![4/2/2018
28
PDP-7 [Unix] (1970)
• Kernel
• Layering and Partitioning
• System Call
• Data Scoping
• Interpreter
• Monolithic Implementation
• Process Management
• Descriptor Management
• Separation of File Metadata from File Naming
• Devices as Files
• File I/O
• Filesystem](https://image.slidesharecdn.com/handoutsv3-230802175411-5bdc810e/85/_-_handouts_v3-pdf-28-320.jpg)
![4/2/2018
44
Data Structure Definition & Reuse
buf.h filsys.h proc.h text.h
conf.h inode.h reg.h tty.h
file.h param.h systm.h user.h
Dynamic Resource Management
int coremap[CMAPSIZ];
int swapmap[SMAPSIZ];
struct map {
char *m_size;
char *m_addr;
};
malloc(mp, size)
struct map *mp;
{
…](https://image.slidesharecdn.com/handoutsv3-230802175411-5bdc810e/85/_-_handouts_v3-pdf-44-320.jpg)
![4/2/2018
45
Device Driver Abstraction
Driver Interface
struct {
int (*d_open)();
int (*d_close)();
int (*d_strategy)();
} bdevsw[];
struct {
int (*d_open)();
int (*d_close)();
int (*d_read)();
int (*d_write)();
int (*d_sgtty)();
} cdevsw[];](https://image.slidesharecdn.com/handoutsv3-230802175411-5bdc810e/85/_-_handouts_v3-pdf-45-320.jpg)
![4/2/2018
46
Buffer Cache
Fourth Edition
#define B_READ 01
#define B_DONE 02
#define B_ERROR 04
#define B_BUSY 010
#define B_XMEM 060
#define B_WANTED 0100
#define B_RELOC 0200
#define B_ASYNC 0400
#define B_DELWRI 01000
FreeBSD 11.1
#define B_ASYNC 0x00000004
/* Start I/O, do not wait.
*/
[…]
#define B_DELWRI 0x00000080
/* Delay I/O until buffer
reused. */
#define B_DONE 0x00000200
/* I/O completed. */
Fifth Research Edition (Jun 1974)
• Command Files
chdir /usr/source/s3
cc -c ctime.c
ar r /lib/liba.a ctime.o
rm ctime.o
chdir /usr/source/s1
cc -s -n date.c
cp a.out /bin/date
cc -s -n dump.c
cp a.out /bin/dump
cc -s -n ls.c
cp a.out /bin/ls
rm a.out](https://image.slidesharecdn.com/handoutsv3-230802175411-5bdc810e/85/_-_handouts_v3-pdf-46-320.jpg)
![4/2/2018
49
Unix as a Virtual Machine
Also, about this time [1973] I had a
fateful discussion with Dennis, in which
he said
“I think it may be easier to port Unix to a
new piece of hardware than to port a
complex application from Unix to a new
OS”
— Steve Johnson
Dynamic User Memory Allocation
• malloc(3), free(3)
• 26 programs: awk cc col cron dc dcheck diff ed
eqn expr graph icheck learn ls m4 neqn nm
quot ratfor spline struct tar tsort uucp xsend
quiz
• stdio(3), mp(3)](https://image.slidesharecdn.com/handoutsv3-230802175411-5bdc810e/85/_-_handouts_v3-pdf-49-320.jpg)
유닉스_헤리티지_handouts_v3.pdf
- 1. 4/2/2018 1 Unix Architecture Evolution: Milestones and Lessons Learned Diomidis Spinellis Department of Management Science and Technology Athens University of Economics and Business www.spinellis.gr @CoolSWEng
- 2. 4/2/2018 2
- 3. 4/2/2018 3
- 4. 4/2/2018 4 Overview • Groundwork • Sources • Important architectural milestones • Evolution in numbers • Evolution in words
- 5. 4/2/2018 5
- 6. 4/2/2018 6
- 7. 4/2/2018 7
- 8. 4/2/2018 8
- 9. 4/2/2018 9
- 10. 4/2/2018 10
- 11. 4/2/2018 11
- 12. 4/2/2018 12
- 13. 4/2/2018 13
- 14. 4/2/2018 14
- 15. 4/2/2018 15 Why Unix is important • Exemplar design • Technical contributions, • Impact • Development model • Widespread use • “unusual simplicity, power, and elegance”
