cachepc-linux

Fork of AMDESE/linux with modifications for CachePC side-channel attack
git clone https://git.sinitax.com/sinitax/cachepc-linux
Log | Files | Refs | README | LICENSE | sfeed.txt

active_mm.rst (3922B)


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