Intervjuavsnitt #14 - @snare

Lyssna

mp3

Innehåll

Detta är det fjortonde intervjuavsnittet av Säkerhetspodcasten, i vilket Peter och Jesper intervjuar @Snare på Sec-T 2014. De diskuterar Thunderbolt DMA-Attacker och mycket mer. Reklam: Swehack har en CTF som går av stapeln den 20e mars. Läs mer på 2015.swectf.se.

Länkar

SweCTF

AI transkribering

AI försöker förstå oss… Ha överseende med galna feltranskriberingar.

1 00:00:00,260 --> 00:00:03,880 Hej, det här är Johan Möller och Jesper Larsson.

2 00:00:05,460 --> 00:00:08,600 Innan avsnittet börjar ska vi pusha för några grejer som händer snart.

3 00:00:08,720 --> 00:00:09,880 Ja, väldigt viktiga grejer.

4 00:00:10,460 --> 00:00:12,700 Ska vi börja med Sveahack?

5 00:00:12,880 --> 00:00:20,880 Yes, forumet sveahack.org ska köra igång en CTF som går stapeln nästa helg,

6 00:00:21,000 --> 00:00:22,680 det vill säga den 20-20 mars.

7 00:00:22,780 --> 00:00:25,140 Precis, den ska man vara med och köra.

8 00:00:25,240 --> 00:00:28,720 Det är en klassisk Jeopardy-style CTF, alltså Capture the Flag.

9 00:00:28,720 --> 00:00:29,840 Yes, öppen på nätet.

10 00:00:29,840 --> 00:00:34,400 Där man kommer kunna köra lite webbexploitering, lite forensik, lite reversing.

11 00:00:34,580 --> 00:00:35,520 Det kommer vara lite punables.

12 00:00:35,780 --> 00:00:37,640 Man kommer få fina priser.

13 00:00:38,020 --> 00:00:43,680 Vi kan locka med bland annat Raspberry Pis, lite Bitcoins.

14 00:00:44,020 --> 00:00:44,420 Yes.

15 00:00:44,800 --> 00:00:45,420 Har du med lite OVP?

16 00:00:45,420 --> 00:00:46,700 Ja, det låter som en sweet deal.

17 00:00:47,060 --> 00:00:49,200 Vi vill promota Sveahack, de gör bra skit.

18 00:00:49,340 --> 00:00:54,820 Så hoppa in på Sveahack och testa din lycka mot alla andra galna herrar.

19 00:00:54,820 --> 00:00:59,820 Precis, och då ska ni självklart gå in på 2015.sveac.org

20 00:00:59,840 --> 00:01:04,160 Yes, och läs mer information där.

21 00:01:04,760 --> 00:01:06,980 Ni kan också gå in på ctftime.org.

22 00:01:07,420 --> 00:01:08,520 Ja, där finns de med också.

23 00:01:08,920 --> 00:01:12,420 Sen har vi en grej till som är viktigare än alla andra grejer som vi precis har sagt.

24 00:01:12,980 --> 00:01:18,220 Och det är Säkerhetspodcastens 100 000 Downloads Party.

25 00:01:18,400 --> 00:01:18,720 Fest.

26 00:01:19,120 --> 00:01:23,520 Det är alltså så att vi har nått en bit över 100 000 nedladdningar.

27 00:01:24,000 --> 00:01:27,140 Och för att fira det så tänkte vi bjuda in alla lyssnare till en här jättejävla fest.

28 00:01:27,140 --> 00:01:29,140 Ja, som kommer gå av stapeln.

29 00:01:29,840 --> 00:01:31,380 I Göteborg.

30 00:01:31,760 --> 00:01:36,200 Så ni från Stockholm får ta på er vandringskängorna och börja gå ner mot Sveriges framsida.

31 00:01:36,300 --> 00:01:37,240 Det kommer vara värt det.

32 00:01:37,640 --> 00:01:38,200 Herregud ja.

33 00:01:38,320 --> 00:01:40,860 Det här kommer gå av stapeln i slutet på april eller början på maj.

34 00:01:41,140 --> 00:01:43,260 Vi håller på att titta på datum nu.

35 00:01:43,360 --> 00:01:45,480 Det kommer vara på Hard Rock Cafe på Soliga Avenyn.

