Sep 5 23:39 2011 README Page 1 Sep 5 23:39 2011 table of contents Page 1

xv6 is a reimplementation of Dennis Ritchies and Ken Thompsons Unix The numbers to the left of the file names in the table are sheet numbers.
Version 6 (v6). xv6 loosely follows the structure and style of v6, The source code has been printed in a double column format with fifty
but is implemented for a modern x86based multiprocessor using ANSI C. lines per column, giving one hundred lines per sheet (or page).
Thus there is a convenient relationship between line numbers and sheet numbers.
ACKNOWLEDGMENTS

xv6 is inspired by John Lionss Commentary on UNIX 6th Edition (Peer # basic headers # system calls # string operations
to Peer Communications; ISBN: 1573980137; 1st edition (June 14, 01 types.h 28 traps.h 59 string.c
2000)). See also http://pdos.csail.mit.edu/6.828/2007/v6.html, which 01 param.h 28 vectors.pl
provides pointers to online resources for v6. 02 memlayout.h 29 trapasm.S # lowlevel hardware
02 defs.h 29 trap.c 61 mp.h
xv6 borrows code from the following sources: 04 x86.h 31 syscall.h 62 mp.c
JOS (asm.h, elf.h, mmu.h, bootasm.S, ide.c, console.c, and others) 06 asm.h 31 syscall.c 64 lapic.c
Plan 9 (entryother.S, mp.h, mp.c, lapic.c) 07 mmu.h 33 sysproc.c 66 ioapic.c
FreeBSD (ioapic.c) 09 elf.h 67 picirq.c
NetBSD (console.c) # file system 68 kbd.h
# entering xv6 34 buf.h 69 kbd.c
The following people made contributions: 10 entry.S 34 fcntl.h 70 console.c
Russ Cox (context switching, locking) 11 entryother.S 35 stat.h 73 timer.c
Cliff Frey (MP) 12 main.c 35 fs.h 74 uart.c
Xiao Yu (MP) 36 file.h
Nickolai Zeldovich # locks 36 ide.c # userlevel
Austin Clements 14 spinlock.h 38 bio.c 75 initcode.S
14 spinlock.c 40 log.c 75 usys.S
In addition, we are grateful for the patches contributed by Greg 42 fs.c 76 init.c
Price, Yandong Mao, and Hitoshi Mitake. # processes 50 file.c 76 sh.c
16 vm.c 52 sysfile.c
The code in the files that constitute xv6 is 20 proc.h 57 exec.c # bootloader
Copyright 20062011 Frans Kaashoek, Robert Morris, and Russ Cox. 21 proc.c 82 bootasm.S
26 swtch.S # pipes 83 bootmain.c
ERROR REPORTS 27 kalloc.c 58 pipe.c

If you spot errors or have suggestions for improvement, please send

email to Frans Kaashoek and Robert Morris (kaashoek,[email protected]).
The source listing is preceded by a crossreference that lists every defined
BUILDING AND RUNNING XV6 constant, struct, global variable, and function in xv6. Each entry gives,
on the same line as the name, the line number (or, in a few cases, numbers)
To build xv6 on an x86 ELF machine (like Linux or FreeBSD), run "make". where the name is defined. Successive lines in an entry list the line
On nonx86 or nonELF machines (like OS X, even on x86), you will numbers where the name is used. For example, this entry:
need to install a crosscompiler gcc suite capable of producing x86 ELF
binaries. See http://pdos.csail.mit.edu/6.828/2011/tools.html. swtch 2658
Then run "make TOOLPREFIX=i386joself". 0374 2428 2466 2657 2658

To run xv6, you can use Bochs or QEMU, both PC simulators. indicates that swtch is defined on line 2658 and is mentioned on five lines
Bochs makes debugging easier, but QEMU is much faster. on sheets 03, 24, and 26.
To run in Bochs, run "make bochs" and then type "c" at the bochs prompt.
To run in QEMU, run "make qemu".

To create a typeset version of the code, run "make xv6.pdf". This

requires the "mpage" utility. See http://www.mesa.nl/pub/mpage/.
Sep 5 23:39 2011 crossreferences Page 1 Sep 5 23:39 2011 crossreferences Page 2

acquire 1474 3829 3968 bwrite 3964 CONSOLE 3639

0377 1474 1478 2160 2323 begin_trans 4125 0264 3964 3967 4079 4111 3639 7321 7322
2358 2417 2474 2518 2533 0334 4125 5071 5174 5323 4175 consoleinit 7316
2566 2579 2766 2781 3016 5413 5511 5556 5574 bzero 4289 0267 1225 7316
3372 3392 3757 3815 3920 bfree 4331 4289 4320 consoleintr 7212
3979 4127 4145 4457 4490 4331 4662 4672 4675 C 6831 7209 0269 6998 7212 7475
4510 4539 4554 4564 5025 bget 3916 6831 6879 6904 6905 6906 consoleread 7251
5041 5056 5863 5884 5905 3916 3946 3956 6907 6908 6910 7209 7219 7251 7322
7060 7216 7258 7306 binit 3889 7222 7229 7240 7269 consolewrite 7301
allocproc 2155 0261 1229 3889 CAPSLOCK 6812 7301 7321
2155 2207 2260 bmap 4610 6812 6845 6986 consputc 7186
allocuvm 1827 4610 4636 4719 4769 cgaputc 7155 7016 7047 7069 7087 7090
0422 1827 1841 2237 5743 bootmain 8317 7155 7198 7094 7095 7186 7226 7232
5753 8268 8317 clearpteu 1903 7239 7308
alltraps 2904 BPB 3588 0431 1903 1909 5755 context 2043
2859 2867 2880 2885 2903 3588 3591 4312 4314 4339 cli 0557 0251 0374 2006 2043 2061
2904 bread 3952 0557 0559 1126 1560 7110 2188 2189 2190 2191 2428
ALT 6810 0262 3952 4076 4077 4089 7189 8212 2466 2628
6810 6838 6840 4104 4173 4282 4293 4313 cmd 7665 copyout 1968
argfd 5213 4338 4411 4432 4517 4626 7665 7677 7686 7687 7692 0430 1968 5763 5774
5213 5256 5271 5283 5294 4668 4719 4769 7693 7698 7702 7706 7715 copyuvm 1916
5306 brelse 3974 7718 7723 7731 7737 7741 0427 1916 1927 1929 2264
argint 3195 0263 3974 3977 4080 4081 7751 7775 7777 7852 7855 cprintf 7052
0395 3195 3208 3224 3332 4096 4112 4176 4284 4296 7857 7858 7859 7860 7863 0268 1222 1264 1841 2626
3356 3370 5218 5271 5283 4319 4324 4345 4417 4420 7864 7866 7868 7869 7870 2630 2632 3040 3053 3058
5508 5576 5577 5626 4441 4525 4632 4674 4722 7871 7872 7873 7874 7875 3285 6319 6339 6511 6662
argptr 3204 4773 7876 7879 7880 7882 7884 7052 7112 7113 7114 7117
0396 3204 5271 5283 5306 BSIZE 3557 7885 7886 7887 7888 7889 cpu 2004
5657 3557 3568 3582 3588 4057 7900 7901 7903 7905 7906 0309 1222 1264 1266 1278
argstr 3221 4078 4174 4294 4719 4720 7907 7908 7909 7910 7913 1406 1466 1487 1508 1546
0397 3221 5318 5408 5508 4721 4765 4769 4770 4771 7914 7916 7918 7919 7920 1561 1562 1570 1572 1618
5557 5575 5606 5626 buf 3400 7921 7922 8012 8013 8014 1631 1637 1772 1773 1774
__attribute__ 1316 0250 0262 0263 0264 0306 8015 8017 8021 8024 8030 1775 2004 2014 2018 2029
0270 0365 1209 1316 0333 1970 1973 1982 1984 8031 8034 8037 8039 8042 2428 2459 2465 2466 2467
BACK 7661 3400 3404 3405 3406 3661 8046 8048 8050 8053 8055 3015 3040 3041 3053 3054
7661 7774 7920 8189 3676 3679 3725 3754 3804 8058 8060 8063 8064 8075 3058 3060 6213 6214 6511
backcmd 7696 7914 3806 3809 3877 3881 3885 8078 8081 8085 8100 8103 7112
7696 7709 7775 7914 7916 3891 3903 3915 3918 3951 8108 8112 8113 8116 8121 cpunum 6501
8042 8155 8190 3954 3964 3974 4005 4076 8122 8128 8137 8138 8144 0324 1256 1288 1624 6501
BACKSPACE 7150 4077 4089 4090 4096 4104 8145 8151 8152 8161 8164 6673 6682
7150 7167 7194 7226 7232 4105 4111 4112 4159 4173 8166 8172 8173 8178 8184 CR0_PE 0727
balloc 4304 4269 4280 4291 4307 4333 8190 8191 8194 0727 1135 1171 8243
4304 4326 4617 4625 4629 4405 4429 4504 4613 4657 COM1 7413 CR0_PG 0737
BBLOCK 3591 4705 4755 7029 7040 7044 7413 7423 7426 7427 7428 0737 1050 1171
3591 4313 4338 7047 7203 7224 7238 7268 7429 7430 7431 7434 7440 CR0_WP 0733
B_BUSY 3409 7301 7308 7784 7787 7788 7441 7457 7459 7467 7469 0733 1050 1171
3409 3808 3926 3927 3938 7789 7803 7815 7816 7819 commit_trans 4136 CR4_PSE 0739
3941 3966 3976 3988 7820 7821 7825 0335 4136 5073 5179 5328 0739 1043 1164
B_DIRTY 3411 B_VALID 3410 5346 5355 5445 5452 5513 create 5457
3411 3737 3766 3771 3810 3410 3770 3810 3829 3957 5558 5562 5579 5583 5457 5477 5490 5494 5512
Sep 5 23:39 2011 crossreferences Page 3 Sep 5 23:39 2011 crossreferences Page 4

5557 5578 6417 6480 6481 fork 2254 IDE_ERR 3666

CRTPORT 7151 exec 5710 0360 2254 3311 7560 7623 3666 3689
7151 7160 7161 7162 7163 0273 5642 5710 7568 7629 7625 7843 7845 ideinit 3701
7178 7179 7180 7181 7630 7726 7727 fork1 7839 0304 1232 3701
CTL 6809 EXEC 7657 7700 7742 7754 7761 7776 ideintr 3752
6809 6835 6839 6985 7657 7722 7859 8165 7824 7839 0305 3024 3752
deallocuvm 1855 execcmd 7669 7853 forkret 2483 idelock 3675
0423 1842 1855 1889 2240 7669 7710 7723 7853 7855 2117 2191 2483 3675 3705 3757 3759 3778
DEVSPACE 0204 8121 8127 8128 8156 8166 freevm 1883 3815 3830 3833
0204 1729 1742 exit 2304 0424 1883 1888 1940 2371 iderw 3804
devsw 3632 0359 2304 2340 3005 3009 5790 5795 0306 3804 3809 3811 3813
3632 3637 4708 4710 4758 3069 3078 3317 7516 7519 gatedesc 0901 3958 3969
4760 5007 7321 7322 7561 7626 7631 7716 7725 0523 0526 0901 2961 idestart 3725
dinode 3572 7735 7780 7828 7835 getcallerpcs 1526 3679 3725 3728 3776 3825
3572 3582 4406 4412 4430 EXTMEM 0202 0378 1488 1526 2628 7115 idewait 3683
4433 4505 4518 0202 0208 getcmd 7784 3683 3708 3730 3766
dirent 3596 fdalloc 5232 7784 7815 idtinit 2979
3596 4815 4855 5364 5404 5232 5258 5526 5662 gettoken 7956 0406 1265 2979
dirlink 4852 fetchint 3167 7956 8041 8045 8057 8070 idup 4488
0286 4822 4852 4867 4875 0398 3167 3197 5633 8071 8107 8111 8133 0289 2280 4488 4961
5339 5489 5493 5494 fetchstr 3179 growproc 2231 iget 4453
dirlookup 4812 0399 3179 3226 5639 0361 2231 3359 4394 4418 4453 4473 4830
0287 4812 4818 4859 4974 file 3600 havedisk1 3678 4959
5421 5467 0252 0276 0277 0278 0280 3678 3714 3812 iinit 4389
DIRSIZ 3594 0281 0282 0351 2064 3600 holding 1544 0290 1231 4389
3594 3598 4805 4872 4928 4271 5004 5010 5020 5023 0379 1477 1504 1544 2457 ilock 4502
4929 4991 5315 5405 5461 5026 5038 5039 5052 5054 ialloc 4402 0291 4502 4508 4528 4964
DPL_USER 0779 5079 5102 5152 5207 5213 0288 4402 4422 5476 5477 5082 5111 5175 5325 5338
0779 1627 1628 2214 2215 5216 5232 5253 5267 5279 IBLOCK 3585 5351 5415 5423 5465 5469
2973 3068 3077 5292 5303 5505 5654 5806 3585 4411 4432 4517 5479 5519 5608 5722 7263
E0ESC 6816 5821 7010 7408 7678 7733 I_BUSY 3627 7283 7310
6816 6970 6974 6975 6977 7734 7864 7872 8072 3627 4511 4513 4536 4540 inb 0453
6980 filealloc 5021 4557 4559 0453 3687 3713 6354 6964
elfhdr 0955 0276 5021 5526 5827 ICRHI 6428 6967 7161 7163 7434 7440
0955 5715 8319 8324 fileclose 5052 6428 6487 6556 6568 7441 7457 7467 7469 8223
ELF_MAGIC 0952 0277 2315 5052 5058 5297 ICRLO 6418 8231 8354
0952 5728 8330 5528 5665 5666 5854 5856 6418 6488 6489 6557 6559 initlock 1462
ELF_PROG_LOAD 0986 filedup 5039 6569 0380 1462 2125 2744 2975
0986 5739 0278 2279 5039 5043 5260 ID 6411 3705 3893 4061 4391 5016
enter_alloc 2725 fileinit 5014 6411 6447 6516 5835 7318 7319
0314 1299 1755 2725 0279 1230 5014 IDE_BSY 3663 initlog 4055
entry 1040 fileread 5102 3663 3687 0332 2494 4055 4058
0961 1036 1039 1040 2731 0280 5102 5117 5273 IDE_CMD_READ 3668 inituvm 1786
2852 2853 5787 6121 8321 filestat 5079 3668 3741 0425 1786 1791 2211
8345 8346 0281 5079 5308 IDE_CMD_WRITE 3669 inode 3613
EOI 6414 filewrite 5152 3669 3738 0253 0286 0287 0288 0289
6414 6484 6525 0282 5152 5184 5189 5285 IDE_DF 3665 0291 0292 0293 0294 0295
ERROR 6435 FL_IF 0710 3665 3689 0297 0298 0299 0300 0301
6435 6477 0710 1562 1568 2218 2463 IDE_DRDY 3664 0426 1803 2065 3606 3613
ESR 6417 6508 3664 3687 3633 3634 4274 4385 4394
Sep 5 23:39 2011 crossreferences Page 5 Sep 5 23:39 2011 crossreferences Page 6

4401 4427 4452 4455 4461 2833 3034 7442 7443 KERNBASE 0207 6525 6556 6557 6559 6568
4487 4488 4502 4534 4552 IRQ_ERROR 2835 0207 0208 0212 0213 0217 6569
4574 4610 4654 4685 4702 2835 6477 0218 0220 0221 1321 1533 lcr3 0590
4752 4811 4812 4852 4856 IRQ_IDE 2834 1832 1889 2730 0590 1764 1779
4953 4956 4988 4995 5316 2834 3023 3027 3706 3707 KERNLINK 0208 lgdt 0512
5361 5403 5456 5460 5506 IRQ_KBD 2832 0208 1727 0512 0520 1133 1633 8241
5554 5569 5604 5716 7251 2832 3030 7325 7326 KEY_DEL 6828 lidt 0526
7301 IRQ_SLAVE 6710 6828 6869 6891 6915 0526 0534 2981
INPUT_BUF 7200 6710 6714 6752 6767 KEY_DN 6822 LINT0 6433
7200 7203 7224 7236 7238 IRQ_SPURIOUS 2836 6822 6865 6887 6911 6433 6468
7240 7268 2836 3039 6457 KEY_END 6820 LINT1 6434
insl 0462 IRQ_TIMER 2831 6820 6868 6890 6914 6434 6469
0462 0464 3767 8373 2831 3014 3073 6464 7380 KEY_HOME 6819 LIST 7660
install_trans 4071 isdirempty 5361 6819 6868 6890 6914 7660 7740 7907 8183
4071 4119 4140 5361 5368 5427 KEY_INS 6827 listcmd 7690 7901
INT_DISABLED 6619 ismp 6215 6827 6869 6891 6915 7690 7711 7741 7901 7903
6619 6667 0338 1233 6215 6312 6320 KEY_LF 6823 8046 8157 8184
ioapic 6627 6340 6343 6655 6675 6823 6867 6889 6913 loadgs 0551
6307 6329 6330 6624 6627 itrunc 4654 KEY_PGDN 6826 0551 1634
6636 6637 6643 6644 6658 4274 4561 4654 6826 6866 6888 6912 loaduvm 1803
IOAPIC 6608 iunlock 4534 KEY_PGUP 6825 0426 1803 1809 1812 5745
6608 6658 0293 4534 4537 4576 4971 6825 6866 6888 6912 log 4040 4050
ioapicenable 6673 5084 5114 5178 5334 5532 KEY_RT 6824 4040 4050 4061 4063 4064
0309 3707 6673 7326 7443 5613 7256 7305 6824 6867 6889 6913 4065 4075 4076 4077 4089
ioapicid 6217 iunlockput 4574 KEY_UP 6821 4092 4093 4094 4104 4107
0310 6217 6330 6347 6661 0294 4574 4966 4975 4978 6821 6865 6887 6911 4108 4109 4120 4127 4128
6662 5327 5340 5343 5354 5428 kfree 2756 4129 4131 4132 4138 4141
ioapicinit 6651 5439 5443 5451 5468 5472 0316 1871 1873 1893 1896 4145 4146 4147 4148 4163
0311 1224 6651 6662 5496 5521 5529 5561 5582 2265 2369 2747 2756 2761 4165 4168 4169 4172 4173
ioapicread 6634 5610 5748 5797 5852 5873 4177 4178
6634 6659 6660 iupdate 4427 kill 2575 logheader 4035
ioapicwrite 6641 0295 4427 4563 4680 4778 0362 2575 3059 3334 7567 4035 4046 4057 4058 4090
6641 6667 6668 6681 6682 5333 5353 5437 5442 5483 kinit 2740 4105
IO_PIC1 6707 5487 0317 1236 2740 LOGSIZE 0160
6707 6720 6735 6744 6747 I_VALID 3628 KSTACKSIZE 0151 0160 4037 4163 5167
6752 6762 6776 6777 3628 4516 4526 4555 0151 1054 1063 1300 1775 log_write 4159
IO_PIC2 6708 kalloc 2777 2177 0333 4159 4295 4318 4344
6708 6721 6736 6765 6766 0315 1792 1794 1839 1846 kvmalloc 1753 4416 4440 4630 4772
6767 6770 6779 6780 1923 1931 1934 2173 2209 0418 1218 1753 ltr 0538
IO_RTC 6535 2777 5731 5829 lapiceoi 6522 0538 0540 1776
6535 6548 6549 KBDATAP 6804 0326 3021 3025 3032 3036 mappages 1679
IO_TIMER1 7359 6804 6967 3042 6522 1679 1745 1794 1846 1934
7359 7368 7378 7379 kbdgetc 6956 lapicinit 6451 MAXARG 0159
IPB 3582 6956 6998 0327 1220 1256 6451 0159 5622 5714 5760
3582 3585 3591 4412 4433 kbdintr 6996 lapicstartap 6540 MAXARGS 7663
4518 0321 3031 6996 0328 1304 6540 7663 7671 7672 8140
iput 4552 KBS_DIB 6803 lapicw 6444 MAXFILE 3569
0292 2320 4552 4558 4577 6803 6965 6444 6457 6463 6464 6465 3569 4765
4860 4982 5072 5344 5614 KBSTATP 6802 6468 6469 6474 6477 6480 memcmp 5965
IRQ_COM1 2833 6802 6964 6481 6484 6487 6488 6493 0386 5965 6245 6288
Sep 5 23:39 2011 crossreferences Page 7 Sep 5 23:39 2011 crossreferences Page 8

