first, an meta-comment: patches mangled quoted-unprintably by MIME are
really hard to read; there's no apparent reason for the message you
sent to have been MIME-encoded.

The patch in question is incorrect in multiple ways.

1) it references a system call argument in the user address space
using memcpy rather than copyin/copyinstr,

2) it may cause a kernel mode page fault in the event that there
aren't PATH_MAX valid bytes after the start of the "path" argument,
and

3) uses UIO_USERSPACE to refer to kernel-space memory (the
newly-allocated "path" variable).   (if this code appears to work *at
all*, it suggests that there's a bug in that platform's copyinstr() --
it should return EFAULT when presented with a kernel-space address).

A more correct way to do this would involve the introduction of a new
namei operational flag, passing that from mkdir(), and if that flag is
set, trimming the trailing slashes off the pathname early in namei().

					- Bill