36 00:01:45,780 --> 00:01:48,000 Yes, alltid sol på Avenyn.

37 00:01:48,160 --> 00:01:50,680 Det kommer vara lite livepodcast, lite föredrag kanske.

38 00:01:51,180 --> 00:01:52,540 Vi kommer bjuda på bash.

39 00:01:53,320 --> 00:01:55,760 Det kommer bli en sjuhälsikers jävla fest helt enkelt.

40 00:01:55,760 --> 00:01:57,160 Där vi kommer ha jävligt roligt.

41 00:01:57,420 --> 00:01:59,220 Vi kommer mingla med en massa säkerhetsfolk helt enkelt.

42 00:01:59,320 --> 00:01:59,680 Yes.

43 00:01:59,840 --> 00:02:06,720 Och då är det så att för att det här ska bli riktigt bra så behöver vi lite sponsorer.

44 00:02:07,400 --> 00:02:07,800 Yes.

45 00:02:07,980 --> 00:02:11,980 För vi vill ju hyra stället, vi vill bjuda på öl, vi vill kanske göra lite giveaways.

46 00:02:12,120 --> 00:02:13,000 Ja, så är det.

47 00:02:13,560 --> 00:02:17,280 Vi har några sponsorer klara men vi skulle behöva någon till.

48 00:02:17,500 --> 00:02:17,580 Ja.

49 00:02:17,880 --> 00:02:21,920 Så vet du med dig att ditt företag vill lägga lite stålar på…

50 00:02:21,920 --> 00:02:25,540 Ja, skulle vilja hävda Sveriges fräckaste säkerhetsfest.

51 00:02:26,720 --> 00:02:27,680 Så hej!

52 00:02:28,420 --> 00:02:29,820 Gå in på säkerhetspodcasten.se.

53 00:02:29,840 --> 00:02:34,100 Klicka in dig på hundratusenfesten och läs våra sponsornivåer där.

54 00:02:34,180 --> 00:02:36,640 Precis, vi har tre sponsornivåer. Den kan ni läsa mer om där.

55 00:02:38,000 --> 00:02:43,320 Gå gärna in och gör detta så snart som möjligt så kan vi boka stället och så vidare och läsa coola grejer.

56 00:02:45,320 --> 00:02:49,420 Och det är även på hundratusenfestensidan som vi kommer gå ut med mer information.

57 00:02:49,560 --> 00:02:50,960 Kommer självklart på Twitter och Facebook också.

58 00:02:50,960 --> 00:02:56,260 Yes, vi kommer bli väldigt virala när vi kommer skicka ut en Eventbrite-länk, tänker jag.

59 00:02:56,340 --> 00:02:56,420 Ja.

60 00:02:56,920 --> 00:02:59,780 Med anmälan för själva eventet.

61 00:02:59,840 --> 00:03:01,580 Ja, så vi vet hur många som dyker upp.

62 00:03:01,800 --> 00:03:02,000 Precis.

63 00:03:02,420 --> 00:03:06,340 Men om du vet med dig att ditt bolag skulle vara intresserade av att sponsra den här festen.

64 00:03:06,340 --> 00:03:13,220 Så gå in och kolla vilka sponsornivåer vi har och vad vi erbjuder på säkerhetspodcasten.se och klicka på hundratusenfesten.

65 00:03:14,140 --> 00:03:16,800 All right, nu kör vi en intervju med Snare.

66 00:03:16,960 --> 00:03:17,300 Yes!

67 00:03:17,500 --> 00:03:18,080 Från Sec-T.

68 00:03:18,420 --> 00:03:19,180 Peace!

69 00:03:21,040 --> 00:03:23,080 We’re doing a Swedish podcast.

70 00:03:23,340 --> 00:03:23,640 Okej.

71 00:03:23,680 --> 00:03:26,040 Involving security, ID security.

72 00:03:27,120 --> 00:03:28,340 And like…

73 00:03:28,960 --> 00:03:29,760 Every sense.

74 00:03:29,840 --> 00:03:33,420 We’re planning to make everyone speak Swedish and dominate.

75 00:03:34,760 --> 00:03:35,620 That’s the thing.

76 00:03:35,820 --> 00:03:36,200 All right.

77 00:03:36,220 --> 00:03:36,760 That’s the thing.

