WEBVTT

1
00:00:45.630 --> 00:00:46.830
William Cheng: Welcome to lecture.

2
00:00:46.890 --> 00:00:56.850
Lalit Gupta: 21st and I have some adults. Sure.

3
00:00:59.010 --> 00:00:59.820
Lalit Gupta: So my

4
00:01:05.850 --> 00:01:16.980
Lalit Gupta: One of my daughters, like I'm bit slightly confused on this cleanup page function and cleanup function because in one of your

5
00:01:18.060 --> 00:01:23.490
Lalit Gupta: Email a you said the key. I think of, maybe, I think in FAQ. It is written.

6
00:01:24.930 --> 00:01:25.350
Lalit Gupta: That

7
00:01:27.210 --> 00:01:43.350
Lalit Gupta: That that there's a I think it says, what does the polymorphous cleaner page me and it will return if a page is dirty, which means that if it is that it is different from what is on the desk and to clean the page frame to write it back and to clean the page frame.

8
00:01:44.580 --> 00:01:46.470
Lalit Gupta: Needs to write it back to the store.

9
00:01:47.460 --> 00:01:48.030
Lalit Gupta: Comes clean

10
00:01:48.420 --> 00:01:48.840
Yes.

11
00:01:49.860 --> 00:01:50.400
Lalit Gupta: Whereas

12
00:01:51.510 --> 00:01:59.280
Lalit Gupta: In the, in the widget artsy directly like biomorphic pointer function.

13
00:02:00.330 --> 00:02:00.840
Lalit Gupta: There is a

14
00:02:02.010 --> 00:02:11.310
Lalit Gupta: comment that says that right the content of the page dream of starting at the address this, but the the cardinal address of the Patriots

15
00:02:15.270 --> 00:02:16.680
William Cheng: Let me see, let me go there.

16
00:02:25.170 --> 00:02:25.770
William Cheng: I see.

17
00:02:26.070 --> 00:02:26.640
Yes.

18
00:02:29.100 --> 00:02:30.780
William Cheng: I don't think there's a fall like that.

19
00:02:32.310 --> 00:02:34.380
William Cheng: I think while

20
00:02:35.460 --> 00:02:36.390
Lalit Gupta: I let me tell you the

21
00:02:38.550 --> 00:02:39.870
Lalit Gupta: Guilt is sponsored

22
00:02:52.770 --> 00:02:53.190
Lalit Gupta: Really

23
00:02:54.540 --> 00:02:55.050
William Cheng: The place.

24
00:02:56.850 --> 00:02:59.850
Lalit Gupta: Maybe had a file. Okay, that's, that's the one.

25
00:03:00.930 --> 00:03:05.730
Lalit Gupta: Yeah, it seems like the content of the page fame starting at that address this.

26
00:03:06.030 --> 00:03:06.420
Lalit Gupta: Which is

27
00:03:06.450 --> 00:03:08.310
Lalit Gupta: I think Cardinal enter for the pitch for him.

28
00:03:09.510 --> 00:03:13.020
Lalit Gupta: To the page identity identified by the object.

29
00:03:13.140 --> 00:03:14.130
Lalit Gupta: And the page number

30
00:03:14.400 --> 00:03:20.460
Lalit Gupta: Then it's sort of, it seems from from looking at this thing. It seems to me we are writing

31
00:03:21.960 --> 00:03:24.120
Lalit Gupta: The writing the content.

32
00:03:25.620 --> 00:03:29.940
Lalit Gupta: Which is identified by the page object with the page number two the cardinal part of the address.

33
00:03:31.290 --> 00:03:31.710
William Cheng: To the

34
00:03:32.340 --> 00:03:40.020
Lalit Gupta: To the cardinal part of the page because we have this PDF add right so that I think correspond to the cardinal part of the address.

35
00:03:41.730 --> 00:03:42.600
Lalit Gupta: Or the page team.

36
00:03:44.970 --> 00:03:49.320
William Cheng: Right, so p of add on. That's the, that's the data for the page frame right

37
00:03:51.240 --> 00:04:02.850
William Cheng: Okay, it says, write it, you know, write the content of the page rain to the page identified by PF LBJ and page nom. So that's, you know, that's, that's the disk, right.

38
00:04:06.480 --> 00:04:13.560
William Cheng: I mean, if we if we have a fighting to address space writing into the address space, the same as writing into the file though right into the desk right

39
00:04:16.680 --> 00:04:20.670
Lalit Gupta: Oh, sorry. Oh, so this p. So, what is this

40
00:04:22.320 --> 00:04:25.650
Lalit Gupta: So this PDF or widget and PDF page number will correspond to disk.

41
00:04:26.490 --> 00:04:27.750
William Cheng: With the corresponding this

42
00:04:29.580 --> 00:04:30.330
Lalit Gupta: Okay, that's

43
00:04:31.290 --> 00:04:35.760
William Cheng: polymorphous them you know for some object and make perfect sense. So object doesn't make sense at all right.

44
00:04:38.280 --> 00:04:38.550
Lalit Gupta: Oh,

45
00:04:39.120 --> 00:04:47.670
William Cheng: And then when you have an array of function pointer. Not every object have to implement all the functions, right, we saw before, you know, like, you know, directory and files.

46
00:04:48.690 --> 00:04:54.300
William Cheng: You know that the file you can read and write, but for the directory those, you know, those, those functional valid right

47
00:04:55.710 --> 00:05:08.550
William Cheng: So same meaning of the page, if you will, you try to clean a page you want to make this page, the same as what's on the desk. What if there's no disk way. There's no death, there's, you know, then then then then what what what do you do, okay.

48
00:05:12.090 --> 00:05:12.720
He

49
00:05:14.070 --> 00:05:15.120
Lalit Gupta: Yeah properties.

50
00:05:15.180 --> 00:05:23.970
William Cheng: Represent for us this object over here right for Colonel three there's three kinds of debtors anonymous object. There's the shadow object and then there's the, the one inside the vino

51
00:05:24.810 --> 00:05:39.540
William Cheng: Rather way inside the vino makes sense that you want to make it the same as what's on the desk right so in that case. Yeah, you will use clean page and the co has been written for you already. Right. So that one is Colonel FS VN right so that's this one is clean page right

52
00:05:40.680 --> 00:05:54.090
William Cheng: Okay, so this one. What it does is that, you know, it just write the data out to the desk. Okay. But before the other two. You know, the comment. The comment is pretty weird right Curnow via that old see

53
00:05:55.830 --> 00:05:56.730
William Cheng: It says

54
00:05:58.200 --> 00:06:00.330
William Cheng: You know, the next two functions are not difficult.

55
00:06:01.380 --> 00:06:03.090
Lalit Gupta: Yeah, there's very

56
00:06:04.230 --> 00:06:12.780
William Cheng: Rapid because again we're using polymorphous them some, you know, some of these talking, you know, doesn't really make sense, but you still have to do something to make it work.

57
00:06:14.520 --> 00:06:17.580
William Cheng: Okay. Similarly, with the anonymous object Kernow

58
00:06:18.750 --> 00:06:23.430
William Cheng: VM Ananda see right at the end he says the following three functions should not be difficult, right.

59
00:06:25.410 --> 00:06:28.710
William Cheng: Okay, because these functions doesn't make sense where these two objects, right.

60
00:06:31.800 --> 00:06:33.150
William Cheng: There's really nothing to clean.

61
00:06:35.250 --> 00:06:37.410
Lalit Gupta: Yes. Seems like that.

62
00:06:38.160 --> 00:06:45.210
William Cheng: So, so it's a, it's a, it's a tiny little puzzle. Well, in that case, what are you supposed to do, right. So anyways, that's what people need to figure out what to do.

63
00:06:46.290 --> 00:06:50.670
William Cheng: Knowing that you know these. I mean, you know, in Colonel to, you know,

64
00:06:51.450 --> 00:07:01.590
William Cheng: You know, some of these functional, they don't use the null pointer and you know you you're required to test the null pointer to make sure that the pointer is not know they can actually call it

65
00:07:02.190 --> 00:07:12.480
William Cheng: But for some reason for mmm LBJ the Brown University will decide to take a different approach. They, they have all these functions, even if the functions are, you know, not not useful for this type of object.

66
00:07:15.930 --> 00:07:17.280
William Cheng: So the question is, why should you do

67
00:07:20.100 --> 00:07:26.910
Lalit Gupta: For the shadow object is it doesn't the shadow object has the ability because in the shape. Okay. We have a

68
00:07:30.540 --> 00:07:47.370
Lalit Gupta: We have a memory area and that will correspond to a chain or that will point to a chain of shadow object. Right, right. Shadow object has shadowed a jacket basically empty object that is that is that manages some pages for the pages.

69
00:07:47.670 --> 00:07:52.740
Lalit Gupta: Bad, and then we want to. And then we want to clean that page, it makes sense to

70
00:07:53.850 --> 00:07:55.050
Lalit Gupta: Write it back to the desk.

71
00:07:55.770 --> 00:07:56.430
William Cheng: Right, yeah.

72
00:07:58.770 --> 00:08:05.640
William Cheng: Right. So, but, but, but, you know, as I mentioned in class shadow object does not have a backing store. So where do you write it to

73
00:08:10.890 --> 00:08:14.820
William Cheng: Anonymous up. Dad, don't have a backing store when you write the patrons do

74
00:08:16.080 --> 00:08:16.890
William Cheng: That's the puzzle.

75
00:08:18.720 --> 00:08:19.620
Lalit Gupta: Okay, I

76
00:08:20.910 --> 00:08:24.480
Lalit Gupta: Mean, it has to be written somewhere. We call it is like important stuff.

77
00:08:25.440 --> 00:08:27.900
William Cheng: There's no backing store. There's no way to write

78
00:08:29.580 --> 00:08:32.250
Lalit Gupta: But if you don't write it, then it will be bad.

79
00:08:33.270 --> 00:08:33.810
William Cheng: Why

80
00:08:34.500 --> 00:08:35.580
Lalit Gupta: Because we have

81
00:08:36.660 --> 00:08:37.860
Lalit Gupta: Because we have a

82
00:08:39.510 --> 00:08:47.040
Lalit Gupta: Shadow Jane. And then when we do copy on write, we actually data point, there is a pointer in the shadow change that is actually pointing to the

83
00:08:48.420 --> 00:08:51.270
Lalit Gupta: I think I can probably from the slide, I can tell you better

84
00:08:51.450 --> 00:08:58.110
William Cheng: And I know. But, but, but, you know, the idea is to make the data, the same as what's on the desk, but there's no desk.

85
00:08:59.550 --> 00:09:03.390
William Cheng: You know, for shadow object reporting on those objects, they don't have a backing store.

86
00:09:06.630 --> 00:09:19.560
Lalit Gupta: Or dance that data will be because we have because when we do copy and right we actually there is a pointer that is pointing to that that that page, right, and that is the

87
00:09:20.130 --> 00:09:24.030
William Cheng: Word before the shadow objects, although you know all the patients are in memory. There's no disk.

88
00:09:25.110 --> 00:09:29.700
William Cheng: The only thing that can be on this is the, the, the white inside the the vino

89
00:09:33.150 --> 00:09:39.930
Lalit Gupta: Yes, but isn't that possible that that the pages which is on memory can be paged out by the page or demon

90
00:09:40.680 --> 00:09:41.700
William Cheng: How is it going to find it.

91
00:09:44.640 --> 00:09:45.210
William Cheng: How is the pig.

92
00:09:46.230 --> 00:09:46.740
William Cheng: Oh,

93
00:09:47.370 --> 00:09:48.840
Lalit Gupta: OK. OK.

94
00:09:50.130 --> 00:09:50.730
William Cheng: OK, yeah.

95
00:09:51.180 --> 00:09:52.650
Lalit Gupta: hide it from the pod.

96
00:09:53.220 --> 00:09:59.610
William Cheng: It's hidden on the page agreement. So there's no harm if you don't write it back to the desk, because there is no corresponding image on the desk.

97
00:10:01.620 --> 00:10:05.790
Lalit Gupta: Okay, so it's yeah look at some point we have to hide it from the page or demon. So then it

98
00:10:05.790 --> 00:10:16.890
William Cheng: can hide it. Because otherwise, when the page document found it the page demon is going to say, oh, is this one dirty, you know, if it's dirty. I need right out to the desk. Oh, there's no desk, then, then it's gonna it doesn't know what to do.

99
00:10:18.300 --> 00:10:19.170
Lalit Gupta: Okay.

100
00:10:20.250 --> 00:10:24.720
William Cheng: Well, I should be at that time. Maybe it will call clean page that hey, write it out to the desk, while in that case, what do you do

101
00:10:27.150 --> 00:10:28.110
Lalit Gupta: Nothing.

