Use wiggle "trappable error" to implement wasi-common (#5279)

* convert wasi-common from defining its own error to using wiggle trappable error

* wasi-common impl crates: switch error strategy

* wasmtime-wasi: error is trappable, and no longer requires UserErrorConversion

* docs

* typo

* readdir: windows fixes

* fix windows scheduler errors

fun fact! the Send and Recv errors here that just had a `.context` on
them were previously not being captured in the downcasting either. They
need to be traps, and would have ended up that way by ommission, but
you'd never actually know that by reading the code!

Author: Pat Hickey

crates/wasi-common/cap-std-sync/src/dir.rs

@@ -138,14 +138,27 @@ impl WasiDir for Dir {
         &self,
         cursor: ReaddirCursor,
     ) -> Result<Box<dyn Iterator<Item = Result<ReaddirEntity, Error>> + Send>, Error> {
+        // We need to keep a full-fidelity io Error around to check for a special failure mode
+        // on windows, but also this function can fail due to an illegal byte sequence in a
+        // filename, which we can't construct an io Error to represent.
+        enum ReaddirError {
+            Io(std::io::Error),
+            IllegalSequence,
+        }
+        impl From<std::io::Error> for ReaddirError {
+            fn from(e: std::io::Error) -> ReaddirError {
+                ReaddirError::Io(e)
+            }
+        }
+
         // cap_std's read_dir does not include . and .., we should prepend these.
         // Why does the Ok contain a tuple? We can't construct a cap_std::fs::DirEntry, and we don't
         // have enough info to make a ReaddirEntity yet.
         let dir_meta = self.0.dir_metadata()?;
         let rd = vec![
             {
                 let name = ".".to_owned();
-                Ok((FileType::Directory, dir_meta.ino(), name))
+                Ok::<_, ReaddirError>((FileType::Directory, dir_meta.ino(), name))
             },
             {
                 let name = "..".to_owned();
@@ -163,24 +176,22 @@ impl WasiDir for Dir {
                 let name = entry
                     .file_name()
                     .into_string()
-                    .map_err(|_| Error::illegal_byte_sequence().context("filename"))?;
+                    .map_err(|_| ReaddirError::IllegalSequence)?;
                 Ok((filetype, inode, name))
             });
 
             // On Windows, filter out files like `C:\DumpStack.log.tmp` which we
             // can't get a full metadata for.
             #[cfg(windows)]
-            let entries = entries.filter(|entry: &Result<_, wasi_common::Error>| {
+            let entries = entries.filter(|entry| {
                 use windows_sys::Win32::Foundation::{
                     ERROR_ACCESS_DENIED, ERROR_SHARING_VIOLATION,
                 };
-                if let Err(err) = entry {
-                    if let Some(err) = err.downcast_ref::<std::io::Error>() {
-                        if err.raw_os_error() == Some(ERROR_SHARING_VIOLATION as i32)
-                            || err.raw_os_error() == Some(ERROR_ACCESS_DENIED as i32)
-                        {
-                            return false;
-                        }
+                if let Err(ReaddirError::Io(err)) = entry {
+                    if err.raw_os_error() == Some(ERROR_SHARING_VIOLATION as i32)
+                        || err.raw_os_error() == Some(ERROR_ACCESS_DENIED as i32)
+                    {
+                        return false;
                     }
                 }
                 true
@@ -197,7 +208,8 @@ impl WasiDir for Dir {
                 inode,
                 name,
             }),
-            Err(e) => Err(e),
+            Err(ReaddirError::Io(e)) => Err(e.into()),
+            Err(ReaddirError::IllegalSequence) => Err(Error::illegal_byte_sequence()),
         })
         .skip(u64::from(cursor) as usize);
 

crates/wasi-common/cap-std-sync/src/sched/unix.rs

@@ -47,7 +47,7 @@ pub async fn poll_oneoff<'a>(poll: &mut Poll<'a>) -> Result<(), Error> {
         match rustix::io::poll(&mut pollfds, poll_timeout) {
             Ok(ready) => break ready,
             Err(rustix::io::Errno::INTR) => continue,
-            Err(err) => return Err(err.into()),
+            Err(err) => return Err(std::io::Error::from(err).into()),
         }
     };
     if ready > 0 {

crates/wasi-common/cap-std-sync/src/sched/windows.rs

@@ -8,7 +8,6 @@
 // We suspect there are bugs in this scheduler, however, we have not
 // taken the time to improve it. See bug #2880.
 
