Multiple broken tests under Miri

[00xc]

Running all tests with Miri enabled:

MIRIFLAGS="-Zmiri-permissive-provenance" cargo +nightly miri test --workspace

The following generate Miri errors:

• insn_decode::insn::tests::test_decode_ins_u8
• insn_decode::insn::tests::test_decode_ins_u16
• insn_decode::insn::tests::test_decode_ins_u32
• insn_decode::insn::tests::test_decode_mov_imm_to_reg
• insn_decode::insn::tests::test_decode_mov_reg_to_moffset
• insn_decode::insn::tests::test_decode_mov_reg_to_rm
• queue::tests::add_buffers_indirect

The following get stuck in an infinite loop (or run extremely slowly):

• device::blk::tests::device_id
• device::blk::tests::flush
• device::blk::tests::read
• device::blk::tests::write

[osteffenrh]

The device::* ones are from the virtio-drivers sub-crate we imported.

I see the same (=hanging) when I run them

MIRIFLAGS="-Zmiri-permissive-provenance" cargo +nightly miri test -p virtio-drivers 

When I Ctrl+C it, I get some messages:

error: interpretation was interrupted                                                                                                                                   130 ↵
    --> /home/osteffen/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/cell.rs:2421:22
     |
2421 |     pub const fn get(&self) -> *mut T {
     |                      ^^^^^ error occurred here
     |
     = note: BACKTRACE on thread `device::blk::te`:
     = note: inside `std::cell::UnsafeCell::<u16>::get` at /home/osteffen/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/cell.rs:2421:22: 2421:27
note: inside `queue::VirtQueue::<hal::fake::FakeHal, 16>::can_pop`
    --> virtio-drivers/src/queue.rs:376:40
     |
 376 |         self.last_used_idx != unsafe { (*self.used.as_ptr()).idx.load(Ordering::Acquire) }
     |                                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
note: inside `queue::VirtQueue::<hal::fake::FakeHal, 16>::add_notify_wait_pop::<transport::fake::FakeTransport<device::blk::BlkConfig>>`
    --> virtio-drivers/src/queue.rs:319:16
     |
 319 |         while !self.can_pop() {
     |                ^^^^^^^^^^^^^^
note: inside `device::blk::VirtIOBlk::<hal::fake::FakeHal, transport::fake::FakeTransport<device::blk::BlkConfig>>::request_read`
    --> virtio-drivers/src/device/blk.rs:131:9
     |
 131 | /         self.queue.add_notify_wait_pop(
 132 | |             &[request.as_bytes()],
 133 | |             &mut [data, resp.as_mut_bytes()],
 134 | |             &mut self.transport,
 135 | |         )?;
     | |_________^
note: inside `device::blk::VirtIOBlk::<hal::fake::FakeHal, transport::fake::FakeTransport<device::blk::BlkConfig>>::device_id`
    --> virtio-drivers/src/device/blk.rs:169:9
     |
 169 | /         self.request_read(
 170 | |             BlkReq {
 171 | |                 type_: ReqType::GetId,
 172 | |                 ..Default::default()
 173 | |             },
 174 | |             id,
 175 | |         )?;
     | |_________^
note: inside `device::blk::tests::device_id`
    --> virtio-drivers/src/device/blk.rs:897:22
     |
 897 |         let length = blk.device_id(&mut id).unwrap();
     |                      ^^^^^^^^^^^^^^^^^^^^^^
note: inside closure
    --> virtio-drivers/src/device/blk.rs:835:19
     |
 834 |     #[test]
     |     ------- in this attribute macro expansion
 835 |     fn device_id() {
     |                   ^

note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace

error: aborting due to 1 previous error; 1 warning emitted

I'll keep looking into it.

[00xc]

Yes, I think I came to the conclusion that the secondary thread spawned in the test is not really executed by miri somehow, causing the main thread to block.