102
00:10:28.410 --> 00:10:34.110
William Cheng: Nothing. I can't really do that. Anyways, that's a little puzzle. People need to figure out what to do. Right.

103
00:10:35.700 --> 00:10:43.440
William Cheng: But nobody's talking about this in the class Google rule but i mean you know you. I mean, I cannot tell you what code right but your classmate can discuss how to solve this puzzle.

104
00:10:44.070 --> 00:10:44.700
Yeah.

105
00:10:46.800 --> 00:10:47.970
Lalit Gupta: No, actually what to right

106
00:10:48.900 --> 00:10:52.260
William Cheng: Yeah, yeah, I cannot say sorry

107
00:10:53.070 --> 00:10:57.540
Lalit Gupta: I think your work, you said is definitely helpful.

108
00:10:58.080 --> 00:11:03.900
William Cheng: All right. Yeah. So just, you know, just think about and you know some of the stuff will make sense.

109
00:11:05.490 --> 00:11:06.540
Lalit Gupta: Oh key.

110
00:11:12.840 --> 00:11:19.860
Lalit Gupta: And the Phoenix is such that we will never run out of memory, right.

111
00:11:20.370 --> 00:11:30.120
William Cheng: As far as I know, again, I don't, I don't know for sure that because I die that I didn't design this I didn't implement this, but I think with 256 megabytes of memory.

112
00:11:30.930 --> 00:11:44.970
William Cheng: I think you know if you are in Colonel three. There's a test called forkball right foreground supposed to run forever if it runs forever and nothing ever dies, then in and then eventually it's going to run out of memory.

113
00:11:46.440 --> 00:11:52.380
William Cheng: Okay, okay. But I think for bomb on you implemented correctly, then

114
00:11:52.890 --> 00:12:03.060
William Cheng: You know, it's going to keep creating child processes and then they, you know, I guess, all these policies, most of them is going to be dead. So, therefore, you get to clean up all the memory that occasion. You should never run out of memory.

115
00:12:04.260 --> 00:12:16.830
Lalit Gupta: Okay but consider this hypothetical scenario in which actually we keep on writing and writing and writing and forking and writing and 14 and writing and these are like a very long change and at some point.

116
00:12:18.840 --> 00:12:20.130
William Cheng: We're gonna run out of memory. Yeah.

117
00:12:21.030 --> 00:12:25.020
Lalit Gupta: Okay, this is a demon cannot you know because there's no backing store for the shadow on. Yeah, right.

118
00:12:25.320 --> 00:12:35.130
William Cheng: Okay, so therefore the page. I even cannot help and eventually going to run out of memory. Okay, so, so the I read the colonel FAQ, there is something called the

119
00:12:37.980 --> 00:12:39.210
William Cheng: Now I can't remember what it's called.

120
00:12:41.610 --> 00:12:44.340
William Cheng: Yeah, there's, there's a comment about something called a shadow demon

121
00:12:46.320 --> 00:12:50.970
William Cheng: Chateau de sh. A WD you can do a stream search as a

122
00:12:52.380 --> 00:12:56.400
Lalit Gupta: Consultant, I think in conflict. I remember there's a Chateau de something like

123
00:12:56.400 --> 00:13:00.030
William Cheng: This is a function called migrate okay so

124
00:13:00.390 --> 00:13:03.420
Lalit Gupta: You know, yeah. The function of migrating keyframe doxey yes

125
00:13:03.630 --> 00:13:08.160
William Cheng: Right, so I so I think we need that function is written for you already.

126
00:13:08.460 --> 00:13:09.120
Lalit Gupta: So to return.

127
00:13:09.480 --> 00:13:24.450
William Cheng: You just have to call it at the right time, because otherwise the chain of chatter object is going to go. It's going to go infinitely long and then eventually get into trouble. So therefore you should do that. You should actually shrink it if you don't, if you are preventable growing

128
00:13:25.920 --> 00:13:35.130
William Cheng: So, so if you get if you get the shadow. If you get to fork bomb and the only bug inside your kernel is that eventually will run out of memory.

129
00:13:36.270 --> 00:13:40.770
William Cheng: Then in that case that you know that that will be the bug that you have to deal with.

130
00:13:42.390 --> 00:13:44.970
William Cheng: But if you haven't got that far yet what, don't worry about it yet.

131
00:13:45.420 --> 00:13:47.310
Lalit Gupta: But we don't have to worry about shadow be right.

132
00:13:49.980 --> 00:14:03.120
William Cheng: Do saturday i think you know there are earlier version of the kind of source code from Brown University that you have to deal with the shadow. The thing I think the recent versions. I think they, you know, they, they, I think they they took care of it.

133
00:14:09.450 --> 00:14:16.650
William Cheng: Yeah, once in a while I get an update from Brown University, they say, Okay, here's the new new stuff and then there might be surprises in them or something.

134
00:14:33.930 --> 00:14:35.400
Just come on

135
00:14:40.380 --> 00:14:41.970
Lalit Gupta: Board. In the meantime, I'm just looking

136
00:14:45.060 --> 00:14:46.260
Hardik Mahipal Surana: Question professor. Good morning.

137
00:14:47.490 --> 00:14:54.120
Hardik Mahipal Surana: I actually wanted to know for phase two of Colonel three or do we really need to implement shadow objects because

138
00:14:54.660 --> 00:15:07.470
Hardik Mahipal Surana: In the weekend discussion video. You mentioned that shared objects are not only needed for phase two. But when you are running the code, we actually hit a point where some functions from shadowed see we're being called well

139
00:15:07.500 --> 00:15:10.260
William Cheng: Only the first one, only the, the, the net.

140
00:15:10.980 --> 00:15:15.360
William Cheng: Okay, so in there, but we know how to write in there, right, because that's the same for any

141
00:15:16.080 --> 00:15:16.440
Hardik Mahipal Surana: Yes.

142
00:15:16.740 --> 00:15:21.360
William Cheng: So that will be the one you have to implementing all the other ones, they will, you know, they will not get call

143
00:15:23.730 --> 00:15:24.030
Hardik Mahipal Surana: Okay.

144
00:15:25.830 --> 00:15:28.650
Hardik Mahipal Surana: Being called for some reason. And we couldn't figure out why that scene which

145
00:15:28.650 --> 00:15:30.660
Hardik Mahipal Surana: One shadow Phil page.

146
00:15:33.720 --> 00:15:36.810
William Cheng: You know, if you so so have you initialize whichever object.

147
00:15:38.310 --> 00:15:42.240
Hardik Mahipal Surana: So in the VM map underscore map function.

148
00:15:44.760 --> 00:15:57.270
Hardik Mahipal Surana: Basically fear initializing a shadow object. If the mapping is private. So, what I am doing what my team is doing is that we are checking if the mapping is private, and if it is, then we are creating a shadow object.

149
00:15:57.360 --> 00:15:58.860
William Cheng: Right, so don't don't don't do that.

150
00:16:01.050 --> 00:16:07.920
William Cheng: Right. Because if you are not implemented, let me go take a look of the colonel VM shadow that see

151
00:16:09.840 --> 00:16:17.820
William Cheng: Right. The first one over here. I think this function is called in an ID call all you know what one of the initialization function, right. So that's why you have to implement it.

152
00:16:18.510 --> 00:16:26.640
William Cheng: If you never return a shadow object right shadow create if you return. Now you should check the return value in return know then then don't create a shout out yet, right.

153
00:16:28.830 --> 00:16:39.240
Hardik Mahipal Surana: Correct. But so when we are mapping of part of the file into the address space and later on when we need to handle copy on write for the for can wait function.

154
00:16:40.140 --> 00:16:49.560
William Cheng: Right then, then you need to shout out your iPad at the beginning when you're doing. Hello, right, you still need to copy on write what you can do copy on write with the bottom of the. You don't need to shout out yet.

155
00:16:52.200 --> 00:16:54.690
Hardik Mahipal Surana: But isn't copyright always which shadow

156
00:16:55.890 --> 00:17:09.510
William Cheng: Is supposed to. Because once you know about shadow. It is a wise. Once you have the copy go into the shadow up there. But now you can actually go to the bottom object right he if the modification, going to the bottom of your last original data right

157
00:17:10.710 --> 00:17:13.860
William Cheng: But you don't need the original data anymore because because you don't have fork.

158
00:17:14.760 --> 00:17:15.150
Hardik Mahipal Surana: Correct.

159
00:17:15.420 --> 00:17:18.720
William Cheng: You're not sharing with anybody, so therefore it's your own private copy

160
00:17:19.800 --> 00:17:20.280
William Cheng: Okay, so

161
00:17:20.340 --> 00:17:22.680
Hardik Mahipal Surana: That could work only for getting face to turn

162
00:17:23.190 --> 00:17:26.310
William Cheng: That will that will only work until you get to fork.

163
00:17:27.120 --> 00:17:27.780
Hardik Mahipal Surana: Okay, okay.

164
00:17:27.810 --> 00:17:33.330
William Cheng: The first time you call for you have to implement shadow shadow object because you know you need to get forward to work with copy. All right.

165
00:17:34.590 --> 00:17:45.390
Hardik Mahipal Surana: Okay. And also we were hitting another bug relating to the loading of the file itself. So in the came in from in sorry

166
00:17:47.130 --> 00:17:51.090
Hardik Mahipal Surana: In gaming actually we call the kernel exec we eat function.

167
00:17:51.300 --> 00:17:59.760
Hardik Mahipal Surana: With a little bit low. So actually we were hitting and into an error in the bin FMT underscore Lord function.

168
00:18:00.120 --> 00:18:07.230
Hardik Mahipal Surana: So that is the function that actually goes and loads the address space into the it proc

169
00:18:09.030 --> 00:18:22.110
Hardik Mahipal Surana: In in in it processes outer space so that so we were actually using GDP to step through that. And what we found is we were able to create VM areas for the tech segment and the data plus BSS segment.

170
00:18:23.610 --> 00:18:32.730
Hardik Mahipal Surana: And we noticed a couple of things. First of all, we were getting the flags. So we use the colonel VM map mapping info GDP command.

171
00:18:34.410 --> 00:18:45.000
Hardik Mahipal Surana: It did that it printed to VM areas for us. So we stepped through the that load function. And that was actually going back into the underscore load function in elf 32 dot c

172
00:18:45.360 --> 00:18:45.750
William Cheng: Nice.

173
00:18:45.900 --> 00:18:57.000
Hardik Mahipal Surana: To that and we were able to create memories segments for the, I think. So we were able to create two segments. So I'm assuming that they were for the tech segment and the

174
00:18:57.360 --> 00:19:14.340
Hardik Mahipal Surana: Data plus BSS segment and when we were trying to create a memory segment for the stack. We were actually hating. And so it was the very end of the function there, we say that there is a comment which says that the now everything is done. We just need to send this to the user space.

175
00:19:15.990 --> 00:19:18.750
William Cheng: Okay, so let's see.

176
00:19:20.610 --> 00:19:23.490
William Cheng: It's called the VM map, map function.

177
00:19:26.250 --> 00:19:28.590
Hardik Mahipal Surana: Yes. So actually, if you could go

178
00:19:29.910 --> 00:19:31.890
William Cheng: Right here, it says allocate a stack right

179
00:19:32.850 --> 00:19:46.410
Hardik Mahipal Surana: Yes, yes. So we were able to do this. And then if you could come a little further down and towards the end of the function. So there was a comment which says now everything is done and we need just need to copy something back to the user space.

180
00:19:47.640 --> 00:19:48.510
William Cheng: Now every

181
00:19:49.230 --> 00:19:52.800
Hardik Mahipal Surana: Within the load function itself I think after we create the stack.

182
00:19:53.460 --> 00:19:56.970
William Cheng: OK, so now the return address will be fixed by the this

183
00:19:57.240 --> 00:19:58.560
Hardik Mahipal Surana: Before this, I feel

184
00:19:58.980 --> 00:19:59.730
William Cheng: Before this

185
00:20:00.360 --> 00:20:03.450
Hardik Mahipal Surana: It's very supposed to copy something you to the user space.

186
00:20:07.410 --> 00:20:08.280
William Cheng: You see here

187
00:20:10.200 --> 00:20:12.930
William Cheng: Okay, copying everything into the user space, right.

188
00:20:14.640 --> 00:20:26.370
Hardik Mahipal Surana: Yes, yes. So here what we are facing is that I see a sec. So we went into this function and we printed our pillow and we used a VM map.