-use anyhow::Context;
 use once_cell::sync::Lazy;
 use std::ops::Deref;
 use std::sync::mpsc::{self, Receiver, RecvTimeoutError, Sender, TryRecvError};
@@ -73,7 +72,7 @@ pub async fn poll_oneoff_<'a>(
     if !stdin_read_subs.is_empty() {
         let state = STDIN_POLL
             .lock()
-            .map_err(|_| Error::trap("failed to take lock of STDIN_POLL"))?
+            .map_err(|_| Error::trap(anyhow::Error::msg("failed to take lock of STDIN_POLL")))?
             .poll(waitmode)?;
         for readsub in stdin_read_subs.into_iter() {
             match state {
@@ -167,34 +166,36 @@ impl StdinPoll {
             // Clean up possibly unread result from previous poll.
             Ok(_) | Err(TryRecvError::Empty) => {}
             Err(TryRecvError::Disconnected) => {
-                return Err(Error::trap("StdinPoll notify_rx channel closed"))
+                return Err(Error::trap(anyhow::Error::msg(
+                    "StdinPoll notify_rx channel closed",
+                )))
             }
         }
 
         // Notify the worker thread to poll stdin
         self.request_tx
             .send(())
-            .context("request_tx channel closed")?;
+            .map_err(|_| Error::trap(anyhow::Error::msg("request_tx channel closed")))?;
 
         // Wait for the worker thread to send a readiness notification
         match wait_mode {
             WaitMode::Timeout(timeout) => match self.notify_rx.recv_timeout(timeout) {
                 Ok(r) => Ok(r),
                 Err(RecvTimeoutError::Timeout) => Ok(PollState::TimedOut),
-                Err(RecvTimeoutError::Disconnected) => {
-                    Err(Error::trap("StdinPoll notify_rx channel closed"))
-                }
+                Err(RecvTimeoutError::Disconnected) => Err(Error::trap(anyhow::Error::msg(
+                    "StdinPoll notify_rx channel closed",
+                ))),
             },
             WaitMode::Infinite => self
                 .notify_rx
                 .recv()
-                .context("StdinPoll notify_rx channel closed"),
+                .map_err(|_| Error::trap(anyhow::Error::msg("StdinPoll notify_rx channel closed"))),
             WaitMode::Immediate => match self.notify_rx.try_recv() {
                 Ok(r) => Ok(r),
                 Err(TryRecvError::Empty) => Ok(PollState::NotReady),
-                Err(TryRecvError::Disconnected) => {
-                    Err(Error::trap("StdinPoll notify_rx channel closed"))
-                }
+                Err(TryRecvError::Disconnected) => Err(Error::trap(anyhow::Error::msg(
+                    "StdinPoll notify_rx channel closed",
+                ))),
             },
         }
     }

crates/wasi-common/cap-std-sync/src/stdio.rs

@@ -128,7 +128,9 @@ macro_rules! wasi_file_write_impl {
             }
             async fn write_vectored<'a>(&mut self, bufs: &[io::IoSlice<'a>]) -> Result<u64, Error> {
                 let n = (&*self.0.as_filelike_view::<File>()).write_vectored(bufs)?;
-                Ok(n.try_into().map_err(|c| Error::range().context(c))?)
+                Ok(n.try_into().map_err(|_| {
+                    Error::range().context("converting write_vectored total length")
+                })?)
             }
             async fn write_vectored_at<'a>(
                 &mut self,

crates/wasi-common/src/error.rs

@@ -1,145 +1,15 @@
-//! `wasi_common::Error` is now `anyhow::Error`.
+//! wasi-common uses an [`Error`] type which represents either a preview 1 [`Errno`] enum, on
+//! [`anyhow::Error`] for trapping execution.
 //!
-//! Snapshots (right now only `wasi_common::snapshots::preview_1`) contains
-//! all of the logic for transforming an `Error` into the snapshot's own
-//! `Errno`. They may do so by downcasting the error into any of:
-//! * `std::io::Error` - these are thrown by `std`, `cap_std`, etc for most of
-//! the operations WASI is concerned with.
-//! * `wasi_common::ErrorKind` - these are a subset of the Errnos, and are
-//! constructed directly by wasi-common or an impl rather than coming from the
-//! OS or some library which doesn't know about WASI.
-//! * `wiggle::GuestError`
-//! * `std::num::TryFromIntError`
-//! * `std::str::Utf8Error`
-//! and then applying specialized logic to translate each of those into
-//! `Errno`s.
-//!
-//! The `wasi_common::ErrorExt` trait provides human-friendly constructors for
-//! the `wasi_common::ErrorKind` variants .
-//!
-//! If you throw an error that does not downcast to one of those, it will turn
-//! into a `wiggle::Trap` and terminate execution.
-//!
-//! The real value of using `anyhow::Error` here is being able to use
-//! `anyhow::Result::context` to aid in debugging of errors.
+//! The user can construct an [`Error`] out of an [`Errno`] using the `From`/`Into` traits.
+//! They may also use [`Error::trap`] to construct an error that traps execution. The contents
+//! can be inspected with [`Error::downcast`] and [`Error::downcast_ref`]. Additional context
+//! can be provided with the [`Error::context`] method. This context is only observable with the
+//! `Display` and `Debug` impls of the error.
 