memmove 5981 6238 6264 6268 6271 2418 2557 2580 2619 5490 5494 7063 7105 7112
0387 1285 1795 1933 1982 multiboot_header 1025 NPTENTRIES 0822 7701 7720 7753 7832 7845
4078 4174 4283 4439 4524 1024 1025 0822 1867 8028 8072 8106 8110 8136
4721 4771 4929 4931 5981 namecmp 4803 NSEGS 2001 8141
6004 7173 0296 4803 4825 5418 1611 2001 2008 panicked 7018
memset 5954 namei 4989 nulterminate 8152 7018 7118 7188
0388 1666 1741 1793 1845 0297 2223 4989 5320 5517 8015 8030 8152 8173 8179 parseblock 8101
2190 2213 2733 2764 4294 5606 5720 8180 8185 8186 8191 8101 8106 8125
4414 5432 5629 5954 7175 nameiparent 4996 NUMLOCK 6813 parsecmd 8018
7787 7858 7869 7885 7906 0298 4954 4969 4981 4996 6813 6846 7702 7825 8018
7919 5336 5410 5463 O_CREATE 3453 parseexec 8117
microdelay 6531 namex 4954 3453 5510 8078 8081 8014 8055 8117
0329 6531 6558 6560 6570 4954 4992 4998 O_RDONLY 3450 parseline 8035
7458 NBUF 0155 3450 5520 8075 8012 8024 8035 8046 8108
min 4273 0155 3881 3903 O_RDWR 3452 parsepipe 8051
4273 4720 4770 ncpu 6216 3452 5538 7614 7616 7807 8013 8039 8051 8058
mp 6102 1222 1287 2019 3707 6216 outb 0471 parseredirs 8064
6102 6208 6237 6244 6245 6318 6319 6323 6324 6325 0471 3711 3720 3731 3732 8064 8112 8131 8142
6246 6255 6260 6264 6265 6345 3733 3734 3735 3736 3738 PCINT 6432
6268 6269 6280 6283 6285 NCPU 0152 3741 6353 6354 6548 6549 6432 6474
6287 6294 6304 6310 6350 0152 2018 6213 6720 6721 6735 6736 6744 pde_t 0103
mpbcpu 6220 NDEV 0157 6747 6752 6762 6765 6766 0103 0420 0421 0422 0423
0339 1220 6220 0157 4708 4758 5007 6767 6770 6776 6777 6779 0424 0425 0426 0427 0430
MPBUS 6152 NDIRECT 3567 6780 7160 7162 7178 7179 0431 1210 1270 1317 1610
6152 6333 3567 3569 3578 3624 4615 7180 7181 7377 7378 7379 1654 1656 1679 1733 1736
mpconf 6113 4620 4624 4625 4660 4667 7423 7426 7427 7428 7429 1739 1786 1803 1827 1855
6113 6279 6282 6287 6305 4668 4675 4676 7430 7431 7459 8228 8236 1883 1903 1915 1916 1918
mpconfig 6280 NELEM 0434 8364 8365 8366 8367 8368 1952 1968 2055 5718
6280 6310 0434 1744 2622 3282 5631 8369 PDX 0812
mpenter 1252 nextpid 2116 outsl 0483 0812 1659
1252 1301 2116 2169 0483 0485 3739 PDXSHIFT 0827
mpinit 6301 NFILE 0154 outw 0477 0812 0818 0827 1321
0340 1219 6301 6319 6339 0154 5010 5026 0477 1181 1183 8274 8276 peek 8001
mpioapic 6139 NINDIRECT 3568 O_WRONLY 3451 8001 8025 8040 8044 8056
6139 6307 6329 6331 3568 3569 4622 4670 3451 5537 5538 8078 8081 8069 8105 8109 8124 8132
MPIOAPIC 6153 NINODE 0156 P2V 0218 PGROUNDDOWN 0830
6153 6328 0156 4385 4461 0218 1726 6262 6550 7152 0830 1685 1686 1975
MPIOINTR 6154 NO 6806 panic 7105 7832 PGROUNDUP 0829
6154 6334 6806 6852 6855 6857 6858 0270 1478 1505 1569 1571 0829 1837 1863 2732 2745
MPLINTR 6155 6859 6860 6862 6874 6877 1691 1743 1778 1791 1809 5752
6155 6335 6879 6880 6881 6882 6884 1812 1871 1888 1909 1927 PGSIZE 0823
mpmain 1262 6902 6903 6905 6906 6907 1929 2210 2310 2340 2458 0823 0829 0830 1316 1666
1209 1239 1257 1262 6908 2460 2462 2464 2506 2509 1695 1696 1741 1790 1793
mpproc 6128 NOFILE 0153 2731 2761 3055 3728 3809 1794 1808 1810 1814 1817
6128 6306 6317 6326 0153 2064 2277 2313 5220 3811 3813 3946 3967 3977 1838 1845 1846 1864 1867
MPPROC 6151 5236 4058 4164 4166 4326 4342 1925 1933 1934 1979 1985
6151 6316 NPDENTRIES 0821 4422 4473 4508 4528 4537 2212 2219 2733 2734 2746
mpsearch 6256 0821 1317 1890 4558 4636 4818 4822 4867 2760 2764 5753 5755
6256 6285 NPROC 0150 4875 5043 5058 5117 5184 PHYSTOP 0203
mpsearch1 6238 0150 2111 2161 2329 2362 5189 5368 5426 5434 5477 0203 1728 1742 1743 2746
Sep 5 23:39 2011 crossreferences Page 9 Sep 5 23:39 2011 crossreferences Page 10

2760 2955 3004 3006 3008 3051 readsb 4278 0365 1267 2006 2408 2428
picenable 6725 3059 3060 3062 3068 3073 0285 4062 4278 4311 4337 2466
0344 3706 6725 7325 7380 3077 3155 3167 3179 3197 4409 SCROLLLOCK 6814
7442 3210 3226 3279 3281 3283 readsect 8360 6814 6847
picinit 6732 3286 3287 3306 3340 3358 8360 8395 SECTSIZE 8312
0345 1223 6732 3375 3657 4267 4961 5205 readseg 8379 8312 8373 8386 8389 8394
picsetmask 6717 5220 5237 5238 5296 5614 8314 8327 8338 8379 SEG 0769
6717 6727 6783 5615 5633 5639 5664 5704 recover_from_log 4116 0769 1625 1626 1627 1628
pinit 2123 5781 5784 5785 5786 5787 4052 4066 4116 1631
0363 1227 2123 5788 5789 5804 5887 5907 REDIR 7658 SEG16 0773
pipe 5811 6211 6306 6317 6318 6319 7658 7730 7870 8171 0773 1772
0254 0352 0353 0354 3605 6322 7013 7261 7410 redircmd 7675 7864 SEG_ASM 0660
5069 5109 5159 5811 5823 procdump 2604 7675 7713 7731 7864 7866 0660 1190 1191 8284 8285
5829 5835 5839 5843 5861 0364 2604 7220 8075 8078 8081 8159 8172 segdesc 0752
5880 5901 7563 7752 7753 proghdr 0974 REG_ID 6610 0509 0512 0752 0769 0773
PIPE 7659 0974 5717 8320 8334 6610 6660 1611 2008
7659 7750 7886 8177 PTE_ADDR 0844 REG_TABLE 6612 seginit 1616
pipealloc 5821 0844 1661 1813 1869 1892 6612 6667 6668 6681 6682 0417 1221 1255 1616
0351 5659 5821 1930 1961 REG_VER 6611 SEG_KCODE 0741
pipeclose 5861 PTE_P 0833 6611 6659 0741 1150 1625 2972 2973
0352 5069 5861 0833 1319 1321 1660 1670 release 1502 8253
pipecmd 7684 7880 1690 1692 1868 1891 1928 0381 1502 1505 2164 2170 SEG_KCPU 0743
7684 7712 7751 7880 7882 1957 2377 2384 2435 2477 2487 0743 1631 1634 2916
8058 8158 8178 PTE_PS 0840 2519 2532 2568 2586 2590 SEG_KDATA 0742
piperead 5901 0840 1319 1321 2770 2785 3019 3376 3381 0742 1154 1626 1774 2913
0353 5109 5901 pte_t 0847 3394 3759 3778 3833 3928 8258
PIPESIZE 5809 0847 1653 1657 1661 1663 3942 3991 4132 4148 4464 SEG_NULLASM 0654
5809 5813 5886 5894 5916 1683 1806 1857 1905 1919 4480 4492 4514 4542 4560 0654 1189 8283
pipewrite 5880 1954 4569 5029 5033 5045 5060 SEG_TSS 0746
0354 5159 5880 PTE_U 0835 5066 5872 5875 5888 5897 0746 1772 1773 1776
popcli 1566 0835 1670 1794 1846 1910 5908 5919 7101 7248 7262 SEG_UCODE 0744
0383 1521 1566 1569 1571 1934 1959 7282 7309 0744 1627 2214
1780 PTE_W 0834 ROOTDEV 0158 SEG_UDATA 0745
printint 7026 0834 1319 1321 1670 1726 0158 4062 4065 4959 0745 1628 2215
7026 7077 7081 1728 1729 1794 1846 1934 ROOTINO 3556 SETGATE 0921
proc 2053 PTX 0815 3556 4959 0921 2972 2973
0255 0358 0398 0399 0428 0815 1672 run 2711 setupkvm 1734
1205 1458 1606 1638 1769 PTXSHIFT 0826 2611 2711 2712 2717 2758 0420 1734 1755 1923 2209
1775 2015 2030 2053 2059 0815 0818 0826 2767 2779 5731
2106 2111 2114 2154 2157 pushcli 1555 runcmd 7706 SHIFT 6808
2161 2204 2235 2237 2240 0382 1476 1555 1771 7706 7720 7737 7743 7745 6808 6836 6837 6985
2243 2244 2257 2264 2270 rcr2 0582 7759 7766 7777 7825 skipelem 4915
2271 2272 2278 2279 2280 0582 3054 3061 RUNNING 2050 4915 4963
2284 2306 2309 2314 2315 readeflags 0544 2050 2427 2461 2611 3073 sleep 2503
2316 2320 2321 2326 2329 0544 1559 1568 2463 6508 safestrcpy 6032 0367 2389 2503 2506 2509
2330 2338 2355 2362 2363 read_head 4087 0389 2222 2284 5781 6032 2609 3379 3830 3931 4129
2383 2389 2410 2418 2425 4087 4118 sched 2453 4512 5892 5911 7266 7579
2428 2433 2461 2466 2475 readi 4702 0366 2339 2453 2458 2460 spinlock 1401
2505 2523 2524 2528 2555 0299 1818 4702 4821 4866 2462 2464 2476 2525 0256 0367 0377 0379 0380
2557 2577 2580 2615 2619 5112 5367 5368 5726 5737 scheduler 2408 0381 0409 1401 1459 1462
Sep 5 23:39 2011 crossreferences Page 11 Sep 5 23:39 2011 crossreferences Page 12

1474 1502 1544 2107 2110 0258 0285 3560 4060 4278 3111 3261 3248 3266 5277
2503 2709 2716 2958 2963 4308 4334 4407 sys_kill 3328 SYS_write 3117
3660 3675 3876 3880 4003 SVR 6415 3237 3256 3328 3117 3266
4041 4268 4384 5005 5009 6415 6457 SYS_kill 3106 taskstate 0851
5807 5812 7008 7021 7202 switchkvm 1762 3106 3256 0851 2007
7406 0429 1254 1756 1762 2429 sys_link 5313 TDCR 6439
STA_R 0669 0786 switchuvm 1769 3238 3269 5313 6439 6463
0669 0786 1190 1625 1627 0428 1769 1778 2244 2426 SYS_link 3120 T_DEV 3502
8284 5789 3120 3269 3502 4707 4757 5578
start 1125 7508 8211 swtch 2658 sys_mkdir 5551 T_DIR 3500
1124 1125 1167 1175 1177 0374 2428 2466 2657 2658 3239 3270 5551 3500 4817 4965 5326 5427
4042 4063 4076 4089 4104 syscall 3275 SYS_mkdir 3121 5435 5485 5520 5557 5609
4173 7507 7508 8210 8211 0400 3007 3157 3275 3121 3270 T_FILE 3501
8267 SYSCALL 7553 7560 7561 7562 7563 75 sys_mknod 5567 3501 5470 5512
startothers 1274 7560 7561 7562 7563 7564 3240 3267 5567 ticks 2964
1208 1235 1274 7565 7566 7567 7568 7569 SYS_mknod 3118 0407 2964 3017 3018 3373
stat 3504 7570 7571 7572 7573 7574 3118 3267 3374 3379 3393
0257 0281 0300 3504 4265 7575 7576 7577 7578 7579 sys_open 5501 tickslock 2963
4685 5079 5203 5304 7603 7580 3241 3265 5501 0409 2963 2975 3016 3019
stati 4685 sys_chdir 5601 SYS_open 3116 3372 3376 3379 3381 3392
0300 4685 5083 3229 3259 5601 3116 3265 3280 3282 3394
STA_W 0668 0785 SYS_chdir 3109 sys_pipe 5651 TICR 6437
0668 0785 1191 1626 1628 3109 3259 3242 3254 5651 6437 6465
1631 8285 sys_close 5289 SYS_pipe 3104 TIMER 6429
STA_X 0665 0782 3230 3271 5289 3104 3254 6429 6464
0665 0782 1190 1625 1627 SYS_close 3122 sys_read 5265 TIMER_16BIT 7371
8284 3122 3271 3243 3255 5265 7371 7377
sti 0563 sys_dup 5251 SYS_read 3105 TIMER_DIV 7366
0563 0565 1573 2414 3231 3260 5251 3105 3255 7366 7378 7379
stosb 0492 SYS_dup 3110 sys_sbrk 3351 TIMER_FREQ 7365
0492 0494 5960 8340 3110 3260 3244 3262 3351 7365 7366
stosl 0501 sys_exec 5620 SYS_sbrk 3112 timerinit 7374
0501 0503 5958 3232 3257 5620 3112 3262 0403 1234 7374
strlen 6051 SYS_exec 3107 sys_sleep 3365 TIMER_MODE 7368
0390 5762 5763 6051 7819 3107 3257 7512 3245 3263 3365 7368 7377
8023 sys_exit 3315 SYS_sleep 3113 TIMER_RATEGEN 7370
strncmp 6008 3233 3252 3315 3113 3263 7370 7377
0391 4805 6008 SYS_exit 3102 sys_unlink 5401 TIMER_SEL0 7369
strncpy 6018 3102 3252 7517 3246 3268 5401 7369 7377
0392 4872 6018 sys_fork 3309 SYS_unlink 3119 T_IRQ0 2829
STS_IG32 0800 3234 3251 3309 3119 3268 2829 3014 3023 3027 3030
0800 0927 SYS_fork 3101 sys_uptime 3388 3034 3038 3039 3073 6457
STS_T32A 0797 3101 3251 3249 3264 3388 6464 6477 6667 6681 6747
0797 1772 sys_fstat 5301 SYS_uptime 3114 6766
STS_TG32 0801 3235 3258 5301 3114 3264 TPR 6413
0801 0927 SYS_fstat 3108 sys_wait 3322 6413 6493
sum 6226 3108 3258 3247 3253 3322 trap 3001
6226 6228 6230 6232 6233 sys_getpid 3338 SYS_wait 3103 2852 2854 2922 3001 3053
6245 6292 3236 3261 3338 3103 3253 3055 3058
superblock 3560 SYS_getpid 3111 sys_write 5277 trapframe 0602
Sep 5 23:39 2011 crossreferences Page 13

0602 2060 2181 3001 VER 6412

trapret 2927 6412 6473
2118 2186 2926 2927 wait 2353
T_SYSCALL 2826 0369 2353 3324 7562 7633
2826 2973 3003 7513 7518 7744 7770 7771 7826
7557 waitdisk 8351
tvinit 2967 8351 8363 8372
0408 1228 2967 wakeup 2564
uart 7415 0370 2564 3018 3772 3989
7415 7436 7455 7465 4147 4541 4566 5866 5869
uartgetc 7463 5891 5896 5918 7242
7463 7475 wakeup1 2553
uartinit 7418 2120 2326 2333 2553 2567
0412 1226 7418 walkpgdir 1654
uartintr 7473 1654 1688 1811 1865 1907
0413 3035 7473 1926 1956
uartputc 7451 write_head 4102
0414 7195 7197 7447 7451 4102 4121 4139 4142
userinit 2202 writei 4752
0368 1237 2202 2210 0301 4752 4874 5176 5433
uva2ka 1952 5434
0421 1952 1976 xchg 0569
V2P 0217 0569 1266 1483 1519
0217 1727 1728 yield 2472
V2P_WO 0220 0371 2472 3074
0220 1036 1046
Sep 5 23:39 2011 xv6/types.h Page 1 Sep 5 23:39 2011 xv6/param.h Page 1

0100 typedef unsigned int uint; 0150 #define NPROC 64 // maximum number of processes
0101 typedef unsigned short ushort; 0151 #define KSTACKSIZE 4096 // size of perprocess kernel stack
0102 typedef unsigned char uchar; 0152 #define NCPU 8 // maximum number of CPUs
0103 typedef uint pde_t; 0153 #define NOFILE 16 // open files per process
0104 0154 #define NFILE 100 // open files per system
0105 0155 #define NBUF 10 // size of disk block cache
0106 0156 #define NINODE 50 // maximum number of active inodes
0107 0157 #define NDEV 10 // maximum major device number
0108 0158 #define ROOTDEV 1 // device number of file system root disk
0109 0159 #define MAXARG 32 // max exec arguments
0110 0160 #define LOGSIZE 10 // max data sectors in ondisk log
0111 0161
0112 0162
0113 0163
0114 0164
0115 0165
0116 0166
0117 0167
0118 0168
0119 0169
0120 0170
0121 0171
0122 0172
0123 0173
0124 0174
0125 0175
0126 0176
0127 0177
0128 0178
0129 0179
0130 0180
0131 0181
0132 0182
0133 0183
0134 0184
0135 0185
0136 0186
0137 0187
0138 0188
0139 0189
0140 0190
0141 0191
0142 0192
0143 0193
0144 0194
0145 0195
0146 0196
0147 0197
0148 0198
0149 0199

Sheet 01 Sheet 01
Sep 5 23:39 2011 xv6/memlayout.h Page 1 Sep 5 23:39 2011 xv6/defs.h Page 1

0200 // Memory layout 0250 struct buf;

0201 0251 struct context;
0202 #define EXTMEM 0x100000 // Start of extended memory 0252 struct file;
0203 #define PHYSTOP 0xE000000 // Top physical memory 0253 struct inode;
0204 #define DEVSPACE 0xFE000000 // Other devices are at high addresses 0254 struct pipe;
0205 0255 struct proc;
0206 // Key addresses for address space layout (see kmap in vm.c for layout) 0256 struct spinlock;
0207 #define KERNBASE 0x80000000 // First kernel virtual address 0257 struct stat;
0208 #define KERNLINK (KERNBASE+EXTMEM) // Address where kernel is linked 0258 struct superblock;
0209 0259
0210 #ifndef __ASSEMBLER__ 0260 // bio.c
0211 0261 void binit(void);
0212 static inline uint v2p(void *a) { return (uint) a KERNBASE; } 0262 struct buf* bread(uint, uint);
0213 static inline void *p2v(uint a) { return (void *) a + KERNBASE; } 0263 void brelse(struct buf*);
0214 0264 void bwrite(struct buf*);
0215 #endif 0265
0216 0266 // console.c
0217 #define V2P(a) ((uint) a KERNBASE) 0267 void consoleinit(void);
0218 #define P2V(a) ((void *) a + KERNBASE) 0268 void cprintf(char*, ...);
0219 0269 void consoleintr(int(*)(void));
0220 #define V2P_WO(x) ((x) KERNBASE) // same as V2P, but without casts 0270 void panic(char*) __attribute__((noreturn));
0221 #define P2V_WO(x) ((x) + KERNBASE) // same as V2P, but without casts 0271
0222 0272 // exec.c
0223 0273 int exec(char*, char**);
0224 0274
0225 0275 // file.c
0226 0276 struct file* filealloc(void);
0227 0277 void fileclose(struct file*);
0228 0278 struct file* filedup(struct file*);
0229 0279 void fileinit(void);
0230 0280 int fileread(struct file*, char*, int n);
0231 0281 int filestat(struct file*, struct stat*);
0232 0282 int filewrite(struct file*, char*, int n);
0233 0283
0234 0284 // fs.c
0235 0285 void readsb(int dev, struct superblock *sb);
0236 0286 int dirlink(struct inode*, char*, uint);
0237 0287 struct inode* dirlookup(struct inode*, char*, uint*);
0238 0288 struct inode* ialloc(uint, short);
0239 0289 struct inode* idup(struct inode*);
0240 0290 void iinit(void);
0241 0291 void ilock(struct inode*);
0242 0292 void iput(struct inode*);
0243 0293 void iunlock(struct inode*);
0244 0294 void iunlockput(struct inode*);
0245 0295 void iupdate(struct inode*);
0246 0296 int namecmp(const char*, const char*);
0247 0297 struct inode* namei(char*);
0248 0298 struct inode* nameiparent(char*, char*);
0249 0299 int readi(struct inode*, char*, uint, uint);

Sheet 02 Sheet 02
Sep 5 23:39 2011 xv6/defs.h Page 2 Sep 5 23:39 2011 xv6/defs.h Page 3