189
00:20:26.790 --> 00:20:40.410
Hardik Mahipal Surana: We did a look up on the in the VM map to find the VM area in which we supposed to copy the information which starts in the address are below. So what we found is that like our VM area we're starting at

190
00:20:41.550 --> 00:21:00.930
Hardik Mahipal Surana: A VF N like 32840 and our glow was three to eight and 33 32399 so we were off by one. And because of that, we were not able to find the VM area. And that's why we sort of hit a case where we could not find the VM area.

191
00:21:01.740 --> 00:21:07.650
William Cheng: Okay, so that's what I just suggested, you know, in the class Google group you know today.

192
00:21:08.880 --> 00:21:10.920
William Cheng: You know, whenever you call BM

193
00:21:11.010 --> 00:21:11.940
William Cheng: Map function.

194
00:21:13.080 --> 00:21:13.500
William Cheng: Okay.

195
00:21:13.620 --> 00:21:21.540
William Cheng: So what I would do is that, you know, I was set a breakpoint like right here, right, or as you know I set a breakpoint right on this line. And before I call

196
00:21:22.500 --> 00:21:31.590
William Cheng: The VM app. I'm sure I will use the GDP command, you know, Colonel info VM and not being for current prod and the last argument will be using this map right

197
00:21:31.950 --> 00:21:38.550
William Cheng: It this way you print the address space. What does it look like before you call this VM app and also as soon as this function returns. So you just do a next

198
00:21:38.580 --> 00:21:44.160
William Cheng: When it goes to the next slide over here. Then again, you use the same command to say what is the I just have a look like afterwards.

199
00:21:44.550 --> 00:21:45.720
Hardik Mahipal Surana: Actually, we did do that.

200
00:21:48.000 --> 00:21:52.320
William Cheng: But you need to post it in the class Google group and ask other people, if that's what they're seeing

201
00:21:52.950 --> 00:21:53.790
Hardik Mahipal Surana: Oh, OK. OK.

202
00:21:53.880 --> 00:21:56.280
William Cheng: This way I can also comment on that. Right.

203
00:21:56.340 --> 00:21:58.470
William Cheng: Okay, I can I tell you what code right

204
00:21:58.710 --> 00:22:02.520
William Cheng: But if you show me the map. I can tell you that hey you know what it is. Look right or not.

205
00:22:04.650 --> 00:22:05.250
Hardik Mahipal Surana: Okay, sure.

206
00:22:05.730 --> 00:22:07.470
William Cheng: It is because, you know, if you only have

207
00:22:07.530 --> 00:22:09.630
William Cheng: A to memory segment or something wrong, right.

208
00:22:10.260 --> 00:22:13.680
Hardik Mahipal Surana: Correct, yeah. After this function is called we should have actually hard to memory segments.

209
00:22:13.710 --> 00:22:21.930
William Cheng: So you should do this for every VM map be a map related function VM M AP underscore something right over here.

210
00:22:22.590 --> 00:22:30.780
William Cheng: There's VMware find range. And then, I guess. Yeah, this is the only VMware function that's inside this function and this is an Elf. Elf 30 to load.

211
00:22:31.590 --> 00:22:40.440
William Cheng: There are other function over here inside elf 32 maps segment. Over here, over here, there's call BMI right here and then it will call the map right here.

212
00:22:40.740 --> 00:22:50.490
William Cheng: Okay, so, so again every time before you call the map you print the outer space, you know, for the first argument and then after the function return you print the outer space again and then we can see

213
00:22:50.520 --> 00:22:52.620
William Cheng: You know whether the address. They look correct or not.

214
00:22:55.320 --> 00:23:04.350
William Cheng: Okay, well, because I I think you're missing one of you know one of the VM map over here. And that's what it doesn't work. Okay. There's also they're doing something weird

215
00:23:04.890 --> 00:23:17.490
William Cheng: For the, the, the data plus BSS segment so often is that at the beginning, you know, you know the the data segment the psi them and there's no initial value. So the data segment come from the desk.

216
00:23:18.480 --> 00:23:23.400
William Cheng: Okay, no. Okay, so if you just map it from there does seem to memory. Why, in that case, you got, you know,

217
00:23:23.940 --> 00:23:30.180
William Cheng: In that case, it's kind of like a read only right just like tech segment. Well, you want to be able to modify your global variable.

218
00:23:30.690 --> 00:23:40.200
William Cheng: So I think the code instead of load over here. That is something funny. The cop. The, the mapping to address space and then what it will do is that they will take the memory segment and they will replace it with the new one.

219
00:23:41.280 --> 00:23:52.200
William Cheng: You know, bye bye bye bye managing and using anonymous object, okay, because we need to do copy on. Right, right. So therefore you cannot use the the demo VJ inside the inside the vino

220
00:23:52.800 --> 00:24:01.290
William Cheng: Okay, so all that code is done somewhere over here again can see the code is very, very complicated over here. So, so maybe you're not doing that process. You lost a memory segment.

221
00:24:03.360 --> 00:24:04.290
Hardik Mahipal Surana: Oh, okay.

222
00:24:04.620 --> 00:24:14.700
William Cheng: I don't know if this is happening in your code, good, but I've seen people that you know when when they finish the loader. They the memory segment looks wrong.

223
00:24:16.560 --> 00:24:29.730
William Cheng: Okay, so, so, so the ways I think the best way is that you you post it in the Google group or your GDP print out know for those property address space and then everybody can look at it to see, you know, if it makes sense or not.

224
00:24:31.860 --> 00:24:40.710
Hardik Mahipal Surana: Okay, so can you help me understand after the just load function completes, how many segments are we supposed to what is expected number three.

225
00:24:42.060 --> 00:24:50.010
William Cheng: Again, this is a reasonable thing to post to the Glass, Google group right but he's definitely the tech segment you need the data plus BSS you need to have a stack.

226
00:24:51.030 --> 00:24:52.590
William Cheng: You know, so those are the one you have to have

227
00:24:53.850 --> 00:25:06.900
William Cheng: That because in order for you to run the hello program. You don't need to heap segment right you don't have memory map file. So, you know, tax data plus BSS and you know and stack. Those are the minimum that you need. Right.

228
00:25:09.240 --> 00:25:09.570
Okay.

229
00:25:11.850 --> 00:25:17.250
William Cheng: So again, it's important to post in the group. So everybody can comment and then we can actually see that

230
00:25:17.760 --> 00:25:18.870
Hardik Mahipal Surana: Sure I'll definitely do that.

231
00:25:18.900 --> 00:25:26.670
William Cheng: Yeah, because I think a lot of people are stuck, but they don't share anything, you know, in the class move so they're missing an opportunity for people to help each other.

232
00:25:30.750 --> 00:25:32.340
Hardik Mahipal Surana: Okay sure that I've been through that.

233
00:25:32.550 --> 00:25:32.940
Okay.

234
00:25:34.020 --> 00:25:34.230
William Cheng: Yeah.

235
00:25:34.260 --> 00:25:35.490
William Cheng: Good as early as possible.

236
00:25:36.150 --> 00:25:43.140
Hardik Mahipal Surana: Sure. Another thing I noticed is that the two memory segments that that get created when we load this program.

237
00:25:43.650 --> 00:25:51.060
Hardik Mahipal Surana: One has read and execute permission and it has private mapping and the second one has read right permission and has again private mapping

238
00:25:51.570 --> 00:26:04.200
Hardik Mahipal Surana: So I was a little confused because the tech segment is supposed to be shared and read only. But I see that it has a Rx flag protection bits and mapping the flags is set as private

239
00:26:04.980 --> 00:26:08.400
William Cheng: Right so private means that you didn't copy on. Right, right.

240
00:26:09.420 --> 00:26:14.700
William Cheng: But this is read only. Never. We never right into as it doesn't hurt.

241
00:26:16.170 --> 00:26:21.120
Hardik Mahipal Surana: Exactly. But if this is tech segment, shouldn't it will just read only and share

242
00:26:21.720 --> 00:26:28.050
William Cheng: Well, I don't know why Brown University people write the code. This way they are the expert assume they know what they're doing.

243
00:26:29.130 --> 00:26:29.430
Hardik Mahipal Surana: Okay.

244
00:26:30.000 --> 00:26:31.170
William Cheng: Why, because it doesn't hurt right

245
00:26:32.190 --> 00:26:32.940
Hardik Mahipal Surana: Okay, yeah, sure.

246
00:26:33.300 --> 00:26:42.360
William Cheng: So maybe they tried to do this, making to make things look more uniform. I think lecture, I have a slide to say that you can actually stick a shadow object in there because it says is private.

247
00:26:43.260 --> 00:26:47.520
William Cheng: Right, but since you you never supposed to write into your tech segment. So therefore you should be OK.

248
00:26:49.350 --> 00:26:49.770
I see.

249
00:26:51.390 --> 00:26:59.280
William Cheng: I think in the lecture, I tried to, you know, be very specific. I say a memory segment is copy on write if it's private. And if it's rewrite

250
00:26:59.700 --> 00:27:00.090
Hardik Mahipal Surana: Yes.

251
00:27:00.360 --> 00:27:07.380
William Cheng: Right. So if it's a read only and private. Well, then in that case if you know that you're really not doing copyright right

252
00:27:09.120 --> 00:27:09.420
Right.

253
00:27:12.540 --> 00:27:25.200
Hardik Mahipal Surana: Yeah, so I had one more doubt. So if you're actually trying to read or write into of a mayor or memories that segment which has the visitor for copy on write basically it has rewrite and private setup.

254
00:27:25.920 --> 00:27:27.420
Hardik Mahipal Surana: And let's say we get a page fault.

255
00:27:28.230 --> 00:27:35.100
Hardik Mahipal Surana: In that case, checking about the mapping and doing all of the things that you have mentioned in the lecture.

256
00:27:36.180 --> 00:27:41.970
Hardik Mahipal Surana: Basically says setting the read write access, creating a copy copying the data from one page to another and then

257
00:27:42.330 --> 00:27:52.020
Hardik Mahipal Surana: Updating the page table entries. All of that stuff. Is that supposed to happen within the shadow look up page and fill page functions or some part of what's supposed to be done inside the page for itself.

258
00:27:53.160 --> 00:28:08.730
William Cheng: Well, that's part of your handle page ball right this is this is Colonel VM page. All right. Okay, this one. There's only one function. Right. Okay, so, so, so they all started off from here I handle page fault. Okay, so I

259
00:28:10.110 --> 00:28:12.090
William Cheng: Had a patient, you have to do all the stuff you just mentioned.

260
00:28:15.570 --> 00:28:15.990
Hardik Mahipal Surana: Yes.

261
00:28:16.260 --> 00:28:27.150
William Cheng: Okay, so what you would do is that you first you walk down that list of VM area, find out which memory segment it belongs to, and then see if you're going to kill the user space program or not. If you don't kill the users based program.

262
00:28:27.600 --> 00:28:33.750
William Cheng: That means that there's a way to fix it. Okay, if it turns out it's copy on write what at that point you need is a sort of perform all this copy. All right.

263
00:28:36.330 --> 00:28:37.440
Hardik Mahipal Surana: Okay, so

264
00:28:38.250 --> 00:28:39.690
William Cheng: They also added this function.

265
00:28:40.590 --> 00:28:44.340
Hardik Mahipal Surana: For the pieces for the steps that we need to do for copyright

266
00:28:45.360 --> 00:28:54.840
Hardik Mahipal Surana: Is it, are these pieces split between like are some pieces done by the page for handle page for function and other pieces done by the shadow functions.

267
00:28:55.890 --> 00:29:02.910
William Cheng: Well, the page ball handler at the pace will handle over here is the one that will call the function inside the shadow object.

268
00:29:03.000 --> 00:29:03.810
Hardik Mahipal Surana: Yes, yes.

269
00:29:04.800 --> 00:29:08.040
William Cheng: Right, so, so everything started out over here and then

270
00:29:08.220 --> 00:29:17.340
William Cheng: You know, if it turns out it's not a, you know, it's not a private member segment and it needs to use some other object. Right, so, so, so all the codes are started out started off from here I to to to to fix the page all

271
00:29:17.640 --> 00:29:26.010
William Cheng: Right, so why don't you do all that copy on write all the patient my setup. The last thing that you do is to change the table entry right that's the function PT map over here.

272
00:29:26.610 --> 00:29:27.000
Hardik Mahipal Surana: Yes.

273
00:29:27.120 --> 00:29:31.860
William Cheng: There's a bunch of PT map. Okay so PT map is basically the last thing that you have to do.

274
00:29:32.790 --> 00:29:46.050
William Cheng: And to to set up a stable entry. And then when you return from Hannah paintball, you're going to go back into the user space program and the user space. Well, I'm going to try again. And, you know, and hopefully you're not gonna, you're not going to get another page right

275
00:29:47.190 --> 00:29:47.850
William Cheng: On this page,

276
00:29:49.200 --> 00:29:49.470
Hardik Mahipal Surana: Agree.