-pub use anyhow::{Context, Error};
+pub use crate::snapshots::preview_1::error::{Errno, Error, ErrorExt};
 use std::fmt;
 
-/// Internal error type for the `wasi-common` crate.
-///
-/// This Contains variants of the WASI `$errno` type that are used internally
-/// by the crate, and which aren't one-to-one with a `std::io::ErrorKind`
-/// error.
-///
-/// When the Rust [io_error_more] feature is stabilized, that will enable
-/// us to replace several more of these codes with `std::io::ErrorKind` codes.
-///
-/// [io_error_more]: https://doc.rust-lang.org/beta/unstable-book/library-features/io-error-more.html
-#[derive(Copy, Clone, Debug, PartialEq, Eq, thiserror::Error)]
-#[non_exhaustive]
-pub enum ErrorKind {
-    /// Errno::TooBig: Argument list too long
-    #[error("TooBig: Argument list too long")]
-    TooBig,
-    /// Errno::Badf: Bad file descriptor
-    #[error("Badf: Bad file descriptor")]
-    Badf,
-    /// Errno::Ilseq: Illegal byte sequence
-    #[error("Ilseq: Illegal byte sequence")]
-    Ilseq,
-    /// Errno::Io: I/O error
-    #[error("Io: I/O error")]
-    Io,
-    /// Errno::Nametoolong: Filename too long
-    #[error("Nametoolong: Filename too long")]
-    Nametoolong,
-    /// Errno::Notdir: Not a directory or a symbolic link to a directory.
-    #[error("Notdir: Not a directory or a symbolic link to a directory")]
-    Notdir,
-    /// Errno::Notsup: Not supported, or operation not supported on socket.
-    #[error("Notsup: Not supported, or operation not supported on socket")]
-    Notsup,
-    /// Errno::Overflow: Value too large to be stored in data type.
-    #[error("Overflow: Value too large to be stored in data type")]
-    Overflow,
-    /// Errno::Range: Result too large
-    #[error("Range: Result too large")]
-    Range,
-    /// Errno::Spipe: Invalid seek
-    #[error("Spipe: Invalid seek")]
-    Spipe,
-    /// Errno::Perm: Permission denied
-    #[error("Permission denied")]
-    Perm,
-}
-
-pub trait ErrorExt {
-    fn trap(msg: impl Into<String>) -> Self;
-    fn not_found() -> Self;
-    fn too_big() -> Self;
-    fn badf() -> Self;
-    fn exist() -> Self;
-    fn illegal_byte_sequence() -> Self;
-    fn invalid_argument() -> Self;
-    fn io() -> Self;
-    fn name_too_long() -> Self;
-    fn not_dir() -> Self;
-    fn not_supported() -> Self;
-    fn overflow() -> Self;
-    fn range() -> Self;
-    fn seek_pipe() -> Self;
-    fn perm() -> Self;
-}
-
-impl ErrorExt for Error {
-    fn trap(msg: impl Into<String>) -> Self {
-        anyhow::anyhow!(msg.into())
-    }
-    fn not_found() -> Self {
-        std::io::Error::from(std::io::ErrorKind::NotFound).into()
-    }
-    fn too_big() -> Self {
-        ErrorKind::TooBig.into()
-    }
-    fn badf() -> Self {
-        ErrorKind::Badf.into()
-    }
-    fn exist() -> Self {
-        std::io::Error::from(std::io::ErrorKind::AlreadyExists).into()
-    }
-    fn illegal_byte_sequence() -> Self {
-        ErrorKind::Ilseq.into()
-    }
-    fn invalid_argument() -> Self {
-        std::io::Error::from(std::io::ErrorKind::InvalidInput).into()
-    }
-    fn io() -> Self {
-        ErrorKind::Io.into()
-    }
-    fn name_too_long() -> Self {
-        ErrorKind::Nametoolong.into()
-    }
-    fn not_dir() -> Self {
-        ErrorKind::Notdir.into()
-    }
-    fn not_supported() -> Self {
-        ErrorKind::Notsup.into()
-    }
-    fn overflow() -> Self {
-        ErrorKind::Overflow.into()
-    }
-    fn range() -> Self {
-        ErrorKind::Range.into()
-    }
-    fn seek_pipe() -> Self {
-        ErrorKind::Spipe.into()
-    }
-    fn perm() -> Self {
-        ErrorKind::Perm.into()
-    }
-}
-
 /// An error returned from the `proc_exit` host syscall.
 ///
 /// Embedders can test if an error returned from wasm is this error, in which

crates/wasi-common/src/lib.rs

@@ -66,7 +66,7 @@ pub use cap_rand::RngCore;
 pub use clocks::{SystemTimeSpec, WasiClocks, WasiMonotonicClock, WasiSystemClock};
 pub use ctx::WasiCtx;
 pub use dir::WasiDir;
-pub use error::{Context, Error, ErrorExt, ErrorKind, I32Exit};
+pub use error::{Error, ErrorExt, I32Exit};
 pub use file::WasiFile;
 pub use sched::{Poll, WasiSched};
 pub use string_array::StringArrayError;