Nov 29, 2020

Uh oh!

bjorn3 commented Nov 29, 2020

Uh oh!

bjorn3 left a comment

Choose a reason for hiding this comment

Uh oh!

bjorn3 Nov 29, 2020

Choose a reason for hiding this comment

Uh oh!

oli-obk Nov 29, 2020

Choose a reason for hiding this comment

Uh oh!

nagisa commented Dec 9, 2020

Uh oh!

bors commented Dec 9, 2020

Uh oh!

bors commented Dec 9, 2020

Uh oh!

bors commented Dec 9, 2020

Uh oh!

bjorn3 commented Dec 9, 2020

Uh oh!

bors commented Dec 9, 2020

Uh oh!

bors commented Dec 9, 2020

Uh oh!

Uh oh!

Directly use raw pointers in `AtomicPtr` store/load #77611

Directly use raw pointers in `AtomicPtr` store/load #77611