[Regression] GlobalValue with DLLImport Storage is dso_local!

[asensio-project]

Hello,

I was trying to compile the uefi crate, but I can't compile it with nightly-2022-09-02. The previous nightly works fine.

I get:
GlobalValue with DLLImport Storage is dso_local!
ptr @_ZN4core7unicode12unicode_data11white_space14WHITESPACE_MAP17h76c794c1a16111d0E
LLVM ERROR: Broken module found, compilation aborted!

Maybe, the pr that broke it is #100537.

[nicholasbishop]

Confirmed with cargo-bisect-rustc --with-src --preserve --start 2022-09-01 -- build --target x86_64-unknown-uefi that #100537 is where it starts failing:

searched nightlies: from nightly-2022-09-01 to nightly-2022-09-03
regressed nightly: nightly-2022-09-02
searched commit range: 9243168...2e35f95
regressed commit: aa857eb

bisected with cargo-bisect-rustc v0.6.4

Host triple: x86_64-unknown-linux-gnu
Reproduce with:

cargo bisect-rustc --preserve --start 2022-09-01 -- build --target x86_64-unknown-uefi 

# Cargo.toml:
[dependencies]
uefi = "0.16.1"
uefi-services = "0.13.1"

# .cargo/config.toml:
[unstable]
build-std = ["core", "compiler_builtins", "alloc"]
build-std-features = ["compiler-builtins-mem"]

// src/main.rs:
#![no_main]
#![no_std]

use uefi::prelude::*;

#[export_name = "efi_main"]
pub extern "C" fn main(_h: *mut core::ffi::c_void, mut st: SystemTable<Boot>) -> usize {
    uefi_services::init(&mut st).expect("Failed to initialize utilities");

    0
}

The uefi-services crate specifically seems to trigger the failure, compiling with just uefi seems OK.

[nicholasbishop]

Here's a slightly reduced test case:

# Cargo.toml:
[dependencies]
log = { version = "0.4.5", default-features = false }

# .cargo/config.toml:
[unstable]
build-std = ["core", "compiler_builtins", "alloc"]
build-std-features = ["compiler-builtins-mem"]

// src/main.rs:
#![no_main]
#![no_std]

use core::ffi::c_void;

#[panic_handler]
fn panic_handler(_info: &core::panic::PanicInfo) -> ! {
    loop {}
}

#[export_name = "efi_main"]
pub extern "C" fn main(_h: *mut c_void, _st: *mut c_void) -> usize {
    log::error!("hello");

    0
}

Build output:

$ cargo +nightly build --target x86_64-unknown-uefi
   Compiling compiler_builtins v0.1.79
   Compiling core v0.0.0 (/var/home/nbishop/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core)
   Compiling log v0.4.17
   Compiling rustc-std-workspace-core v1.99.0 (/var/home/nbishop/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/rustc-std-workspace-core)
   Compiling alloc v0.0.0 (/var/home/nbishop/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/alloc)
   Compiling cfg-if v1.0.0
   Compiling bug v0.1.0 (/var/home/nbishop/src/rust-broken-link)
GlobalValue with DLLImport Storage is dso_local!
ptr @_ZN3log20MAX_LOG_LEVEL_FILTER17h159617aa6246edb9E
LLVM ERROR: Broken module found, compilation aborted!
error: could not compile `bug`

It seems the issue here is the switch from PIC to static for the reloc model in the UEFI targets. I will put up a PR to switch it back.

[dvdhrm]

It feels wrong to produce executables with DLLImport annotations, knowing that UEFI loaders will never read .idata sections. While technically they could, I am not aware of any such implementations and the spec clearly states All other linkage to and from an UEFI image is done programmatically.

So I think there is some bug hiding somewhere. However, reverting to PIC to get things working is preferable to the status quo.

The linked PR looks good to me!