- 16. 4/2/2018 16 System technology • Hierarchical file system • Compatible file, device, networking, and inter- process I/O • Pipes and filters architecture • Virtual file systems • The shell as a user-selectable regular process
- 17. 4/2/2018 17 Associated Technologies • C and C++ • Parser and lexical analyzer generators • Software development environments • Document preparation tools and declarative markup • Scripting languages • TCP/IP networking • Configuration management systems
- 18. 4/2/2018 18
- 19. 4/2/2018 19
- 20. 4/2/2018 20 Motivation • Explore evolution of programming style • Consolidate digital artifacts of historical importance • Collect and record history that is fading away • Provide a data set of digital archeology and repository mining
- 21. 4/2/2018 21 In Numbers … Metric Unix history Linux history Start date 30/06/1970 17/09/1991 Start files 43 92 Start lines 11,500 917,812 End files 63,049 51,396 End lines 27,388,943 21,525,436 Data set size (.git) 1.1GB 1.0GB Number of commits 495,622 611,735 Number of merges 2,523 48,821 Number of authors 973 18,465 Days with activity 13,004 5,126
- 22. 4/2/2018 22
- 24. 4/2/2018 24
- 25. 4/2/2018 25
- 27. 4/2/2018 27 Things to Take Away … • Architectural evolution in practice • 1.1GB Git repository – github.com/dspinellis/unix-history-repo • Open source project – github.com/dspinellis/unix-history-make • Architectural lessons to apply • Joint work with Paris Avgeriou Professor (chair of Software Engineering) and Head of the SEARCH (Software Engineering and Architecture) Group of the Johann Bernoulli Institute for Mathematics and Computer Science at the University of Groningen (RuG). @CoolSWEng
- 28. 4/2/2018 28 PDP-7 [Unix] (1970) • Kernel • Layering and Partitioning • System Call • Data Scoping • Interpreter • Monolithic Implementation • Process Management • Descriptor Management • Separation of File Metadata from File Naming • Devices as Files • File I/O • Filesystem
- 29. 4/2/2018 29 Kernel • 2584 lines • Loads and executes user-level commands • Provides the file abstraction • Virtualizes the hardware interfaces • Establishes ownership of files
- 30. 4/2/2018 30 Layering and Partitioning adm.s cat.s dskio.s init.s s6.s ald.s check.s dskres.s lcase.b s7.s apr.s chmod.s dsksav.s maksys.s s8.s as.s chown.s ds.s s1.s s9.s bc.s chrm.s dsw.s s2.s scope.v bi.s cp.s ed1.s s3.s sop.s bl.s db.s ed2.s s4.s trysys.s cas.s dmabs.s ind.b s5.s Process Management (fork)
- 31. 4/2/2018 31 Descriptor Management Separation of File Metadata from File Naming
- 32. 4/2/2018 32 Devices as Files File I/O • open • read • write • seek • tell • close
- 33. 4/2/2018 33 Filesystem • creat • rename • link • unlink Interpreter
- 34. 4/2/2018 34 First Research Edition (Nov 1971) • System Calls • Binary-Code API • Abstraction of Standard I/O • Generic File I/O Layer • User-Contributed Tools and Games • The Shell as a User Program • Interoperability through Documented File Formats • Tree Directory Structure • Mountable Filesystem Interface
- 35. 4/2/2018 35
- 36. 4/2/2018 36
- 37. 4/2/2018 37 First Edition Unix 1972 FreeBSD 11.1 2018 The Shell as a User Program
- 38. 4/2/2018 38 Abstraction of Standard I/O Interoperability through Documented File Formats