277
00:29:53.070 --> 00:29:54.420
Hardik Mahipal Surana: I agree to the fact that

278
00:29:55.560 --> 00:30:04.620
Hardik Mahipal Surana: The shadow functions do it called from here and we do the necessary checks to identify whether this is something that requires the shadow of jet or not based on copyright or not copy and they

279
00:30:04.920 --> 00:30:13.950
William Cheng: Were so the basic logic over here. Right. Is there anything you want on the list of VM area, find out which VM area belongs to write and inside the VM era, there's a link list.

280
00:30:14.220 --> 00:30:26.460
William Cheng: Of shadow object at the end, there's a bottom object right here. So what you do is that you need to go to that list and then find the right page rank. Okay. So will you try to say I want to find the right page right well that include copy on. Right, right.

281
00:30:27.960 --> 00:30:33.270
William Cheng: Yes. Yes. Right. Because if you're supposed to do copy on write, then the patient that you find should be the one that's copied.

282
00:30:34.410 --> 00:30:34.770
William Cheng: But, but

283
00:30:36.180 --> 00:30:41.640
William Cheng: Now it's private. Right, so that should be the one that that patient with that. That should be the one I get return

284
00:30:46.500 --> 00:30:47.220
Okay, okay.

285
00:30:48.420 --> 00:30:56.730
William Cheng: Right. I mean, at a high level, the concept is is very simple. Right. You go to that VM area, you get the page frame one. How do you get the patient was complicated

286
00:30:58.890 --> 00:30:59.490
Hardik Mahipal Surana: Yes.

287
00:30:59.850 --> 00:31:00.330
Okay.

288
00:31:01.860 --> 00:31:07.230
William Cheng: Robin, the high level idea is that, you know, you just get the page RAM and you let all the other code take care of itself. Right.

289
00:31:09.450 --> 00:31:12.120
Hardik Mahipal Surana: Yeah, so I had one more question regarding op Jacob.

290
00:31:13.200 --> 00:31:22.620
Hardik Mahipal Surana: So if we were to just look at that object dump and then figure out what the actual VM areas are supposed to look like is there a way for us to do that.

291
00:31:24.600 --> 00:31:30.210
Hardik Mahipal Surana: I'm looking at the addresses and figuring out what the ranges of the VM areas should be

292
00:31:30.870 --> 00:31:40.140
William Cheng: Yeah, because you know like like even the first thing confuses me, you know, because the, you know, they have a we do the object. Damn. There's a memory segment called read only data.

293
00:31:41.130 --> 00:31:49.230
William Cheng: Okay, that our own data. So I think what we need does is that it combine that together with the text segments to make it into one segment.

294
00:31:50.310 --> 00:32:03.600
William Cheng: Okay, so I don't really know you know exactly what the winnings loader is doing. But, you know, if it's combining one segment. What does it do with the other segments and where do they come from. I think it's quite difficult to figure that out.

295
00:32:05.700 --> 00:32:17.130
William Cheng: Okay, so that that's why you know I I sort of mentioned that we should just trust that all the CO coming from Brian University. They're all perfect they know what they're doing. And then we just need to get our function. So work.

296
00:32:19.110 --> 00:32:19.800
Yes, yes.

297
00:32:22.860 --> 00:32:33.480
Hardik Mahipal Surana: Okay, so I had one last question. You mentioned that there's a good process good chance that we might face these off by one errors in the VM map functions. Right.

298
00:32:34.110 --> 00:32:52.320
Hardik Mahipal Surana: Here off by one where like if you're comparing the start VF N and the end VF N, which he started sending a notice of the VM area and four functions like his range empty or find range or remove, we need to do a lot of comparisons between a given address

299
00:32:53.340 --> 00:32:58.410
Hardik Mahipal Surana: Given starting address and ending address and the start and end of the of the VM area.

300
00:32:58.740 --> 00:32:59.820
William Cheng: Okay, so

301
00:33:00.030 --> 00:33:01.290
Hardik Mahipal Surana: In these functions.

302
00:33:02.520 --> 00:33:04.800
Hardik Mahipal Surana: What would be a good way for us to validate whether

303
00:33:05.910 --> 00:33:10.020
Hardik Mahipal Surana: You know, we have used the right comparisons between the addresses.

304
00:33:10.050 --> 00:33:16.290
William Cheng: Well, so, so the interval, always the left side is closed and the right side is open, right. You just need. You just need to be very careful. Right.

305
00:33:16.650 --> 00:33:24.000
William Cheng: I mean, some people, I think that what it is that they have, they have a team of, you know, for students, three of them think one way, one of them think. Either way they're going to be in trouble.

306
00:33:25.200 --> 00:33:27.360
William Cheng: Yes, you got it. You got to talk to each other.

307
00:33:29.700 --> 00:33:36.840
Hardik Mahipal Surana: I agree. We've been doing that and like we we aren't able to get our head around one place on how it's supposed to be.

308
00:33:37.290 --> 00:33:45.600
William Cheng: Right, so, so again, one thing that you should do is that you can post in the class, Google, this is what I address based look why we found this thing here. Does that make sense. You know, it started discussion.

309
00:33:46.950 --> 00:33:53.340
William Cheng: Okay, sure. Right. Because, you know, if I see message like that if its reasonable for me to answer. I will also answer that right

310
00:33:54.960 --> 00:33:55.560
Hardik Mahipal Surana: Okay, okay.

311
00:33:57.960 --> 00:34:00.690
William Cheng: Now, because if I don't see any questions. I mean, how

312
00:34:02.940 --> 00:34:04.470
William Cheng: Do I start a conversation

313
00:34:05.940 --> 00:34:06.600
William Cheng: It's difficult.

314
00:34:09.750 --> 00:34:10.830
Hardik Mahipal Surana: Sure, definitely do that.

315
00:34:11.190 --> 00:34:12.150
William Cheng: Thank you. So please. Yeah.

316
00:34:22.380 --> 00:34:31.020
Lalit Gupta: I have a question on this block. Can you talk about like what blocking means and things like that as something as something blocks.

317
00:34:32.100 --> 00:34:34.890
Lalit Gupta: For instance, we have some operation and this will block. What does it mean

318
00:34:35.280 --> 00:34:45.300
William Cheng: Yeah, that means that you know so so you could call a function right you're calling the function using a colonel threat right there's kind of a fall asleep, waiting for something that's called blocking

319
00:34:50.220 --> 00:34:57.840
William Cheng: Remember the, the, the state transition diagram for threat right you started out in the running state, you can go into a waiting state or the blocking state.

320
00:34:59.190 --> 00:35:02.940
William Cheng: So that means that you're waiting permutations, you're waiting for it. Oh, you're waiting for something, right.

321
00:35:04.830 --> 00:35:11.880
Lalit Gupta: Okay, so then we are blocking. We actually are not running some from someone else's running. And at some point, that some operation will be finished. And we will

322
00:35:13.770 --> 00:35:20.490
William Cheng: Go to sleep right at some point somebody is going to change your state into the renewables day bye bye bye picking you up from from wherever you are.

323
00:35:20.730 --> 00:35:24.750
William Cheng: Moving into the run queue and then you wait for your chance to start running out of CPU. Yeah.

324
00:35:31.530 --> 00:35:37.920
William Cheng: Yeah, so if you simply give up the CPU add yourself to the rank you you are not in a blocking state you're you know you're in a reasonable state.

325
00:35:42.870 --> 00:35:43.080
Oh,

326
00:35:46.290 --> 00:35:46.650
Lalit Gupta: And

327
00:35:48.690 --> 00:35:52.950
Lalit Gupta: What's the difference between the page 14 Colonel one in color too because we

328
00:35:55.050 --> 00:35:59.880
Lalit Gupta: One can get to a fault in Colonel 102 by two. And what's the difference between the page for editing Colonel three

329
00:36:00.780 --> 00:36:12.150
William Cheng: Wow, so, so, so kind of one into the distance, there is no user space program, the only kind of pace or you can get as a colonel baseball right and Colonel baseball is a kernel panic. So therefore, you're kind of crashed.

330
00:36:12.840 --> 00:36:16.620
William Cheng: Okay and Colonel three, there are two kinds of pace while you can get a user space page wall.

331
00:36:16.950 --> 00:36:25.170
William Cheng: Or Colonel page for the colonel peaceful and kernels, who is the same as current one and twos. Colonel page fall. You're not supposed to get a colonel patient, you're, you're never supposed to get a cardinal baseball

332
00:36:26.190 --> 00:36:32.190
William Cheng: Okay, so therefore in Canarsie. There's also a user space page up have a page fall. So when there's a user space page, you should get here.

333
00:36:36.480 --> 00:36:42.000
William Cheng: Well as you can actually look for the code over here. So let's say that I'm going to do a graph right grab handle baseball

334
00:36:43.170 --> 00:36:49.680
William Cheng: Star that see Colonel mm page table does see

335
00:36:52.530 --> 00:36:59.220
William Cheng: Where, so this is in the pristine corner. So this is the only place that it calls the you know the handle baseball

336
00:36:59.880 --> 00:37:02.580
William Cheng: So we here. So this is the page ball handler. Right.

337
00:37:03.060 --> 00:37:08.580
William Cheng: That comes in over here. So what does it do, is that he checked the cause, to see if the cause is a user page fall

338
00:37:08.820 --> 00:37:14.670
William Cheng: It was a user space page. Well, then in that case it called a page or handler, because you're you're supposed to handle the paywall

339
00:37:15.090 --> 00:37:27.180
William Cheng: If it turns out the user ID is not said that means that it's a colonel page fall they nutcase it cause a kernel panic right so this is the Colonel, one and two kind of went into the always always get here, Colonel. So you can get here if we get a colonel page far

340
00:37:29.310 --> 00:37:42.810
Lalit Gupta: And wide and why does this function, underscore, underscore fault underscore handler is called like what's what could be the one of the reason that this should be called

341
00:37:43.230 --> 00:37:47.250
William Cheng: You know V zero you try to write to a read only. No.

342
00:37:47.610 --> 00:37:49.920
Lalit Gupta: No in the in the in the cardinal one into

343
00:37:51.600 --> 00:37:57.750
William Cheng: Including one and two, you know, because the for the pace table, the top, you know, the top

344
00:37:58.230 --> 00:38:11.250
William Cheng: Three quarters are not being used. Right. So the top one quarter, they're all equal to zero then the bottom part over here is going to be used by the Colonel. So if you try to access memory location, the user space one in this case, again, you're gonna you're gonna be here.

345
00:38:13.080 --> 00:38:26.400
Lalit Gupta: Okay, so, so I think this one is because I remember this one because this is the that happens when we, when we do some record. Another thing is when you do recursive thing and the Stack Overflow, then I think this is quiet because we are actually. Oh.

346
00:38:26.610 --> 00:38:38.970
William Cheng: Yeah, so I don't, I don't know, where does it go when the stack overflows, you know, because this, this page far I think some people actually use that term yesterday in the class Google group. They call it the mm you interrupt.

347
00:38:40.290 --> 00:38:52.500
William Cheng: So when MMU when they have a new detect a fault right because we go to zero because you had to write to a read only. Things like that. So, in that case it will generate a hardware interrupt. So this is the interrupt handler for, you know, for the MMU fault.

348
00:38:55.560 --> 00:38:57.150
William Cheng: Okay, so it's kind of like a hardware interrupt.

349
00:39:03.540 --> 00:39:06.540
Lalit Gupta: And vendors. This. Mm. Your fault renders me you

350
00:39:07.560 --> 00:39:08.730
Lalit Gupta: Try to do a fault.

351
00:39:09.330 --> 00:39:12.480
William Cheng: Yeah, when you perform address translation. Now you try to convert a virtual

352
00:39:13.530 --> 00:39:18.360
William Cheng: virtual address to a physical address. So you can do that for the colonel and also for the user space program. Right.

353
00:39:18.630 --> 00:39:30.960
William Cheng: So we can probably address translation, if it turns out the page table entry is invalid because we go to zero or you try to write to a read only page. So in that case, you know, you end up getting this interrupt.

354
00:39:34.050 --> 00:39:52.140
Lalit Gupta: Key in the FAQ, I think you mentioned, like how to how to fix the page fault. And so, in, in some cases, it is possible to actually figure out at which, at what at what place. Actually, the page fault occur and sometimes it

355
00:39:52.290 --> 00:40:02.520
William Cheng: So, okay. So this is, you know, when a call I know. Baseball is going to come with the DDR right so VAD tell you which virtual address is the one that caused the page, while during address translation.