0300 void stati(struct inode*, struct stat*); 0350 // pipe.c

0301 int writei(struct inode*, char*, uint, uint); 0351 int pipealloc(struct file**, struct file**);
0302 0352 void pipeclose(struct pipe*, int);
0303 // ide.c 0353 int piperead(struct pipe*, char*, int);
0304 void ideinit(void); 0354 int pipewrite(struct pipe*, char*, int);
0305 void ideintr(void); 0355
0306 void iderw(struct buf*); 0356
0307 0357 // proc.c
0308 // ioapic.c 0358 struct proc* copyproc(struct proc*);
0309 void ioapicenable(int irq, int cpu); 0359 void exit(void);
0310 extern uchar ioapicid; 0360 int fork(void);
0311 void ioapicinit(void); 0361 int growproc(int);
0312 0362 int kill(int);
0313 // kalloc.c 0363 void pinit(void);
0314 char* enter_alloc(void); 0364 void procdump(void);
0315 char* kalloc(void); 0365 void scheduler(void) __attribute__((noreturn));
0316 void kfree(char*); 0366 void sched(void);
0317 void kinit(void); 0367 void sleep(void*, struct spinlock*);
0318 uint detect_memory(void); 0368 void userinit(void);
0319 0369 int wait(void);
0320 // kbd.c 0370 void wakeup(void*);
0321 void kbdintr(void); 0371 void yield(void);
0322 0372
0323 // lapic.c 0373 // swtch.S
0324 int cpunum(void); 0374 void swtch(struct context**, struct context*);
0325 extern volatile uint* lapic; 0375
0326 void lapiceoi(void); 0376 // spinlock.c
0327 void lapicinit(int); 0377 void acquire(struct spinlock*);
0328 void lapicstartap(uchar, uint); 0378 void getcallerpcs(void*, uint*);
0329 void microdelay(int); 0379 int holding(struct spinlock*);
0330 0380 void initlock(struct spinlock*, char*);
0331 // log.c 0381 void release(struct spinlock*);
0332 void initlog(void); 0382 void pushcli(void);
0333 void log_write(struct buf*); 0383 void popcli(void);
0334 void begin_trans(); 0384
0335 void commit_trans(); 0385 // string.c
0336 0386 int memcmp(const void*, const void*, uint);
0337 // mp.c 0387 void* memmove(void*, const void*, uint);
0338 extern int ismp; 0388 void* memset(void*, int, uint);
0339 int mpbcpu(void); 0389 char* safestrcpy(char*, const char*, int);
0340 void mpinit(void); 0390 int strlen(const char*);
0341 void mpstartthem(void); 0391 int strncmp(const char*, const char*, uint);
0342 0392 char* strncpy(char*, const char*, int);
0343 // picirq.c 0393
0344 void picenable(int); 0394 // syscall.c
0345 void picinit(void); 0395 int argint(int, int*);
0346 0396 int argptr(int, char**, int);
0347 0397 int argstr(int, char**);
0348 0398 int fetchint(struct proc*, uint, int*);
0349 0399 int fetchstr(struct proc*, uint, char**);

Sheet 03 Sheet 03
Sep 5 23:39 2011 xv6/defs.h Page 4 Sep 5 23:39 2011 xv6/x86.h Page 1

0400 void syscall(void); 0450 // Routines to let C code use special x86 instructions.
0401 0451
0402 // timer.c 0452 static inline uchar
0403 void timerinit(void); 0453 inb(ushort port)
0404 0454 {
0405 // trap.c 0455 uchar data;
0406 void idtinit(void); 0456
0407 extern uint ticks; 0457 asm volatile("in %1,%0" : "=a" (data) : "d" (port));
0408 void tvinit(void); 0458 return data;
0409 extern struct spinlock tickslock; 0459 }
0410 0460
0411 // uart.c 0461 static inline void
0412 void uartinit(void); 0462 insl(int port, void *addr, int cnt)
0413 void uartintr(void); 0463 {
0414 void uartputc(int); 0464 asm volatile("cld; rep insl" :
0415 0465 "=D" (addr), "=c" (cnt) :
0416 // vm.c 0466 "d" (port), "0" (addr), "1" (cnt) :
0417 void seginit(void); 0467 "memory", "cc");
0418 void kvmalloc(void); 0468 }
0419 void vmenable(void); 0469
0420 pde_t* setupkvm(char* (*alloc)()); 0470 static inline void
0421 char* uva2ka(pde_t*, char*); 0471 outb(ushort port, uchar data)
0422 int allocuvm(pde_t*, uint, uint); 0472 {
0423 int deallocuvm(pde_t*, uint, uint); 0473 asm volatile("out %0,%1" : : "a" (data), "d" (port));
0424 void freevm(pde_t*); 0474 }
0425 void inituvm(pde_t*, char*, uint); 0475
0426 int loaduvm(pde_t*, char*, struct inode*, uint, uint); 0476 static inline void
0427 pde_t* copyuvm(pde_t*, uint); 0477 outw(ushort port, ushort data)
0428 void switchuvm(struct proc*); 0478 {
0429 void switchkvm(void); 0479 asm volatile("out %0,%1" : : "a" (data), "d" (port));
0430 int copyout(pde_t*, uint, void*, uint); 0480 }
0431 void clearpteu(pde_t *pgdir, char *uva); 0481
0432 0482 static inline void
0433 // number of elements in fixedsize array 0483 outsl(int port, const void *addr, int cnt)
0434 #define NELEM(x) (sizeof(x)/sizeof((x)[0])) 0484 {
0435 0485 asm volatile("cld; rep outsl" :
0436 0486 "=S" (addr), "=c" (cnt) :
0437 0487 "d" (port), "0" (addr), "1" (cnt) :
0438 0488 "cc");
0439 0489 }
0440 0490
0441 0491 static inline void
0442 0492 stosb(void *addr, int data, int cnt)
0443 0493 {
0444 0494 asm volatile("cld; rep stosb" :
0445 0495 "=D" (addr), "=c" (cnt) :
0446 0496 "0" (addr), "1" (cnt), "a" (data) :
0447 0497 "memory", "cc");
0448 0498 }
0449 0499

Sheet 04 Sheet 04
Sep 5 23:39 2011 xv6/x86.h Page 2 Sep 5 23:39 2011 xv6/x86.h Page 3

0500 static inline void 0550 static inline void

0501 stosl(void *addr, int data, int cnt) 0551 loadgs(ushort v)
0502 { 0552 {
0503 asm volatile("cld; rep stosl" : 0553 asm volatile("movw %0, %%gs" : : "r" (v));
0504 "=D" (addr), "=c" (cnt) : 0554 }
0505 "0" (addr), "1" (cnt), "a" (data) : 0555
0506 "memory", "cc"); 0556 static inline void
0507 } 0557 cli(void)
0508 0558 {
0509 struct segdesc; 0559 asm volatile("cli");
0510 0560 }
0511 static inline void 0561
0512 lgdt(struct segdesc *p, int size) 0562 static inline void
0513 { 0563 sti(void)
0514 volatile ushort pd[3]; 0564 {
0515 0565 asm volatile("sti");
0516 pd[0] = size1; 0566 }
0517 pd[1] = (uint)p; 0567
0518 pd[2] = (uint)p >> 16; 0568 static inline uint
0519 0569 xchg(volatile uint *addr, uint newval)
0520 asm volatile("lgdt (%0)" : : "r" (pd)); 0570 {
0521 } 0571 uint result;
0522 0572
0523 struct gatedesc; 0573 // The + in "+m" denotes a readmodifywrite operand.
0524 0574 asm volatile("lock; xchgl %0, %1" :
0525 static inline void 0575 "+m" (*addr), "=a" (result) :
0526 lidt(struct gatedesc *p, int size) 0576 "1" (newval) :
0527 { 0577 "cc");
0528 volatile ushort pd[3]; 0578 return result;
0529 0579 }
0530 pd[0] = size1; 0580
0531 pd[1] = (uint)p; 0581 static inline uint
0532 pd[2] = (uint)p >> 16; 0582 rcr2(void)
0533 0583 {
0534 asm volatile("lidt (%0)" : : "r" (pd)); 0584 uint val;
0535 } 0585 asm volatile("movl %%cr2,%0" : "=r" (val));
0536 0586 return val;
0537 static inline void 0587 }
0538 ltr(ushort sel) 0588
0539 { 0589 static inline void
0540 asm volatile("ltr %0" : : "r" (sel)); 0590 lcr3(uint val)
0541 } 0591 {
0542 0592 asm volatile("movl %0,%%cr3" : : "r" (val));
0543 static inline uint 0593 }
0544 readeflags(void) 0594
0545 { 0595
0546 uint eflags; 0596
0547 asm volatile("pushfl; popl %0" : "=r" (eflags)); 0597
0548 return eflags; 0598
0549 } 0599

Sheet 05 Sheet 05
Sep 5 23:39 2011 xv6/x86.h Page 4 Sep 5 23:39 2011 xv6/asm.h Page 1

0600 // Layout of the trap frame built on the stack by the 0650 //
0601 // hardware and by trapasm.S, and passed to trap(). 0651 // assembler macros to create x86 segments
0602 struct trapframe { 0652 //
0603 // registers as pushed by pusha 0653
0604 uint edi; 0654 #define SEG_NULLASM \
0605 uint esi; 0655 .word 0, 0; \
0606 uint ebp; 0656 .byte 0, 0, 0, 0
0607 uint oesp; // useless & ignored 0657
0608 uint ebx; 0658 // The 0xC0 means the limit is in 4096byte units
0609 uint edx; 0659 // and (for executable segments) 32bit mode.
0610 uint ecx; 0660 #define SEG_ASM(type,base,lim) \
0611 uint eax; 0661 .word (((lim) >> 12) & 0xffff), ((base) & 0xffff); \
0612 0662 .byte (((base) >> 16) & 0xff), (0x90 | (type)), \
0613 // rest of trap frame 0663 (0xC0 | (((lim) >> 28) & 0xf)), (((base) >> 24) & 0xff)
0614 ushort gs; 0664
0615 ushort padding1; 0665 #define STA_X 0x8 // Executable segment
0616 ushort fs; 0666 #define STA_E 0x4 // Expand down (nonexecutable segments)
0617 ushort padding2; 0667 #define STA_C 0x4 // Conforming code segment (executable only)
0618 ushort es; 0668 #define STA_W 0x2 // Writeable (nonexecutable segments)
0619 ushort padding3; 0669 #define STA_R 0x2 // Readable (executable segments)
0620 ushort ds; 0670 #define STA_A 0x1 // Accessed
0621 ushort padding4; 0671
0622 uint trapno; 0672
0623 0673
0624 // below here defined by x86 hardware 0674
0625 uint err; 0675
0626 uint eip; 0676
0627 ushort cs; 0677
0628 ushort padding5; 0678
0629 uint eflags; 0679
0630 0680
0631 // below here only when crossing rings, such as from user to kernel 0681
0632 uint esp; 0682
0633 ushort ss; 0683
0634 ushort padding6; 0684
0635 }; 0685
0636 0686
0637 0687
0638 0688
0639 0689
0640 0690
0641 0691
0642 0692
0643 0693
0644 0694
0645 0695
0646 0696
0647 0697
0648 0698
0649 0699

Sheet 06 Sheet 06
Sep 5 23:39 2011 xv6/mmu.h Page 1 Sep 5 23:39 2011 xv6/mmu.h Page 2

0700 // This file contains definitions for the 0750 #ifndef __ASSEMBLER__
0701 // x86 memory management unit (MMU). 0751 // Segment Descriptor
0702 0752 struct segdesc {
0703 // Eflags register 0753 uint lim_15_0 : 16; // Low bits of segment limit
0704 #define FL_CF 0x00000001 // Carry Flag 0754 uint base_15_0 : 16; // Low bits of segment base address
0705 #define FL_PF 0x00000004 // Parity Flag 0755 uint base_23_16 : 8; // Middle bits of segment base address
0706 #define FL_AF 0x00000010 // Auxiliary carry Flag 0756 uint type : 4; // Segment type (see STS_ constants)
0707 #define FL_ZF 0x00000040 // Zero Flag 0757 uint s : 1; // 0 = system, 1 = application
0708 #define FL_SF 0x00000080 // Sign Flag 0758 uint dpl : 2; // Descriptor Privilege Level
0709 #define FL_TF 0x00000100 // Trap Flag 0759 uint p : 1; // Present
0710 #define FL_IF 0x00000200 // Interrupt Enable 0760 uint lim_19_16 : 4; // High bits of segment limit
0711 #define FL_DF 0x00000400 // Direction Flag 0761 uint avl : 1; // Unused (available for software use)
0712 #define FL_OF 0x00000800 // Overflow Flag 0762 uint rsv1 : 1; // Reserved
0713 #define FL_IOPL_MASK 0x00003000 // I/O Privilege Level bitmask 0763 uint db : 1; // 0 = 16bit segment, 1 = 32bit segment
0714 #define FL_IOPL_0 0x00000000 // IOPL == 0 0764 uint g : 1; // Granularity: limit scaled by 4K when set
0715 #define FL_IOPL_1 0x00001000 // IOPL == 1 0765 uint base_31_24 : 8; // High bits of segment base address
0716 #define FL_IOPL_2 0x00002000 // IOPL == 2 0766 };
0717 #define FL_IOPL_3 0x00003000 // IOPL == 3 0767
0718 #define FL_NT 0x00004000 // Nested Task 0768 // Normal segment
0719 #define FL_RF 0x00010000 // Resume Flag 0769 #define SEG(type, base, lim, dpl) (struct segdesc) \
0720 #define FL_VM 0x00020000 // Virtual 8086 mode 0770 { ((lim) >> 12) & 0xffff, (uint)(base) & 0xffff, \
0721 #define FL_AC 0x00040000 // Alignment Check 0771 ((uint)(base) >> 16) & 0xff, type, 1, dpl, 1, \
0722 #define FL_VIF 0x00080000 // Virtual Interrupt Flag 0772 (uint)(lim) >> 28, 0, 0, 1, 1, (uint)(base) >> 24 }
0723 #define FL_VIP 0x00100000 // Virtual Interrupt Pending 0773 #define SEG16(type, base, lim, dpl) (struct segdesc) \
0724 #define FL_ID 0x00200000 // ID flag 0774 { (lim) & 0xffff, (uint)(base) & 0xffff, \
0725 0775 ((uint)(base) >> 16) & 0xff, type, 1, dpl, 1, \
0726 // Control Register flags 0776 (uint)(lim) >> 16, 0, 0, 1, 0, (uint)(base) >> 24 }
0727 #define CR0_PE 0x00000001 // Protection Enable 0777 #endif
0728 #define CR0_MP 0x00000002 // Monitor coProcessor 0778
0729 #define CR0_EM 0x00000004 // Emulation 0779 #define DPL_USER 0x3 // User DPL
0730 #define CR0_TS 0x00000008 // Task Switched 0780
0731 #define CR0_ET 0x00000010 // Extension Type 0781 // Application segment type bits
0732 #define CR0_NE 0x00000020 // Numeric Errror 0782 #define STA_X 0x8 // Executable segment
0733 #define CR0_WP 0x00010000 // Write Protect 0783 #define STA_E 0x4 // Expand down (nonexecutable segments)
0734 #define CR0_AM 0x00040000 // Alignment Mask 0784 #define STA_C 0x4 // Conforming code segment (executable only)
0735 #define CR0_NW 0x20000000 // Not Writethrough 0785 #define STA_W 0x2 // Writeable (nonexecutable segments)
0736 #define CR0_CD 0x40000000 // Cache Disable 0786 #define STA_R 0x2 // Readable (executable segments)
0737 #define CR0_PG 0x80000000 // Paging 0787 #define STA_A 0x1 // Accessed
0738 0788
0739 #define CR4_PSE 0x00000010 // Page size extension 0789 // System segment type bits
0740 0790 #define STS_T16A 0x1 // Available 16bit TSS
0741 #define SEG_KCODE 1 // kernel code 0791 #define STS_LDT 0x2 // Local Descriptor Table
0742 #define SEG_KDATA 2 // kernel data+stack 0792 #define STS_T16B 0x3 // Busy 16bit TSS
0743 #define SEG_KCPU 3 // kernel percpu data 0793 #define STS_CG16 0x4 // 16bit Call Gate
0744 #define SEG_UCODE 4 // user code 0794 #define STS_TG 0x5 // Task Gate / Coum Transmitions
0745 #define SEG_UDATA 5 // user data+stack 0795 #define STS_IG16 0x6 // 16bit Interrupt Gate
0746 #define SEG_TSS 6 // this processs task state 0796 #define STS_TG16 0x7 // 16bit Trap Gate
0747 0797 #define STS_T32A 0x9 // Available 32bit TSS
0748 0798 #define STS_T32B 0xB // Busy 32bit TSS
0749 0799 #define STS_CG32 0xC // 32bit Call Gate

Sheet 07 Sheet 07
Sep 5 23:39 2011 xv6/mmu.h Page 3 Sep 5 23:39 2011 xv6/mmu.h Page 4

0800 #define STS_IG32 0xE // 32bit Interrupt Gate 0850 // Task state segment format
0801 #define STS_TG32 0xF // 32bit Trap Gate 0851 struct taskstate {
0802 0852 uint link; // Old ts selector
0803 // A virtual address la has a threepart structure as follows: 0853 uint esp0; // Stack pointers and segment selectors
0804 // 0854 ushort ss0; // after an increase in privilege level
0805 // +10+10+12+ 0855 ushort padding1;
0806 // | Page Directory | Page Table | Offset within Page | 0856 uint *esp1;
0807 // | Index | Index | | 0857 ushort ss1;
0808 // ++++ 0858 ushort padding2;
0809 // \ PDX(va) / \ PTX(va) / 0859 uint *esp2;
0810 0860 ushort ss2;
0811 // page directory index 0861 ushort padding3;
0812 #define PDX(va) (((uint)(va) >> PDXSHIFT) & 0x3FF) 0862 void *cr3; // Page directory base
0813 0863 uint *eip; // Saved state from last task switch
0814 // page table index 0864 uint eflags;
0815 #define PTX(va) (((uint)(va) >> PTXSHIFT) & 0x3FF) 0865 uint eax; // More saved state (registers)
0816 0866 uint ecx;
0817 // construct virtual address from indexes and offset 0867 uint edx;
0818 #define PGADDR(d, t, o) ((uint)((d) << PDXSHIFT | (t) << PTXSHIFT | (o))) 0868 uint ebx;
0819 0869 uint *esp;
0820 // Page directory and page table constants. 0870 uint *ebp;
0821 #define NPDENTRIES 1024 // # directory entries per page directory 0871 uint esi;
0822 #define NPTENTRIES 1024 // # PTEs per page table 0872 uint edi;
0823 #define PGSIZE 4096 // bytes mapped by a page 0873 ushort es; // Even more saved state (segment selectors)
0824 0874 ushort padding4;
0825 #define PGSHIFT 12 // log2(PGSIZE) 0875 ushort cs;
0826 #define PTXSHIFT 12 // offset of PTX in a linear address 0876 ushort padding5;
0827 #define PDXSHIFT 22 // offset of PDX in a linear address 0877 ushort ss;
0828 0878 ushort padding6;
0829 #define PGROUNDUP(sz) (((sz)+PGSIZE1) & ~(PGSIZE1)) 0879 ushort ds;
0830 #define PGROUNDDOWN(a) (((a)) & ~(PGSIZE1)) 0880 ushort padding7;
0831 0881 ushort fs;
0832 // Page table/directory entry flags. 0882 ushort padding8;
0833 #define PTE_P 0x001 // Present 0883 ushort gs;
0834 #define PTE_W 0x002 // Writeable 0884 ushort padding9;
0835 #define PTE_U 0x004 // User 0885 ushort ldt;
0836 #define PTE_PWT 0x008 // WriteThrough 0886 ushort padding10;
0837 #define PTE_PCD 0x010 // CacheDisable 0887 ushort t; // Trap on task switch
0838 #define PTE_A 0x020 // Accessed 0888 ushort iomb; // I/O map base address
0839 #define PTE_D 0x040 // Dirty 0889 };
0840 #define PTE_PS 0x080 // Page Size 0890
0841 #define PTE_MBZ 0x180 // Bits must be zero 0891
0842 0892
0843 // Address in page table or page directory entry 0893
0844 #define PTE_ADDR(pte) ((uint)(pte) & ~0xFFF) 0894
0845 0895
0846 #ifndef __ASSEMBLER__ 0896
0847 typedef uint pte_t; 0897
0848 0898
0849 0899

Sheet 08 Sheet 08
Sep 5 23:39 2011 xv6/mmu.h Page 5 Sep 5 23:39 2011 xv6/elf.h Page 1

0900 // Gate descriptors for interrupts and traps 0950 // Format of an ELF executable file
0901 struct gatedesc { 0951
0902 uint off_15_0 : 16; // low 16 bits of offset in segment 0952 #define ELF_MAGIC 0x464C457FU // "\x7FELF" in little endian
0903 uint cs : 16; // code segment selector 0953
0904 uint args : 5; // # args, 0 for interrupt/trap gates 0954 // File header
0905 uint rsv1 : 3; // reserved(should be zero I guess) 0955 struct elfhdr {
0906 uint type : 4; // type(STS_{TG,IG32,TG32}) 0956 uint magic; // must equal ELF_MAGIC
0907 uint s : 1; // must be 0 (system) 0957 uchar elf[12];
0908 uint dpl : 2; // descriptor(meaning new) privilege level 0958 ushort type;
0909 uint p : 1; // Present 0959 ushort machine;
0910 uint off_31_16 : 16; // high bits of offset in segment 0960 uint version;
0911 }; 0961 uint entry;
0912 0962 uint phoff;
0913 // Set up a normal interrupt/trap gate descriptor. 0963 uint shoff;
0914 // istrap: 1 for a trap (= exception) gate, 0 for an interrupt gate. 0964 uint flags;
0915 // interrupt gate clears FL_IF, trap gate leaves FL_IF alone 0965 ushort ehsize;
0916 // sel: Code segment selector for interrupt/trap handler 0966 ushort phentsize;
0917 // off: Offset in code segment for interrupt/trap handler 0967 ushort phnum;
0918 // dpl: Descriptor Privilege Level 0968 ushort shentsize;
0919 // the privilege level required for software to invoke 0969 ushort shnum;
0920 // this interrupt/trap gate explicitly using an int instruction. 0970 ushort shstrndx;
0921 #define SETGATE(gate, istrap, sel, off, d) \ 0971 };
0922 { \ 0972
0923 (gate).off_15_0 = (uint)(off) & 0xffff; \ 0973 // Program section header
0924 (gate).cs = (sel); \ 0974 struct proghdr {
0925 (gate).args = 0; \ 0975 uint type;
0926 (gate).rsv1 = 0; \ 0976 uint off;
0927 (gate).type = (istrap) ? STS_TG32 : STS_IG32; \ 0977 uint vaddr;
0928 (gate).s = 0; \ 0978 uint paddr;
0929 (gate).dpl = (d); \ 0979 uint filesz;
0930 (gate).p = 1; \ 0980 uint memsz;
0931 (gate).off_31_16 = (uint)(off) >> 16; \ 0981 uint flags;
0932 } 0982 uint align;
0933 0983 };
0934 #endif 0984
0935 0985 // Values for Proghdr type
0936 0986 #define ELF_PROG_LOAD 1
0937 0987
0938 0988 // Flag bits for Proghdr flags
0939 0989 #define ELF_PROG_FLAG_EXEC 1
0940 0990 #define ELF_PROG_FLAG_WRITE 2
0941 0991 #define ELF_PROG_FLAG_READ 4
0942 0992
0943 0993
0944 0994
0945 0995
0946 0996
0947 0997
0948 0998
0949 0999

