1=========
2Active MM
3=========
4
5Note, the mm_count refcount may no longer include the "lazy" users
6(running tasks with ->active_mm == mm && ->mm == NULL) on kernels
7with CONFIG_MMU_LAZY_TLB_REFCOUNT=n. Taking and releasing these lazy
8references must be done with mmgrab_lazy_tlb() and mmdrop_lazy_tlb()
9helpers, which abstract this config option.
10
11::
12
13 List:       linux-kernel
14 Subject:    Re: active_mm
15 From:       Linus Torvalds <torvalds () transmeta ! com>
16 Date:       1999-07-30 21:36:24
17
18 Cc'd to linux-kernel, because I don't write explanations all that often,
19 and when I do I feel better about more people reading them.
20
21 On Fri, 30 Jul 1999, David Mosberger wrote:
22 >
23 > Is there a brief description someplace on how "mm" vs. "active_mm" in
24 > the task_struct are supposed to be used?  (My apologies if this was
25 > discussed on the mailing lists---I just returned from vacation and
26 > wasn't able to follow linux-kernel for a while).
27
28 Basically, the new setup is:
29
30  - we have "real address spaces" and "anonymous address spaces". The
31    difference is that an anonymous address space doesn't care about the
32    user-level page tables at all, so when we do a context switch into an
33    anonymous address space we just leave the previous address space
34    active.
35
36    The obvious use for a "anonymous address space" is any thread that
37    doesn't need any user mappings - all kernel threads basically fall into
38    this category, but even "real" threads can temporarily say that for
39    some amount of time they are not going to be interested in user space,
40    and that the scheduler might as well try to avoid wasting time on
41    switching the VM state around. Currently only the old-style bdflush
42    sync does that.
43
44  - "tsk->mm" points to the "real address space". For an anonymous process,
45    tsk->mm will be NULL, for the logical reason that an anonymous process
46    really doesn't _have_ a real address space at all.
47
48  - however, we obviously need to keep track of which address space we
49    "stole" for such an anonymous user. For that, we have "tsk->active_mm",
50    which shows what the currently active address space is.
51
52    The rule is that for a process with a real address space (ie tsk->mm is
53    non-NULL) the active_mm obviously always has to be the same as the real
54    one.
55
56    For a anonymous process, tsk->mm == NULL, and tsk->active_mm is the
57    "borrowed" mm while the anonymous process is running. When the
58    anonymous process gets scheduled away, the borrowed address space is
59    returned and cleared.
60
61 To support all that, the "struct mm_struct" now has two counters: a
62 "mm_users" counter that is how many "real address space users" there are,
63 and a "mm_count" counter that is the number of "lazy" users (ie anonymous
64 users) plus one if there are any real users.
65
66 Usually there is at least one real user, but it could be that the real
67 user exited on another CPU while a lazy user was still active, so you do
68 actually get cases where you have a address space that is _only_ used by
69 lazy users. That is often a short-lived state, because once that thread
70 gets scheduled away in favour of a real thread, the "zombie" mm gets
71 released because "mm_count" becomes zero.
72
73 Also, a new rule is that _nobody_ ever has "init_mm" as a real MM any
74 more. "init_mm" should be considered just a "lazy context when no other
75 context is available", and in fact it is mainly used just at bootup when
76 no real VM has yet been created. So code that used to check
77
78 	if (current->mm == &init_mm)
79
80 should generally just do
81
82 	if (!current->mm)
83
84 instead (which makes more sense anyway - the test is basically one of "do
85 we have a user context", and is generally done by the page fault handler
86 and things like that).
87
88 Anyway, I put a pre-patch-2.3.13-1 on ftp.kernel.org just a moment ago,
89 because it slightly changes the interfaces to accommodate the alpha (who
90 would have thought it, but the alpha actually ends up having one of the
91 ugliest context switch codes - unlike the other architectures where the MM
92 and register state is separate, the alpha PALcode joins the two, and you
93 need to switch both together).
94
95 (From http://marc.info/?l=linux-kernel&m=93337278602211&w=2)
96