356
00:40:03.870 --> 00:40:14.160
William Cheng: Okay. So, therefore, you know, that why you know has a corresponding page table entry. That means that the page table entry over there has something that's incompatible and that that that's the table. Enter, you have to fix.

357
00:40:16.050 --> 00:40:30.030
Lalit Gupta: Okay, but sometimes it is not possible. Like sometimes some sometimes it gets Douglas, the place where the page for the function or it can, but sometimes it just couldn't find the place where pitch world.

358
00:40:30.120 --> 00:40:38.430
William Cheng: Well, forget, for example, right. You be add or equal to zero, right. So if you're using a space program, try to access memory location zero. Well, then in that case you have to kill the user space program.

359
00:40:40.230 --> 00:40:43.650
William Cheng: Right. So, therefore, in that case this handle page should not return

360
00:40:46.740 --> 00:40:48.150
Lalit Gupta: Okay. Oh, so it's okay.

361
00:40:48.210 --> 00:40:48.780
Lalit Gupta: So when

362
00:40:48.900 --> 00:40:55.320
William Cheng: He returned to us. We're gonna try again. Right. If it's access the memory location zero then then there shouldn't try again.

363
00:40:56.760 --> 00:41:02.790
Lalit Gupta: Okay, so if the really the idea is corrupted. Then he can't even figure out where the page for a

364
00:41:03.270 --> 00:41:15.840
William Cheng: While so will you will. Yeah. So, so this is the VAD Dr. So can you try, you try to walk down that linear list of VM area. Okay, if it turns out this the DVR is in the gap. Well, that means that you have to kill the user space program. Right.

365
00:41:19.200 --> 00:41:25.230
William Cheng: Okay, if it's within a within a VM area. What that means that there is a way for you to fix it. Hopefully,

366
00:41:27.570 --> 00:41:35.940
William Cheng: Good. But this now within the VM area I want in that case it clearly like memory location zero. It's not inside the VM area. Well, then in that case, you'll kill the program.

367
00:41:38.970 --> 00:41:40.500
William Cheng: And then that that's a segmentation call

368
00:41:41.490 --> 00:41:57.390
Lalit Gupta: Yes, definitely understand trying to like make sense of like why it's why sometimes, it, the, the method you described with the FAQ, and he actually figure out where the pitch, what has occurred, and why, in some cases, it can't tell you where to pitch, what has occurred.

369
00:41:58.470 --> 00:42:01.080
William Cheng: When you get a paper, paper right at the DDR

370
00:42:04.350 --> 00:42:15.900
William Cheng: The hardware knows exactly where be a DDR is right over here and over here. It says, you know, move car to register so car to register, I guess, according to this code is the page fall dress.

371
00:42:17.580 --> 00:42:32.220
William Cheng: Okay, okay. So, therefore, what it does is that it copied his value into, you know, this local variable called the DDR right and then it passes to handle baseball. So now you know this is the the virtual address that caused the page ball. So go ahead and fix it, right.

372
00:42:34.230 --> 00:42:37.860
William Cheng: And if it's unfixable then you do something about it. Yeah.

373
00:42:40.170 --> 00:42:41.190
William Cheng: Go key.

374
00:43:02.700 --> 00:43:04.440
Lalit Gupta: Can you go to this lecture 16 slide.

375
00:43:04.440 --> 00:43:13.680
2016

376
00:43:15.360 --> 00:43:16.170
Lalit Gupta: Your quantities.

377
00:43:27.300 --> 00:43:27.870
William Cheng: This one.

378
00:43:29.880 --> 00:43:30.600
Lalit Gupta: Quantity.

379
00:43:30.690 --> 00:43:32.040
Two 323

380
00:43:35.010 --> 00:43:35.670
William Cheng: This one.

381
00:43:35.820 --> 00:43:41.250
Lalit Gupta: Yes, so why there's a gap between this step one, step two, step three, like this.

382
00:43:42.330 --> 00:43:47.790
William Cheng: Probably knows space, you know, I just draw that way, doesn't it doesn't really matter. I mean, the, you know,

383
00:43:49.320 --> 00:43:52.560
Lalit Gupta: But it would be like overlapping make for for one function.

384
00:43:53.580 --> 00:43:54.030
Lalit Gupta: Packages

385
00:43:54.150 --> 00:43:56.250
William Cheng: On another memories that one can never overlap.

386
00:43:58.200 --> 00:44:00.930
Lalit Gupta: No, I don't think the SEC one sec, one sec to instruct tree.

387
00:44:01.500 --> 00:44:03.120
William Cheng: Right. They can be right next to each other.

388
00:44:03.810 --> 00:44:09.690
William Cheng: Okay, just like text and data and be there right next to each other. The there can be gaps.

389
00:44:10.740 --> 00:44:12.870
William Cheng: Okay, but that they can overlap.

390
00:44:15.300 --> 00:44:22.020
Lalit Gupta: Okay. And where does the heat comes from the reversal order of the heap in this one.

391
00:44:23.160 --> 00:44:28.860
William Cheng: Well, again, the compiler decides, right. So the heat. Typically, you know, the heat typically goes right here. Right.

392
00:44:30.690 --> 00:44:45.060
William Cheng: Okay, okay, actually for our Colonel Simon the heap starts at the end of the BSS segment over here at this picture says BSS a dynamic. So typically we say be data plus BSS right here. So here's the dynamics, like that.

393
00:44:47.520 --> 00:44:49.230
Lalit Gupta: Oh ok dynamic

394
00:44:49.740 --> 00:44:56.640
William Cheng: Right, so I think I have a typo right here because the dynamic, you know, if you don't call Malik and free. You don't need a dynamic region right so in that case it doesn't have to exist.

395
00:44:59.130 --> 00:45:14.370
William Cheng: Okay, so, so in the Linux kernel the dynamic region starts at the end of the BSS segment. So if data plus BSS if it's not multiple have, you know, a full four kilobytes. So in that case, the dynamic region. Start in the middle of a pain.

396
00:45:15.390 --> 00:45:18.510
William Cheng: Starts in the middle of the page rank. So it's kind of weird.

397
00:45:21.030 --> 00:45:23.010
Lalit Gupta: Can you just you just one more time.

398
00:45:23.730 --> 00:45:34.500
William Cheng: Yeah. So let's say that you know we have a patient like this, right. So, so this is you know data plus BSS right so what if data plus BSS is only one kilobytes in size, right. So here's one kilobyte over here, right.

399
00:45:34.710 --> 00:45:40.620
William Cheng: And this entire data structure over here is what kilobytes. Yes. So in that case, your dynamic region starts right here.

400
00:45:41.400 --> 00:45:42.600
Lalit Gupta: What why

401
00:45:43.890 --> 00:45:45.060
William Cheng: It's an internal FAQ

402
00:45:46.800 --> 00:45:53.010
Lalit Gupta: You actually i i actually looking into this rubric and it's actually mentioned that, that this

403
00:45:54.300 --> 00:45:55.740
Lalit Gupta: That this page.

404
00:45:56.760 --> 00:45:57.090
Lalit Gupta: This

405
00:45:59.040 --> 00:46:00.750
Lalit Gupta: Page Break in the rock.

406
00:46:01.260 --> 00:46:07.800
William Cheng: Right at the end of this is called the process break why this is the br k, right. This is the end is over here so you can actually change it.

407
00:46:08.670 --> 00:46:10.440
Lalit Gupta: Yeah, and also

408
00:46:10.860 --> 00:46:14.370
William Cheng: Like what the star Greg is always at the end of the BSS segment.

409
00:46:15.990 --> 00:46:17.940
William Cheng: So the star break is not page aligned

410
00:46:24.210 --> 00:46:26.070
Lalit Gupta: Why did you select to do that like

411
00:46:26.580 --> 00:46:29.370
William Cheng: I don't know i again i did is i mean exactly

412
00:46:30.300 --> 00:46:31.920
Lalit Gupta: makes things more complicated.

413
00:46:32.580 --> 00:46:33.390
William Cheng: I don't know.

414
00:46:36.420 --> 00:46:40.350
William Cheng: Maybe that's how it's done in you know all the units and the Linux operating system.

415
00:46:41.520 --> 00:46:41.970
Lalit Gupta: Okay.

416
00:46:44.460 --> 00:46:58.800
William Cheng: I mean, you know, in the good old days memory is very, very precious you want to save as much memory as possible. So, you know, wasting memory instead of four kilobytes. I mean, in the good old days every, every page is happy kilobyte right so they're trying to sort of minimize internal fragmentation.

417
00:47:01.290 --> 00:47:04.170
Lalit Gupta: Okay, so is this dynamic dynamic region.

418
00:47:06.510 --> 00:47:08.370
Lalit Gupta: Read, write and private

419
00:47:09.150 --> 00:47:09.930
William Cheng: Well, yes. Yeah.

420
00:47:12.690 --> 00:47:14.880
Lalit Gupta: So then what about what about this data and

421
00:47:15.120 --> 00:47:17.640
Lalit Gupta: The second segment is this Detroit and private

422
00:47:18.180 --> 00:47:19.830
William Cheng: Yes, right.

423
00:47:20.190 --> 00:47:23.640
William Cheng: And again, again, the data segment is funny because originally the comp on the desk.

424
00:47:24.180 --> 00:47:31.860
William Cheng: Okay, so first we map the data from the test into memory and then we replace it with the private you know memory segment.

425
00:47:35.760 --> 00:47:42.030
William Cheng: Right at the beginning. It's kind of like read only. Right. You know, because the data over here contained global variable. But you want to make make make them into private

426
00:47:42.300 --> 00:47:51.420
William Cheng: So the way that's done in weeks is that you map into memory. So you can actually copy the data and then you switch the memory segue into into the one that's backed by the anonymous object.

427
00:47:52.530 --> 00:47:53.850
William Cheng: Because you want to do copy on write

428
00:48:09.480 --> 00:48:14.880
William Cheng: But again, you know, if you asked me, why did they do it that way. I don't know. That's how it's done.

429
00:48:17.550 --> 00:48:20.670
Lalit Gupta: Yeah, so when I'm not even that this is what this actually

430
00:48:24.630 --> 00:48:34.440
Lalit Gupta: Does a hypothetical question. Will. Will this strategy work if. Both. Both. Both. The, the data was BSS and the dynamic has different

431
00:48:35.640 --> 00:48:40.350
Lalit Gupta: Permissions like one is only one is read on, let's say data plus pieces are read only.

432
00:48:41.820 --> 00:48:43.170
William Cheng: The programming model.

433
00:48:43.500 --> 00:48:54.180
William Cheng: Why I mean you know that that we have learned for all the programming language, they all have the same thing, right, you know, the data plus B says it was the copy on write the dynamics of the copyright the stack supposed to be happy. All right.

434
00:48:55.830 --> 00:49:03.690
William Cheng: So as on until one day somebody actually completely change the programming model so far, you know, the others pretty good approach that seems to work.

435
00:49:10.830 --> 00:49:15.810
Lalit Gupta: And I think there's some, I think there's some point, you were saying, that's this big can increase or decrease

436
00:49:17.340 --> 00:49:18.750
William Cheng: Or increase or decrease this

437
00:49:19.110 --> 00:49:21.330
Lalit Gupta: Just break this VR ke

438
00:49:22.050 --> 00:49:31.650
William Cheng: Y. So the big A CAN CAN CAN CAN GO BACK TO, YOU KNOW, where's the star break. By the way, again, that's the end of the PSS that man so that part can never change.

439
00:49:33.840 --> 00:49:36.810
William Cheng: Right so PR the PR cake, and I should go all the way to the to the beginning.

440
00:49:39.450 --> 00:49:41.130
Lalit Gupta: There's like a relic does that

441
00:49:42.240 --> 00:49:47.670
William Cheng: I don't think anybody would do that. Right. But this is this your heat right you have a heat, we feel calling Malcolm free

442
00:49:47.880 --> 00:49:54.750
William Cheng: If you're calling Malik I'm free. Well, then you need memory. So, you know, even though it can happen. It doesn't make sense that it will ever happen.

443
00:49:55.680 --> 00:50:03.060
Lalit Gupta: Why, because let's say I do Malik Malik Malik continuously says that I am reached the end of the break should or shouldn't that

444
00:50:04.140 --> 00:50:07.560
Lalit Gupta: Like at some point, increase, increase the uptake.

445
00:50:08.220 --> 00:50:19.380
William Cheng: Yeah, so, so, so you will increase the break right but you probably going to do it much, much earlier. Right. You can't really wait until you run out of brave because at that time, you're gonna get a segmentation fall and the opposite is going to kill your program. So it'll be too late.