- 39. 4/2/2018 39 Format Description Clients a.out Assembler and linker output as, ld, strip, nm, un Archive Object code libraries ar, ld Core Crashed program image Kernel, db Directory File system directories du, find, ls, ln, mkdir, rmdir File system File system format check, dump,* mkfs, restor* Ident GECOD ident card format opr Password User accounts and passwords chown, find, getpw,* login,* ls, passwd* Tape* DECtape file format mt,* tap* Uid User identifier to name map chown utmp Logged in users init, login,* who,* write* wtmp* Users login history acct, date, init, login, tacct, who User-Contributed Tools and Games
- 40. 4/2/2018 40 Tree Directory Structure • mkdir(II) • chdir(II) • chdir(I) • find(I) • ln(I) • ls(I) • stat(I) • mkdir(I) • mv(I) • rm(I) • rmdir(I) Mountable Filesystem Interface • mount(II) • umount(II) • mount(I) • umount(I)
- 41. 4/2/2018 41 Second Research Edition (Jun 1972) • Software Library Third Research Edition (Feb 1973) • Pipes and Filters
- 42. 4/2/2018 42 Fourth Research Edition (Nov 1973) • Structured Programming • User Groups • Language-Independent API • Data Structure Definition Reuse • Dynamic Resource Management • Device Driver Abstraction • Buffer Cache
- 43. 4/2/2018 43 Structured Programming • Kernel implemented in “New B” • 6373 lines New B • 768 lines PDP-11 assembly • 105 functions + 50 assembly symbols • vs 248 global symbols in the First Edition Language-Independent API
- 44. 4/2/2018 44 Data Structure Definition & Reuse buf.h filsys.h proc.h text.h conf.h inode.h reg.h tty.h file.h param.h systm.h user.h Dynamic Resource Management int coremap[CMAPSIZ]; int swapmap[SMAPSIZ]; struct map { char *m_size; char *m_addr; }; malloc(mp, size) struct map *mp; { …
- 45. 4/2/2018 45 Device Driver Abstraction Driver Interface struct { int (*d_open)(); int (*d_close)(); int (*d_strategy)(); } bdevsw[]; struct { int (*d_open)(); int (*d_close)(); int (*d_read)(); int (*d_write)(); int (*d_sgtty)(); } cdevsw[];
- 46. 4/2/2018 46 Buffer Cache Fourth Edition #define B_READ 01 #define B_DONE 02 #define B_ERROR 04 #define B_BUSY 010 #define B_XMEM 060 #define B_WANTED 0100 #define B_RELOC 0200 #define B_ASYNC 0400 #define B_DELWRI 01000 FreeBSD 11.1 #define B_ASYNC 0x00000004 /* Start I/O, do not wait. */ […] #define B_DELWRI 0x00000080 /* Delay I/O until buffer reused. */ #define B_DONE 0x00000200 /* I/O completed. */ Fifth Research Edition (Jun 1974) • Command Files chdir /usr/source/s3 cc -c ctime.c ar r /lib/liba.a ctime.o rm ctime.o chdir /usr/source/s1 cc -s -n date.c cp a.out /bin/date cc -s -n dump.c cp a.out /bin/dump cc -s -n ls.c cp a.out /bin/ls rm a.out
- 47. 4/2/2018 47 Sixth Research Edition (May 1975) • Portable C Library alloc.c clenf.c makbuf.c scan1.c calloc.c copen.c maktab.c scan2.c cclose.c cputc.c nexch.c scan3.c ceof.c cwrd.c nodig.c system.c cerror.c dummy.s printf.c tmpnam.c cexit.c ftoa.c putch.c unget.c cflush.c getch.c puts.c unprnt.s cfree.c gets.c relvec.c wdleng.c cgetc.c getvec.c revput.c ciodec.c iehzap.c run
- 48. 4/2/2018 48 Seventh Research Edition (Jan 1979) • Unix as a Virtual Machine • Dynamic Memory Allocation • Static Analysis • Environment Variables • Language Development Tools • Domain-Specific Languages • Filesystem Directory Hierarchy
