Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
3 state-of-the-art technologies in Linux and fu...
Search
KONDO Uchio
February 21, 2021
Technology
690
2
Share
3 state-of-the-art technologies in Linux and future of the containers #SECKUN
2021.02.21 「新しいセキュリティビジネスキャリア」シンポジウム
KONDO Uchio
February 21, 2021
More Decks by KONDO Uchio
See All by KONDO Uchio
大規模レガシーテストを 倒すための CI基盤の作り方 / #CICD2023
udzura
5
2.5k
Ruby x BPF in Action / RubyKaigi 2022
udzura
0
300
Narrative of Ruby & Rust
udzura
0
260
開発者生産性指標の可視化 / pepabo-four-keys
udzura
3
1.8k
Talk of RBS
udzura
0
490
Re: みなさん最近どうですか? / FGN tech meetup in 2021
udzura
0
850
Dockerとやわらかい仮想化 - ProSec-IT/SECKUN 2021 edition -
udzura
2
810
Device access filtering in cgroup v2
udzura
1
1k
"Story of Rucy" on RubyKaigi takeout 2021
udzura
0
910
Other Decks in Technology
See All in Technology
マルチプロダクトの信頼性を効率良く保っていくために
kworkdev
PRO
0
160
インターネットの技術 / Internet technology
ks91
PRO
0
210
最近の技術系の話題で気になったもの色々(IoT系以外も) / IoTLT 花見予定会(たぶんBBQ) @都立潮風公園バーベキュー広場
you
PRO
1
240
MLOps導入のための組織作りの第一歩
akasan
0
330
Data Hubグループ 紹介資料
sansan33
PRO
0
2.9k
AIが書いたコードを信じられない問題 〜レビュー負荷を下げるために変えたこと〜 / The AI Code Trust Gap: Reducing the Review Burden
bitkey
PRO
7
1.3k
自分のハンドルは自分で握れ! ― 自分のケイパビリティを増やし、メンバーのケイパビリティ獲得を支援する ― / Take the wheel yourself
takaking22
1
910
AWS認定資格は本当に意味があるのか?
nrinetcom
PRO
2
270
20年前の「OSS革命」に学ぶ AI時代の生存戦略
samakada
0
430
20260423_執筆の工夫と裏側 技術書の企画から刊行まで / From the planning to the publication of technical book
nash_efp
3
400
Pure Intonation on Browser: Building a Sequencer with Ruby
nagachika
0
110
Oracle AI Database@AWS:サービス概要のご紹介
oracle4engineer
PRO
4
2.3k
Featured
See All Featured
Principles of Awesome APIs and How to Build Them.
keavy
128
17k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
162
16k
Designing Powerful Visuals for Engaging Learning
tmiket
1
340
Paper Plane (Part 1)
katiecoart
PRO
0
6.6k
Reality Check: Gamification 10 Years Later
codingconduct
0
2.1k
Digital Projects Gone Horribly Wrong (And the UX Pros Who Still Save the Day) - Dean Schuster
