0001 .. _active_mm:
0002
0003 =========
0004 Active MM
0005 =========
0006
0007 ::
0008
0009 List: linux-kernel
0010 Subject: Re: active_mm
0011 From: Linus Torvalds <torvalds () transmeta ! com>
0012 Date: 1999-07-30 21:36:24
0013
0014 Cc'd to linux-kernel, because I don't write explanations all that often,
0015 and when I do I feel better about more people reading them.
0016
0017 On Fri, 30 Jul 1999, David Mosberger wrote:
0018 >
0019 > Is there a brief description someplace on how "mm" vs. "active_mm" in
0020 > the task_struct are supposed to be used? (My apologies if this was
0021 > discussed on the mailing lists---I just returned from vacation and
0022 > wasn't able to follow linux-kernel for a while).
0023
0024 Basically, the new setup is:
0025
0026 - we have "real address spaces" and "anonymous address spaces". The
0027 difference is that an anonymous address space doesn't care about the
0028 user-level page tables at all, so when we do a context switch into an
0029 anonymous address space we just leave the previous address space
0030 active.
0031
0032 The obvious use for a "anonymous address space" is any thread that
0033 doesn't need any user mappings - all kernel threads basically fall into
0034 this category, but even "real" threads can temporarily say that for
0035 some amount of time they are not going to be interested in user space,
0036 and that the scheduler might as well try to avoid wasting time on
0037 switching the VM state around. Currently only the old-style bdflush
0038 sync does that.
0039
0040 - "tsk->mm" points to the "real address space". For an anonymous process,
0041 tsk->mm will be NULL, for the logical reason that an anonymous process
0042 really doesn't _have_ a real address space at all.
0043
0044 - however, we obviously need to keep track of which address space we
0045 "stole" for such an anonymous user. For that, we have "tsk->active_mm",
0046 which shows what the currently active address space is.
0047
0048 The rule is that for a process with a real address space (ie tsk->mm is
0049 non-NULL) the active_mm obviously always has to be the same as the real
0050 one.
0051
0052 For a anonymous process, tsk->mm == NULL, and tsk->active_mm is the
0053 "borrowed" mm while the anonymous process is running. When the
0054 anonymous process gets scheduled away, the borrowed address space is
0055 returned and cleared.
0056
0057 To support all that, the "struct mm_struct" now has two counters: a
0058 "mm_users" counter that is how many "real address space users" there are,
0059 and a "mm_count" counter that is the number of "lazy" users (ie anonymous
0060 users) plus one if there are any real users.
0061
0062 Usually there is at least one real user, but it could be that the real
0063 user exited on another CPU while a lazy user was still active, so you do
0064 actually get cases where you have a address space that is _only_ used by
0065 lazy users. That is often a short-lived state, because once that thread
0066 gets scheduled away in favour of a real thread, the "zombie" mm gets
0067 released because "mm_count" becomes zero.
0068
0069 Also, a new rule is that _nobody_ ever has "init_mm" as a real MM any
0070 more. "init_mm" should be considered just a "lazy context when no other
0071 context is available", and in fact it is mainly used just at bootup when
0072 no real VM has yet been created. So code that used to check
0073
0074 if (current->mm == &init_mm)
0075
0076 should generally just do
0077
0078 if (!current->mm)
0079
0080 instead (which makes more sense anyway - the test is basically one of "do
0081 we have a user context", and is generally done by the page fault handler
0082 and things like that).
0083
0084 Anyway, I put a pre-patch-2.3.13-1 on ftp.kernel.org just a moment ago,
0085 because it slightly changes the interfaces to accommodate the alpha (who
0086 would have thought it, but the alpha actually ends up having one of the
0087 ugliest context switch codes - unlike the other architectures where the MM
0088 and register state is separate, the alpha PALcode joins the two, and you
0089 need to switch both together).
0090
0091 (From http://marc.info/?l=linux-kernel&m=93337278602211&w=2)
