Rollup merge of #106661 - mjguzik:linux_statx, r=Mark-Simulacrum

Stop probing for statx unless necessary

As is the current toy program:
fn main() -> std::io::Result<()> {
    use std::fs;

    let metadata = fs::metadata("foo.txt")?;

    assert!(!metadata.is_dir());
    Ok(())
}

... observed under strace will issue:
[snip]
statx(0, NULL, AT_STATX_SYNC_AS_STAT, STATX_ALL, NULL) = -1 EFAULT (Bad address) statx(AT_FDCWD, "foo.txt", AT_STATX_SYNC_AS_STAT, STATX_ALL, {stx_mask=STATX_ALL|STATX_MNT_ID, stx_attributes=0, stx_mode=S_IFREG|0644, stx_size=0, ...}) = 0

While statx is not necessarily always present, checking for it can be delayed to the first error condition. Said condition may very well never happen, in which case the check got avoided altogether.

Note this is still suboptimal as there still will be programs issuing it, but bulk of the problem is removed.

Tested by forbidding the syscall for the binary and observing it correctly falls back to newfstatat.

While here tidy up the commentary, in particular by denoting some problems with the current approach.

Author: matthiaskrgr

tests/pass-dep/shims/libc-fs-with-isolation.rs

@@ -22,7 +22,8 @@ fn main() {
     }
 
     // test `stat`
-    assert_eq!(fs::metadata("foo.txt").unwrap_err().kind(), ErrorKind::PermissionDenied);
+    let err = fs::metadata("foo.txt").unwrap_err();
+    assert_eq!(err.kind(), ErrorKind::PermissionDenied);
     // check that it is the right kind of `PermissionDenied`
-    assert_eq!(Error::last_os_error().raw_os_error(), Some(libc::EACCES));
+    assert_eq!(err.raw_os_error(), Some(libc::EACCES));
 }