uxyall
0
1.1k
Self-Hosted WebAssembly Runtime for Runtime-Neutral Checkpoint/Restore in Edge–Cloud Continuum
chikuwait
0
490
Java REST API Framework Comparison - PWX 2021
mraible
34
9.3k
The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs
inesmontani
PRO
3
3.4k
Build your cross-platform service in a week with App Engine
jlugia
234
18k
End of SEO as We Know It (SMX Advanced Version)
ipullrank
3
4.1k
Have SEOs Ruined the Internet? - User Awareness of SEO in 2025
akashhashmi
0
320
Transcript
ཁૉٕज़ͷstate of the art͔Βߟ͑Δ ۙ౻Ӊஐ࿕ / GMO Pepabo, Inc. 2021.02.21
ʮ৽͍͠ηΩϡϦςΟϏδωεΩϟϦΞʯγϯϙδϜ Linuxίϯςφͷະདྷ
GMOϖύϘגࣜձࣾ γχΞϓϦϯγύϧ ٕज़෦ٕज़ج൫νʔϜॴଐ ۙ౻Ӊஐ࿕ (@udzura)
ۙ౻Ӊஐ࿕ ུྺ • ࡾՏᅳͷਓɻچ٢ాൡߍ࣌शؗߴߍΛଔۀɺ౦ژେֶจֶ෦ຊޠຊจֶઐम՝ఔ ͷֶ࢜ଔʢ2007ʣɻ • ϚείϛͷࣾSEɺECαΠτ։ൃɺΦϯϥΠϯήʔϜ ։ൃͳͲΛܦͯ2013ΑΓݱ৬ɺಉʹԬҠॅɻ • RubyɺίϯςφɺΫϥυωΠςΟϒٕज़ͳͲͷίϛϡχςΟͰ
׆ಈɻஶॻʹʮWebͰ͑ΔmrubyγεςϜϓϩάϥϛϯάೖʯʢC&Rݚڀॴʣ • ͖ͳγεςϜίʔϧʢ࠷ۙʣ socketpair(2) ɻ
ࠓͷ͓ •ίϯςφͷཁૉٕज़ʹ͍ͭͯɺࢲͷߨٛͰʢͬ͘͟Γʣཧղ͞Εͨํ ͚ͷ༰Ͱ͢ɻཁૉٕज़ͷղઆࢀߟࢿྉΛͲ͏ͧɻ •ࢀߟ1: https://container-security.dev/ •ࢀߟ2: ʰίϯςφܕԾԽ֓ʱʢޱ, ΧοτγεςϜʣ •ۙͷΧʔωϧʹؚ·ΕΔ৽ٕज़ͷ͏ͪɺίϯςφʹؔ͢ΔͷΛ հ͠·͢ɻ
•͕࣌ؒ͋Εɺͦͷ্ͰίϯςφͷະདྷΛߟ͑·͢ɻ
cgroup v2
cgroup ͷ͓͞Β͍ •Linux Kernelͷجຊٕज़ͷҰͭɻ •ϓϩηεΛάϧʔϐϯά͠ɺͦͷάϧʔϓ୯ҐͰϦιʔεར༻ͷ੍ݶ Λ͔͚Δٕज़ɻCPUɺϝϞϦɺIOɺϓϩηε... IUUQTHJIZPKQBENJOTFSJBMMJOVY@DPOUBJOFST
cgroup v1ͷྫ •cgroupfs ͱ͍͏ϑΝΠϧγεςϜʹmkdir(2) read(2) write(2)ͳͲΛ࣮ ߦ͠ɺૢ࡞Λߦ͏
cgroup v2 •v1 ͷ͍͔ͭ͘ͷܽ - ओʹ੍ޚରʢίϯτϩʔϥʣ͝ͱʹσΟϨΫ τϦΛ͚ͳ͚Ε͍͚ͳ͍༷ - Λࠀ͘͢։ൃ͞Εͨ •େ͖ͳҧ͍ͱͯ͠ɺ
v1 ͰίϯτϩʔϥผʹσΟϨΫτϦ ΛϚϯ τɺݸผʹάϧʔϓʹॴଐͰ͖ͨͷʹର͠ɺv2ͰશίϯτϩʔϥΛ ·ͱΊͨҰͭͷσΟϨΫτϦͷΈΛϚϯτ͠ɺ·ͱΊͯάϧʔϓΛ ࡞͢ΔڍಈʹͳΔɻ •ίϯςφͱͯͪ͜͠Βͷํ͕߹͕͍͍ɻ
Unified hierarchy /sys/fs/cgroup /sys/fs/cgroup /group-a /group-b /cpu.* /memory.* /io.* ...
/cpu.* /memory.* /io.* ... /cpu /memory /blkio /group-a /group-b /group-a /group-c
ίϯςφϥϯλΠϜͰͷcgroupͷར༻ •ʢOCIܥͷʣϥϯλΠϜͰҎԼͷ2ͭͷઃఆ߲͕͋Δ •Cgroup Driver: ίϯςφʹׂΓͯΔcgroupΛͲ͏ίϯτϩʔϧ͢Δ͔ •cgroupfs: cgroupfsͷͷϑΝΠϧૢ࡞ •systemd: systemdʹΑΔཧ •Cgroup
Version: Ϧιʔε੍ݶʹ v1/v2 ͲͪΒΛར༻͢Δ͔ •/sys/fs/cgroup ʹͲͷϑΝΠϧγεςϜ͕Ϛϯτ͞ΕͯΔ͔Ͱఆ
v2 ͷ৽ػೳ •Unified Hierarchy •PSI(Pressure Stall Information) •eBPFͰcgoup IDͷऔಘ͕Մೳʹ •nsdelegate
(ඇಛݖίϯςφʹॏཁ) •clone3(2) Ͱಛఆͷcgroup෦ʹϓϩηε࡞͕Մೳʹ •ͳͲͳͲ...