78 00:03:37,100 --> 00:03:39,960 So yeah, we are five guys, work at the same company.

79 00:03:40,700 --> 00:03:44,660 This is a hobby of us just to get together and talk security.

80 00:03:44,860 --> 00:03:45,060 Cool.

81 00:03:45,680 --> 00:03:51,020 None of us, I think Peter has the most experience in like this low-level hardware stuff.

82 00:03:51,300 --> 00:03:54,620 But we’ll just have a talk, maybe ten minutes.

83 00:03:55,120 --> 00:03:55,860 And then we will break.

84 00:03:56,140 --> 00:03:56,500 No worries.

85 00:03:56,840 --> 00:03:57,020 Cool.

86 00:03:58,320 --> 00:03:59,440 So are we there, Peter?

87 00:03:59,440 --> 00:04:00,360 Yeah, we’re recording.

88 00:04:00,560 --> 00:04:01,180 Yeah, cool.

89 00:04:01,620 --> 00:04:08,600 So there was sort of this big time jump because we were suddenly talking about FPGAs and microplays.

90 00:04:08,980 --> 00:04:12,160 And that is exactly what I was doing in 2003.

91 00:04:12,380 --> 00:04:14,260 And since then I haven’t touched FPGAs.

92 00:04:14,640 --> 00:04:14,940 Right.

93 00:04:15,380 --> 00:04:23,440 So I was working with one of the first development boards, one of the first development SDKs for microplays.

94 00:04:23,700 --> 00:04:23,840 Oh, cool.

95 00:04:23,840 --> 00:04:28,660 So it was very strange to suddenly hear it.

96 00:04:28,760 --> 00:04:29,420 Someone talking.

97 00:04:29,560 --> 00:04:34,200 Yeah, I mean, I only touched it for the first time like two, maybe two or three years ago.

98 00:04:34,840 --> 00:04:40,900 And yeah, from a software perspective, I was kind of like, holy shit, this is, you know, there’s logic and gates.

99 00:04:41,060 --> 00:04:47,000 And I, this is going back to university and trying to relearn all this electrical engineering stuff that I never really learned properly in the first place.

100 00:04:47,240 --> 00:04:52,560 And then apply that to sort of pseudo software in the form of, you know, hardware description languages.

101 00:04:52,740 --> 00:04:54,300 So yeah, it was.

102 00:04:54,600 --> 00:04:55,920 It’s a very different world.

103 00:04:56,040 --> 00:04:56,880 Very different world.

104 00:04:56,940 --> 00:04:58,680 Yeah, I mean, it looks like code, but it’s not really code.

105 00:04:58,680 --> 00:04:59,120 It’s.

106 00:04:59,440 --> 00:05:01,660 You know, it’s actually a circuit, right?

107 00:05:01,840 --> 00:05:01,960 So.

108 00:05:02,400 --> 00:05:02,500 Yeah.

109 00:05:02,820 --> 00:05:03,020 Mm.

110 00:05:03,140 --> 00:05:03,940 Just to sum up here.

111 00:05:04,000 --> 00:05:05,680 What, what, who is Snare?

112 00:05:06,300 --> 00:05:07,600 Tell us a brief bio.

113 00:05:07,920 --> 00:05:13,660 I, um, I work for a company called Azimuth Security, which is sort of based all around the world, but primarily in Australia.

114 00:05:14,040 --> 00:05:19,420 Um, I do research stuff and, you know, code audits and all that kind of stuff, I guess.

115 00:05:19,500 --> 00:05:22,580 You know, lots of typical consulting gigs.

116 00:05:22,960 --> 00:05:24,960 Um, I like firmware.

117 00:05:25,480 --> 00:05:27,280 You know, I mess with UEFI quite a bit.

118 00:05:27,280 --> 00:05:29,020 Um, Mac kernel stuff.

119 00:05:29,120 --> 00:05:29,220 I’m.

120 00:05:29,440 --> 00:05:35,320 I’m sort of a Mac OS guy from way back, um, I like metal and, um, yeah.

121 00:05:35,980 --> 00:05:44,260 So, uh, so that, uh, where does the, where does the work stop and where does the fun begin?

122 00:05:44,260 --> 00:05:45,040 Wow.

123 00:05:45,040 --> 00:05:46,560 Or, or is it everything mixed together?