Sheet 09 Sheet 09
Sep 5 23:39 2011 xv6/entry.S Page 1 Sep 5 23:39 2011 xv6/entry.S Page 2

1000 # Multiboot header, for multiboot boot loaders like GNU Grub. 1050 orl $(CR0_PG|CR0_WP), %eax
1001 # http://www.gnu.org/software/grub/manual/multiboot/multiboot.html 1051 movl %eax, %cr0
1002 # 1052
1003 # Using GRUB 2, you can boot xv6 from a file stored in a 1053 # Set up the stack pointer.
1004 # Linux file system by copying kernel or kernelmemfs to /boot 1054 movl $(stack + KSTACKSIZE), %esp
1005 # and then adding this menu entry: 1055
1006 # 1056 # Jump to main(), and switch to executing at
1007 # menuentry "xv6" { 1057 # high addresses. The indirect call is needed because
1008 # insmod ext2 1058 # the assembler produces a PCrelative instruction
1009 # set root=(hd0,msdos1) 1059 # for a direct jump.
1010 # set kernel=/boot/kernel 1060 mov $main, %eax
1011 # echo "Loading ${kernel}..." 1061 jmp *%eax
1012 # multiboot ${kernel} ${kernel} 1062
1013 # boot 1063 .comm stack, KSTACKSIZE
1014 # } 1064
1015 1065
1016 #include "asm.h" 1066
1017 #include "memlayout.h" 1067
1018 #include "mmu.h" 1068
1019 #include "param.h" 1069
1020 1070
1021 # Multiboot header. Data to direct multiboot loader. 1071
1022 .p2align 2 1072
1023 .text 1073
1024 .globl multiboot_header 1074
1025 multiboot_header: 1075
1026 #define magic 0x1badb002 1076
1027 #define flags 0 1077
1028 .long magic 1078
1029 .long flags 1079
1030 .long (magicflags) 1080
1031 1081
1032 # By convention, the _start symbol specifies the ELF entry point. 1082
1033 # Since we havent set up virtual memory yet, our entry point is 1083
1034 # the physical address of entry. 1084
1035 .globl _start 1085
1036 _start = V2P_WO(entry) 1086
1037 1087
1038 # Entering xv6 on boot processor. Machine is mostly set up. 1088
1039 .globl entry 1089
1040 entry: 1090
1041 # Turn on page size extension for 4Mbyte pages 1091
1042 movl %cr4, %eax 1092
1043 orl $(CR4_PSE), %eax 1093
1044 movl %eax, %cr4 1094
1045 # Set page directory 1095
1046 movl $(V2P_WO(entrypgdir)), %eax 1096
1047 movl %eax, %cr3 1097
1048 # Turn on paging. 1098
1049 movl %cr0, %eax 1099

Sheet 10 Sheet 10
Sep 5 23:39 2011 xv6/entryother.S Page 1 Sep 5 23:39 2011 xv6/entryother.S Page 2

1100 #include "asm.h" 1150 ljmpl $(SEG_KCODE<<3), $(start32)

1101 #include "memlayout.h" 1151
1102 #include "mmu.h" 1152 .code32
1103 1153 start32:
1104 # Each nonboot CPU ("AP") is started up in response to a STARTUP 1154 movw $(SEG_KDATA<<3), %ax
1105 # IPI from the boot CPU. Section B.4.2 of the MultiProcessor 1155 movw %ax, %ds
1106 # Specification says that the AP will start in real mode with CS:IP 1156 movw %ax, %es
1107 # set to XY00:0000, where XY is an 8bit value sent with the 1157 movw %ax, %ss
1108 # STARTUP. Thus this code must start at a 4096byte boundary. 1158 movw $0, %ax
1109 # 1159 movw %ax, %fs
1110 # Because this code sets DS to zero, it must sit 1160 movw %ax, %gs
1111 # at an address in the low 2^16 bytes. 1161
1112 # 1162 # Turn on page size extension for 4Mbyte pages
1113 # Startothers (in main.c) sends the STARTUPs one at a time. 1163 movl %cr4, %eax
1114 # It copies this code (start) at 0x7000. It puts the address of 1164 orl $(CR4_PSE), %eax
1115 # a newly allocated percore stack in start4,the address of the 1165 movl %eax, %cr4
1116 # place to jump to (mpenter) in start8, and the physical address 1166 # Use enterpgdir as our initial page table
1117 # of entrypgdir in start12. 1167 movl (start12), %eax
1118 # 1168 movl %eax, %cr3
1119 # This code is identical to bootasm.S except: 1169 # Turn on paging.
1120 # it does not need to enable A20 1170 movl %cr0, %eax
1121 # it uses the address at start4, start8, and start12 1171 orl $(CR0_PE|CR0_PG|CR0_WP), %eax
1122 1172 movl %eax, %cr0
1123 .code16 1173
1124 .globl start 1174 # Switch to the stack allocated by startothers()
1125 start: 1175 movl (start4), %esp
1126 cli 1176 # Call mpenter()
1127 1177 call *(start8)
1128 xorw %ax,%ax 1178
1129 movw %ax,%ds 1179 movw $0x8a00, %ax
1130 movw %ax,%es 1180 movw %ax, %dx
1131 movw %ax,%ss 1181 outw %ax, %dx
1132 1182 movw $0x8ae0, %ax
1133 lgdt gdtdesc 1183 outw %ax, %dx
1134 movl %cr0, %eax 1184 spin:
1135 orl $CR0_PE, %eax 1185 jmp spin
1136 movl %eax, %cr0 1186
1137 1187 .p2align 2
1138 1188 gdt:
1139 1189 SEG_NULLASM
1140 1190 SEG_ASM(STA_X|STA_R, 0, 0xffffffff)
1141 1191 SEG_ASM(STA_W, 0, 0xffffffff)
1142 1192
1143 1193
1144 1194 gdtdesc:
1145 1195 .word (gdtdesc gdt 1)
1146 1196 .long gdt
1147 1197
1148 1198
1149 1199

Sheet 11 Sheet 11
Sep 5 23:39 2011 xv6/main.c Page 1 Sep 5 23:39 2011 xv6/main.c Page 2

1200 #include "types.h" 1250 // Other CPUs jump here from entryother.S.
1201 #include "defs.h" 1251 static void
1202 #include "param.h" 1252 mpenter(void)
1203 #include "memlayout.h" 1253 {
1204 #include "mmu.h" 1254 switchkvm();
1205 #include "proc.h" 1255 seginit();
1206 #include "x86.h" 1256 lapicinit(cpunum());
1207 1257 mpmain();
1208 static void startothers(void); 1258 }
1209 static void mpmain(void) __attribute__((noreturn)); 1259
1210 extern pde_t *kpgdir; 1260 // Common CPU setup code.
1211 1261 static void
1212 // Bootstrap processor starts running C code here. 1262 mpmain(void)
1213 // Allocate a real stack and switch to it, first 1263 {
1214 // doing some setup required for memory allocator to work. 1264 cprintf("cpu%d: starting\n", cpu>id);
1215 int 1265 idtinit(); // load idt register
1216 main(void) 1266 xchg(&cpu>started, 1); // tell startothers() were up
1217 { 1267 scheduler(); // start running processes
1218 kvmalloc(); // kernel page table 1268 }
1219 mpinit(); // collect info about this machine 1269
1220 lapicinit(mpbcpu()); 1270 pde_t entrypgdir[]; // For entry.S
1221 seginit(); // set up segments 1271
1222 cprintf("\ncpu%d: starting xv6\n\n", cpu>id); 1272 // Start the nonboot (AP) processors.
1223 picinit(); // interrupt controller 1273 static void
1224 ioapicinit(); // another interrupt controller 1274 startothers(void)
1225 consoleinit(); // I/O devices & their interrupts 1275 {
1226 uartinit(); // serial port 1276 extern uchar _binary_entryother_start[], _binary_entryother_size[];
1227 pinit(); // process table 1277 uchar *code;
1228 tvinit(); // trap vectors 1278 struct cpu *c;
1229 binit(); // buffer cache 1279 char *stack;
1230 fileinit(); // file table 1280
1231 iinit(); // inode cache 1281 // Write entry code to unused memory at 0x7000.
1232 ideinit(); // disk 1282 // The linker has placed the image of entryother.S in
1233 if(!ismp) 1283 // _binary_entryother_start.
1234 timerinit(); // uniprocessor timer 1284 code = p2v(0x7000);
1235 startothers(); // start other processors (must come before kinit) 1285 memmove(code, _binary_entryother_start, (uint)_binary_entryother_size);
1236 kinit(); // initialize memory allocator 1286
1237 userinit(); // first user process (must come after kinit) 1287 for(c = cpus; c < cpus+ncpu; c++){
1238 // Finish setting up this processor in mpmain. 1288 if(c == cpus+cpunum()) // Weve started already.
1239 mpmain(); 1289 continue;
1240 } 1290
1241 1291 // Tell entryother.S what stack to use, where to enter, and what
1242 1292 // pgdir to use. We cannot use kpgdir yet, because the AP processor
1243 1293 // is running in low memory, so we use entrypgdir for the APs too.
1244 1294 // kalloc can return addresses above 4Mbyte (the machine may have
1245 1295 // much more physical memory than 4Mbyte), which arent mapped by
1246 1296 // entrypgdir, so we must allocate a stack using enter_alloc();
1247 1297 // this introduces the constraint that xv6 cannot use kalloc until
1248 1298 // after these last enter_alloc invocations.
1249 1299 stack = enter_alloc();

Sheet 12 Sheet 12
Sep 5 23:39 2011 xv6/main.c Page 3 Sep 5 23:39 2011 xv6/main.c Page 4

1300 *(void**)(code4) = stack + KSTACKSIZE; 1350 // Blank page.

1301 *(void**)(code8) = mpenter; 1351
1302 *(int**)(code12) = (void *) v2p(entrypgdir); 1352
1303 1353
1304 lapicstartap(c>id, v2p(code)); 1354
1305 1355
1306 // wait for cpu to finish mpmain() 1356
1307 while(c>started == 0) 1357
1308 ; 1358
1309 } 1359
1310 } 1360
1311 1361
1312 // Boot page table used in entry.S and entryother.S. 1362
1313 // Page directories (and page tables), must start on a page boundary, 1363
1314 // hence the "__aligned__" attribute. 1364
1315 // Use PTE_PS in page directory entry to enable 4Mbyte pages. 1365
1316 __attribute__((__aligned__(PGSIZE))) 1366
1317 pde_t entrypgdir[NPDENTRIES] = { 1367
1318 // Map VAs [0, 4MB) to PAs [0, 4MB) 1368
1319 [0] = (0) + PTE_P + PTE_W + PTE_PS, 1369
1320 // Map VAs [KERNBASE, KERNBASE+4MB) to PAs [0, 4MB) 1370
1321 [KERNBASE>>PDXSHIFT] = (0) + PTE_P + PTE_W + PTE_PS, 1371
1322 }; 1372
1323 1373
1324 1374
1325 1375
1326 1376
1327 1377
1328 1378
1329 1379
1330 1380
1331 1381
1332 1382
1333 1383
1334 1384
1335 1385
1336 1386
1337 1387
1338 1388
1339 1389
1340 1390
1341 1391
1342 1392
1343 1393
1344 1394
1345 1395
1346 1396
1347 1397
1348 1398
1349 1399

Sheet 13 Sheet 13
Sep 5 23:39 2011 xv6/spinlock.h Page 1 Sep 5 23:39 2011 xv6/spinlock.c Page 1

1400 // Mutual exclusion lock. 1450 // Mutual exclusion spin locks.

1401 struct spinlock { 1451
1402 uint locked; // Is the lock held? 1452 #include "types.h"
1403 1453 #include "defs.h"
1404 // For debugging: 1454 #include "param.h"
1405 char *name; // Name of lock. 1455 #include "x86.h"
1406 struct cpu *cpu; // The cpu holding the lock. 1456 #include "memlayout.h"
1407 uint pcs[10]; // The call stack (an array of program counters) 1457 #include "mmu.h"
1408 // that locked the lock. 1458 #include "proc.h"
1409 }; 1459 #include "spinlock.h"
1410 1460
1411 1461 void
1412 1462 initlock(struct spinlock *lk, char *name)
1413 1463 {
1414 1464 lk>name = name;
1415 1465 lk>locked = 0;
1416 1466 lk>cpu = 0;
1417 1467 }
1418 1468
1419 1469 // Acquire the lock.
1420 1470 // Loops (spins) until the lock is acquired.
1421 1471 // Holding a lock for a long time may cause
1422 1472 // other CPUs to waste time spinning to acquire it.
1423 1473 void
1424 1474 acquire(struct spinlock *lk)
1425 1475 {
1426 1476 pushcli(); // disable interrupts to avoid deadlock.
1427 1477 if(holding(lk))
1428 1478 panic("acquire");
1429 1479
1430 1480 // The xchg is atomic.
1431 1481 // It also serializes, so that reads after acquire are not
1432 1482 // reordered before it.
1433 1483 while(xchg(&lk>locked, 1) != 0)
1434 1484 ;
1435 1485
1436 1486 // Record info about lock acquisition for debugging.
1437 1487 lk>cpu = cpu;
1438 1488 getcallerpcs(&lk, lk>pcs);
1439 1489 }
1440 1490
1441 1491
1442 1492
1443 1493
1444 1494
1445 1495
1446 1496
1447 1497
1448 1498
1449 1499

Sheet 14 Sheet 14
Sep 5 23:39 2011 xv6/spinlock.c Page 2 Sep 5 23:39 2011 xv6/spinlock.c Page 3

1500 // Release the lock. 1550 // Pushcli/popcli are like cli/sti except that they are matched:
1501 void 1551 // it takes two popcli to undo two pushcli. Also, if interrupts
1502 release(struct spinlock *lk) 1552 // are off, then pushcli, popcli leaves them off.
1503 { 1553
1504 if(!holding(lk)) 1554 void
1505 panic("release"); 1555 pushcli(void)
1506 1556 {
1507 lk>pcs[0] = 0; 1557 int eflags;
1508 lk>cpu = 0; 1558
1509 1559 eflags = readeflags();
1510 // The xchg serializes, so that reads before release are 1560 cli();
1511 // not reordered after it. The 1996 PentiumPro manual (Volume 3, 1561 if(cpu>ncli++ == 0)
1512 // 7.2) says reads can be carried out speculatively and in 1562 cpu>intena = eflags & FL_IF;
1513 // any order, which implies we need to serialize here. 1563 }
1514 // But the 2007 Intel 64 Architecture Memory Ordering White 1564
1515 // Paper says that Intel 64 and IA32 will not move a load 1565 void
1516 // after a store. So lock>locked = 0 would work here. 1566 popcli(void)
1517 // The xchg being asm volatile ensures gcc emits it after 1567 {
1518 // the above assignments (and after the critical section). 1568 if(readeflags()&FL_IF)
1519 xchg(&lk>locked, 0); 1569 panic("popcli interruptible");
1520 1570 if(cpu>ncli < 0)
1521 popcli(); 1571 panic("popcli");
1522 } 1572 if(cpu>ncli == 0 && cpu>intena)
1523 1573 sti();
1524 // Record the current call stack in pcs[] by following the %ebp chain. 1574 }
1525 void 1575
1526 getcallerpcs(void *v, uint pcs[]) 1576
1527 { 1577
1528 uint *ebp; 1578
1529 int i; 1579
1530 1580
1531 ebp = (uint*)v 2; 1581
1532 for(i = 0; i < 10; i++){ 1582
1533 if(ebp == 0 || ebp < (uint*)KERNBASE || ebp == (uint*)0xffffffff) 1583
1534 break; 1584
1535 pcs[i] = ebp[1]; // saved %eip 1585
1536 ebp = (uint*)ebp[0]; // saved %ebp 1586
1537 } 1587
1538 for(; i < 10; i++) 1588
1539 pcs[i] = 0; 1589
1540 } 1590
1541 1591
1542 // Check whether this cpu is holding the lock. 1592
1543 int 1593
1544 holding(struct spinlock *lock) 1594
1545 { 1595
1546 return lock>locked && lock>cpu == cpu; 1596
1547 } 1597
1548 1598
1549 1599

Sheet 15 Sheet 15
Sep 5 23:39 2011 xv6/vm.c Page 1 Sep 5 23:39 2011 xv6/vm.c Page 2

1600 #include "param.h" 1650 // Return the address of the PTE in page table pgdir
1601 #include "types.h" 1651 // that corresponds to virtual address va. If alloc!=0,
1602 #include "defs.h" 1652 // create any required page table pages.
1603 #include "x86.h" 1653 static pte_t *
1604 #include "memlayout.h" 1654 walkpgdir(pde_t *pgdir, const void *va, char* (*alloc)(void))
1605 #include "mmu.h" 1655 {
1606 #include "proc.h" 1656 pde_t *pde;
1607 #include "elf.h" 1657 pte_t *pgtab;
1608 1658
1609 extern char data[]; // defined by kernel.ld 1659 pde = &pgdir[PDX(va)];
1610 pde_t *kpgdir; // for use in scheduler() 1660 if(*pde & PTE_P){
1611 struct segdesc gdt[NSEGS]; 1661 pgtab = (pte_t*)p2v(PTE_ADDR(*pde));
1612 1662 } else {
1613 // Set up CPUs kernel segment descriptors. 1663 if(!alloc || (pgtab = (pte_t*)alloc()) == 0)
1614 // Run once on entry on each CPU. 1664 return 0;
1615 void 1665 // Make sure all those PTE_P bits are zero.
1616 seginit(void) 1666 memset(pgtab, 0, PGSIZE);
1617 { 1667 // The permissions here are overly generous, but they can
1618 struct cpu *c; 1668 // be further restricted by the permissions in the page table
1619 1669 // entries, if necessary.
1620 // Map "logical" addresses to virtual addresses using identity map. 1670 *pde = v2p(pgtab) | PTE_P | PTE_W | PTE_U;
1621 // Cannot share a CODE descriptor for both kernel and user 1671 }
1622 // because it would have to have DPL_USR, but the CPU forbids 1672 return &pgtab[PTX(va)];
1623 // an interrupt from CPL=0 to DPL=3. 1673 }
1624 c = &cpus[cpunum()]; 1674
1625 c>gdt[SEG_KCODE] = SEG(STA_X|STA_R, 0, 0xffffffff, 0); 1675 // Create PTEs for virtual addresses starting at va that refer to
1626 c>gdt[SEG_KDATA] = SEG(STA_W, 0, 0xffffffff, 0); 1676 // physical addresses starting at pa. va and size might not
1627 c>gdt[SEG_UCODE] = SEG(STA_X|STA_R, 0, 0xffffffff, DPL_USER); 1677 // be pagealigned.
1628 c>gdt[SEG_UDATA] = SEG(STA_W, 0, 0xffffffff, DPL_USER); 1678 static int
1629 1679 mappages(pde_t *pgdir, void *va, uint size, uint pa,
1630 // Map cpu, and curproc 1680 int perm, char* (*alloc)(void))
1631 c>gdt[SEG_KCPU] = SEG(STA_W, &c>cpu, 8, 0); 1681 {
1632 1682 char *a, *last;
1633 lgdt(c>gdt, sizeof(c>gdt)); 1683 pte_t *pte;
1634 loadgs(SEG_KCPU << 3); 1684
1635 1685 a = (char*)PGROUNDDOWN((uint)va);
1636 // Initialize cpulocal storage. 1686 last = (char*)PGROUNDDOWN(((uint)va) + size 1);
1637 cpu = c; 1687 for(;;){
1638 proc = 0; 1688 if((pte = walkpgdir(pgdir, a, alloc)) == 0)
1639 } 1689 return 1;
1640 1690 if(*pte & PTE_P)
1641 1691 panic("remap");
1642 1692 *pte = pa | perm | PTE_P;
1643 1693 if(a == last)
1644 1694 break;
1645 1695 a += PGSIZE;
1646 1696 pa += PGSIZE;
1647 1697 }
1648 1698 return 0;
1649 1699 }

Sheet 16 Sheet 16
Sep 5 23:39 2011 xv6/vm.c Page 3 Sep 5 23:39 2011 xv6/vm.c Page 4