446
00:50:22.470 --> 00:50:28.860
Lalit Gupta: Why, why, because I have. I have. I have a gap that is the difference. I have, I have

447
00:50:30.360 --> 00:50:33.810
Lalit Gupta: A gap available. That is the great minus to start break

448
00:50:34.830 --> 00:50:38.790
William Cheng: If this is your, this is your heat memory. Uh huh.

449
00:50:39.030 --> 00:50:46.200
William Cheng: Okay, the heat memory. You gonna give it to the first member member alligator to allocate so at some point the First Amendment allocate is gonna say I'm out of memory.

450
00:50:47.580 --> 00:50:54.060
William Cheng: Okay. So supposedly one thing they should do at that point is to increase the break over here to be larger. What if you can do it.

451
00:50:56.460 --> 00:51:00.000
Lalit Gupta: Yeah, then you have to kill the program probably

452
00:51:01.200 --> 00:51:03.990
William Cheng: You know i mean i think i think i think

453
00:51:05.250 --> 00:51:09.060
Lalit Gupta: It should be actually kind of turn on that we don't have any memory.

454
00:51:10.170 --> 00:51:14.550
William Cheng: Right then they'll, then what is the problem. Gonna do you try to allocate memory. There's no memory.

455
00:51:20.400 --> 00:51:24.930
William Cheng: When you put a message to tell the tell the user that the amount of memory. Look, what can you do

456
00:51:25.980 --> 00:51:26.580
William Cheng: Yes.

457
00:51:26.940 --> 00:51:35.640
Lalit Gupta: But should we also die either, but that should we know that would be very right. Like, if we have no memory and we say, okay, you have no memory and that seems very bad.

458
00:51:37.260 --> 00:51:38.430
William Cheng: Not dependent upon your point of view.

459
00:51:41.100 --> 00:51:42.930
William Cheng: Depends on what your program is right.

460
00:51:43.830 --> 00:51:46.890
William Cheng: For the banking application they use a while, probably not.

461
00:51:49.170 --> 00:51:53.640
Lalit Gupta: Okay. Works. Works, the worst model for this genomics.

462
00:51:54.780 --> 00:51:56.880
Lalit Gupta: What is the point of view of Unix

463
00:51:57.150 --> 00:51:58.020
William Cheng: That should never happen.

464
00:51:59.970 --> 00:52:01.230
Lalit Gupta: Or we should never know not only

465
00:52:02.040 --> 00:52:09.990
William Cheng: By. I don't think there's any code to test this condition we run out of actually I think in in the in the textbook and map.

466
00:52:11.250 --> 00:52:15.210
William Cheng: Men cares. There's a user function called men. Men test.

467
00:52:16.230 --> 00:52:20.670
William Cheng: That I think he mentors. One of them has actually used up the entire address space.

468
00:52:22.200 --> 00:52:22.620
William Cheng: I think

469
00:52:24.900 --> 00:52:31.350
William Cheng: So in that case, even if you try to, you know, allocate one more page or try to do, you know, I tried to

470
00:52:32.160 --> 00:52:40.860
William Cheng: Call the function to say fine range and you just want to find one page and he will return failure because there's nothing available in the address space with the entire address basis map.

471
00:52:44.790 --> 00:52:46.140
William Cheng: Yeah, so I think there's a test for that.

472
00:52:47.400 --> 00:52:56.130
Lalit Gupta: Okay, so. Okay, so, so, so are you saying that this time. This test will should be able to tell us whether we should just say no memory, or rather, should we die.

473
00:52:56.850 --> 00:53:03.480
William Cheng: Well, the this test, test that condition. Right. So in that case, they will make sure that when you try to find, you know, when we try to create another page.

474
00:53:03.690 --> 00:53:08.940
William Cheng: I will we try to create another memories that man is that why there's no way to create this memory segment, because the entire address basis us

475
00:53:11.700 --> 00:53:15.660
Lalit Gupta: Yes, we can tell that. But should we like kill, kill the program.

476
00:53:16.440 --> 00:53:18.930
William Cheng: No, no, no. You're supposed to, you know, pass the test.

477
00:53:20.040 --> 00:53:27.810
William Cheng: Okay, so in that case you return now and then. It's a great huge return because that's what's expected because you supposed to run out of memory, right. And then, then in that case you pass the test.

478
00:53:29.670 --> 00:53:33.150
Lalit Gupta: Okay, so in this in this this dynamic region there.

479
00:53:34.530 --> 00:53:40.290
Lalit Gupta: Are two kind of elevator. One is the elevator and the other one is the body applicator.

480
00:53:41.040 --> 00:53:46.920
William Cheng: No no no no no no user space. They always use first fed the buddy system is using the Colonel.

481
00:53:48.690 --> 00:53:52.200
Lalit Gupta: Okay, so okay so this dynamic region is actually first one

482
00:53:52.500 --> 00:53:53.010
Yes.

483
00:53:58.350 --> 00:54:08.940
William Cheng: So it is possible that you're here is because of, you know, external fragmentation that in the end he will return. No, not because you run out of memory, but because you know the fragments are too small to be useful.

484
00:54:10.440 --> 00:54:19.800
William Cheng: Right. Or maybe the you know the the the the the the we can call Malik, you try to call them out on 2 billion or something like that. Well, then in that case. Yeah. If you fail.

485
00:54:21.090 --> 00:54:24.450
Lalit Gupta: Okay, so this dynamic region is is this dynamic region.

486
00:54:25.500 --> 00:54:26.730
Lalit Gupta: Managed by Kamala

487
00:54:27.390 --> 00:54:30.900
William Cheng: Know this, this is managed by the memory alligator, which is Malaga free

488
00:54:34.860 --> 00:54:35.820
William Cheng: This is user space.

489
00:54:36.960 --> 00:54:41.520
Lalit Gupta: Oh, yes. Oh, yes, it's Malik, but in the US, the users.

490
00:54:42.570 --> 00:54:45.300
William Cheng: Right now, like is done by the first memory allocated.

491
00:54:47.340 --> 00:54:50.580
Lalit Gupta: And what about the Kamala in the corner space.

492
00:54:51.750 --> 00:54:53.580
Lalit Gupta: Because they don't see Malik in the corner space.

493
00:54:53.640 --> 00:55:07.050
William Cheng: I don't know exactly what came out does because, you know, usually in the kernel, you use the slab alligator, will you use the the the page frame alligator alligator one page at a time.

494
00:55:08.730 --> 00:55:11.700
William Cheng: Yeah, I think, hey Malik is being used somewhere. I don't. Yes.

495
00:55:11.730 --> 00:55:13.980
Lalit Gupta: It does. I pretty, pretty sure

496
00:55:15.870 --> 00:55:21.480
William Cheng: Yeah, so I think came out like just call you know the the buddy system to allocate, you know, multiple pages.

497
00:55:22.440 --> 00:55:23.490
William Cheng: Oh so came out again.

498
00:55:23.850 --> 00:55:26.790
William Cheng: So you're gonna end up with the with the internal fragmentation.

499
00:55:28.380 --> 00:55:32.220
Lalit Gupta: Okay, so okay so okay Malik behind the scene is actually body electrician only

500
00:55:32.370 --> 00:55:39.660
William Cheng: Yeah, so I think if you say came out five kilobytes, it will return to you, you know, two pages that are contiguous or you end up with eight kilobytes or

501
00:55:42.300 --> 00:55:47.850
William Cheng: Maybe in the device driver, they call came out when they need to allocate a buffer I because there's no snap, snap alligator for them.

502
00:55:56.460 --> 00:56:00.360
William Cheng: And also just call came Allah. You want a four kilobytes. Right, so you can also do that.

503
00:56:02.940 --> 00:56:10.230
Lalit Gupta: Yeah, but then it's probably due to initialize us. It's better to initialize a slab alligator. If we know already that we

504
00:56:11.250 --> 00:56:15.030
William Cheng: Will use the slab alligator only if it's a commonly used Colonel data structure.

505
00:56:15.480 --> 00:56:16.290
Oh,

506
00:56:19.170 --> 00:56:30.360
Lalit Gupta: Actually, I was thinking I was thinking why in this idea because now you know you told me that came out like his body alligator. But I had previously thought that came alligator alligator.

507
00:56:30.990 --> 00:56:34.350
William Cheng: So you said the Colonel, you can, you know, you

508
00:56:35.430 --> 00:56:42.120
William Cheng: Have to use a memory allocated as good performance for us to a memory allocated order n. That's not acceptable inside a carnal.

509
00:56:48.750 --> 00:56:55.320
Lalit Gupta: I have, can you go to this have a similar similar question on on this thing a vertex in

510
00:56:56.640 --> 00:57:00.000
Lalit Gupta: The function elf category 32 dot see

511
00:57:00.810 --> 00:57:02.520
William Cheng: Okay, let me switch back

512
00:57:12.480 --> 00:57:15.090
Lalit Gupta: Yes, so I think

513
00:57:20.190 --> 00:57:22.110
William Cheng: We have a line number you can give me a line number

514
00:57:23.070 --> 00:57:24.090
Lalit Gupta: Yeah, I should have written there.

515
00:57:30.330 --> 00:57:31.200
Lalit Gupta: Okay, thank you.

516
00:58:03.300 --> 00:58:04.560
William Cheng: So what does that line look like

517
00:58:06.480 --> 00:58:12.330
Lalit Gupta: It's like a car Brock arrow be underscore be RK s i think

518
00:58:13.110 --> 00:58:15.090
Lalit Gupta: The arcade. Yeah.

519
00:58:16.170 --> 00:58:18.600
William Cheng: Okay, are we here. Yeah. So this right here.

520
00:58:19.050 --> 00:58:20.100
Lalit Gupta: You're here it is.

521
00:58:21.810 --> 00:58:22.200
Lalit Gupta: So,

522
00:58:24.240 --> 00:58:30.750
Lalit Gupta: So x via I was basically wondering why it seems like we have no keep

523
00:58:33.630 --> 00:58:38.010
William Cheng: Looking at this, yeah. So star break over here is the, you know, it's the end of the

524
00:58:38.040 --> 00:58:39.150
BSS right

525
00:58:40.170 --> 00:58:42.330
Lalit Gupta: Yes, and the br K is also the same thing.

526
00:58:43.380 --> 00:58:43.800
William Cheng: Right.

527
00:58:45.270 --> 00:58:47.580
William Cheng: Was at the beginning. The, the, he was empty.

528
00:58:56.640 --> 00:59:10.380
Lalit Gupta: Okay, so I what is if you can. Do you know what is this data. Data variable Prague, an alert digital. Hi. There's also probably know this is probably pirate, there is a problem variable. So, in this same function.

529
00:59:11.010 --> 00:59:11.400
Okay.

530
00:59:12.450 --> 00:59:13.080
William Cheng: Hi.

531
00:59:13.890 --> 00:59:15.990
Lalit Gupta: Tom. Hi, is their problem. The problem also

532
00:59:17.070 --> 00:59:25.590
William Cheng: Right, so it says elf 32 characters. I mean, you know, we just have to trust that this code works. You know, we can't really figure out how everything works. So be they'll take too long.

533
00:59:26.460 --> 00:59:30.960
Lalit Gupta: Okay. All right, can, can you go to this function. This calc prog bands.

534
00:59:32.580 --> 00:59:33.480
William Cheng: Well, I mean,

535
00:59:34.230 --> 00:59:36.690
Lalit Gupta: There's a comment that I want to make sure

536
00:59:39.420 --> 00:59:41.070
William Cheng: I mean this is readable code.

537
00:59:41.310 --> 00:59:41.970
Yeah.

538
00:59:43.140 --> 00:59:46.020
William Cheng: So why try to figure this out, you know,

539
00:59:47.790 --> 00:59:53.070
Lalit Gupta: I'm just trying to a non content and not trying to understand this code. I'm trying to understand the comment.

540
00:59:54.300 --> 01:00:00.210
William Cheng: Yeah, I mean, I, I've never read this. I don't think we should care which you just trust that it works. And that's that.

541
01:00:03.360 --> 01:00:07.620
Lalit Gupta: Okay, what okay so even then, what is this program header table.

542
01:00:10.230 --> 01:00:13.800
William Cheng: You all have 32 right the executable files in the elf 32 format.

543
01:00:14.130 --> 01:00:27.810
William Cheng: So, you know, again, the binary file is represented some kind of a data structure that describes the address space, right. So, this we also know that this code is for parsing the the elf 32 file format.

544
01:00:29.130 --> 01:00:36.150
William Cheng: If it's a different file format, there's a different way to calculate all these kinds of stuff. Right. So again, this is, you know, specific to the elf 32 file format.

545
01:00:37.500 --> 01:00:37.830
Lalit Gupta: Okay.

546
01:00:38.010 --> 01:00:40.530
William Cheng: And again, anybody should care what the elf 32 pounds.