124 00:05:46,560 --> 00:05:48,120 That’s kind of everything mixed together these days.

125 00:05:48,120 --> 00:05:52,540 I mean, it’s, you know, yeah, it’s this project you’ve been talking about right now with the DMA attack.

126 00:05:52,540 --> 00:05:55,820 Is it actually a thing that you will, uh, productify?

127 00:05:55,840 --> 00:05:56,500 Can you say that?

128 00:05:56,560 --> 00:05:59,400 No, no, this is, this is just like a research project.

129 00:05:59,400 --> 00:06:00,420 Yeah, yeah, yeah.

130 00:06:00,420 --> 00:06:05,900 I mean, um, my friend Sam and I were working on this for, um, you know, for the, for the last couple of years on and off.

131 00:06:06,060 --> 00:06:07,960 Um, but yeah, it’s, it’s just a research project really.

132 00:06:08,460 --> 00:06:08,600 Yeah.

133 00:06:09,060 --> 00:06:10,000 Super awesome though.

134 00:06:10,100 --> 00:06:10,760 Cool, thank you.

135 00:06:11,240 --> 00:06:11,620 Super awesome.

136 00:06:11,620 --> 00:06:12,180 I enjoyed it.

137 00:06:12,180 --> 00:06:22,940 So, uh, so for the people who haven’t heard the talk, you, you were basically talking about how the old DMA attacks against firmware worked.

138 00:06:23,520 --> 00:06:29,180 And then how you could extend those and run them on, uh, firmware.

139 00:06:29,400 --> 00:06:31,080 Underbolt with a connection cable.

140 00:06:31,320 --> 00:06:43,560 And as the last step, you talked about your research into making it natively through FireWire and getting into the very PCI express bridge.

141 00:06:43,560 --> 00:06:55,500 And, uh, what do you say, uh, emulate a device which gave you rights to, uh, read, write the entire memory.

142 00:06:55,680 --> 00:06:56,180 That’s right.

143 00:06:56,180 --> 00:06:56,400 Yeah.

144 00:06:56,400 --> 00:06:56,640 Yeah.

145 00:06:56,940 --> 00:06:57,180 Yeah.

146 00:06:57,180 --> 00:06:58,900 And lots of fun tricks you can do with that.

147 00:06:58,900 --> 00:06:59,280 Like.

148 00:06:59,400 --> 00:07:16,840 You know, uh, overriding code in the kernel or in frameworks that are loaded and unlocking, you know, unlocking systems and doing full memory acquisition that same sort of thing you can do with FireWire, um, DMA attacks, but without the limitations of those, those attacks, like the, you know, uh, 32 bit addressing and, and, and all that sort of stuff.

149 00:07:16,840 --> 00:07:29,140 So, yeah, I don’t think, I think people, uh, you still come across people today when they hear about, hear about FireWire DMA attacks, they’re surprised.

150 00:07:29,140 --> 00:07:30,680 And says, you can’t be right.

151 00:07:31,120 --> 00:07:31,440 Yeah.

152 00:07:31,440 --> 00:07:32,220 You must be wrong.

153 00:07:32,220 --> 00:07:33,600 There’s not possible.

154 00:07:33,820 --> 00:07:34,080 Yeah.

155 00:07:34,080 --> 00:07:35,100 That it works that way.

156 00:07:35,580 --> 00:07:35,840 Yeah.

157 00:07:35,840 --> 00:07:47,820 But basically, yeah, basically the protocols are designed so that they expect a bus master device to access the memory it’s supposed to access.

158 00:07:47,820 --> 00:07:48,980 That’s, that’s exactly right.

159 00:07:48,980 --> 00:07:49,260 Yeah.

160 00:07:49,380 --> 00:07:53,820 And, uh, you know, putting more restrictions in place, um, slows it down.

161 00:07:53,820 --> 00:07:54,080 Right.

162 00:07:54,140 --> 00:07:59,040 So, I mean, that’s the, I guess that’s kind of why Thunderbolt exists because rather than having.

163 00:07:59,140 --> 00:08:14,860 These multiple layers of, you know, indirection between the, the PCI express bus and the device like, you know, FireWire or whatever other fast, you know, USB three or whatever other fast protocols, um, you just, just have PCI express, you know, it goes in one side gets, you know, turned into Thunderbolt comes out the other side.