1700 // The mappings from logical to virtual are one to one (i.e., 1750 // Allocate one page table for the machine for the kernel address
1701 // segmentation doesnt do anything). There is one page table per 1751 // space for scheduler processes.
1702 // process, plus one thats used when a CPU is not running any process 1752 void
1703 // (kpgdir). A user process uses the same page table as the kernel; the 1753 kvmalloc(void)
1704 // page protection bits prevent it from accessing kernel memory. 1754 {
1705 // 1755 kpgdir = setupkvm(enter_alloc);
1706 // setupkvm() and exec() set up every page table like this: 1756 switchkvm();
1707 // 0..KERNBASE: user memory (text+data+stack+heap), mapped to some free 1757 }
1708 // phys memory 1758
1709 // KERNBASE..KERNBASE+EXTMEM: mapped to 0..EXTMEM (for I/O space) 1759 // Switch h/w page table register to the kernelonly page table,
1710 // KERNBASE+EXTMEM..KERNBASE+end: mapped to EXTMEM..end kernel, 1760 // for when no process is running.
1711 // w. no write permission 1761 void
1712 // KERNBASE+end..KERBASE+PHYSTOP: mapped to end..PHYSTOP, 1762 switchkvm(void)
1713 // rw data + free memory 1763 {
1714 // 0xfe000000..0: mapped direct (devices such as ioapic) 1764 lcr3(v2p(kpgdir)); // switch to the kernel page table
1715 // 1765 }
1716 // The kernel allocates memory for its heap and for user memory 1766
1717 // between KERNBASE+end and the end of physical memory (PHYSTOP). 1767 // Switch TSS and h/w page table to correspond to process p.
1718 // The user program sits in the bottom of the address space, and the 1768 void
1719 // kernel at the top at KERNBASE. 1769 switchuvm(struct proc *p)
1720 static struct kmap { 1770 {
1721 void *virt; 1771 pushcli();
1722 uint phys_start; 1772 cpu>gdt[SEG_TSS] = SEG16(STS_T32A, &cpu>ts, sizeof(cpu>ts)1, 0);
1723 uint phys_end; 1773 cpu>gdt[SEG_TSS].s = 0;
1724 int perm; 1774 cpu>ts.ss0 = SEG_KDATA << 3;
1725 } kmap[] = { 1775 cpu>ts.esp0 = (uint)proc>kstack + KSTACKSIZE;
1726 { P2V(0), 0, 1024*1024, PTE_W}, // I/O space 1776 ltr(SEG_TSS << 3);
1727 { (void*)KERNLINK, V2P(KERNLINK), V2P(data), 0}, // kernel text+rodata 1777 if(p>pgdir == 0)
1728 { data, V2P(data), PHYSTOP, PTE_W}, // kernel data, memory 1778 panic("switchuvm: no pgdir");
1729 { (void*)DEVSPACE, DEVSPACE, 0, PTE_W}, // more devices 1779 lcr3(v2p(p>pgdir)); // switch to new address space
1730 }; 1780 popcli();
1731 1781 }
1732 // Set up kernel part of a page table. 1782
1733 pde_t* 1783 // Load the initcode into address 0 of pgdir.
1734 setupkvm(char* (*alloc)(void)) 1784 // sz must be less than a page.
1735 { 1785 void
1736 pde_t *pgdir; 1786 inituvm(pde_t *pgdir, char *init, uint sz)
1737 struct kmap *k; 1787 {
1738 1788 char *mem;
1739 if((pgdir = (pde_t*)alloc()) == 0) 1789
1740 return 0; 1790 if(sz >= PGSIZE)
1741 memset(pgdir, 0, PGSIZE); 1791 panic("inituvm: more than a page");
1742 if (p2v(PHYSTOP) > (void*)DEVSPACE) 1792 mem = kalloc();
1743 panic("PHYSTOP too high"); 1793 memset(mem, 0, PGSIZE);
1744 for(k = kmap; k < &kmap[NELEM(kmap)]; k++) 1794 mappages(pgdir, 0, PGSIZE, v2p(mem), PTE_W|PTE_U, kalloc);
1745 if(mappages(pgdir, k>virt, k>phys_end k>phys_start, 1795 memmove(mem, init, sz);
1746 (uint)k>phys_start, k>perm, alloc) < 0) 1796 }
1747 return 0; 1797
1748 return pgdir; 1798
1749 } 1799

Sheet 17 Sheet 17
Sep 5 23:39 2011 xv6/vm.c Page 5 Sep 5 23:39 2011 xv6/vm.c Page 6

1800 // Load a program segment into pgdir. addr must be pagealigned 1850 // Deallocate user pages to bring the process size from oldsz to
1801 // and the pages from addr to addr+sz must already be mapped. 1851 // newsz. oldsz and newsz need not be pagealigned, nor does newsz
1802 int 1852 // need to be less than oldsz. oldsz can be larger than the actual
1803 loaduvm(pde_t *pgdir, char *addr, struct inode *ip, uint offset, uint sz) 1853 // process size. Returns the new process size.
1804 { 1854 int
1805 uint i, pa, n; 1855 deallocuvm(pde_t *pgdir, uint oldsz, uint newsz)
1806 pte_t *pte; 1856 {
1807 1857 pte_t *pte;
1808 if((uint) addr % PGSIZE != 0) 1858 uint a, pa;
1809 panic("loaduvm: addr must be page aligned"); 1859
1810 for(i = 0; i < sz; i += PGSIZE){ 1860 if(newsz >= oldsz)
1811 if((pte = walkpgdir(pgdir, addr+i, 0)) == 0) 1861 return oldsz;
1812 panic("loaduvm: address should exist"); 1862
1813 pa = PTE_ADDR(*pte); 1863 a = PGROUNDUP(newsz);
1814 if(sz i < PGSIZE) 1864 for(; a < oldsz; a += PGSIZE){
1815 n = sz i; 1865 pte = walkpgdir(pgdir, (char*)a, 0);
1816 else 1866 if(!pte)
1817 n = PGSIZE; 1867 a += (NPTENTRIES 1) * PGSIZE;
1818 if(readi(ip, p2v(pa), offset+i, n) != n) 1868 else if((*pte & PTE_P) != 0){
1819 return 1; 1869 pa = PTE_ADDR(*pte);
1820 } 1870 if(pa == 0)
1821 return 0; 1871 panic("kfree");
1822 } 1872 char *v = p2v(pa);
1823 1873 kfree(v);
1824 // Allocate page tables and physical memory to grow process from oldsz to 1874 *pte = 0;
1825 // newsz, which need not be page aligned. Returns new size or 0 on error. 1875 }
1826 int 1876 }
1827 allocuvm(pde_t *pgdir, uint oldsz, uint newsz) 1877 return newsz;
1828 { 1878 }
1829 char *mem; 1879
1830 uint a; 1880 // Free a page table and all the physical memory pages
1831 1881 // in the user part.
1832 if(newsz >= KERNBASE) 1882 void
1833 return 0; 1883 freevm(pde_t *pgdir)
1834 if(newsz < oldsz) 1884 {
1835 return oldsz; 1885 uint i;
1836 1886
1837 a = PGROUNDUP(oldsz); 1887 if(pgdir == 0)
1838 for(; a < newsz; a += PGSIZE){ 1888 panic("freevm: no pgdir");
1839 mem = kalloc(); 1889 deallocuvm(pgdir, KERNBASE, 0);
1840 if(mem == 0){ 1890 for(i = 0; i < NPDENTRIES; i++){
1841 cprintf("allocuvm out of memory\n"); 1891 if(pgdir[i] & PTE_P){
1842 deallocuvm(pgdir, newsz, oldsz); 1892 char * v = p2v(PTE_ADDR(pgdir[i]));
1843 return 0; 1893 kfree(v);
1844 } 1894 }
1845 memset(mem, 0, PGSIZE); 1895 }
1846 mappages(pgdir, (char*)a, PGSIZE, v2p(mem), PTE_W|PTE_U, kalloc); 1896 kfree((char*)pgdir);
1847 } 1897 }
1848 return newsz; 1898
1849 } 1899

Sheet 18 Sheet 18
Sep 5 23:39 2011 xv6/vm.c Page 7 Sep 5 23:39 2011 xv6/vm.c Page 8

1900 // Clear PTE_U on a page. Used to create an inaccessible 1950 // Map user virtual address to kernel address.
1901 // page beneath the user stack. 1951 char*
1902 void 1952 uva2ka(pde_t *pgdir, char *uva)
1903 clearpteu(pde_t *pgdir, char *uva) 1953 {
1904 { 1954 pte_t *pte;
1905 pte_t *pte; 1955
1906 1956 pte = walkpgdir(pgdir, uva, 0);
1907 pte = walkpgdir(pgdir, uva, 0); 1957 if((*pte & PTE_P) == 0)
1908 if(pte == 0) 1958 return 0;
1909 panic("clearpteu"); 1959 if((*pte & PTE_U) == 0)
1910 *pte &= ~PTE_U; 1960 return 0;
1911 } 1961 return (char*)p2v(PTE_ADDR(*pte));
1912 1962 }
1913 // Given a parent processs page table, create a copy 1963
1914 // of it for a child. 1964 // Copy len bytes from p to user address va in page table pgdir.
1915 pde_t* 1965 // Most useful when pgdir is not the current page table.
1916 copyuvm(pde_t *pgdir, uint sz) 1966 // uva2ka ensures this only works for PTE_U pages.
1917 { 1967 int
1918 pde_t *d; 1968 copyout(pde_t *pgdir, uint va, void *p, uint len)
1919 pte_t *pte; 1969 {
1920 uint pa, i; 1970 char *buf, *pa0;
1921 char *mem; 1971 uint n, va0;
1922 1972
1923 if((d = setupkvm(kalloc)) == 0) 1973 buf = (char*)p;
1924 return 0; 1974 while(len > 0){
1925 for(i = 0; i < sz; i += PGSIZE){ 1975 va0 = (uint)PGROUNDDOWN(va);
1926 if((pte = walkpgdir(pgdir, (void *) i, 0)) == 0) 1976 pa0 = uva2ka(pgdir, (char*)va0);
1927 panic("copyuvm: pte should exist"); 1977 if(pa0 == 0)
1928 if(!(*pte & PTE_P)) 1978 return 1;
1929 panic("copyuvm: page not present"); 1979 n = PGSIZE (va va0);
1930 pa = PTE_ADDR(*pte); 1980 if(n > len)
1931 if((mem = kalloc()) == 0) 1981 n = len;
1932 goto bad; 1982 memmove(pa0 + (va va0), buf, n);
1933 memmove(mem, (char*)p2v(pa), PGSIZE); 1983 len = n;
1934 if(mappages(d, (void*)i, PGSIZE, v2p(mem), PTE_W|PTE_U, kalloc) < 0) 1984 buf += n;
1935 goto bad; 1985 va = va0 + PGSIZE;
1936 } 1986 }
1937 return d; 1987 return 0;
1938 1988 }
1939 bad: 1989
1940 freevm(d); 1990
1941 return 0; 1991
1942 } 1992
1943 1993
1944 1994
1945 1995
1946 1996
1947 1997
1948 1998
1949 1999

Sheet 19 Sheet 19
Sep 5 23:39 2011 xv6/proc.h Page 1 Sep 5 23:39 2011 xv6/proc.h Page 2

2000 // Segments in proc>gdt. 2050 enum procstate { UNUSED, EMBRYO, SLEEPING, RUNNABLE, RUNNING, ZOMBIE };
2001 #define NSEGS 7 2051
2002 2052 // Perprocess state
2003 // PerCPU state 2053 struct proc {
2004 struct cpu { 2054 uint sz; // Size of process memory (bytes)
2005 uchar id; // Local APIC ID; index into cpus[] below 2055 pde_t* pgdir; // Page table
2006 struct context *scheduler; // swtch() here to enter scheduler 2056 char *kstack; // Bottom of kernel stack for this process
2007 struct taskstate ts; // Used by x86 to find stack for interrupt 2057 enum procstate state; // Process state
2008 struct segdesc gdt[NSEGS]; // x86 global descriptor table 2058 volatile int pid; // Process ID
2009 volatile uint started; // Has the CPU started? 2059 struct proc *parent; // Parent process
2010 int ncli; // Depth of pushcli nesting. 2060 struct trapframe *tf; // Trap frame for current syscall
2011 int intena; // Were interrupts enabled before pushcli? 2061 struct context *context; // swtch() here to run process
2012 2062 void *chan; // If nonzero, sleeping on chan
2013 // Cpulocal storage variables; see below 2063 int killed; // If nonzero, have been killed
2014 struct cpu *cpu; 2064 struct file *ofile[NOFILE]; // Open files
2015 struct proc *proc; // The currentlyrunning process. 2065 struct inode *cwd; // Current directory
2016 }; 2066 char name[16]; // Process name (debugging)
2017 2067 };
2018 extern struct cpu cpus[NCPU]; 2068
2019 extern int ncpu; 2069 // Process memory is laid out contiguously, low addresses first:
2020 2070 // text
2021 // PerCPU variables, holding pointers to the 2071 // original data and bss
2022 // current cpu and to the current process. 2072 // fixedsize stack
2023 // The asm suffix tells gcc to use "%gs:0" to refer to cpu 2073 // expandable heap
2024 // and "%gs:4" to refer to proc. seginit sets up the 2074
2025 // %gs segment register so that %gs refers to the memory 2075
2026 // holding those two variables in the local cpus struct cpu. 2076
2027 // This is similar to how threadlocal variables are implemented 2077
2028 // in thread libraries such as Linux pthreads. 2078
2029 extern struct cpu *cpu asm("%gs:0"); // &cpus[cpunum()] 2079
2030 extern struct proc *proc asm("%gs:4"); // cpus[cpunum()].proc 2080
2031 2081
2032 2082
2033 // Saved registers for kernel context switches. 2083
2034 // Dont need to save all the segment registers (%cs, etc), 2084
2035 // because they are constant across kernel contexts. 2085
2036 // Dont need to save %eax, %ecx, %edx, because the 2086
2037 // x86 convention is that the caller has saved them. 2087
2038 // Contexts are stored at the bottom of the stack they 2088
2039 // describe; the stack pointer is the address of the context. 2089
2040 // The layout of the context matches the layout of the stack in swtch.S 2090
2041 // at the "Switch stacks" comment. Switch doesnt save eip explicitly, 2091
2042 // but it is on the stack and allocproc() manipulates it. 2092
2043 struct context { 2093
2044 uint edi; 2094
2045 uint esi; 2095
2046 uint ebx; 2096
2047 uint ebp; 2097
2048 uint eip; 2098
2049 }; 2099

Sheet 20 Sheet 20
Sep 5 23:39 2011 xv6/proc.c Page 1 Sep 5 23:39 2011 xv6/proc.c Page 2

2100 #include "types.h" 2150 // Look in the process table for an UNUSED proc.
2101 #include "defs.h" 2151 // If found, change state to EMBRYO and initialize
2102 #include "param.h" 2152 // state required to run in the kernel.
2103 #include "memlayout.h" 2153 // Otherwise return 0.
2104 #include "mmu.h" 2154 static struct proc*
2105 #include "x86.h" 2155 allocproc(void)
2106 #include "proc.h" 2156 {
2107 #include "spinlock.h" 2157 struct proc *p;
2108 2158 char *sp;
2109 struct { 2159
2110 struct spinlock lock; 2160 acquire(&ptable.lock);
2111 struct proc proc[NPROC]; 2161 for(p = ptable.proc; p < &ptable.proc[NPROC]; p++)
2112 } ptable; 2162 if(p>state == UNUSED)
2113 2163 goto found;
2114 static struct proc *initproc; 2164 release(&ptable.lock);
2115 2165 return 0;
2116 int nextpid = 1; 2166
2117 extern void forkret(void); 2167 found:
2118 extern void trapret(void); 2168 p>state = EMBRYO;
2119 2169 p>pid = nextpid++;
2120 static void wakeup1(void *chan); 2170 release(&ptable.lock);
2121 2171
2122 void 2172 // Allocate kernel stack.
2123 pinit(void) 2173 if((p>kstack = kalloc()) == 0){
2124 { 2174 p>state = UNUSED;
2125 initlock(&ptable.lock, "ptable"); 2175 return 0;
2126 } 2176 }
2127 2177 sp = p>kstack + KSTACKSIZE;
2128 2178
2129 2179 // Leave room for trap frame.
2130 2180 sp = sizeof *p>tf;
2131 2181 p>tf = (struct trapframe*)sp;
2132 2182
2133 2183 // Set up new context to start executing at forkret,
2134 2184 // which returns to trapret.
2135 2185 sp = 4;
2136 2186 *(uint*)sp = (uint)trapret;
2137 2187
2138 2188 sp = sizeof *p>context;
2139 2189 p>context = (struct context*)sp;
2140 2190 memset(p>context, 0, sizeof *p>context);
2141 2191 p>context>eip = (uint)forkret;
2142 2192
2143 2193 return p;
2144 2194 }
2145 2195
2146 2196
2147 2197
2148 2198
2149 2199

Sheet 21 Sheet 21
Sep 5 23:39 2011 xv6/proc.c Page 3 Sep 5 23:39 2011 xv6/proc.c Page 4

2200 // Set up first user process. 2250 // Create a new process copying p as the parent.
2201 void 2251 // Sets up stack to return as if from system call.
2202 userinit(void) 2252 // Caller must set state of returned proc to RUNNABLE.
2203 { 2253 int
2204 struct proc *p; 2254 fork(void)
2205 extern char _binary_initcode_start[], _binary_initcode_size[]; 2255 {
2206 2256 int i, pid;
2207 p = allocproc(); 2257 struct proc *np;
2208 initproc = p; 2258
2209 if((p>pgdir = setupkvm(kalloc)) == 0) 2259 // Allocate process.
2210 panic("userinit: out of memory?"); 2260 if((np = allocproc()) == 0)
2211 inituvm(p>pgdir, _binary_initcode_start, (int)_binary_initcode_size); 2261 return 1;
2212 p>sz = PGSIZE; 2262
2213 memset(p>tf, 0, sizeof(*p>tf)); 2263 // Copy process state from p.
2214 p>tf>cs = (SEG_UCODE << 3) | DPL_USER; 2264 if((np>pgdir = copyuvm(proc>pgdir, proc>sz)) == 0){
2215 p>tf>ds = (SEG_UDATA << 3) | DPL_USER; 2265 kfree(np>kstack);
2216 p>tf>es = p>tf>ds; 2266 np>kstack = 0;
2217 p>tf>ss = p>tf>ds; 2267 np>state = UNUSED;
2218 p>tf>eflags = FL_IF; 2268 return 1;
2219 p>tf>esp = PGSIZE; 2269 }
2220 p>tf>eip = 0; // beginning of initcode.S 2270 np>sz = proc>sz;
2221 2271 np>parent = proc;
2222 safestrcpy(p>name, "initcode", sizeof(p>name)); 2272 *np>tf = *proc>tf;
2223 p>cwd = namei("/"); 2273
2224 2274 // Clear %eax so that fork returns 0 in the child.
2225 p>state = RUNNABLE; 2275 np>tf>eax = 0;
2226 } 2276
2227 2277 for(i = 0; i < NOFILE; i++)
2228 // Grow current processs memory by n bytes. 2278 if(proc>ofile[i])
2229 // Return 0 on success, 1 on failure. 2279 np>ofile[i] = filedup(proc>ofile[i]);
2230 int 2280 np>cwd = idup(proc>cwd);
2231 growproc(int n) 2281
2232 { 2282 pid = np>pid;
2233 uint sz; 2283 np>state = RUNNABLE;
2234 2284 safestrcpy(np>name, proc>name, sizeof(proc>name));
2235 sz = proc>sz; 2285 return pid;
2236 if(n > 0){ 2286 }
2237 if((sz = allocuvm(proc>pgdir, sz, sz + n)) == 0) 2287
2238 return 1; 2288
2239 } else if(n < 0){ 2289
2240 if((sz = deallocuvm(proc>pgdir, sz, sz + n)) == 0) 2290
2241 return 1; 2291
2242 } 2292
2243 proc>sz = sz; 2293
2244 switchuvm(proc); 2294
2245 return 0; 2295
2246 } 2296
2247 2297
2248 2298
2249 2299

Sheet 22 Sheet 22
Sep 5 23:39 2011 xv6/proc.c Page 5 Sep 5 23:39 2011 xv6/proc.c Page 6

2300 // Exit the current process. Does not return. 2350 // Wait for a child process to exit and return its pid.
2301 // An exited process remains in the zombie state 2351 // Return 1 if this process has no children.
2302 // until its parent calls wait() to find out it exited. 2352 int
2303 void 2353 wait(void)
2304 exit(void) 2354 {
2305 { 2355 struct proc *p;
2306 struct proc *p; 2356 int havekids, pid;
2307 int fd; 2357
2308 2358 acquire(&ptable.lock);
2309 if(proc == initproc) 2359 for(;;){
2310 panic("init exiting"); 2360 // Scan through table looking for zombie children.
2311 2361 havekids = 0;
2312 // Close all open files. 2362 for(p = ptable.proc; p < &ptable.proc[NPROC]; p++){
2313 for(fd = 0; fd < NOFILE; fd++){ 2363 if(p>parent != proc)
2314 if(proc>ofile[fd]){ 2364 continue;
2315 fileclose(proc>ofile[fd]); 2365 havekids = 1;
2316 proc>ofile[fd] = 0; 2366 if(p>state == ZOMBIE){
2317 } 2367 // Found one.
2318 } 2368 pid = p>pid;
2319 2369 kfree(p>kstack);
2320 iput(proc>cwd); 2370 p>kstack = 0;
2321 proc>cwd = 0; 2371 freevm(p>pgdir);
2322 2372 p>state = UNUSED;
2323 acquire(&ptable.lock); 2373 p>pid = 0;
2324 2374 p>parent = 0;
2325 // Parent might be sleeping in wait(). 2375 p>name[0] = 0;
2326 wakeup1(proc>parent); 2376 p>killed = 0;
2327 2377 release(&ptable.lock);
2328 // Pass abandoned children to init. 2378 return pid;
2329 for(p = ptable.proc; p < &ptable.proc[NPROC]; p++){ 2379 }
2330 if(p>parent == proc){ 2380 }
2331 p>parent = initproc; 2381
2332 if(p>state == ZOMBIE) 2382 // No point waiting if we dont have any children.
2333 wakeup1(initproc); 2383 if(!havekids || proc>killed){
2334 } 2384 release(&ptable.lock);
2335 } 2385 return 1;
2336 2386 }
2337 // Jump into the scheduler, never to return. 2387
2338 proc>state = ZOMBIE; 2388 // Wait for children to exit. (See wakeup1 call in proc_exit.)
2339 sched(); 2389 sleep(proc, &ptable.lock);
2340 panic("zombie exit"); 2390 }
2341 } 2391 }
2342 2392
2343 2393
2344 2394
2345 2395
2346 2396
2347 2397
2348 2398
2349 2399