547
01:00:40.530 --> 01:00:44.400
William Cheng: Format is unless somebody really wants to learn and deployment.

548
01:00:45.990 --> 01:00:47.910
Lalit Gupta: In general, what is heritable means

549
01:00:49.350 --> 01:00:57.900
William Cheng: What it's a, you know, the fall is a data structure. I mean, whatever they want to call it, you know, they all have 32 they call it header for, I don't know. I mean, doesn't really matter.

550
01:00:58.590 --> 01:00:58.980
Okay.

551
01:01:00.060 --> 01:01:03.780
William Cheng: I mean, it's the terminology specific to the elf 32 format.

552
01:01:05.070 --> 01:01:11.520
William Cheng: Okay, so if you really want to become an expert in yellow 32 format. Well then you try to understand this. But, but, you know,

553
01:01:12.540 --> 01:01:15.330
William Cheng: Some clouds were not interested in you know binary format.

554
01:01:17.220 --> 01:01:23.130
William Cheng: Almost like you know you want to go to a JPEG file right JPG file, what impact. What is the header look like, Well, I don't know.

555
01:01:24.240 --> 01:01:28.380
William Cheng: We didn't we are if you use the JPEG harsher you trust that it will do the right thing, right.

556
01:01:41.760 --> 01:01:48.990
Lalit Gupta: I'm basically trying to like in from from this. Actually, I was trying to basically understand like, who is actually changing this br

557
01:01:50.850 --> 01:02:05.160
William Cheng: Yeah, it's the user space program. Right, so, so, so if you look for be okay right so grab the RK user user been start RC. Okay. Nothing there is doing that. So how about user live

558
01:02:06.210 --> 01:02:07.080
William Cheng: User

559
01:02:08.220 --> 01:02:17.190
William Cheng: Lip. Lip see Star Darcy. Oh, okay. So inside Malik. So when you call Malik. That's when it called break

560
01:02:19.200 --> 01:02:19.710
Lalit Gupta: Okay.

561
01:02:19.770 --> 01:02:25.080
William Cheng: Well, it kind of makes sense, right, because you never call mallard. That means that you don't need a heat, but the first time you call Malik.

562
01:02:25.980 --> 01:02:34.380
William Cheng: That's when the heat is get created. Yes. So in that case, you will make the br cases some calls right at the beginning, you know, the heap size is equal to zero.

563
01:02:34.920 --> 01:02:42.180
William Cheng: So now, you would just, you know, make it bigger. Okay, so, so, so, again, the first time you're gonna you're gonna call Malik is in s been in it.

564
01:02:43.110 --> 01:02:54.000
William Cheng: Okay, so if you follow the grading guideline right you pass on that program where you get to spin in it as been in is the one that will call Malik. OK. OK. So, at that time, you need to get the process break to work.

565
01:02:55.500 --> 01:02:55.890
Lalit Gupta: Oh,

566
01:02:58.770 --> 01:03:09.900
William Cheng: Yeah. So I guess the breakdown to work and also the do em map function has to work so you so you need to handle the brake system call need to handle the end map system call you know when you try to get the s been in it to work.

567
01:03:14.250 --> 01:03:14.670
Okay.

568
01:03:18.000 --> 01:03:21.030
Lalit Gupta: And can you also go to this a peek m dot V

569
01:03:24.390 --> 01:03:24.690
William Cheng: Okay.

570
01:03:25.710 --> 01:03:29.190
Lalit Gupta: Yeah and there's a functioning PC get resident

571
01:03:32.490 --> 01:03:35.730
Lalit Gupta: Underscore, underscore, underscore,

572
01:03:37.410 --> 01:03:38.130
William Cheng: Okay, right.

573
01:03:39.300 --> 01:03:55.950
Lalit Gupta: Yeah, so I'm not able to understand this. So if condition. So the same if not been then send it to look at. But I thought if it should. I thought this is like a redundant, I, I'm not able to invite why we ever need this.

574
01:03:59.850 --> 01:04:02.700
William Cheng: Why, why do the pin business.

575
01:04:03.180 --> 01:04:04.380
Lalit Gupta: Yes, the spin business.

576
01:04:04.590 --> 01:04:05.640
William Cheng: Yeah, I don't know.

577
01:04:07.830 --> 01:04:08.820
Okay, so

578
01:04:09.960 --> 01:04:14.880
William Cheng: So it removed from the list and then it so it's like the most recently used thing.

579
01:04:17.700 --> 01:04:31.170
William Cheng: If it's not pin that means that it's in the list that's that that's that's used by the page on demand. So, therefore, this page frame is most recently used so therefore they removed it from wherever it is, and then added to the end. So, so it says are you

580
01:04:37.140 --> 01:04:37.380
Lalit Gupta: Well,

581
01:04:39.570 --> 01:04:53.340
Lalit Gupta: Well, I thought I think you said something, I think, which makes sense, but I'm not able to understand right now i right now. What I my previous thought was that if the page has not been then it has to be allocated bucket list.

582
01:04:53.460 --> 01:04:58.680
Lalit Gupta: Yes. And then it makes no sense to actually remove from the liquid list and insert to the request list.

583
01:04:59.400 --> 01:05:09.600
William Cheng: Well, so remember my lecture that I talked about the LR you a replacement policy where I have a, you know, I, the example that I use it, you have a rack of DVDs.

584
01:05:09.960 --> 01:05:17.190
William Cheng: And you want to sort of implement the value algorithm. So what whichever DVD that you watch you took it out and you put it at the end of the list.

585
01:05:18.570 --> 01:05:24.630
William Cheng: That's exactly what this code is doing right they remove from, you know, from from the link list and then put it at the end.

586
01:05:26.040 --> 01:05:27.900
Lalit Gupta: Which link list. Is it coming from

587
01:05:29.160 --> 01:05:32.370
William Cheng: You know, you made this. As you know, in the allocated list over here. I

588
01:05:33.690 --> 01:05:37.290
Lalit Gupta: Know, yeah, that's, I know that it is putting into the

589
01:05:38.640 --> 01:05:39.420
Lalit Gupta: Ah,

590
01:05:39.510 --> 01:05:40.110
William Cheng: Pick up on the

591
01:05:40.860 --> 01:05:41.640
William Cheng: Move it to the end.

592
01:05:44.100 --> 01:05:47.670
William Cheng: Business will be sorted according to the according to the last time they were accessed.

593
01:05:48.630 --> 01:05:51.450
Lalit Gupta: Oh, OK. OK. I know exactly what it does. No.

594
01:05:51.540 --> 01:05:51.870
Thank you.

595
01:06:05.430 --> 01:06:15.060
Lalit Gupta: Yeah, there is a in in shadow dot c day I'm not able to like understand the difference between this look up and fill page.

596
01:06:18.840 --> 01:06:21.180
Lalit Gupta: There is a fun to function. This look up and fill page.

597
01:06:21.720 --> 01:06:27.090
Lalit Gupta: Yeah. Seems. What's the difference

598
01:06:28.980 --> 01:06:31.680
William Cheng: Yeah, so the Phil page is the one that does copy on. Right, right.

599
01:06:34.290 --> 01:06:36.060
William Cheng: Right over here. It says, you know,

600
01:06:41.550 --> 01:06:47.310
William Cheng: Write the copy on write magic is not here right so be here. You don't do copyright. There's just wanted to just try to find a page.

601
01:06:48.210 --> 01:06:57.840
William Cheng: So will you tried to look up a page for writing. It's always the first shadow object, right, we need to look up the page for reading you need to go down the link lives over there and find, you know, find the object.

602
01:07:00.660 --> 01:07:06.240
William Cheng: Right. Remember the algorithm. I will you try to look up a you know a shadow object. If it's well for writing it has to be the first one in the link address

603
01:07:06.690 --> 01:07:16.980
William Cheng: And if it's a if you're trying to, you know, look for a shadow patient needs I shot object for for reading you need to iterate through the list of Java object until you find it.

604
01:07:17.850 --> 01:07:33.660
Lalit Gupta: But if it is a writing and if, how about let's say if it is for a field in the local page for writing, then if it is not in the past shadow object, then it shouldn't be. Go to the next next punter

605
01:07:33.810 --> 01:07:37.080
William Cheng: Down the to know for for writing it has to be in the first one right

606
01:07:39.390 --> 01:07:44.520
William Cheng: So with that in the first one that you have to create it. Right. I mean, you need to get ready for copying. Right, right.

607
01:07:46.290 --> 01:07:51.630
Lalit Gupta: Yes, then we have to get it get it for, copy, copy on write and then Dan. We have to call this still page function.

608
01:07:52.650 --> 01:07:53.340
William Cheng: Right, yeah.

609
01:07:55.620 --> 01:07:57.300
Lalit Gupta: Then it basically we are

610
01:07:57.930 --> 01:07:59.130
William Cheng: Whereas over here.

611
01:08:04.890 --> 01:08:05.940
William Cheng: It's important.

612
01:08:07.050 --> 01:08:08.340
William Cheng: Okay. Yeah, over here, right.

613
01:08:13.680 --> 01:08:17.670
William Cheng: So they feel page is the function below it. Right. So this is where where it gets

614
01:08:18.840 --> 01:08:19.680
William Cheng: You know, copy. All right.

615
01:08:22.890 --> 01:08:23.820
Lalit Gupta: Yes, so if

616
01:08:28.890 --> 01:08:39.570
Lalit Gupta: Yes, I suppose we want to do do a copy and right and we started from the first shadow object and, three and, and, and the page, we want to see is not there.

617
01:08:39.930 --> 01:08:41.040
William Cheng: Right. So what you would do is

618
01:08:41.700 --> 01:08:52.410
William Cheng: When you say you tried to do a lookup right you try to look it up for reading or writing right if you want to look up for writing, then you will end up with the pace rain. That's it. That's the first object. Right. The first object.

619
01:08:53.340 --> 01:08:59.340
William Cheng: Yes. Okay. And this one, you know, maybe it's empty, because at the beginning you create an empty one

620
01:09:00.960 --> 01:09:14.580
William Cheng: Okay. And then when you get to shadow Phil page, you know, later on. That's why you will. What will you do the copy on. Right, right. You get you need to find the one that's that's to reframe, and then you write it into the copy right

621
01:09:29.790 --> 01:09:34.080
Lalit Gupta: Why, why, why does it even creative me. Why does this

622
01:09:35.250 --> 01:09:49.560
Lalit Gupta: Okay. So you are saying when we are doing for writing, then when the shadow page can look of page can create an empty page, but the job of job to actually fill that page to the data is the job of the film page.

623
01:09:51.030 --> 01:09:55.920
William Cheng: Okay. Over here says, write the foreword arguments to the page look for writing for this.

624
01:09:59.460 --> 01:10:06.270
William Cheng: One for right is false, find the first other object in the chain, which has given who has given page resident

625
01:10:07.620 --> 01:10:09.270
William Cheng: World for right is true.

626
01:10:11.460 --> 01:10:15.210
William Cheng: Is handle in the Phil page right not here.

627
01:10:20.070 --> 01:10:23.130
William Cheng: Yeah, so, so this function supposed to return a page frame right

628
01:10:24.210 --> 01:10:26.940
William Cheng: Okay, Jeff, what if you try to look up a page to him for right

629
01:10:27.000 --> 01:10:34.800
William Cheng: It has to be, it has to be belong to the first shadow object. Right. Yes. So you do whatever it takes to make sure that that's true.

630
01:10:39.750 --> 01:10:40.380
William Cheng: So, so you

631
01:10:40.620 --> 01:10:43.950
William Cheng: Shot object is empty, you need to create a, you know, an empty object.

632
01:10:44.160 --> 01:10:57.330
William Cheng: And then you need to trust that later on. Phil page will get called and then you will copy the data from, you know, from the right place to the, you know, to, to, to, to this particular page for him that's returned by by look up page. Oh.

633
01:10:59.070 --> 01:11:03.300
William Cheng: Right. So yeah, maybe because of polymorphous them. This is kind of, you know, we're way of writing code.

634
01:11:06.810 --> 01:11:13.560
Lalit Gupta: Yes, yes. You like that, eventually, it seems like that sort of pages calling this film page on

635
01:11:13.650 --> 01:11:23.730
William Cheng: Right. So again, this is the Brown University design you know that that's all we have to deal with they they they divide this into two steps. One is look at page. The otherwise about the you know the full page.

636
01:11:26.670 --> 01:11:31.440
William Cheng: So you can look up page, you'll find a page frame and then, you know, fill page you fill the page with data.

637
01:11:34.800 --> 01:11:35.610
Lalit Gupta: And

638
01:11:37.950 --> 01:11:39.120
Lalit Gupta: The pages for

639
01:11:40.230 --> 01:11:48.240
Lalit Gupta: Reading and then we what we have to do is we have to find the data we have to go down the chain and find where the data, is data first.