164 00:08:14,860 --> 00:08:22,600 So, I mean, it’s, um, there’s, there’s not a lot of things getting in the way of the speed, um, or, you know, the, the throughput.

165 00:08:22,600 --> 00:08:26,240 So I guess the, uh, putting more restrictions on that slows it all down.

166 00:08:26,240 --> 00:08:28,760 So, you know, the, the hardware designers don’t want that.

167 00:08:29,140 --> 00:08:32,140 Yeah, until enough people complain about the security implications of those features.

168 00:08:32,140 --> 00:08:44,540 Yeah, but, but when they, when they built, uh, Thunderbolt and designed that, at that time, they already know from FireWire that they needed to do it properly.

169 00:08:44,540 --> 00:08:45,540 Absolutely.

170 00:08:45,540 --> 00:08:53,940 So, so they, I think, I think they can’t blame we didn’t know about it because with FireWire it was new.

171 00:08:53,940 --> 00:08:54,940 Yeah.

172 00:08:54,940 --> 00:08:55,940 Yeah.

173 00:08:55,940 --> 00:08:56,940 That’s my question as well.

174 00:08:56,940 --> 00:08:58,940 How did you, how in this research, have you been trying?

175 00:08:58,940 --> 00:09:02,340 Obviously, you’ve been reached out for Intel and, and so forth.

176 00:09:02,340 --> 00:09:04,740 What has been the, your communication level?

177 00:09:04,740 --> 00:09:08,940 All the others, like, led on, I won’t talk to this guy, this guy is dangerous, lock him out.

178 00:09:08,940 --> 00:09:13,940 No, I mean, they, they, they all know what the, the implications of extending the PCI Express bus out of the box are.

179 00:09:13,940 --> 00:09:17,940 I mean, it’s the same as, it’s the same as what it was with, um, with Express Card.

180 00:09:17,940 --> 00:09:18,940 Yeah.

181 00:09:18,940 --> 00:09:23,940 Um, and, you know, similar to FireWire and the same as, you know, what it was with PCMCIA, like CardBus as well.

182 00:09:23,940 --> 00:09:26,940 It could do exactly the same thing with CardBus on a PCI system.

183 00:09:26,940 --> 00:09:27,940 So they know what the implications are.

184 00:09:27,940 --> 00:09:34,940 Um, uh, and, you know, yeah, they’re, they’re more than happy to, to talk about it, but they, you know, they, they, they know exactly what the implications are, so.

185 00:09:34,940 --> 00:09:35,940 Yeah, yeah, yeah.

186 00:09:35,940 --> 00:09:36,940 Yeah.

187 00:09:36,940 --> 00:09:45,940 It’s funny though, it’s like, yeah, that it actually works though, but it’s like, yeah, as you said, it’s, it’s, it’s designed to be fairly simple or for, like.

188 00:09:45,940 --> 00:09:50,940 But, uh, no, no, there, there are, uh, mitigations being put in place.

189 00:09:50,940 --> 00:09:51,940 Yeah, that’s right.

190 00:09:51,940 --> 00:09:53,940 What’s the primary defense?

191 00:09:53,940 --> 00:09:57,940 Yeah, VCD, I mean, is, is, uh, a virtualization technology for, uh.

192 00:09:57,940 --> 00:10:06,940 Um, for partitioning device, you know, the sort of the IO device space into, so you can assign a, a PCI Express device to a, to a virtualization guest.

193 00:10:06,940 --> 00:10:12,940 And then not allow that device to talk to the memory of the internet or to talk to any of the other guests, um, or the host memory.

194 00:10:12,940 --> 00:10:16,940 Um, and then only allow, um, that guest to talk to the device.

195 00:10:16,940 --> 00:10:24,940 So that in itself sort of allows for, um, uh, access control around, around memory.

196 00:10:24,940 --> 00:10:26,940 So, so when a, you know, when a kernel buffer gets allocated.

197 00:10:26,940 --> 00:10:27,940 Yeah.

198 00:10:27,940 --> 00:10:32,940 The, the VCD unit says, well, you know, this, this is owned by this device and only this device can write to it and read from it.