- 49. 4/2/2018 49 Unix as a Virtual Machine Also, about this time [1973] I had a fateful discussion with Dennis, in which he said “I think it may be easier to port Unix to a new piece of hardware than to port a complex application from Unix to a new OS” — Steve Johnson Dynamic User Memory Allocation • malloc(3), free(3) • 26 programs: awk cc col cron dc dcheck diff ed eqn expr graph icheck learn ls m4 neqn nm quot ratfor spline struct tar tsort uucp xsend quiz • stdio(3), mp(3)
- 50. 4/2/2018 50 @CoolSWEng Environment Variables • KEY=value • Kernel • Shell • C Library
- 51. 4/2/2018 51 Language Development Tools • lex(1) • yacc(1) • 12 clients: awk bc cpp egrep eqn lex m4 make pcc neqn struct • Domain-Specific Languages • sh • awk • sed • find • expr • egrep • m4 • make
- 53. 4/2/2018 53 First and Second Berkeley Software Distributions (1978) • Software Packages – csh – ex – Mail – Pascal – termlib 3BSD (1979) • Virtual Memory Paging – vm_*.c – 2808 out of 16039 C source code – 17% of kernel source code – vread(2), vwrite(2), vfork(2)
- 54. 4/2/2018 54 4BSD (Oct 1980) • Regular Expression Library: regex(3) • Optimized Screen Handling
- 55. 4/2/2018 55 Regular Expression Library: regex(3) – 5 implementations: awk, sed, ed, grep, expr – 1 client: more(1) – 2 more by 4.3: dbx(1), rdist(1) – 4 replacements in FreeBSD: ed, grep, sed, expr Optimized Screen Handling – curses(3) – termcap(5)
- 56. 4/2/2018 56
- 57. 4/2/2018 57 4.2BSD (Sep 1983) • Internet Protocol Family – ARP, IP, TCP, UDP, ICMP • Local and Remote Interprocess Communication – socket(2), etc. • Network and User Database Access – getfsent(3x), getgrent(3), gethostent(3n), getnetent(3n), getprotoent(3n), getpwent(3), getservent(3n) • Pseudo-Terminal Driver – pty(4) System call Uses bind 23 connect 15 accept 13 select 12 listen 11 sendto 10 shutdown 9 recvfrom 8 getsockname 6 recv 2 send 2 sendmsg 1 getsockopt 0 recvmsg 0 socketpair 0 • 2 library functions – rcmd(3x) – rexec(3x) • 11 system daemons – comsat(8c) – ftpd(8c) – gettable(8c) – implogd(8c) – rexecd(8c) – rlogind(8c) – af(8c) – rshd(8c) – rwhod(8c) – telnetd(8c) – tftpd(8c) • 8 user-mode programs – ftp(1c) – rlogin(1c) – rsh(1c) – talk(1c) – telnet(1c) – tftp(1c) – whois(1c) – sendmail(1c)
- 58. 4/2/2018 58 4.3BSD Tahoe (Jun 1988) • Multiple CPU Architecture Support – VAX – CCI Power 6/32 (Tahoe) • Timezone Handling – Community contribution – Separation of timezone rules from code Machine independent 33% VAX 48% Tahoe 19%
- 59. 4/2/2018 59 4.3BSD Reno (Jun 1990) • Kernel Packet Forwarding Database – route(4) – routed(8), XNSrouted(8) • Virtual Filesystem Interface – … vnode /* * Operations on vnodes. */ struct vnodeops { int (*vn_lookup)( /* ndp */ ); int (*vn_create)( /* ndp, fflags, vap, cred */ ); int (*vn_mknod)( /* ndp, vap, cred */ ); int (*vn_open)( /* vp, fflags, cred */ ); int (*vn_close)( /* vp, fflags, cred */ ); int (*vn_access)( /* vp, fflags, cred */ ); int (*vn_getattr)( /* vp, vap, cred */ ); int (*vn_setattr)( /* vp, vap, cred */ ); int (*vn_read)( /* vp, uiop, offp, ioflag, cred */ ); int (*vn_write)( /* vp, uiop, offp, ioflag, cred */ ); int (*vn_ioctl)( /* vp, com, data, fflag, cred */ ); int (*vn_select)( /* vp, which, cred */ ); int (*vn_mmap)( /* vp, ..., cred */ ); int (*vn_fsync)( /* vp, fflags, cred */ ); int (*vn_seek)( /* vp, (old)offp, off, whence */ ); ….