Sheet 23 Sheet 23
Sep 5 23:39 2011 xv6/proc.c Page 7 Sep 5 23:39 2011 xv6/proc.c Page 8

2400 // PerCPU process scheduler. 2450 // Enter scheduler. Must hold only ptable.lock
2401 // Each CPU calls scheduler() after setting itself up. 2451 // and have changed proc>state.
2402 // Scheduler never returns. It loops, doing: 2452 void
2403 // choose a process to run 2453 sched(void)
2404 // swtch to start running that process 2454 {
2405 // eventually that process transfers control 2455 int intena;
2406 // via swtch back to the scheduler. 2456
2407 void 2457 if(!holding(&ptable.lock))
2408 scheduler(void) 2458 panic("sched ptable.lock");
2409 { 2459 if(cpu>ncli != 1)
2410 struct proc *p; 2460 panic("sched locks");
2411 2461 if(proc>state == RUNNING)
2412 for(;;){ 2462 panic("sched running");
2413 // Enable interrupts on this processor. 2463 if(readeflags()&FL_IF)
2414 sti(); 2464 panic("sched interruptible");
2415 2465 intena = cpu>intena;
2416 // Loop over process table looking for process to run. 2466 swtch(&proc>context, cpu>scheduler);
2417 acquire(&ptable.lock); 2467 cpu>intena = intena;
2418 for(p = ptable.proc; p < &ptable.proc[NPROC]; p++){ 2468 }
2419 if(p>state != RUNNABLE) 2469
2420 continue; 2470 // Give up the CPU for one scheduling round.
2421 2471 void
2422 // Switch to chosen process. It is the processs job 2472 yield(void)
2423 // to release ptable.lock and then reacquire it 2473 {
2424 // before jumping back to us. 2474 acquire(&ptable.lock);
2425 proc = p; 2475 proc>state = RUNNABLE;
2426 switchuvm(p); 2476 sched();
2427 p>state = RUNNING; 2477 release(&ptable.lock);
2428 swtch(&cpu>scheduler, proc>context); 2478 }
2429 switchkvm(); 2479
2430 2480 // A fork childs very first scheduling by scheduler()
2431 // Process is done running for now. 2481 // will swtch here. "Return" to user space.
2432 // It should have changed its p>state before coming back. 2482 void
2433 proc = 0; 2483 forkret(void)
2434 } 2484 {
2435 release(&ptable.lock); 2485 static int first = 1;
2436 2486 // Still holding ptable.lock from scheduler.
2437 } 2487 release(&ptable.lock);
2438 } 2488
2439 2489 if (first) {
2440 2490 // Some initialization functions must be run in the context
2441 2491 // of a regular process (e.g., they call sleep), and thus cannot
2442 2492 // be run from main().
2443 2493 first = 0;
2444 2494 initlog();
2445 2495 }
2446 2496
2447 2497 // Return to "caller", actually trapret (see allocproc).
2448 2498 }
2449 2499

Sheet 24 Sheet 24
Sep 5 23:39 2011 xv6/proc.c Page 9 Sep 5 23:39 2011 xv6/proc.c Page 10

2500 // Atomically release lock and sleep on chan. 2550 // Wake up all processes sleeping on chan.
2501 // Reacquires lock when awakened. 2551 // The ptable lock must be held.
2502 void 2552 static void
2503 sleep(void *chan, struct spinlock *lk) 2553 wakeup1(void *chan)
2504 { 2554 {
2505 if(proc == 0) 2555 struct proc *p;
2506 panic("sleep"); 2556
2507 2557 for(p = ptable.proc; p < &ptable.proc[NPROC]; p++)
2508 if(lk == 0) 2558 if(p>state == SLEEPING && p>chan == chan)
2509 panic("sleep without lk"); 2559 p>state = RUNNABLE;
2510 2560 }
2511 // Must acquire ptable.lock in order to 2561
2512 // change p>state and then call sched. 2562 // Wake up all processes sleeping on chan.
2513 // Once we hold ptable.lock, we can be 2563 void
2514 // guaranteed that we wont miss any wakeup 2564 wakeup(void *chan)
2515 // (wakeup runs with ptable.lock locked), 2565 {
2516 // so its okay to release lk. 2566 acquire(&ptable.lock);
2517 if(lk != &ptable.lock){ 2567 wakeup1(chan);
2518 acquire(&ptable.lock); 2568 release(&ptable.lock);
2519 release(lk); 2569 }
2520 } 2570
2521 2571 // Kill the process with the given pid.
2522 // Go to sleep. 2572 // Process wont exit until it returns
2523 proc>chan = chan; 2573 // to user space (see trap in trap.c).
2524 proc>state = SLEEPING; 2574 int
2525 sched(); 2575 kill(int pid)
2526 2576 {
2527 // Tidy up. 2577 struct proc *p;
2528 proc>chan = 0; 2578
2529 2579 acquire(&ptable.lock);
2530 // Reacquire original lock. 2580 for(p = ptable.proc; p < &ptable.proc[NPROC]; p++){
2531 if(lk != &ptable.lock){ 2581 if(p>pid == pid){
2532 release(&ptable.lock); 2582 p>killed = 1;
2533 acquire(lk); 2583 // Wake process from sleep if necessary.
2534 } 2584 if(p>state == SLEEPING)
2535 } 2585 p>state = RUNNABLE;
2536 2586 release(&ptable.lock);
2537 2587 return 0;
2538 2588 }
2539 2589 }
2540 2590 release(&ptable.lock);
2541 2591 return 1;
2542 2592 }
2543 2593
2544 2594
2545 2595
2546 2596
2547 2597
2548 2598
2549 2599

Sheet 25 Sheet 25
Sep 5 23:39 2011 xv6/proc.c Page 11 Sep 5 23:39 2011 xv6/swtch.S Page 1

2600 // Print a process listing to console. For debugging. 2650 # Context switch
2601 // Runs when user types ^P on console. 2651 #
2602 // No lock to avoid wedging a stuck machine further. 2652 # void swtch(struct context **old, struct context *new);
2603 void 2653 #
2604 procdump(void) 2654 # Save current register context in old
2605 { 2655 # and then load register context from new.
2606 static char *states[] = { 2656
2607 [UNUSED] "unused", 2657 .globl swtch
2608 [EMBRYO] "embryo", 2658 swtch:
2609 [SLEEPING] "sleep ", 2659 movl 4(%esp), %eax
2610 [RUNNABLE] "runble", 2660 movl 8(%esp), %edx
2611 [RUNNING] "run ", 2661
2612 [ZOMBIE] "zombie" 2662 # Save old calleesave registers
2613 }; 2663 pushl %ebp
2614 int i; 2664 pushl %ebx
2615 struct proc *p; 2665 pushl %esi
2616 char *state; 2666 pushl %edi
2617 uint pc[10]; 2667
2618 2668 # Switch stacks
2619 for(p = ptable.proc; p < &ptable.proc[NPROC]; p++){ 2669 movl %esp, (%eax)
2620 if(p>state == UNUSED) 2670 movl %edx, %esp
2621 continue; 2671
2622 if(p>state >= 0 && p>state < NELEM(states) && states[p>state]) 2672 # Load new calleesave registers
2623 state = states[p>state]; 2673 popl %edi
2624 else 2674 popl %esi
2625 state = "???"; 2675 popl %ebx
2626 cprintf("%d %s %s", p>pid, state, p>name); 2676 popl %ebp
2627 if(p>state == SLEEPING){ 2677 ret
2628 getcallerpcs((uint*)p>context>ebp+2, pc); 2678
2629 for(i=0; i<10 && pc[i] != 0; i++) 2679
2630 cprintf(" %p", pc[i]); 2680
2631 } 2681
2632 cprintf("\n"); 2682
2633 } 2683
2634 } 2684
2635 2685
2636 2686
2637 2687
2638 2688
2639 2689
2640 2690
2641 2691
2642 2692
2643 2693
2644 2694
2645 2695
2646 2696
2647 2697
2648 2698
2649 2699

Sheet 26 Sheet 26
Sep 5 23:39 2011 xv6/kalloc.c Page 1 Sep 5 23:39 2011 xv6/kalloc.c Page 2

2700 // Physical memory allocator, intended to allocate 2750

2701 // memory for user processes, kernel stacks, page table pages, 2751 // Free the page of physical memory pointed at by v,
2702 // and pipe buffers. Allocates 4096byte pages. 2752 // which normally should have been returned by a
2703 2753 // call to kalloc(). (The exception is when
2704 #include "types.h" 2754 // initializing the allocator; see kinit above.)
2705 #include "defs.h" 2755 void
2706 #include "param.h" 2756 kfree(char *v)
2707 #include "memlayout.h" 2757 {
2708 #include "mmu.h" 2758 struct run *r;
2709 #include "spinlock.h" 2759
2710 2760 if((uint)v % PGSIZE || v < end || v2p(v) >= PHYSTOP)
2711 struct run { 2761 panic("kfree");
2712 struct run *next; 2762
2713 }; 2763 // Fill with junk to catch dangling refs.
2714 2764 memset(v, 1, PGSIZE);
2715 struct { 2765
2716 struct spinlock lock; 2766 acquire(&kmem.lock);
2717 struct run *freelist; 2767 r = (struct run*)v;
2718 } kmem; 2768 r>next = kmem.freelist;
2719 2769 kmem.freelist = r;
2720 extern char end[]; // first address after kernel loaded from ELF file 2770 release(&kmem.lock);
2721 static char *newend; 2771 }
2722 2772
2723 // A simple page allocator to get off the ground during entry 2773 // Allocate one 4096byte page of physical memory.
2724 char * 2774 // Returns a pointer that the kernel can use.
2725 enter_alloc(void) 2775 // Returns 0 if the memory cannot be allocated.
2726 { 2776 char*
2727 if (newend == 0) 2777 kalloc(void)
2728 newend = end; 2778 {
2729 2779 struct run *r;
2730 if ((uint) newend >= KERNBASE + 0x400000) 2780
2731 panic("only first 4Mbyte are mapped during entry"); 2781 acquire(&kmem.lock);
2732 void *p = (void*)PGROUNDUP((uint)newend); 2782 r = kmem.freelist;
2733 memset(p, 0, PGSIZE); 2783 if(r)
2734 newend = newend + PGSIZE; 2784 kmem.freelist = r>next;
2735 return p; 2785 release(&kmem.lock);
2736 } 2786 return (char*)r;
2737 2787 }
2738 // Initialize free list of physical pages. 2788
2739 void 2789
2740 kinit(void) 2790
2741 { 2791
2742 char *p; 2792
2743 2793
2744 initlock(&kmem.lock, "kmem"); 2794
2745 p = (char*)PGROUNDUP((uint)newend); 2795
2746 for(; p + PGSIZE <= (char*)p2v(PHYSTOP); p += PGSIZE) 2796
2747 kfree(p); 2797
2748 } 2798
2749 2799

Sheet 27 Sheet 27
Sep 5 23:39 2011 xv6/traps.h Page 1 Sep 5 23:39 2011 xv6/vectors.pl Page 1

2800 // x86 trap and interrupt constants. 2850 #!/usr/bin/perl w

2801 2851
2802 // Processordefined: 2852 # Generate vectors.S, the trap/interrupt entry points.
2803 #define T_DIVIDE 0 // divide error 2853 # There has to be one entry point per interrupt number
2804 #define T_DEBUG 1 // debug exception 2854 # since otherwise theres no way for trap() to discover
2805 #define T_NMI 2 // nonmaskable interrupt 2855 # the interrupt number.
2806 #define T_BRKPT 3 // breakpoint 2856
2807 #define T_OFLOW 4 // overflow 2857 print "# generated by vectors.pl do not edit\n";
2808 #define T_BOUND 5 // bounds check 2858 print "# handlers\n";
2809 #define T_ILLOP 6 // illegal opcode 2859 print ".globl alltraps\n";
2810 #define T_DEVICE 7 // device not available 2860 for(my $i = 0; $i < 256; $i++){
2811 #define T_DBLFLT 8 // double fault 2861 print ".globl vector$i\n";
2812 // #define T_COPROC 9 // reserved (not used since 486) 2862 print "vector$i:\n";
2813 #define T_TSS 10 // invalid task switch segment 2863 if(!($i == 8 || ($i >= 10 && $i <= 14) || $i == 17)){
2814 #define T_SEGNP 11 // segment not present 2864 print " pushl \$0\n";
2815 #define T_STACK 12 // stack exception 2865 }
2816 #define T_GPFLT 13 // general protection fault 2866 print " pushl \$$i\n";
2817 #define T_PGFLT 14 // page fault 2867 print " jmp alltraps\n";
2818 // #define T_RES 15 // reserved 2868 }
2819 #define T_FPERR 16 // floating point error 2869
2820 #define T_ALIGN 17 // aligment check 2870 print "\n# vector table\n";
2821 #define T_MCHK 18 // machine check 2871 print ".data\n";
2822 #define T_SIMDERR 19 // SIMD floating point error 2872 print ".globl vectors\n";
2823 2873 print "vectors:\n";
2824 // These are arbitrarily chosen, but with care not to overlap 2874 for(my $i = 0; $i < 256; $i++){
2825 // processor defined exceptions or interrupt vectors. 2875 print " .long vector$i\n";
2826 #define T_SYSCALL 64 // system call 2876 }
2827 #define T_DEFAULT 500 // catchall 2877
2828 2878 # sample output:
2829 #define T_IRQ0 32 // IRQ 0 corresponds to int T_IRQ 2879 # # handlers
2830 2880 # .globl alltraps
2831 #define IRQ_TIMER 0 2881 # .globl vector0
2832 #define IRQ_KBD 1 2882 # vector0:
2833 #define IRQ_COM1 4 2883 # pushl $0
2834 #define IRQ_IDE 14 2884 # pushl $0
2835 #define IRQ_ERROR 19 2885 # jmp alltraps
2836 #define IRQ_SPURIOUS 31 2886 # ...
2837 2887 #
2838 2888 # # vector table
2839 2889 # .data
2840 2890 # .globl vectors
2841 2891 # vectors:
2842 2892 # .long vector0
2843 2893 # .long vector1
2844 2894 # .long vector2
2845 2895 # ...
2846 2896
2847 2897
2848 2898
2849 2899

Sheet 28 Sheet 28
Sep 5 23:39 2011 xv6/trapasm.S Page 1 Sep 5 23:39 2011 xv6/trap.c Page 1

2900 #include "mmu.h" 2950 #include "types.h"

2901 2951 #include "defs.h"
2902 # vectors.S sends all traps here. 2952 #include "param.h"
2903 .globl alltraps 2953 #include "memlayout.h"
2904 alltraps: 2954 #include "mmu.h"
2905 # Build trap frame. 2955 #include "proc.h"
2906 pushl %ds 2956 #include "x86.h"
2907 pushl %es 2957 #include "traps.h"
2908 pushl %fs 2958 #include "spinlock.h"
2909 pushl %gs 2959
2910 pushal 2960 // Interrupt descriptor table (shared by all CPUs).
2911 2961 struct gatedesc idt[256];
2912 # Set up data and percpu segments. 2962 extern uint vectors[]; // in vectors.S: array of 256 entry pointers
2913 movw $(SEG_KDATA<<3), %ax 2963 struct spinlock tickslock;
2914 movw %ax, %ds 2964 uint ticks;
2915 movw %ax, %es 2965
2916 movw $(SEG_KCPU<<3), %ax 2966 void
2917 movw %ax, %fs 2967 tvinit(void)
2918 movw %ax, %gs 2968 {
2919 2969 int i;
2920 # Call trap(tf), where tf=%esp 2970
2921 pushl %esp 2971 for(i = 0; i < 256; i++)
2922 call trap 2972 SETGATE(idt[i], 0, SEG_KCODE<<3, vectors[i], 0);
2923 addl $4, %esp 2973 SETGATE(idt[T_SYSCALL], 1, SEG_KCODE<<3, vectors[T_SYSCALL], DPL_USER);
2924 2974
2925 # Return falls through to trapret... 2975 initlock(&tickslock, "time");
2926 .globl trapret 2976 }
2927 trapret: 2977
2928 popal 2978 void
2929 popl %gs 2979 idtinit(void)
2930 popl %fs 2980 {
2931 popl %es 2981 lidt(idt, sizeof(idt));
2932 popl %ds 2982 }
2933 addl $0x8, %esp # trapno and errcode 2983
2934 iret 2984
2935 2985
2936 2986
2937 2987
2938 2988
2939 2989
2940 2990
2941 2991
2942 2992
2943 2993
2944 2994
2945 2995
2946 2996
2947 2997
2948 2998
2949 2999

Sheet 29 Sheet 29
Sep 5 23:39 2011 xv6/trap.c Page 2 Sep 5 23:39 2011 xv6/trap.c Page 3

3000 void 3050 default:

3001 trap(struct trapframe *tf) 3051 if(proc == 0 || (tf>cs&3) == 0){
3002 { 3052 // In kernel, it must be our mistake.
3003 if(tf>trapno == T_SYSCALL){ 3053 cprintf("unexpected trap %d from cpu %d eip %x (cr2=0x%x)\n",
3004 if(proc>killed) 3054 tf>trapno, cpu>id, tf>eip, rcr2());
3005 exit(); 3055 panic("trap");
3006 proc>tf = tf; 3056 }
3007 syscall(); 3057 // In user space, assume process misbehaved.
3008 if(proc>killed) 3058 cprintf("pid %d %s: trap %d err %d on cpu %d "
3009 exit(); 3059 "eip 0x%x addr 0x%xkill proc\n",
3010 return; 3060 proc>pid, proc>name, tf>trapno, tf>err, cpu>id, tf>eip,
3011 } 3061 rcr2());
3012 3062 proc>killed = 1;
3013 switch(tf>trapno){ 3063 }
3014 case T_IRQ0 + IRQ_TIMER: 3064
3015 if(cpu>id == 0){ 3065 // Force process exit if it has been killed and is in user space.
3016 acquire(&tickslock); 3066 // (If it is still executing in the kernel, let it keep running
3017 ticks++; 3067 // until it gets to the regular system call return.)
3018 wakeup(&ticks); 3068 if(proc && proc>killed && (tf>cs&3) == DPL_USER)
3019 release(&tickslock); 3069 exit();
3020 } 3070
3021 lapiceoi(); 3071 // Force process to give up CPU on clock tick.
3022 break; 3072 // If interrupts were on while locks held, would need to check nlock.
3023 case T_IRQ0 + IRQ_IDE: 3073 if(proc && proc>state == RUNNING && tf>trapno == T_IRQ0+IRQ_TIMER)
3024 ideintr(); 3074 yield();
3025 lapiceoi(); 3075
3026 break; 3076 // Check if the process has been killed since we yielded
3027 case T_IRQ0 + IRQ_IDE+1: 3077 if(proc && proc>killed && (tf>cs&3) == DPL_USER)
3028 // Bochs generates spurious IDE1 interrupts. 3078 exit();
3029 break; 3079 }
3030 case T_IRQ0 + IRQ_KBD: 3080
3031 kbdintr(); 3081
3032 lapiceoi(); 3082
3033 break; 3083
3034 case T_IRQ0 + IRQ_COM1: 3084
3035 uartintr(); 3085
3036 lapiceoi(); 3086
3037 break; 3087
3038 case T_IRQ0 + 7: 3088
3039 case T_IRQ0 + IRQ_SPURIOUS: 3089
3040 cprintf("cpu%d: spurious interrupt at %x:%x\n", 3090
3041 cpu>id, tf>cs, tf>eip); 3091
3042 lapiceoi(); 3092
3043 break; 3093
3044 3094
3045 3095
3046 3096
3047 3097
3048 3098
3049 3099

Sheet 30 Sheet 30
Sep 5 23:39 2011 xv6/syscall.h Page 1 Sep 5 23:39 2011 xv6/syscall.c Page 1

3100 // System call numbers 3150 #include "types.h"