199 00:10:32,940 --> 00:10:40,940 Um, so if you connect another device to the PCI Express bus, then it, it, it won’t be able to write to, uh, to, to the memory that’s been allocated for other purposes.

200 00:10:40,940 --> 00:10:49,940 And if I understood you correctly, it was fairly well, uh, uh, implemented, uh, the VCD protections on Linux.

201 00:10:49,940 --> 00:10:56,940 Uh, on the latest Mac OS with, uh, uh, with, uh, support.

202 00:10:56,940 --> 00:10:58,940 With a supporting chipset.

203 00:10:58,940 --> 00:11:04,940 Uh, it would also implement other defenses, but you had to do configurations to make it enabled?

204 00:11:04,940 --> 00:11:12,940 No, the, the kernel does all the configuration, um, but OS X only, like, configures a single, um, VCD domain with all the devices in it.

205 00:11:12,940 --> 00:11:15,940 Um, so there’s, there’s still access controls between devices within the domain.

206 00:11:15,940 --> 00:11:21,940 Um, whereas Linux actually configures a separate domain for each device.

207 00:11:21,940 --> 00:11:23,940 All right, so you can steal data from the other devices.

208 00:11:23,940 --> 00:11:25,940 Um, yeah, there’s still access controls between the devices.

209 00:11:25,940 --> 00:11:28,940 So it’s not really, there’s not really a lack in security.

210 00:11:28,940 --> 00:11:31,940 It’s just a different approach to, to partitioning the devices.

211 00:11:31,940 --> 00:11:35,940 So I, I, I assume that the, I, I didn’t actually look at the code for the Linux one.

212 00:11:35,940 --> 00:11:37,940 My, my research partner, Sam, did.

213 00:11:37,940 --> 00:11:43,940 Um, but I assume that the reason for that is to make it easier to integrate with KVM and XAN and stuff like that.

214 00:11:43,940 --> 00:11:47,940 Um, so that there’s an API for assigning devices, you know, through to guests.

215 00:11:47,940 --> 00:11:50,940 Whereas OS X, there’s no provision for doing that at all.

216 00:11:50,940 --> 00:11:54,940 So if it, like, VMware on Mac OS can’t do, uh, VM direct path.

217 00:11:55,940 --> 00:11:58,940 So I assume that that’s, that’s the reason for that on Linux.

218 00:12:01,940 --> 00:12:07,940 Yeah, and tell us about this, this, this device, this lab equipment that you showed up with right now.

219 00:12:07,940 --> 00:12:08,940 It’s quite big.

220 00:12:08,940 --> 00:12:09,940 Yeah.

221 00:12:09,940 --> 00:12:11,940 But you’re talking about, uh, doing it in a small.

222 00:12:11,940 --> 00:12:12,940 And power hungry.

223 00:12:12,940 --> 00:12:13,940 And power hungry.

224 00:12:13,940 --> 00:12:14,940 Yeah, yeah, yeah.

225 00:12:14,940 --> 00:12:18,940 Well, there’s, um, the, the board that we’re using is, uh, you know, a quite fully featured FPGA development board.

226 00:12:18,940 --> 00:12:21,940 So it’s, um, you know, it’s, it’s a, it’s a big board.

227 00:12:21,940 --> 00:12:24,940 It’s a good, you know, sort of six inches by, by 10, 10, 11 inches.

228 00:12:25,940 --> 00:12:29,940 Um, and then there’s a, a, a Thunderbolt 2 PCI Express board that connects to it.

229 00:12:29,940 --> 00:12:34,940 Um, but the, the, the FPGA development board itself has, you know, like a whole bunch of networking hardware.

230 00:12:34,940 --> 00:12:38,940 Um, video hardware, like it’s got a DVI output and an HDMI output.

231 00:12:38,940 --> 00:12:39,940 Yeah, yeah.

232 00:12:39,940 --> 00:12:40,940 It’s got a gigabit ethernet.

233 00:12:40,940 --> 00:12:45,940 It’s got, you know, uh, a ton of GPIO stuff connected to a, to a mezzanine connector.

234 00:12:45,940 --> 00:12:49,940 And, um, um, yeah, so it’s a very full featured device.

235 00:12:49,940 --> 00:12:51,940 So it’s quite big and there’s a lot of hardware on it.

236 00:12:51,940 --> 00:12:54,940 Um, and then, yeah, there’s a separate board for, for the Thunderbolt adapter.

