Tracing updates for 6.9:

Main user visible change:

- Add ring_buffer memory mappings

  The tracing ring buffer was created based on being mostly used with the
  splice system call. It is broken up into page ordered sub-buffers and the
  reader swaps a new sub-buffer with an existing sub-buffer that's part
  of the write buffer. It then has total access to the swapped out
  sub-buffer and can do copyless movements of the memory into other mediums
  (file system, network, etc).

  The buffer is great for passing around the ring buffer contents in the
  kernel, but is not so good for when the consumer is the user space task
  itself.

  A new interface is added that allows user space to memory map the ring
  buffer. It will get all the write sub-buffers as well as reader sub-buffer
  (that is not written to). It can send an ioctl to change which sub-buffer
  is the new reader sub-buffer.

  The ring buffer is read only to user space. It only needs to call the
  ioctl when it is finished with a sub-buffer and needs a new sub-buffer
  that the writer will not write over.

  A self test program was also created for testing and can be used as
  an example for the interface to user space. The libtracefs (external
  to the kernel) also has code that interacts with this, although it is
  disabled until the interface is in a official release. It can be enabled
  by compiling the library with a special flag. This was used for testing
  applications that perform better with the buffer being mapped.

  Memory mapped buffers have limitations. The main one is that it can not be
  used with the snapshot logic. If the buffer is mapped, snapshots will be
  disabled. If any logic is set to trigger snapshots on a buffer, that
  buffer will not be allowed to be mapped.

- User events can now have "multi formats"

  The current user events have a single format. If another event is created
  with a different format, it will fail to be created. That is, once an
  event name is used, it cannot be used again with a different format. This
  can cause issues if a library is using an event and updates its format.
  An application using the older format will prevent an application using
  the new library from registering its event.

  A task could also DOS another application if it knows the event names, and
  it creates events with different formats.

  The multi-format event is in a different name space from the single
  format. Both the event name and its format are the unique identifier.
  This will allow two different applications to use the same user event name
  but with different payloads.

- Added support to have ftrace_dump_on_oops dump out instances and
  not just the main top level tracing buffer.

Other changes:

- Add eventfs_root_inode

  Only the root inode has a dentry that is static (never goes away) and
  stores it upon creation. There's no reason that the thousands of other
  eventfs inodes should have a pointer that never gets set in its
  descriptor. Create a eventfs_root_inode desciptor that has a eventfs_inode
  descriptor and a dentry pointer, and only the root inode will use this.

- Added WARN_ON()s in eventfs

  There's some conditionals remaining in eventfs that should never be hit,
  but instead of removing them, add WARN_ON() around them to make sure that
  they are never hit.

- Have saved_cmdlines allocation also include the map_cmdline_to_pid array

  The saved_cmdlines structure allocates a large amount of data to hold its
  mappings. Within it, it has three arrays. Two are already apart of it:
  map_pid_to_cmdline[] and saved_cmdlines[]. More memory can be saved by
  also including the map_cmdline_to_pid[] array as well.

- Restructure __string() and __assign_str() macros used in TRACE_EVENT().

  Dynamic strings in TRACE_EVENT() are declared with:

      __string(name, source)

  And assigned with:

     __assign_str(name, source)

  In the tracepoint callback of the event, the __string() is used to get the
  size needed to allocate on the ring buffer and __assign_str() is used to
  copy the string into the ring buffer. There's a helper structure that is
  created in the TRACE_EVENT() macro logic that will hold the string length
  and its position in the ring buffer which is created by __string().

  There are several trace events that have a function to create the string
  to save. This function is executed twice. Once for __string() and again
  for __assign_str(). There's no reason for this. The helper structure could
  also save the string it used in __string() and simply copy that into
  __assign_str() (it also already has its length).

  By using the structure to store the source string for the assignment, it
  means that the second argument to __assign_str() is no longer needed.

  It will be removed in the next merge window, but for now add a warning if
  the source string given to __string() is different than the source string
  given to __assign_str(), as the source to __assign_str() isn't even used
  and will be going away.

- Other minor clean ups and fixes
-----BEGIN PGP SIGNATURE-----

iIoEABYIADIWIQRRSw7ePDh/lE+zeZMp5XQQmuv6qgUCZfRerhQccm9zdGVkdEBn
b29kbWlzLm9yZwAKCRAp5XQQmuv6qn/yAP4r0kDqm45yUctkzgR1CVhWLAqR+Icl
LcN7Ql+8B/McEwEAhErSDIwuJ7mYqsjw3v6j/0vyTFZFw5YlOj1uV1d0uAA=
=7UM2
-----END PGP SIGNATURE-----