e.g. PSI(Pressure Stall Information) •γεςϜશମɺ·ͨcgroup୯ҐͰར༻Ͱ͖Δෛՙͷࢦඪ •CPU, ϝϞϦ, IO Ͱ stall
ͨ͠୯Ґ࣌ؒͰͷׂ߹ ΛܭଌͰ͖Δ •e.g. 1ؒͰ45ඵؒɺάϧʔϓͷ ͋Δϓϩηε͕CPUىҼͰ Ԇͨ͠߹ɺcpu some: 75.00
e.g. eBPFͰͷτϥοΩϯάใ •bpf_get_current_cgroup_id(void) ϔϧύʔ •eBPFͷΠϕϯτ͕ى͖ͨλεΫ͕Ͳͷcgroup(v2)ʹॴଐ͍ͯ͠Δ͔ɺ ͦͷIDΛฦ͢ɻ
How cgroup-v2 and PSI Impacts Cloud Native? Uchio Kondo /
GMO Pepabo, Inc. 2019.07.23 CloudNative Days Tokyo 2019 Image from pixabay: https://pixabay.com/images/id-3193865/
eBPF per containers
eBPFٕज़ͱ •ϢʔβۭؒͰ࡞ͬͨϓϩάϥϜΛΧʔωϧͰಈ͔ٕ͢ज़ͷͻͱͭ •2012ʹseccompͷಋೖɺ2013ʹLinuxͷSDNͰͷԠ༻͕࣮͞ ΕɺͦΕҎ߱ख़͢Δ •ϑΟϧλϦϯά͕ಘҙʢtcpdump, seccomp, bpftraceʣ •ΧʔωϧͷใʹΞΫηεͰ͖Δ͕ɺةݥͳίʔυಈ͔ͳ͍ͳͲ ҆શੑ͕͋Δఔ୲อ͞Ε͍ͯΔ
eBPFͷԠ༻ྫ
ίϯςφͷeBPFτϨʔεઓུ •ઓུ͕͍͔ͭ͋͘Δ •Linux Namespace·ͨcgroup (v2)ͷใ͕ར༻Ͱ͖Δ
ྫ1: task_struct ͷใΛḷΔ •task_struct→nsproxy ͔Β namespaceͷใΛ औಘͯ͠ϑΟϧλ͢Δ ʢcxrayʣ IUUQTHJUIVCDPNNSUDDYSBZCMPCNBTUFSQLHUSBDFSPQFOPQFOHP--
ྫ2: NS/ϗετͰͷPIDΛൺֱ •BPFϓϩάϥϜͰऔಘͰ͖ͨ tidͱɺϗετͰͷtidΛ ൺֱ͠ɺҰக͠ͳ͚Ε ίϯςφͱఆ͢Δ ʢTraceeʣ • task_structґଘ IUUQTHJUIVCDPNBRVBTFDVSJUZUSBDFFCMPCNBJOUSBDFFUSBDFFCQGD-ɹ
ྫ3: cgroup helperΛར༻ IUUQTHJUIVCDPNVE[VSBDPQFODMPTFCMPCNBTUFSTSDCQGDPQFODMPTFCQGD
࣮ྫ •copenclose(8) •ۙ౻ͷPoC (BPF+Rust) •ϑϥάͰtask_struct/ cgroup v2 ID ΛΓସ͑
bpf_get_current_cgroup_id(void) を添えて Uchio Kondo / Container Runtime Meetup #3 ランタイムとcgroupの
xxxな関係 * Photo by Fukuoka City
seccomp
