vfs: document that read/write & friends are called with file.f_pos_lock held

Commit 9c225f26 ("vfs: atomic f_pos accesses as per POSIX") added
locking on read/write/lseek for regular files so that the file position
is updated atomically.

Commit 10dce8af ("fs: stream_open - opener for stream-like files so
that read and write can run simultaneously without deadlock") removed
that locking if a file is stream-like, so that e.g. read don't deadlock
simultaneous write.

Document that file.f_pos_lock is held on read/write & co for regular
files.

In the documentation we can ignore the fact that f_pos_lock is not taken
if there is only one single thread using the file, because it is just an
optimization and does not affect semantic, by which multiple read /
write cannot run simultaneously.
Signed-off-by: Kirill Smelkov <kirr@nexedi.com>

Documentation/filesystems/vfs.txt

@@ -893,15 +893,20 @@ struct file_operations {
 Again, all methods are called without any locks being held, unless
 otherwise noted.
 
-  llseek: called when the VFS needs to move the file position index
+  llseek: called when the VFS needs to move the file position index.
+        file.f_pos_lock is held for regular files.
 
-  read: called by read(2) and related system calls
+  read: called by read(2) and related system calls.
+        file.f_pos_lock is held for regular files.
 
-  read_iter: possibly asynchronous read with iov_iter as destination
+  read_iter: possibly asynchronous read with iov_iter as destination.
+        file.f_pos_lock is held for regular files.
 
-  write: called by write(2) and related system calls
+  write: called by write(2) and related system calls.
+        file.f_pos_lock is held for regular files.
 
-  write_iter: possibly asynchronous write with iov_iter as source
+  write_iter: possibly asynchronous write with iov_iter as source.
+        file.f_pos_lock is held for regular files.
 
   iopoll: called when aio wants to poll for completions on HIPRI iocbs