237 00:12:54,940 --> 00:13:05,940 So, you know, hopefully we’ll be able to, um, uh, develop something that’s a little bit smaller and, you know, integrates all the necessary components without any of the, any of the extra unnecessary components just onto one board.

238 00:13:05,940 --> 00:13:09,940 And yeah, so there’s, there’s a separate power supply for the FPGA board and for the Thunderbolt board.

239 00:13:09,940 --> 00:13:16,940 So we’d also like to make it, um, bus powered off the Thunderbolt bus, um, and just require that one, you know, power supply.

240 00:13:16,940 --> 00:13:17,940 Yeah.

241 00:13:17,940 --> 00:13:19,940 Cool.

242 00:13:19,940 --> 00:13:20,940 It’s so awesome.

243 00:13:20,940 --> 00:13:21,940 It’s like, yeah.

244 00:13:21,940 --> 00:13:22,940 Yeah.

245 00:13:22,940 --> 00:13:24,940 This is hardware development stuff that I’m quite new to.

246 00:13:24,940 --> 00:13:28,940 So we’re, you know, it’s taken a little while, but Sam’s quite, uh, quite, quite experienced in that, in that area.

247 00:13:28,940 --> 00:13:30,940 So he’s, he’s, uh, working on that.

248 00:13:30,940 --> 00:13:31,940 Yeah.

249 00:13:31,940 --> 00:13:51,940 I remember the old days of developing VHDL and Verilog and, uh, and the, and the worst thing possible is when you’re, you load it onto a card after you have taken all this time to, uh, synthesize your code into, uh, into an FPGA description.

250 00:13:51,940 --> 00:13:53,940 And you load FPGA descriptions.

251 00:13:53,940 --> 00:13:54,940 Mm.

252 00:13:54,940 --> 00:13:55,940 And it’s completely dead.

253 00:13:55,940 --> 00:13:56,940 Yep.

254 00:13:56,940 --> 00:14:01,940 And you know, you’re, uh, only the steps you, you did before took at least an hour.

255 00:14:01,940 --> 00:14:02,940 Yeah.

256 00:14:02,940 --> 00:14:03,940 At least.

257 00:14:03,940 --> 00:14:08,940 So, you know, even before you can start trying to debug the problem, you need to waste another hour.

258 00:14:08,940 --> 00:14:09,940 Absolutely.

259 00:14:09,940 --> 00:14:15,940 I think we’re up to about, um, 92 or 93% utilization on the, on the FPGA.

260 00:14:15,940 --> 00:14:18,940 So it takes a really long time to synthesize at the moment.

261 00:14:18,940 --> 00:14:19,940 Yeah.

262 00:14:19,940 --> 00:14:20,940 Because it’s got a route.

263 00:14:20,940 --> 00:14:21,940 Yeah, yeah.

264 00:14:21,940 --> 00:14:22,940 It’s got some routing problems.

265 00:14:22,940 --> 00:14:23,940 Routing algorithms.

266 00:14:23,940 --> 00:14:24,940 Horrible.

267 00:14:24,940 --> 00:14:35,940 Hoppas det blir annorlunda med Avado på de nya FPGAs, men det här är en Spartan 6, så vi använder den gamla ISE-säten och det är ganska lätt.

268 00:14:35,940 --> 00:14:43,940 Och du förstår inte att du inte har mött tidsnivån innan det har gått en halv dag och du tänker att det tar för lång tid och det kommer inte att fungera.

269 00:14:43,940 --> 00:14:47,940 Ja, det är bara att träna hårdare och hårdare och hårdare.

270 00:14:47,940 --> 00:14:54,940 Ja, vi hade att sätta upp flera optimiseringsflaggor för att faktiskt sätta på FPGA och möta tidsnivån.

271 00:14:54,940 --> 00:15:04,940 Jag utvecklade en EFINET-karta och jag frågade om hur länge någon har kunnat göra en EFINET-karta.

272 00:15:04,940 --> 00:15:10,940 Och de trodde att jag var en idiot, för de ville bara göra sakerna gå snabbare.

273 00:15:10,940 --> 00:15:14,940 Men varför borde en EFINET-karta vara snabb?

