[tip:core/urgent] lib/iov_iter: Document _copy_to_iter_mcsafe()

[tip-bot for Dan Williams]

Commit-ID:  bf3eeb9b5f2a1a05b3a68c6d82112babd58d6a39
Gitweb:     https://git.kernel.org/tip/bf3eeb9b5f2a1a05b3a68c6d82112babd58d6a39
Author:     Dan Williams <dan.j.williams@xxxxxxxxx>
AuthorDate: Sun, 8 Jul 2018 13:46:02 -0700
Committer:  Ingo Molnar <mingo@xxxxxxxxxx>
CommitDate: Mon, 16 Jul 2018 00:05:05 +0200

lib/iov_iter: Document _copy_to_iter_mcsafe()

Add some theory of operation documentation to _copy_to_iter_mcsafe().

Reported-by: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>
Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: Andy Lutomirski <luto@xxxxxxxxxxxxxx>
Cc: Borislav Petkov <bp@xxxxxxxxx>
Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Tony Luck <tony.luck@xxxxxxxxx>
Link: http://lkml.kernel.org/r/153108276256.37979.1689794213845539316.stgit@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: Ingo Molnar <mingo@xxxxxxxxxx>
---
 lib/iov_iter.c | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index 7e43cd54c84c..94fa361be7bb 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -596,6 +596,32 @@ static unsigned long memcpy_mcsafe_to_page(struct page *page, size_t offset,
 	return ret;
 }
 
+/**
+ * _copy_to_iter_mcsafe - copy to user with source-read error exception handling
+ * @addr: source kernel address
+ * @bytes: total transfer length
+ * @iter: destination iterator
+ *
+ * The pmem driver arranges for filesystem-dax to use this facility via
+ * dax_copy_to_iter() for protecting read/write to persistent memory.
+ * Unless / until an architecture can guarantee identical performance
+ * between _copy_to_iter_mcsafe() and _copy_to_iter() it would be a
+ * performance regression to switch more users to the mcsafe version.
+ *
+ * Otherwise, the main differences between this and typical _copy_to_iter().
+ *
+ * * Typical tail/residue handling after a fault retries the copy
+ *   byte-by-byte until the fault happens again. Re-triggering machine
+ *   checks is potentially fatal so the implementation uses source
+ *   alignment and poison alignment assumptions to avoid re-triggering
+ *   hardware exceptions.
+ *
+ * * ITER_KVEC, ITER_PIPE, and ITER_BVEC can return short copies.
+ *   Compare to copy_to_iter() where only ITER_IOVEC attempts might return
+ *   a short copy.
+ *
+ * See MCSAFE_TEST for self-test.
+ */
 size_t _copy_to_iter_mcsafe(const void *addr, size_t bytes, struct iov_iter *i)
 {
 	const char *from = addr;