1List:       linux-kernel
   2Subject:    Re: active_mm
   3From:       Linus Torvalds <torvalds () transmeta ! com>
   4Date:       1999-07-30 21:36:24
   6Cc'd to linux-kernel, because I don't write explanations all that often,
   7and when I do I feel better about more people reading them.
   9On Fri, 30 Jul 1999, David Mosberger wrote:
  11> Is there a brief description someplace on how "mm" vs. "active_mm" in
  12> the task_struct are supposed to be used?  (My apologies if this was
  13> discussed on the mailing lists---I just returned from vacation and
  14> wasn't able to follow linux-kernel for a while).
  16Basically, the new setup is:
  18 - we have "real address spaces" and "anonymous address spaces". The
  19   difference is that an anonymous address space doesn't care about the
  20   user-level page tables at all, so when we do a context switch into an
  21   anonymous address space we just leave the previous address space
  22   active.
  24   The obvious use for a "anonymous address space" is any thread that
  25   doesn't need any user mappings - all kernel threads basically fall into
  26   this category, but even "real" threads can temporarily say that for
  27   some amount of time they are not going to be interested in user space,
  28   and that the scheduler might as well try to avoid wasting time on
  29   switching the VM state around. Currently only the old-style bdflush
  30   sync does that.
  32 - "tsk->mm" points to the "real address space". For an anonymous process,
  33   tsk->mm will be NULL, for the logical reason that an anonymous process
  34   really doesn't _have_ a real address space at all.
  36 - however, we obviously need to keep track of which address space we
  37   "stole" for such an anonymous user. For that, we have "tsk->active_mm",
  38   which shows what the currently active address space is.
  40   The rule is that for a process with a real address space (ie tsk->mm is
  41   non-NULL) the active_mm obviously always has to be the same as the real
  42   one.
  44   For a anonymous process, tsk->mm == NULL, and tsk->active_mm is the
  45   "borrowed" mm while the anonymous process is running. When the
  46   anonymous process gets scheduled away, the borrowed address space is
  47   returned and cleared.
  49To support all that, the "struct mm_struct" now has two counters: a
  50"mm_users" counter that is how many "real address space users" there are,
  51and a "mm_count" counter that is the number of "lazy" users (ie anonymous
  52users) plus one if there are any real users.
  54Usually there is at least one real user, but it could be that the real
  55user exited on another CPU while a lazy user was still active, so you do
  56actually get cases where you have a address space that is _only_ used by
  57lazy users. That is often a short-lived state, because once that thread
  58gets scheduled away in favour of a real thread, the "zombie" mm gets
  59released because "mm_users" becomes zero.
  61Also, a new rule is that _nobody_ ever has "init_mm" as a real MM any
  62more. "init_mm" should be considered just a "lazy context when no other
  63context is available", and in fact it is mainly used just at bootup when
  64no real VM has yet been created. So code that used to check
  66        if (current->mm == &init_mm)
  68should generally just do
  70        if (!current->mm)
  72instead (which makes more sense anyway - the test is basically one of "do
  73we have a user context", and is generally done by the page fault handler
  74and things like that).
  76Anyway, I put a pre-patch-2.3.13-1 on just a moment ago,
  77because it slightly changes the interfaces to accommodate the alpha (who
  78would have thought it, but the alpha actually ends up having one of the
  79ugliest context switch codes - unlike the other architectures where the MM
  80and register state is separate, the alpha PALcode joins the two, and you
  81need to switch both together).