- 60. 4/2/2018 60 4.3BSD Net/2 (Jun 1991) • Stream I/O Functions – funopen(3) – GNU funopencookie(3) added in FreeBSD 11 4.4BSD (Jun 1994) • Stackable Filesystems – mount_null(8) – mount_union(8) • Generic System Control Interface (MIB) – sysctl(1) – sysctl(3) – sysctl(9)
- 61. 4/2/2018 61 386BSD Patch Kit (1992-1993) • Organized Community Contributions – From open source software … – … to an open source project • Patch metadata – title – author – description – prerequisites
- 62. 4/2/2018 62 FreeBSD 1.1 (May 1994) • Package Manager – Patch – Compile – Install – Uninstall – Handling of dependencies
- 63. 4/2/2018 63 FreeBSD 2.0 (Nov 1994) • Process Filesystem – procfs(5) • Dynamically Loadable Kernel Modules – lkm(4), then kld(4) – device drivers – file systems – emulators – system calls – 992 modules in FreeBSD 11.1 FreeBSD 3.0.0 (Jan 1999) • Common Access Method I/O Subsystem (CAM)
- 64. 4/2/2018 64 FreeBSD 3.4.0 (Dec 1999) • Graph-based Kernel Networking and User Library (NETGRAPH) FreeBSD 4.0.0 (Mar 2000) • OpenSSL Secure Sockets Layer and Transport Layer Security framework – Version 0.9.4 – 1127 files, 227118 lines – libssl(3), libcrypto(3), openssl(1) • Jail: Isolate a process and its descendants
- 65. 4/2/2018 65 FreeBSD 5.0.0 (May 2006) • Modular Disk I/O Request Transformation Framework (GEOM) FreeBSD 5.3.0 (Nov 2004) • Streaming Archive Access Library • Miniport Driver Wrapper
- 66. 4/2/2018 66 FreeBSD 7.0.0 (Feb 2008) • ZFS Filesystems and Storage Pools FreeBSD 7.1.0 (Dec 2008) • Dynamic Tracing
- 67. 4/2/2018 67 FreeBSD 8.0.0 (Nov 2009) • Packet Capture Library – pcap(3) FreeBSD 9.0.0 (Jan 2012) • Infiniband / RDMA High-Speed Low-Latency Switched-Fabric Interconnect Library
- 69. 4/2/2018 69 Modularity increases with code size Increase in number of static declarations / statement static short splice;
- 70. 4/2/2018 70 Modularity increases with code size Increase in number of #include directives / line #include "if_uba.h"
- 71. 4/2/2018 71 Software complexity evolution follows self correction Mean lines / function { …. …. …. } Software complexity evolution follows self correction Mean statement nesting if (a) while (b) for (;;) if (d())
- 72. 4/2/2018 72 Software complexity evolution follows self correction Density of C preprocessor non- include directives #define #if
- 73. 4/2/2018 73 Software complexity evolution follows self correction goto keyword density Cyclomatic Complexity 0 1 2 3 4 5 6 7 8 9 Releases (1973-2016) Kernel
- 74. 4/2/2018 74 Cyclomatic Complexity 0 1 2 3 4 5 6 7 8 9 Releases (1973-2016) Library Cyclomatic Complexity 0 1 2 3 4 5 6 7 8 9 10 Releases (1973-2016) Tools
- 75. 4/2/2018 75 Agreement with Lehman’s Laws • Increasing Complexity • Conservation of Familiarity • Declining Quality • Feedback System Growth in Facilities
- 76. 4/2/2018 76
- 77. 4/2/2018 77
- 78. 4/2/2018 78
- 85. 4/2/2018 85
- 88. 4/2/2018 88 Thank you! github.com/dspinellis/unix-history-repo www.spinellis.gr @CoolSWEng Funding Credit The research described has been partially carried out as part of the CROSSMINER Project, which has received funding from the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No. 732223.
- 89. 4/2/2018 89 Image Credits • Decades: Wikipedia (1970,80,90,2000) • ASR-33 Teletype: Rama & Musée Bolo • VT100: Jason Scott • VAX 11/780: Joe Mabel • PDP 11/20: Image courtesy of Computer History Museum • VAX in use: Photo courtesy of Berkeley Lab © 2010 The Regents of the University of California, through the Lawrence Berkeley National Laboratory. • Pentium: Iorsh • Hypotheses: Niklas Morberg • Modules: Suatu Ketika • Reading glasses: Walt Stoneburner • Cables: christof tof • Chemical flasks: Joe Sullivan • Snake Oil cover: Clark Stanley • Haswell Chip: Intel Free Press • Sparcstation 10: Thomas Kaiser • Gold coins: Anonimski • Go to statement considered harmful: Edsger W. Dijkstra and ACM • Manny Lehman: © Copyright 2009 Imperial College London • Saladin and Guy de Lusignan after battle of Hattin in 1187: Said Tahsine (Creative commons licenses) • PDP11/40: Stefan_Kögl, CC BY-SA 3.0 • Digital VAX 11/780: Emiliano Russo, PD • SPARCstation,Fourdee, PD • Hologous bones: Волков Владислав Петрович, CC BY-SA 4.0 • Skeletal evolution: H. Zell, CC BY-SA 3.0 • Company Shocked at a Lady Getting up to Ring the Bell: James Gillray, PD • SLOW: Monyo Kararan, CC BY-SA-ND 2.0 • Bond: JHerbstman, PD • SSL mixer console: Rebecca Wilson CC-BY-2.0 • Ancient Indian observatory at Delhi: Thomas Daniell, PD • Roman milestone: Júlio Reis , CC BY-SA 3.0 • Florence Cathedral: Bruce Stokes, CC BY-SA 2.0