Rollup merge of rust-lang#91153 - birkenfeld:kernel_copy_fallback, r=the8472

kernel_copy: avoid panic on unexpected OS error

According to documentation, the listed errnos should only occur
if the `copy_file_range` call cannot be made at all, so the
assert be correct.  However, since in practice file system
drivers (incl. FUSE etc.) can return any errno they want, we
should not panic here.

Fixes rust-lang#91152

Author: matthiaskrgr

library/std/src/sys/unix/kernel_copy.rs

@@ -576,20 +576,22 @@ pub(super) fn copy_regular_files(reader: RawFd, writer: RawFd, max_len: u64) ->
                 return match err.raw_os_error() {
                     // when file offset + max_length > u64::MAX
                     Some(EOVERFLOW) => CopyResult::Fallback(written),
-                    Some(ENOSYS | EXDEV | EINVAL | EPERM | EOPNOTSUPP | EBADF) => {
+                    Some(ENOSYS | EXDEV | EINVAL | EPERM | EOPNOTSUPP | EBADF) if written == 0 => {
                         // Try fallback io::copy if either:
                         // - Kernel version is < 4.5 (ENOSYS¹)
                         // - Files are mounted on different fs (EXDEV)
                         // - copy_file_range is broken in various ways on RHEL/CentOS 7 (EOPNOTSUPP)
                         // - copy_file_range file is immutable or syscall is blocked by seccomp¹ (EPERM)
                         // - copy_file_range cannot be used with pipes or device nodes (EINVAL)
                         // - the writer fd was opened with O_APPEND (EBADF²)
+                        // and no bytes were written successfully yet.  (All these errnos should
+                        // not be returned if something was already written, but they happen in
+                        // the wild, see #91152.)
                         //
                         // ¹ these cases should be detected by the initial probe but we handle them here
                         //   anyway in case syscall interception changes during runtime
                         // ² actually invalid file descriptors would cause this too, but in that case
                         //   the fallback code path is expected to encounter the same error again
-                        assert_eq!(written, 0);
                         CopyResult::Fallback(0)
                     }
                     _ => CopyResult::Error(err, written),