seccompͷ͓͞Β͍ •ϓϩάϥϜʹ͓͚ΔγεςϜίʔϧݺͼग़͠ΛϑΟϧλϦϯά͢Δ •γεςϜίʔϧͷҾͷ݅ʹΑͬͯࢦఆΛม͑ΒΕΔ •blacklist(denylist), whitelist(allowlist) ͳͲΛ࣮Ͱ͖Δ •ϑϥά͕ࡉ͔͘ଘࡏ͠ɺྫ͑γεςϜίʔϧͷauditϩάͷΈɺҙ ͷerrnoΛฦͤ͞ΔɺͳͲͷࢦఆ͕Ͱ͖Δ
seccompͷར༻(mruby)
User space notification •seccompʹΑΓγεςϜίʔϧݺͼग़͠Λݕ͠ɺͦͷڍಈΛϢʔβ ϥϯυͷϓϩάϥϜʹҕͶΔ͜ͱ͕Ͱ͖Δٕज़ •Linux 5.0 (2019/3) ͔Βͷಋೖ •அ͢Δ·ͰɺͦͷγεςϜίʔϧϒϩοΫ͢Δ
•e.g. LXCͰͷσόΠεΞΫηεͷ੍ޚ IUUQTHJIZPKQBENJOTFSJBMMJOVY@DPOUBJOFSTɹ
User space notification IUUQTHJIZPKQBENJOTFSJBMMJOVY@DPOUBJOFSTɹ • LXCͰֶͿίϯςφೖ ୈ47ճɹඇಛݖίϯςφͷՄೳੑΛ͛Δseccomp notifyػೳ ΑΓ
࣮ྫʢmrubyར༻ʣ •ҎԼͷΑ͏ͳ acceptor.rbΛ ༻ҙ͢Δ
࣮ྫʢmrubyར༻ʣ •ҎԼͷinvokerΛܦ༝ͯ͠ϓϩάϥϜΛ ىಈɺ listen(3) ΛݺͿ
listen(2) ͷىಈݕ •acceptor.rb ଆͷίϯιʔϧͰڐՄ/ېࢭΛ੍ޚՄೳɻ •ېࢭͨ͠Βͦͷ··ىಈࣦഊͯ͠invokerϓϩηε͕མͪΔ •ڐՄͨ͠ΒԿͳ͔͔ͬͨͷΑ͏ʹɺىಈΛܧଓͯ͠Ϧοεϯɻ
listen(2) ͷىಈݕ ېࢭ࣌ͷग़ྗ ڐՄ࣌ͷग़ྗ
Ԡ༻ʁ •ʮҙͷϥΠϒϥϦؔݺͼग़͠ʯͰϓϩηεΛఀࢭɺCRIU(*)ʹΑΓ ϓϩηεμϯϓΛ࡞͢Δ࣮ݧΛߦͬͨɻ •LD_PRELOAD + ϥούؔ + ʮԿ͠ͳ͍ʯsyscall + seccomp
IUUQTVE[VSBIBUFOBCMPHKQFOUSZ $IFDLQPJOUBOE3FTUPSF*O6TFSTQBDF ϓϩηεͷঢ়ଶΛอଘɺ͔ͦ͜Β࠶ੜ͢Δٕज़ IUUQTDSJVPSH.BJO@1BHF
ߟ
৽ٕज़ʹΑΓͰ͖Δ͜ͱ૿͑Δ͕... •৽ٕज़ͷʮग़ݱʯͱʮීٴʯͷλΠϛϯάζϨΔ •ͨͱ͑ cgroup v2ͷॳग़2013ɻ •2019 ~ 2020 ʹϥϯλΠϜͰͷରԠ͕ਐΜͩଆ໘ •पลͷπʔϧ͕ग़ݱ͢ΔͷͬͱઌͰ͋Ζ͏
•ग़ݱظʹ୯ମͰٕज़Λݕূ͠ɺʢηΩϡϦςΟؚΊʣͲ͏͍͏͕ ͋Δ͔ɺͲ͏͍͏Մೳੑ͕͋Δ͔ݕূ͢Δҙٛେ͖͍
eBPF Linuxͷجຊٕज़ʹͳΓͭͭ͋Δ •ద༻ൣғ͕ͲΜͲΜ·͍ͬͯΔ •τϨʔγϯάɺଳҬ੍ޚωοτϫʔΩϯάͷ΄͔ɺcgroup(v2) deviceͷఆɺLSM BPF programͳͲͳͲ... •ηΩϡϦςΟͷจ຺ͰτϨʔεɺࠪɺҟৗݕͱ͔ܽͤͳ͍ٕज़ ʹͳΔ͜ͱ͕૾͞ΕΔ •Ұͭͷprog
typeʹ৮͓͚ͬͯͩ͘Ͱײ͔֮Γͦ͏