274 00:15:14,940 --> 00:15:16,940 Om det kan…

275 00:15:16,940 --> 00:15:17,940 Om det kan…

276 00:15:17,940 --> 00:15:22,940 Om det kan pusha ut data på samma snabbhet som det är nödvändigt.

277 00:15:22,940 --> 00:15:26,940 Det är bara bra om det kan klockas snabbare, för då blir allt lättare.

278 00:15:26,940 --> 00:15:28,940 Ja, absolut.

279 00:15:32,940 --> 00:15:39,940 En av mina stora minnen från att göra min masterfysik är när jag hade designat min EFINET-karta.

280 00:15:39,940 --> 00:15:46,940 Och jag trodde att jag trodde att jag borde få en EFINET-karta och testa den lokalt.

281 00:15:46,940 --> 00:15:49,940 Och då tänkte jag, vad är det värsta som kan hända?

282 00:15:49,940 --> 00:15:55,940 Och jag kopplade min masterfysik till skolans nätverk.

283 00:15:55,940 --> 00:16:01,940 Och då började jag debugga och försöka hitta varför jag inte har en EFINET-karta längre.

284 00:16:01,940 --> 00:16:07,940 Och då började jag höra ljudet och att folk blev irriterade i korridorerna.

285 00:16:07,940 --> 00:16:13,940 Och jag gick över till nätverkarna och frågade dem om det fanns några problem.

286 00:16:13,940 --> 00:16:22,940 Och de berättade att alla CISCO-platser är online och fungerar, men att ingen av dem kan tala med varandra.

287 00:16:22,940 --> 00:16:29,940 Så de var i processen av att reboota alla CISCO-platser i skolan.

288 00:16:29,940 --> 00:16:32,940 Och jag bara, okej, bra.

289 00:16:32,940 --> 00:16:33,940 Och så pluggade du bara den här?

290 00:16:33,940 --> 00:16:36,940 Ja, jag pluggade den direkt.

291 00:16:36,940 --> 00:16:39,940 Jag trodde att jag skrattade min förhållning eller något sådant.

292 00:16:39,940 --> 00:16:42,940 Så då gick jag till dem och tog en liten EFINET-karta.

293 00:16:43,940 --> 00:16:47,940 Och så gjorde resten av min fysisk jobb.

294 00:16:47,940 --> 00:16:50,940 Inblandad i skolnätverket.

295 00:16:50,940 --> 00:16:52,940 Bra saker.

296 00:16:57,940 --> 00:16:59,940 Det har blivit intressant.

297 00:16:59,940 --> 00:17:01,940 Jag tror att vi tar det kort.

298 00:17:01,940 --> 00:17:03,940 Om du har något intressant att tillägga?

299 00:17:03,940 --> 00:17:06,940 Och du är på snär på Twitter?

300 00:17:06,940 --> 00:17:08,940 Ja, det är rätt.

301 00:17:08,940 --> 00:17:12,940 Det är många som kommer upp på Twitter.

302 00:17:12,940 --> 00:17:14,940 Ja, det är många som kommer upp på Twitter.

303 00:17:14,940 --> 00:17:19,940 Du ser dem tala med dig och bli retweetade och så vidare.

304 00:17:19,940 --> 00:17:24,940 Och det är en stor sekret bakom vår podcast.

305 00:17:24,940 --> 00:17:27,940 Vi berättar alla vad som händer på Twitter.

306 00:17:27,940 --> 00:17:29,940 Det ser ut som en tillgänglig lösning.

307 00:17:29,940 --> 00:17:31,940 En lösning för de kommande eventen.

308 00:17:31,940 --> 00:17:32,940 Ja.

309 00:17:32,940 --> 00:17:36,940 Och genom Twitter lärde vi oss att Taylor Swift är bra på skolan.

310 00:17:36,940 --> 00:17:40,940 Ja, det är en bra lösning.

311 00:17:40,940 --> 00:17:43,940 Ja, det har varit en glädje att lyssna på dina berättelser.

312 00:17:43,940 --> 00:17:45,940 Och hoppas att vi hittar en beer senare.

313 00:17:45,940 --> 00:17:47,940 Ja, absolut. Tack för att du var med.

314 00:17:47,940 --> 00:17:49,940 Tack så mycket.

315 00:17:49,940 --> 00:17:51,940 Det var en stor glädje.