640
01:11:49.500 --> 01:11:52.380
Lalit Gupta: You have to find the first memory object where the data is

641
01:11:52.560 --> 01:11:52.920
Yep.

642
01:11:54.090 --> 01:11:55.140
Lalit Gupta: What if we

643
01:11:56.370 --> 01:12:01.320
Lalit Gupta: What if we go down to the bottom and the bottom does not even have it because it is not as urgent.

644
01:12:01.770 --> 01:12:04.020
William Cheng: And then, then the bottom better created that

645
01:12:06.540 --> 01:12:13.470
Lalit Gupta: Okay, so yes. Yes. What you're saying is making sense to me. But damn the comment is ok is saying that

646
01:12:14.310 --> 01:12:29.880
Lalit Gupta: Find a fresh out of debt in the chain, which has the given page resident. So, it should. It seems like it's interesting, interesting to me that it has to be resident, but it's from your command. It seems like in the bottom. It may not be resident and we have to create them.

647
01:12:30.570 --> 01:12:38.820
William Cheng: Right, and also in the first one, right, like in the, in the beginning, you know, when we, you know, first are running the shadow of the shot. Rochelle for celebrities always empty right

648
01:12:39.990 --> 01:12:44.820
William Cheng: So therefore, the first thing that you need to do is to write to a shadow object. Well, in that case, you have to create a page rank right

649
01:12:47.490 --> 01:12:53.940
William Cheng: I mean this, the semantics of this look up, you know, page function is that if it's successful, it will always return a page rank.

650
01:12:55.680 --> 01:13:01.860
William Cheng: Right. So if you want to create a payment if you need a payment for writing and you're supposed to return a page, right, and then you have to create it.

651
01:13:15.060 --> 01:13:18.780
Lalit Gupta: Okay, so. Okay, so look up just fine. But how about the still page.

652
01:13:21.300 --> 01:13:27.870
William Cheng: Right, so, so once you call look at page over here, right, if it turns out it was a festival reading, then you will not call Phil page right

653
01:13:29.280 --> 01:13:36.150
William Cheng: Because it is for reading that and there is there already. But if it's for writing that in this case, the patient will point to an empty page full of zero.

654
01:13:36.360 --> 01:13:47.580
William Cheng: And then you will get to fill page right when you get the pillar page over here. This argument is going to be the one that's returned from the proper look up page and now you're going to fill this page with data by copying from where you're supposed to copy from

655
01:13:54.240 --> 01:13:59.010
William Cheng: Well, we can actually look for Bill page Colonel start. I see.

656
01:14:07.290 --> 01:14:13.470
William Cheng: Let's see, Colonel. So where should Where should we look, we noticed the

657
01:14:23.310 --> 01:14:27.960
William Cheng: Okay, so this one's not very useful colonel and frank, I see.

658
01:14:33.120 --> 01:14:37.740
William Cheng: Okay, so P frame, Phil. Maybe we should look for P friend, Phil.

659
01:14:47.700 --> 01:14:49.560
Lalit Gupta: Should we do up front.

660
01:14:50.190 --> 01:14:57.450
William Cheng: It's right there. But nobody's calling it right so maybe one of your one of the the code that that you need to write will have to call it.

661
01:15:01.320 --> 01:15:11.310
Lalit Gupta: Yeah. But the thing is, be faithful is like is just a pointer, it just eventually, it has to reach our function or the function that is already given to us.

662
01:15:12.300 --> 01:15:14.010
William Cheng: Right, I'm just trying to

663
01:15:45.510 --> 01:15:49.080
William Cheng: So there's the lookup function. And then there's the fill function.

664
01:15:54.780 --> 01:16:03.300
William Cheng: This function P for until it's a static function. Right. Okay. It's not being used anywhere. So, so, so I think one of the function that you have that

665
01:16:03.390 --> 01:16:04.260
Lalit Gupta: Yeah, I know.

666
01:16:05.370 --> 01:16:06.750
Lalit Gupta: I know I think

667
01:16:07.980 --> 01:16:09.030
Lalit Gupta: I can tell you if you want

668
01:16:09.900 --> 01:16:10.920
William Cheng: To get right

669
01:16:11.040 --> 01:16:12.060
Lalit Gupta: Yes, they're speaking. Good.

670
01:16:12.270 --> 01:16:17.190
William Cheng: Okay, so if you look at a comment of P frame get right. What does it say, Phil.

671
01:16:18.720 --> 01:16:20.490
William Cheng: Okay, it says

672
01:16:22.950 --> 01:16:23.370
William Cheng: You know if

673
01:16:24.000 --> 01:16:29.640
William Cheng: Not already resume memory, you know, otherwise we allocate a new page and fill it right

674
01:16:30.930 --> 01:16:37.830
William Cheng: Okay so allocate a new a new page that's done by the lookup function and then still page is done by the Philippines, the Philippines function right

675
01:16:39.240 --> 01:16:41.970
William Cheng: So it's actually give you a hand how they implement paper and get I

676
01:16:44.100 --> 01:16:48.270
William Cheng: You know, if it already exists, you should use it if it doesn't exist. You create a new page and then you fill it.

677
01:16:52.530 --> 01:16:53.460
Lalit Gupta: Okay.

678
01:16:55.320 --> 01:17:09.810
William Cheng: Yes, so, so, so, so if you are you know you've been doing copy on write. And then, you know, when you are calling the lookup for writing why nutcase. Yeah, that's where it is right allocate a new page in our function.

679
01:17:17.160 --> 01:17:26.010
William Cheng: I mean you know the code over here says otherwise we allocate a new page, it doesn't mean that you allocate in a new page inside this function is that you need to call another function that will allocate a new page.

680
01:17:29.160 --> 01:17:31.860
William Cheng: So I think that's the, you know, that's the look on page function.

681
01:17:34.650 --> 01:17:37.080
William Cheng: Okay, I think that's what it's referring to

682
01:17:38.910 --> 01:17:44.190
William Cheng: Right, because look up, look up a page function will create a page frame. It is not there.

683
01:17:45.990 --> 01:17:57.060
Lalit Gupta: Okay, okay. So, if it is there it will just look up it look it up and there and otherwise it will actually if it is not relevant at I will actually create a page and make it resident and fill it with

684
01:17:57.960 --> 01:18:08.040
William Cheng: For for. Yeah, the more than one reason that you know either. This page is not resident or you are doing copy on write for writing right so there are multiple reason why you need to allocate a new page.

685
01:18:13.140 --> 01:18:15.390
William Cheng: So it's all summarizing half a sentence right here.

686
01:18:17.130 --> 01:18:18.360
Lalit Gupta: Yeah, there's a

687
01:18:19.320 --> 01:18:19.920
William Cheng: Yeah, there we go.

688
01:18:23.040 --> 01:18:23.490
Okay.

689
01:18:27.600 --> 01:18:30.450
Lalit Gupta: So, okay, let's can, can you go to the slide.

690
01:18:31.560 --> 01:18:37.980
Lalit Gupta: Lecture 16th like 66

691
01:18:51.510 --> 01:18:52.140
William Cheng: This is the one.

692
01:18:52.680 --> 01:18:53.220
Yes.

693
01:18:56.190 --> 01:19:06.660
Lalit Gupta: So it thing Venus does this into step. It does not change the this pretty to our office. What is that the action it and map that you entire user space table.

694
01:19:07.140 --> 01:19:07.920
Lalit Gupta: And then it's

695
01:19:08.040 --> 01:19:09.390
Lalit Gupta: In the next page fault.

696
01:19:11.250 --> 01:19:12.180
Lalit Gupta: Yes, this is

697
01:19:12.300 --> 01:19:16.140
Lalit Gupta: Late what, how does this next fridge for like finally ordered

698
01:19:17.070 --> 01:19:22.110
William Cheng: Yeah, so, so for all these pages table entries over here, right, if you just said, Be equal to zero of all them right

699
01:19:22.800 --> 01:19:29.190
William Cheng: Okay, and then you can go into the user space right landing page right away when it will address translation because be equal to zero.

700
01:19:30.000 --> 01:19:37.680
William Cheng: Yes. Okay, so. So it's as if you're doing equal to zero. So, next time when you come into the kernel, you said all these entry back to read only. Right.

701
01:19:41.220 --> 01:19:51.330
William Cheng: I mean, as it turns out, if you're writing to one in this case you can set it to rewrite or again you're fixing a baseball bat. So using the in the data settings that external you can figure out what you know what you have to do.

702
01:19:53.130 --> 01:19:57.510
Lalit Gupta: Okay, okay. So depending on the conditions.

703
01:19:58.050 --> 01:19:59.490
William Cheng: Yeah, they just like the first time you

704
01:20:01.050 --> 01:20:10.110
William Cheng: Get over here when we fork it as if you're starting from scratch, right when you thought it sounded much glad you're doing on demand paging so therefore all the data stuck inside of kernel.

705
01:20:11.520 --> 01:20:14.100
William Cheng: memory map to tell you how to fix the page fall

706
01:20:15.630 --> 01:20:16.050
Okay.

707
01:20:21.300 --> 01:20:26.850
Lalit Gupta: Okay. So setting up this thing is is a job of this pitch for Tango.

708
01:20:27.540 --> 01:20:29.310
William Cheng: No, no, this is this is do fork.

709
01:20:31.890 --> 01:20:32.460
Lalit Gupta: Or this is

710
01:20:33.630 --> 01:20:34.020
William Cheng: Rather than

711
01:20:34.620 --> 01:20:37.920
William Cheng: I do for you need to reset for copyright for both the parent of the child. Right.

712
01:20:38.130 --> 01:20:44.700
William Cheng: So you do all this kind of stuff. And now both plan and a child, they will return to the users baseball bat and both of them will get paid fall and then you start fixing

713
01:20:45.960 --> 01:20:46.860
William Cheng: Fixing the page table.

714
01:20:48.900 --> 01:20:56.880
Lalit Gupta: Okay, so when when both of them go back to this after after from the page fall to the Colonel, then we have to set it to this.

715
01:20:58.080 --> 01:21:00.150
Lalit Gupta: This field to our right.

716
01:21:00.390 --> 01:21:05.970
William Cheng: Yeah, so, so as soon as so you know as soon as I say that the parent process called called do for a cold call forth, right.

717
01:21:06.180 --> 01:21:08.940
William Cheng: You're gonna come he's out a colonel in do for you do all these kind of stuff.

718
01:21:09.180 --> 01:21:13.950
William Cheng: And then when the parent process returned back into the user space. The first thing that will happen is is a paywall

719
01:21:16.860 --> 01:21:22.890
William Cheng: Okay, and then the patient is going to be at the at the parents is tech segment because you know the tech side when they try to execute code.

720
01:21:23.100 --> 01:21:30.960
William Cheng: And it turns out address translation fail because equal to zero, you're gonna come into the kernel, you're going to fix it and you go back again and now it's going to you know you gotta, you gotta get a few pages faults.

721
01:21:31.200 --> 01:21:38.040
William Cheng: Because, because you are you know you're doing all kinds of stuff in the parent process. And similarly, when the child processes, go back to the user space, the same things gonna happen.

722
01:21:39.060 --> 01:21:44.100
William Cheng: Okay, okay. So when you are debugging the code for for can wait a fork and wait

723
01:21:45.660 --> 01:21:48.150
William Cheng: So this is where you're going to notice that you're gonna get a lot of page faults.

724
01:21:49.950 --> 01:21:56.820
William Cheng: You know, after you make the fourth system called and the parents going to paste all the child's gonna paste fall so gotta, you gotta get all those code to work.

725
01:22:25.320 --> 01:22:30.150
William Cheng: So, Angela. Do you have any question that you would like to ask

726
01:22:32.880 --> 01:22:38.310
Andrew Strimaitis: Honestly, so many questions were asked in the last hour, hour and a half, that it kind of covered what I was going to ask.

727
01:22:38.490 --> 01:22:42.000
Andrew Strimaitis: Okay. Um, so I don't really need to chime in.

728
01:22:42.390 --> 01:22:42.780
Okay.

729
01:22:50.250 --> 01:22:53.490
Andrew Strimaitis: Thank you both. Okay, thanks. Yeah.

730
01:22:53.550 --> 01:22:57.180
Lalit Gupta: I think this yeah I think this course, like my notes.

731
01:22:57.900 --> 01:22:59.280
Okay, thank you.

732
01:23:03.750 --> 01:23:06.270
William Cheng: Alright, so nobody has any more questions that then

733
01:23:06.510 --> 01:23:09.360
William Cheng: I'm going to, I'm going to shut this down.