3101 #define SYS_fork 1 3151 #include "defs.h"
3102 #define SYS_exit 2 3152 #include "param.h"
3103 #define SYS_wait 3 3153 #include "memlayout.h"
3104 #define SYS_pipe 4 3154 #include "mmu.h"
3105 #define SYS_read 5 3155 #include "proc.h"
3106 #define SYS_kill 6 3156 #include "x86.h"
3107 #define SYS_exec 7 3157 #include "syscall.h"
3108 #define SYS_fstat 8 3158
3109 #define SYS_chdir 9 3159 // User code makes a system call with INT T_SYSCALL.
3110 #define SYS_dup 10 3160 // System call number in %eax.
3111 #define SYS_getpid 11 3161 // Arguments on the stack, from the user call to the C
3112 #define SYS_sbrk 12 3162 // library system call function. The saved user %esp points
3113 #define SYS_sleep 13 3163 // to a saved program counter, and then the first argument.
3114 #define SYS_uptime 14 3164
3115 3165 // Fetch the int at addr from process p.
3116 #define SYS_open 15 3166 int
3117 #define SYS_write 16 3167 fetchint(struct proc *p, uint addr, int *ip)
3118 #define SYS_mknod 17 3168 {
3119 #define SYS_unlink 18 3169 if(addr >= p>sz || addr+4 > p>sz)
3120 #define SYS_link 19 3170 return 1;
3121 #define SYS_mkdir 20 3171 *ip = *(int*)(addr);
3122 #define SYS_close 21 3172 return 0;
3123 3173 }
3124 3174
3125 3175 // Fetch the nulterminated string at addr from process p.
3126 3176 // Doesnt actually copy the string just sets *pp to point at it.
3127 3177 // Returns length of string, not including nul.
3128 3178 int
3129 3179 fetchstr(struct proc *p, uint addr, char **pp)
3130 3180 {
3131 3181 char *s, *ep;
3132 3182
3133 3183 if(addr >= p>sz)
3134 3184 return 1;
3135 3185 *pp = (char*)addr;
3136 3186 ep = (char*)p>sz;
3137 3187 for(s = *pp; s < ep; s++)
3138 3188 if(*s == 0)
3139 3189 return s *pp;
3140 3190 return 1;
3141 3191 }
3142 3192
3143 3193 // Fetch the nth 32bit system call argument.
3144 3194 int
3145 3195 argint(int n, int *ip)
3146 3196 {
3147 3197 return fetchint(proc, proc>tf>esp + 4 + 4*n, ip);
3148 3198 }
3149 3199

Sheet 31 Sheet 31
Sep 5 23:39 2011 xv6/syscall.c Page 2 Sep 5 23:39 2011 xv6/syscall.c Page 3

3200 // Fetch the nth wordsized system call argument as a pointer 3250 static int (*syscalls[])(void) = {
3201 // to a block of memory of size n bytes. Check that the pointer 3251 [SYS_fork] sys_fork,
3202 // lies within the process address space. 3252 [SYS_exit] sys_exit,
3203 int 3253 [SYS_wait] sys_wait,
3204 argptr(int n, char **pp, int size) 3254 [SYS_pipe] sys_pipe,
3205 { 3255 [SYS_read] sys_read,
3206 int i; 3256 [SYS_kill] sys_kill,
3207 3257 [SYS_exec] sys_exec,
3208 if(argint(n, &i) < 0) 3258 [SYS_fstat] sys_fstat,
3209 return 1; 3259 [SYS_chdir] sys_chdir,
3210 if((uint)i >= proc>sz || (uint)i+size > proc>sz) 3260 [SYS_dup] sys_dup,
3211 return 1; 3261 [SYS_getpid] sys_getpid,
3212 *pp = (char*)i; 3262 [SYS_sbrk] sys_sbrk,
3213 return 0; 3263 [SYS_sleep] sys_sleep,
3214 } 3264 [SYS_uptime] sys_uptime,
3215 3265 [SYS_open] sys_open,
3216 // Fetch the nth wordsized system call argument as a string pointer. 3266 [SYS_write] sys_write,
3217 // Check that the pointer is valid and the string is nulterminated. 3267 [SYS_mknod] sys_mknod,
3218 // (There is no shared writable memory, so the string cant change 3268 [SYS_unlink] sys_unlink,
3219 // between this check and being used by the kernel.) 3269 [SYS_link] sys_link,
3220 int 3270 [SYS_mkdir] sys_mkdir,
3221 argstr(int n, char **pp) 3271 [SYS_close] sys_close,
3222 { 3272 };
3223 int addr; 3273
3224 if(argint(n, &addr) < 0) 3274 void
3225 return 1; 3275 syscall(void)
3226 return fetchstr(proc, addr, pp); 3276 {
3227 } 3277 int num;
3228 3278
3229 extern int sys_chdir(void); 3279 num = proc>tf>eax;
3230 extern int sys_close(void); 3280 if(num >= 0 && num < SYS_open && syscalls[num]) {
3231 extern int sys_dup(void); 3281 proc>tf>eax = syscalls[num]();
3232 extern int sys_exec(void); 3282 } else if (num >= SYS_open && num < NELEM(syscalls) && syscalls[num]) {
3233 extern int sys_exit(void); 3283 proc>tf>eax = syscalls[num]();
3234 extern int sys_fork(void); 3284 } else {
3235 extern int sys_fstat(void); 3285 cprintf("%d %s: unknown sys call %d\n",
3236 extern int sys_getpid(void); 3286 proc>pid, proc>name, num);
3237 extern int sys_kill(void); 3287 proc>tf>eax = 1;
3238 extern int sys_link(void); 3288 }
3239 extern int sys_mkdir(void); 3289 }
3240 extern int sys_mknod(void); 3290
3241 extern int sys_open(void); 3291
3242 extern int sys_pipe(void); 3292
3243 extern int sys_read(void); 3293
3244 extern int sys_sbrk(void); 3294
3245 extern int sys_sleep(void); 3295
3246 extern int sys_unlink(void); 3296
3247 extern int sys_wait(void); 3297
3248 extern int sys_write(void); 3298
3249 extern int sys_uptime(void); 3299

Sheet 32 Sheet 32
Sep 5 23:39 2011 xv6/sysproc.c Page 1 Sep 5 23:39 2011 xv6/sysproc.c Page 2

3300 #include "types.h" 3350 int

3301 #include "x86.h" 3351 sys_sbrk(void)
3302 #include "defs.h" 3352 {
3303 #include "param.h" 3353 int addr;
3304 #include "memlayout.h" 3354 int n;
3305 #include "mmu.h" 3355
3306 #include "proc.h" 3356 if(argint(0, &n) < 0)
3307 3357 return 1;
3308 int 3358 addr = proc>sz;
3309 sys_fork(void) 3359 if(growproc(n) < 0)
3310 { 3360 return 1;
3311 return fork(); 3361 return addr;
3312 } 3362 }
3313 3363
3314 int 3364 int
3315 sys_exit(void) 3365 sys_sleep(void)
3316 { 3366 {
3317 exit(); 3367 int n;
3318 return 0; // not reached 3368 uint ticks0;
3319 } 3369
3320 3370 if(argint(0, &n) < 0)
3321 int 3371 return 1;
3322 sys_wait(void) 3372 acquire(&tickslock);
3323 { 3373 ticks0 = ticks;
3324 return wait(); 3374 while(ticks ticks0 < n){
3325 } 3375 if(proc>killed){
3326 3376 release(&tickslock);
3327 int 3377 return 1;
3328 sys_kill(void) 3378 }
3329 { 3379 sleep(&ticks, &tickslock);
3330 int pid; 3380 }
3331 3381 release(&tickslock);
3332 if(argint(0, &pid) < 0) 3382 return 0;
3333 return 1; 3383 }
3334 return kill(pid); 3384
3335 } 3385 // return how many clock tick interrupts have occurred
3336 3386 // since start.
3337 int 3387 int
3338 sys_getpid(void) 3388 sys_uptime(void)
3339 { 3389 {
3340 return proc>pid; 3390 uint xticks;
3341 } 3391
3342 3392 acquire(&tickslock);
3343 3393 xticks = ticks;
3344 3394 release(&tickslock);
3345 3395 return xticks;
3346 3396 }
3347 3397
3348 3398
3349 3399

Sheet 33 Sheet 33
Sep 5 23:39 2011 xv6/buf.h Page 1 Sep 5 23:39 2011 xv6/fcntl.h Page 1

3400 struct buf { 3450 #define O_RDONLY 0x000

3401 int flags; 3451 #define O_WRONLY 0x001
3402 uint dev; 3452 #define O_RDWR 0x002
3403 uint sector; 3453 #define O_CREATE 0x200
3404 struct buf *prev; // LRU cache list 3454
3405 struct buf *next; 3455
3406 struct buf *qnext; // disk queue 3456
3407 uchar data[512]; 3457
3408 }; 3458
3409 #define B_BUSY 0x1 // buffer is locked by some process 3459
3410 #define B_VALID 0x2 // buffer has been read from disk 3460
3411 #define B_DIRTY 0x4 // buffer needs to be written to disk 3461
3412 3462
3413 3463
3414 3464
3415 3465
3416 3466
3417 3467
3418 3468
3419 3469
3420 3470
3421 3471
3422 3472
3423 3473
3424 3474
3425 3475
3426 3476
3427 3477
3428 3478
3429 3479
3430 3480
3431 3481
3432 3482
3433 3483
3434 3484
3435 3485
3436 3486
3437 3487
3438 3488
3439 3489
3440 3490
3441 3491
3442 3492
3443 3493
3444 3494
3445 3495
3446 3496
3447 3497
3448 3498
3449 3499

Sheet 34 Sheet 34
Sep 5 23:39 2011 xv6/stat.h Page 1 Sep 5 23:39 2011 xv6/fs.h Page 1

3500 #define T_DIR 1 // Directory 3550 // Ondisk file system format.

3501 #define T_FILE 2 // File 3551 // Both the kernel and user programs use this header file.
3502 #define T_DEV 3 // Special device 3552
3503 3553 // Block 0 is unused. Block 1 is super block.
3504 struct stat { 3554 // Inodes start at block 2.
3505 short type; // Type of file 3555
3506 int dev; // Device number 3556 #define ROOTINO 1 // root inumber
3507 uint ino; // Inode number on device 3557 #define BSIZE 512 // block size
3508 short nlink; // Number of links to file 3558
3509 uint size; // Size of file in bytes 3559 // File system super block
3510 }; 3560 struct superblock {
3511 3561 uint size; // Size of file system image (blocks)
3512 3562 uint nblocks; // Number of data blocks
3513 3563 uint ninodes; // Number of inodes.
3514 3564 uint nlog; // Number of log blocks
3515 3565 };
3516 3566
3517 3567 #define NDIRECT 12
3518 3568 #define NINDIRECT (BSIZE / sizeof(uint))
3519 3569 #define MAXFILE (NDIRECT + NINDIRECT)
3520 3570
3521 3571 // Ondisk inode structure
3522 3572 struct dinode {
3523 3573 short type; // File type
3524 3574 short major; // Major device number (T_DEV only)
3525 3575 short minor; // Minor device number (T_DEV only)
3526 3576 short nlink; // Number of links to inode in file system
3527 3577 uint size; // Size of file (bytes)
3528 3578 uint addrs[NDIRECT+1]; // Data block addresses
3529 3579 };
3530 3580
3531 3581 // Inodes per block.
3532 3582 #define IPB (BSIZE / sizeof(struct dinode))
3533 3583
3534 3584 // Block containing inode i
3535 3585 #define IBLOCK(i) ((i) / IPB + 2)
3536 3586
3537 3587 // Bitmap bits per block
3538 3588 #define BPB (BSIZE*8)
3539 3589
3540 3590 // Block containing bit for block b
3541 3591 #define BBLOCK(b, ninodes) (b/BPB + (ninodes)/IPB + 3)
3542 3592
3543 3593 // Directory is a file containing a sequence of dirent structures.
3544 3594 #define DIRSIZ 14
3545 3595
3546 3596 struct dirent {
3547 3597 ushort inum;
3548 3598 char name[DIRSIZ];
3549 3599 };

Sheet 35 Sheet 35
Sep 5 23:39 2011 xv6/file.h Page 1 Sep 5 23:39 2011 xv6/ide.c Page 1

3600 struct file { 3650 // Simple PIObased (nonDMA) IDE driver code.
3601 enum { FD_NONE, FD_PIPE, FD_INODE } type; 3651
3602 int ref; // reference count 3652 #include "types.h"
3603 char readable; 3653 #include "defs.h"
3604 char writable; 3654 #include "param.h"
3605 struct pipe *pipe; 3655 #include "memlayout.h"
3606 struct inode *ip; 3656 #include "mmu.h"
3607 uint off; 3657 #include "proc.h"
3608 }; 3658 #include "x86.h"
3609 3659 #include "traps.h"
3610 3660 #include "spinlock.h"
3611 // incore file system types 3661 #include "buf.h"
3612 3662
3613 struct inode { 3663 #define IDE_BSY 0x80
3614 uint dev; // Device number 3664 #define IDE_DRDY 0x40
3615 uint inum; // Inode number 3665 #define IDE_DF 0x20
3616 int ref; // Reference count 3666 #define IDE_ERR 0x01
3617 int flags; // I_BUSY, I_VALID 3667
3618 3668 #define IDE_CMD_READ 0x20
3619 short type; // copy of disk inode 3669 #define IDE_CMD_WRITE 0x30
3620 short major; 3670
3621 short minor; 3671 // idequeue points to the buf now being read/written to the disk.
3622 short nlink; 3672 // idequeue>qnext points to the next buf to be processed.
3623 uint size; 3673 // You must hold idelock while manipulating queue.
3624 uint addrs[NDIRECT+1]; 3674
3625 }; 3675 static struct spinlock idelock;
3626 3676 static struct buf *idequeue;
3627 #define I_BUSY 0x1 3677
3628 #define I_VALID 0x2 3678 static int havedisk1;
3629 3679 static void idestart(struct buf*);
3630 // device implementations 3680
3631 3681 // Wait for IDE disk to become ready.
3632 struct devsw { 3682 static int
3633 int (*read)(struct inode*, char*, int); 3683 idewait(int checkerr)
3634 int (*write)(struct inode*, char*, int); 3684 {
3635 }; 3685 int r;
3636 3686
3637 extern struct devsw devsw[]; 3687 while(((r = inb(0x1f7)) & (IDE_BSY|IDE_DRDY)) != IDE_DRDY)
3638 3688 ;
3639 #define CONSOLE 1 3689 if(checkerr && (r & (IDE_DF|IDE_ERR)) != 0)
3640 3690 return 1;
3641 3691 return 0;
3642 3692 }
3643 3693
3644 3694
3645 3695
3646 3696
3647 3697
3648 3698
3649 3699

Sheet 36 Sheet 36
Sep 5 23:39 2011 xv6/ide.c Page 2 Sep 5 23:39 2011 xv6/ide.c Page 3

3700 void 3750 // Interrupt handler.

3701 ideinit(void) 3751 void
3702 { 3752 ideintr(void)
3703 int i; 3753 {
3704 3754 struct buf *b;
3705 initlock(&idelock, "ide"); 3755
3706 picenable(IRQ_IDE); 3756 // Take first buffer off queue.
3707 ioapicenable(IRQ_IDE, ncpu 1); 3757 acquire(&idelock);
3708 idewait(0); 3758 if((b = idequeue) == 0){
3709 3759 release(&idelock);
3710 // Check if disk 1 is present 3760 // cprintf("spurious IDE interrupt\n");
3711 outb(0x1f6, 0xe0 | (1<<4)); 3761 return;
3712 for(i=0; i<1000; i++){ 3762 }
3713 if(inb(0x1f7) != 0){ 3763 idequeue = b>qnext;
3714 havedisk1 = 1; 3764
3715 break; 3765 // Read data if needed.
3716 } 3766 if(!(b>flags & B_DIRTY) && idewait(1) >= 0)
3717 } 3767 insl(0x1f0, b>data, 512/4);
3718 3768
3719 // Switch back to disk 0. 3769 // Wake process waiting for this buf.
3720 outb(0x1f6, 0xe0 | (0<<4)); 3770 b>flags |= B_VALID;
3721 } 3771 b>flags &= ~B_DIRTY;
3722 3772 wakeup(b);
3723 // Start the request for b. Caller must hold idelock. 3773
3724 static void 3774 // Start disk on next buf in queue.
3725 idestart(struct buf *b) 3775 if(idequeue != 0)
3726 { 3776 idestart(idequeue);
3727 if(b == 0) 3777
3728 panic("idestart"); 3778 release(&idelock);
3729 3779 }
3730 idewait(0); 3780
3731 outb(0x3f6, 0); // generate interrupt 3781
3732 outb(0x1f2, 1); // number of sectors 3782
3733 outb(0x1f3, b>sector & 0xff); 3783
3734 outb(0x1f4, (b>sector >> 8) & 0xff); 3784
3735 outb(0x1f5, (b>sector >> 16) & 0xff); 3785
3736 outb(0x1f6, 0xe0 | ((b>dev&1)<<4) | ((b>sector>>24)&0x0f)); 3786
3737 if(b>flags & B_DIRTY){ 3787
3738 outb(0x1f7, IDE_CMD_WRITE); 3788
3739 outsl(0x1f0, b>data, 512/4); 3789
3740 } else { 3790
3741 outb(0x1f7, IDE_CMD_READ); 3791
3742 } 3792
3743 } 3793
3744 3794
3745 3795
3746 3796
3747 3797
3748 3798
3749 3799

Sheet 37 Sheet 37
Sep 5 23:39 2011 xv6/ide.c Page 4 Sep 5 23:39 2011 xv6/bio.c Page 1

3800 // Sync buf with disk. 3850 // Buffer cache.

3801 // If B_DIRTY is set, write buf to disk, clear B_DIRTY, set B_VALID. 3851 //
3802 // Else if B_VALID is not set, read buf from disk, set B_VALID. 3852 // The buffer cache is a linked list of buf structures holding
3803 void 3853 // cached copies of disk block contents. Caching disk blocks
3804 iderw(struct buf *b) 3854 // in memory reduces the number of disk reads and also provides
3805 { 3855 // a synchronization point for disk blocks used by multiple processes.
3806 struct buf **pp; 3856 //
3807 3857 // Interface:
3808 if(!(b>flags & B_BUSY)) 3858 // * To get a buffer for a particular disk block, call bread.
3809 panic("iderw: buf not busy"); 3859 // * After changing buffer data, call bwrite to flush it to disk.
3810 if((b>flags & (B_VALID|B_DIRTY)) == B_VALID) 3860 // * When done with the buffer, call brelse.
3811 panic("iderw: nothing to do"); 3861 // * Do not use the buffer after calling brelse.
3812 if(b>dev != 0 && !havedisk1) 3862 // * Only one process at a time can use a buffer,
3813 panic("iderw: ide disk 1 not present"); 3863 // so do not keep them longer than necessary.
3814 3864 //
3815 acquire(&idelock); // DOC:acquirelock 3865 // The implementation uses three state flags internally:
3816 3866 // * B_BUSY: the block has been returned from bread
3817 // Append b to idequeue. 3867 // and has not been passed back to brelse.
3818 b>qnext = 0; 3868 // * B_VALID: the buffer data has been initialized
3819 for(pp=&idequeue; *pp; pp=&(*pp)>qnext) // DOC:insertqueue 3869 // with the associated disk block contents.
3820 ; 3870 // * B_DIRTY: the buffer data has been modified
3821 *pp = b; 3871 // and needs to be written to disk.
3822 3872
3823 // Start disk if necessary. 3873 #include "types.h"
3824 if(idequeue == b) 3874 #include "defs.h"
3825 idestart(b); 3875 #include "param.h"
3826 3876 #include "spinlock.h"
3827 // Wait for request to finish. 3877 #include "buf.h"
3828 // Assuming will not sleep too long: ignore proc>killed. 3878
3829 while((b>flags & (B_VALID|B_DIRTY)) != B_VALID){ 3879 struct {
3830 sleep(b, &idelock); 3880 struct spinlock lock;
3831 } 3881 struct buf buf[NBUF];
3832 3882
3833 release(&idelock); 3883 // Linked list of all buffers, through prev/next.
3834 } 3884 // head.next is most recently used.
3835 3885 struct buf head;
3836 3886 } bcache;
3837 3887
3838 3888 void
3839 3889 binit(void)
3840 3890 {
3841 3891 struct buf *b;
3842 3892
3843 3893 initlock(&bcache.lock, "bcache");
3844 3894
3845 3895
3846 3896
3847 3897
3848 3898
3849 3899

Sheet 38 Sheet 38
Sep 5 23:39 2011 xv6/bio.c Page 2 Sep 5 23:39 2011 xv6/bio.c Page 3

3900 // Create linked list of buffers 3950 // Return a B_BUSY buf with the contents of the indicated disk sector.
3901 bcache.head.prev = &bcache.head; 3951 struct buf*
3902 bcache.head.next = &bcache.head; 3952 bread(uint dev, uint sector)
3903 for(b = bcache.buf; b < bcache.buf+NBUF; b++){ 3953 {
3904 b>next = bcache.head.next; 3954 struct buf *b;
3905 b>prev = &bcache.head; 3955
3906 b>dev = 1; 3956 b = bget(dev, sector);
3907 bcache.head.next>prev = b; 3957 if(!(b>flags & B_VALID))
3908 bcache.head.next = b; 3958 iderw(b);
3909 } 3959 return b;
3910 } 3960 }
3911 3961
3912 // Look through buffer cache for sector on device dev. 3962 // Write bs contents to disk. Must be locked.
3913 // If not found, allocate fresh block. 3963 void
3914 // In either case, return locked buffer. 3964 bwrite(struct buf *b)
3915 static struct buf* 3965 {
3916 bget(uint dev, uint sector) 3966 if((b>flags & B_BUSY) == 0)
3917 { 3967 panic("bwrite");
3918 struct buf *b; 3968 b>flags |= B_DIRTY;
3919 3969 iderw(b);
3920 acquire(&bcache.lock); 3970 }
3921 3971
3922 loop: 3972 // Release the buffer b.
3923 // Try for cached block. 3973 void
3924 for(b = bcache.head.next; b != &bcache.head; b = b>next){ 3974 brelse(struct buf *b)
3925 if(b>dev == dev && b>sector == sector){ 3975 {
3926 if(!(b>flags & B_BUSY)){ 3976 if((b>flags & B_BUSY) == 0)
3927 b>flags |= B_BUSY; 3977 panic("brelse");
3928 release(&bcache.lock); 3978
3929 return b; 3979 acquire(&bcache.lock);
3930 } 3980
3931 sleep(b, &bcache.lock); 3981 b>next>prev = b>prev;
3932 goto loop; 3982 b>prev>next = b>next;
3933 } 3983 b>next = bcache.head.next;
3934 } 3984 b>prev = &bcache.head;
3935 3985 bcache.head.next>prev = b;
3936 // Allocate fresh block. 3986 bcache.head.next = b;
3937 for(b = bcache.head.prev; b != &bcache.head; b = b>prev){ 3987
3938 if((b>flags & B_BUSY) == 0){ 3988 b>flags &= ~B_BUSY;
3939 b>dev = dev; 3989 wakeup(b);
3940 b>sector = sector; 3990
3941 b>flags = B_BUSY; 3991 release(&bcache.lock);
3942 release(&bcache.lock); 3992 }
3943 return b; 3993
3944 } 3994
3945 } 3995
3946 panic("bget: no buffers"); 3996
3947 } 3997
3948 3998
3949 3999

