summaryrefslogtreecommitdiff
path: root/service_solahart_jakarta_selatan__082122541663/mondriaan_memory_protection.mdwn
blob: 2c7b9ba1966769555c5c1a5a1955920bb076a33b (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
[[!meta copyright="Copyright © 2013 Free Software Foundation, Inc."]]

[[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable
id="license" text="Permission is granted to copy, distribute and/or modify this
document under the terms of the GNU Free Documentation License, Version 1.2 or
any later version published by the Free Software Foundation; with no Invariant
Sections, no Front-Cover Texts, and no Back-Cover Texts.  A copy of the license
is included in the section entitled [[GNU Free Documentation
License|/fdl]]."]]"""]]

<http://scale.eecs.berkeley.edu/mondriaan/>.


# IRC, freenode, #hurd, 2013-07-02

    <xscript> in any case, what I wanted to check is if current hurd support
      PIE
    <xscript`> I just saw samuel posted some fixes to have PIE working in hurd
    <xscript`> are those included in the official image?
    <youpi> sure
    <youpi> it's just a trivial fixup in some address calculation code
    <xscript> youpi: nice
    <xscript> and does anyone know how complex would it be to implement some
      hackish support to have non-overlapping virtual addresses for
      applications supporting PIE?
    <braunr> not too difficult
    <xscript> really? I didn't expect such an answer XD
    <xscript> I'd like to have something similar to a SASOS
    <xscript> (single address space os)
    <braunr> ?
    <braunr> you mean an sasos on top of mach ?
    <xscript> yes, but only for a few apps I want to evaluate
    <braunr> i see
    <xscript> the optimal would be to have all of hurd's servers on that mode
    <braunr> you'l probably need to implement a small allocator but other than
      that it shouldn't be too hard, yes
    <braunr> uh ??
    <xscript> but running on 32 bits can be a problem here
    <braunr> and not hurdish at all
    <xscript> what's not hurdish?
    <braunr> we do want address space separation
    <xscript> well, you can have multiple address spaces (page tables), but
      without overlapping addresses between them
    <xscript> that's exactly what I'm looking for
    <braunr> sorry i don't see what you mean
    <braunr> if you run several servers in the same address space, they can
      corrupt each other
    <braunr> we don't want that
    <braunr> it's that simple
    <xscript> yes, sorry, I didn't explain myself
    <xscript> I want a separate address space on each server
    <xscript> but I want all memory allocations to be on addresses unique to
      the whole OS
    <braunr> that still doesn't make sense
    <xscript> well, it will still be secure
    <xscript> but I know it does not make sense per se
    <xscript> I want to do some experiments with a simulator
    <braunr> why do you want them non overlapping if they're separate ?
    <xscript> well, in my simulator I wouldn't need to change the page tables,
      protection is provided through other means
    <braunr> segmentation ?
    <xscript> that's one possibility
    <xscript> (small address spaces)
    <braunr> what do you have in mind ?
    <braunr> it wouldn't be on top of mach anyway then
    <braunr> mach implements paging
    <xscript> what I'm simulating is something of the likes of Mondriaan
      (http://www.cs.utexas.edu/~witchel/pubs/mmp-asplos2002.pdf)
    <xscript> paging is ok for me
    <braunr> 19:06 < xscript> well, in my simulator I wouldn't need to change
      the page tables, protection is provided through other means
    <braunr> it didn't sound so
    <xscript> I meant switching page tables (cr3, etc)
    <braunr> mach does that
    <xscript> I know, I know, I can just ignore that part for the moment
    <braunr> ok
    <xscript> for now, I'd like to morph hurd into a SASOS using one page table
      per process
    <xscript> I just wanted to know how hard that would be, without starting
      with a full dive into the code
    <xscript> there are other options (OSes, microkernels), but none of them
      provides as many readily-available applications as hurd
    <xscript> I suppose MINIX would also be easy to modify, but there's less
      apps there, and I also would like to tamper with MIG
    <xscript> I just wonder how hard it would be to modify MIG