Sheet 39 Sheet 39
Sep 5 23:39 2011 xv6/log.c Page 1 Sep 5 23:39 2011 xv6/log.c Page 2

4000 #include "types.h" 4050 struct log log;

4001 #include "defs.h" 4051
4002 #include "param.h" 4052 static void recover_from_log(void);
4003 #include "spinlock.h" 4053
4004 #include "fs.h" 4054 void
4005 #include "buf.h" 4055 initlog(void)
4006 4056 {
4007 // Simple logging. Each system call that might write the file system 4057 if (sizeof(struct logheader) >= BSIZE)
4008 // should be surrounded with begin_trans() and commit_trans() calls. 4058 panic("initlog: too big logheader");
4009 // 4059
4010 // The log holds at most one transaction at a time. Commit forces 4060 struct superblock sb;
4011 // the log (with commit record) to disk, then installs the affected 4061 initlock(&log.lock, "log");
4012 // blocks to disk, then erases the log. begin_trans() ensures that 4062 readsb(ROOTDEV, &sb);
4013 // only one system call can be in a transaction; others must wait. 4063 log.start = sb.size sb.nlog;
4014 // 4064 log.size = sb.nlog;
4015 // Allowing only one transaction at a time means that the file 4065 log.dev = ROOTDEV;
4016 // system code doesnt have to worry about the possibility of 4066 recover_from_log();
4017 // one transaction reading a block that another one has modified, 4067 }
4018 // for example an inode block. 4068
4019 // 4069 // Copy committed blocks from log to their home location
4020 // Readonly system calls dont need to use transactions, though 4070 static void
4021 // this means that they may observe uncommitted data. Inode and 4071 install_trans(void)
4022 // buffer locks prevent readonly calls from seeing inconsistent data. 4072 {
4023 // 4073 int tail;
4024 // The log is a physical redo log containing disk blocks. 4074
4025 // The ondisk log format: 4075 for (tail = 0; tail < log.lh.n; tail++) {
4026 // header block, containing sector #s for block A, B, C, ... 4076 struct buf *lbuf = bread(log.dev, log.start+tail+1); // read log block
4027 // block A 4077 struct buf *dbuf = bread(log.dev, log.lh.sector[tail]); // read dst
4028 // block B 4078 memmove(dbuf>data, lbuf>data, BSIZE); // copy block to dst
4029 // block C 4079 bwrite(dbuf); // flush dst to disk
4030 // ... 4080 brelse(lbuf);
4031 // Log appends are synchronous. 4081 brelse(dbuf);
4032 4082 }
4033 // Contents of the header block, used for both the ondisk header block 4083 }
4034 // and to keep track in memory of logged sector #s before commit. 4084
4035 struct logheader { 4085 // Read the log header from disk into the inmemory log header
4036 int n; 4086 static void
4037 int sector[LOGSIZE]; 4087 read_head(void)
4038 }; 4088 {
4039 4089 struct buf *buf = bread(log.dev, log.start);
4040 struct log { 4090 struct logheader *lh = (struct logheader *) (buf>data);
4041 struct spinlock lock; 4091 int i;
4042 int start; 4092 log.lh.n = lh>n;
4043 int size; 4093 for (i = 0; i < log.lh.n; i++) {
4044 int intrans; 4094 log.lh.sector[i] = lh>sector[i];
4045 int dev; 4095 }
4046 struct logheader lh; 4096 brelse(buf);
4047 }; 4097 }
4048 4098
4049 4099

Sheet 40 Sheet 40
Sep 5 23:39 2011 xv6/log.c Page 3 Sep 5 23:39 2011 xv6/log.c Page 4

4100 // Write inmemory log header to disk, committing log entries till head 4150 // Caller has modified b>data and is done with the buffer.
4101 static void 4151 // Append the block to the log and record the block number,
4102 write_head(void) 4152 // but dont write the log header (which would commit the write).
4103 { 4153 // log_write() replaces bwrite(); a typical use is:
4104 struct buf *buf = bread(log.dev, log.start); 4154 // bp = bread(...)
4105 struct logheader *hb = (struct logheader *) (buf>data); 4155 // modify bp>data[]
4106 int i; 4156 // log_write(bp)
4107 hb>n = log.lh.n; 4157 // brelse(bp)
4108 for (i = 0; i < log.lh.n; i++) { 4158 void
4109 hb>sector[i] = log.lh.sector[i]; 4159 log_write(struct buf *b)
4110 } 4160 {
4111 bwrite(buf); 4161 int i;
4112 brelse(buf); 4162
4113 } 4163 if (log.lh.n >= LOGSIZE || log.lh.n >= log.size 1)
4114 4164 panic("too big a transaction");
4115 static void 4165 if (!log.intrans)
4116 recover_from_log(void) 4166 panic("write outside of trans");
4117 { 4167
4118 read_head(); 4168 for (i = 0; i < log.lh.n; i++) {
4119 install_trans(); // if committed, copy from log to disk 4169 if (log.lh.sector[i] == b>sector) // log absorbtion?
4120 log.lh.n = 0; 4170 break;
4121 write_head(); // clear the log 4171 }
4122 } 4172 log.lh.sector[i] = b>sector;
4123 4173 struct buf *lbuf = bread(b>dev, log.start+i+1);
4124 void 4174 memmove(lbuf>data, b>data, BSIZE);
4125 begin_trans(void) 4175 bwrite(lbuf);
4126 { 4176 brelse(lbuf);
4127 acquire(&log.lock); 4177 if (i == log.lh.n)
4128 while (log.intrans) { 4178 log.lh.n++;
4129 sleep(&log, &log.lock); 4179 }
4130 } 4180
4131 log.intrans = 1; 4181
4132 release(&log.lock); 4182
4133 } 4183
4134 4184
4135 void 4185
4136 commit_trans(void) 4186
4137 { 4187
4138 if (log.lh.n > 0) { 4188
4139 write_head(); // Causes all blocks till log.head to be commited 4189
4140 install_trans(); // Install all the transactions till head 4190
4141 log.lh.n = 0; 4191
4142 write_head(); // Reclaim log 4192
4143 } 4193
4144 4194
4145 acquire(&log.lock); 4195
4146 log.intrans = 0; 4196
4147 wakeup(&log); 4197
4148 release(&log.lock); 4198
4149 } 4199

Sheet 41 Sheet 41
Sep 5 23:39 2011 xv6/log.c Page 5 Sep 5 23:39 2011 xv6/fs.c Page 1

4200 // Blank page. 4250 // File system implementation. Four layers:

4201 4251 // + Blocks: allocator for raw disk blocks.
4202 4252 // + Files: inode allocator, reading, writing, metadata.
4203 4253 // + Directories: inode with special contents (list of other inodes!)
4204 4254 // + Names: paths like /usr/rtm/xv6/fs.c for convenient naming.
4205 4255 //
4206 4256 // Disk layout is: superblock, inodes, block inuse bitmap, data blocks.
4207 4257 //
4208 4258 // This file contains the lowlevel file system manipulation
4209 4259 // routines. The (higherlevel) system call implementations
4210 4260 // are in sysfile.c.
4211 4261
4212 4262 #include "types.h"
4213 4263 #include "defs.h"
4214 4264 #include "param.h"
4215 4265 #include "stat.h"
4216 4266 #include "mmu.h"
4217 4267 #include "proc.h"
4218 4268 #include "spinlock.h"
4219 4269 #include "buf.h"
4220 4270 #include "fs.h"
4221 4271 #include "file.h"
4222 4272
4223 4273 #define min(a, b) ((a) < (b) ? (a) : (b))
4224 4274 static void itrunc(struct inode*);
4225 4275
4226 4276 // Read the super block.
4227 4277 void
4228 4278 readsb(int dev, struct superblock *sb)
4229 4279 {
4230 4280 struct buf *bp;
4231 4281
4232 4282 bp = bread(dev, 1);
4233 4283 memmove(sb, bp>data, sizeof(*sb));
4234 4284 brelse(bp);
4235 4285 }
4236 4286
4237 4287 // Zero a block.
4238 4288 static void
4239 4289 bzero(int dev, int bno)
4240 4290 {
4241 4291 struct buf *bp;
4242 4292
4243 4293 bp = bread(dev, bno);
4244 4294 memset(bp>data, 0, BSIZE);
4245 4295 log_write(bp);
4246 4296 brelse(bp);
4247 4297 }
4248 4298
4249 4299

Sheet 42 Sheet 42
Sep 5 23:39 2011 xv6/fs.c Page 2 Sep 5 23:39 2011 xv6/fs.c Page 3

4300 // Blocks. 4350 // Inodes.

4301 4351 //
4302 // Allocate a zeroed disk block. 4352 // An inode is a single, unnamed file in the file system.
4303 static uint 4353 // The inode disk structure holds metadata (the type, device numbers,
4304 balloc(uint dev) 4354 // and data size) along with a list of blocks where the associated
4305 { 4355 // data can be found.
4306 int b, bi, m; 4356 //
4307 struct buf *bp; 4357 // The inodes are laid out sequentially on disk immediately after
4308 struct superblock sb; 4358 // the superblock. The kernel keeps a cache of the inuse
4309 4359 // ondisk structures to provide a place for synchronizing access
4310 bp = 0; 4360 // to inodes shared between multiple processes.
4311 readsb(dev, &sb); 4361 //
4312 for(b = 0; b < sb.size; b += BPB){ 4362 // ip>ref counts the number of pointer references to this cached
4313 bp = bread(dev, BBLOCK(b, sb.ninodes)); 4363 // inode; references are typically kept in struct file and in proc>cwd.
4314 for(bi = 0; bi < BPB && bi < (sb.size b); bi++){ 4364 // When ip>ref falls to zero, the inode is no longer cached.
4315 m = 1 << (bi % 8); 4365 // It is an error to use an inode without holding a reference to it.
4316 if((bp>data[bi/8] & m) == 0){ // Is block free? 4366 //
4317 bp>data[bi/8] |= m; // Mark block in use on disk. 4367 // Processes are only allowed to read and write inode
4318 log_write(bp); 4368 // metadata and contents when holding the inodes lock,
4319 brelse(bp); 4369 // represented by the I_BUSY flag in the inmemory copy.
4320 bzero(dev, b + bi); 4370 // Because inode locks are held during disk accesses,
4321 return b + bi; 4371 // they are implemented using a flag rather than with
4322 } 4372 // spin locks. Callers are responsible for locking
4323 } 4373 // inodes before passing them to routines in this file; leaving
4324 brelse(bp); 4374 // this responsibility with the caller makes it possible for them
4325 } 4375 // to create arbitrarilysized atomic operations.
4326 panic("balloc: out of blocks"); 4376 //
4327 } 4377 // To give maximum control over locking to the callers,
4328 4378 // the routines in this file that return inode pointers
4329 // Free a disk block. 4379 // return pointers to *unlocked* inodes. It is the callers
4330 static void 4380 // responsibility to lock them before using them. A nonzero
4331 bfree(int dev, uint b) 4381 // ip>ref keeps these unlocked inodes in the cache.
4332 { 4382
4333 struct buf *bp; 4383 struct {
4334 struct superblock sb; 4384 struct spinlock lock;
4335 int bi, m; 4385 struct inode inode[NINODE];
4336 4386 } icache;
4337 readsb(dev, &sb); 4387
4338 bp = bread(dev, BBLOCK(b, sb.ninodes)); 4388 void
4339 bi = b % BPB; 4389 iinit(void)
4340 m = 1 << (bi % 8); 4390 {
4341 if((bp>data[bi/8] & m) == 0) 4391 initlock(&icache.lock, "icache");
4342 panic("freeing free block"); 4392 }
4343 bp>data[bi/8] &= ~m; // Mark block free on disk. 4393
4344 log_write(bp); 4394 static struct inode* iget(uint dev, uint inum);
4345 brelse(bp); 4395
4346 } 4396
4347 4397
4348 4398
4349 4399

Sheet 43 Sheet 43
Sep 5 23:39 2011 xv6/fs.c Page 4 Sep 5 23:39 2011 xv6/fs.c Page 5

4400 // Allocate a new inode with the given type on device dev. 4450 // Find the inode with number inum on device dev
4401 struct inode* 4451 // and return the inmemory copy.
4402 ialloc(uint dev, short type) 4452 static struct inode*
4403 { 4453 iget(uint dev, uint inum)
4404 int inum; 4454 {
4405 struct buf *bp; 4455 struct inode *ip, *empty;
4406 struct dinode *dip; 4456
4407 struct superblock sb; 4457 acquire(&icache.lock);
4408 4458
4409 readsb(dev, &sb); 4459 // Try for cached inode.
4410 for(inum = 1; inum < sb.ninodes; inum++){ // loop over inode blocks 4460 empty = 0;
4411 bp = bread(dev, IBLOCK(inum)); 4461 for(ip = &icache.inode[0]; ip < &icache.inode[NINODE]; ip++){
4412 dip = (struct dinode*)bp>data + inum%IPB; 4462 if(ip>ref > 0 && ip>dev == dev && ip>inum == inum){
4413 if(dip>type == 0){ // a free inode 4463 ip>ref++;
4414 memset(dip, 0, sizeof(*dip)); 4464 release(&icache.lock);
4415 dip>type = type; 4465 return ip;
4416 log_write(bp); // mark it allocated on the disk 4466 }
4417 brelse(bp); 4467 if(empty == 0 && ip>ref == 0) // Remember empty slot.
4418 return iget(dev, inum); 4468 empty = ip;
4419 } 4469 }
4420 brelse(bp); 4470
4421 } 4471 // Allocate fresh inode.
4422 panic("ialloc: no inodes"); 4472 if(empty == 0)
4423 } 4473 panic("iget: no inodes");
4424 4474
4425 // Copy inode, which has changed, from memory to disk. 4475 ip = empty;
4426 void 4476 ip>dev = dev;
4427 iupdate(struct inode *ip) 4477 ip>inum = inum;
4428 { 4478 ip>ref = 1;
4429 struct buf *bp; 4479 ip>flags = 0;
4430 struct dinode *dip; 4480 release(&icache.lock);
4431 4481
4432 bp = bread(ip>dev, IBLOCK(ip>inum)); 4482 return ip;
4433 dip = (struct dinode*)bp>data + ip>inum%IPB; 4483 }
4434 dip>type = ip>type; 4484
4435 dip>major = ip>major; 4485 // Increment reference count for ip.
4436 dip>minor = ip>minor; 4486 // Returns ip to enable ip = idup(ip1) idiom.
4437 dip>nlink = ip>nlink; 4487 struct inode*
4438 dip>size = ip>size; 4488 idup(struct inode *ip)
4439 memmove(dip>addrs, ip>addrs, sizeof(ip>addrs)); 4489 {
4440 log_write(bp); 4490 acquire(&icache.lock);
4441 brelse(bp); 4491 ip>ref++;
4442 } 4492 release(&icache.lock);
4443 4493 return ip;
4444 4494 }
4445 4495
4446 4496
4447 4497
4448 4498
4449 4499

Sheet 44 Sheet 44
Sep 5 23:39 2011 xv6/fs.c Page 6 Sep 5 23:39 2011 xv6/fs.c Page 7

4500 // Lock the given inode. 4550 // Caller holds reference to unlocked ip. Drop reference.
4501 void 4551 void
4502 ilock(struct inode *ip) 4552 iput(struct inode *ip)
4503 { 4553 {
4504 struct buf *bp; 4554 acquire(&icache.lock);
4505 struct dinode *dip; 4555 if(ip>ref == 1 && (ip>flags & I_VALID) && ip>nlink == 0){
4506 4556 // inode is no longer used: truncate and free inode.
4507 if(ip == 0 || ip>ref < 1) 4557 if(ip>flags & I_BUSY)
4508 panic("ilock"); 4558 panic("iput busy");
4509 4559 ip>flags |= I_BUSY;
4510 acquire(&icache.lock); 4560 release(&icache.lock);
4511 while(ip>flags & I_BUSY) 4561 itrunc(ip);
4512 sleep(ip, &icache.lock); 4562 ip>type = 0;
4513 ip>flags |= I_BUSY; 4563 iupdate(ip);
4514 release(&icache.lock); 4564 acquire(&icache.lock);
4515 4565 ip>flags = 0;
4516 if(!(ip>flags & I_VALID)){ 4566 wakeup(ip);
4517 bp = bread(ip>dev, IBLOCK(ip>inum)); 4567 }
4518 dip = (struct dinode*)bp>data + ip>inum%IPB; 4568 ip>ref;
4519 ip>type = dip>type; 4569 release(&icache.lock);
4520 ip>major = dip>major; 4570 }
4521 ip>minor = dip>minor; 4571
4522 ip>nlink = dip>nlink; 4572 // Common idiom: unlock, then put.
4523 ip>size = dip>size; 4573 void
4524 memmove(ip>addrs, dip>addrs, sizeof(ip>addrs)); 4574 iunlockput(struct inode *ip)
4525 brelse(bp); 4575 {
4526 ip>flags |= I_VALID; 4576 iunlock(ip);
4527 if(ip>type == 0) 4577 iput(ip);
4528 panic("ilock: no type"); 4578 }
4529 } 4579
4530 } 4580
4531 4581
4532 // Unlock the given inode. 4582
4533 void 4583
4534 iunlock(struct inode *ip) 4584
4535 { 4585
4536 if(ip == 0 || !(ip>flags & I_BUSY) || ip>ref < 1) 4586
4537 panic("iunlock"); 4587
4538 4588
4539 acquire(&icache.lock); 4589
4540 ip>flags &= ~I_BUSY; 4590
4541 wakeup(ip); 4591
4542 release(&icache.lock); 4592
4543 } 4593
4544 4594
4545 4595
4546 4596
4547 4597
4548 4598
4549 4599

Sheet 45 Sheet 45
Sep 5 23:39 2011 xv6/fs.c Page 8 Sep 5 23:39 2011 xv6/fs.c Page 9

4600 // Inode contents 4650 // Truncate inode (discard contents).

4601 // 4651 // Only called after the last dirent referring
4602 // The contents (data) associated with each inode is stored 4652 // to this inode has been erased on disk.
4603 // in a sequence of blocks on the disk. The first NDIRECT blocks 4653 static void
4604 // are listed in ip>addrs[]. The next NINDIRECT blocks are 4654 itrunc(struct inode *ip)
4605 // listed in the block ip>addrs[NDIRECT]. 4655 {
4606 4656 int i, j;
4607 // Return the disk block address of the nth block in inode ip. 4657 struct buf *bp;
4608 // If there is no such block, bmap allocates one. 4658 uint *a;
4609 static uint 4659
4610 bmap(struct inode *ip, uint bn) 4660 for(i = 0; i < NDIRECT; i++){
4611 { 4661 if(ip>addrs[i]){
4612 uint addr, *a; 4662 bfree(ip>dev, ip>addrs[i]);
4613 struct buf *bp; 4663 ip>addrs[i] = 0;
4614 4664 }
4615 if(bn < NDIRECT){ 4665 }
4616 if((addr = ip>addrs[bn]) == 0) 4666
4617 ip>addrs[bn] = addr = balloc(ip>dev); 4667 if(ip>addrs[NDIRECT]){
4618 return addr; 4668 bp = bread(ip>dev, ip>addrs[NDIRECT]);
4619 } 4669 a = (uint*)bp>data;
4620 bn = NDIRECT; 4670 for(j = 0; j < NINDIRECT; j++){
4621 4671 if(a[j])
4622 if(bn < NINDIRECT){ 4672 bfree(ip>dev, a[j]);
4623 // Load indirect block, allocating if necessary. 4673 }
4624 if((addr = ip>addrs[NDIRECT]) == 0) 4674 brelse(bp);
4625 ip>addrs[NDIRECT] = addr = balloc(ip>dev); 4675 bfree(ip>dev, ip>addrs[NDIRECT]);
4626 bp = bread(ip>dev, addr); 4676 ip>addrs[NDIRECT] = 0;
4627 a = (uint*)bp>data; 4677 }
4628 if((addr = a[bn]) == 0){ 4678
4629 a[bn] = addr = balloc(ip>dev); 4679 ip>size = 0;
4630 log_write(bp); 4680 iupdate(ip);
4631 } 4681 }
4632 brelse(bp); 4682
4633 return addr; 4683 // Copy stat information from inode.
4634 } 4684 void
4635 4685 stati(struct inode *ip, struct stat *st)
4636 panic("bmap: out of range"); 4686 {
4637 } 4687 st>dev = ip>dev;
4638 4688 st>ino = ip>inum;
4639 4689 st>type = ip>type;
4640 4690 st>nlink = ip>nlink;
4641 4691 st>size = ip>size;
4642 4692 }
4643 4693
4644 4694
4645 4695
4646 4696
4647 4697
4648 4698
4649 4699

Sheet 46 Sheet 46
Sep 5 23:39 2011 xv6/fs.c Page 10 Sep 5 23:39 2011 xv6/fs.c Page 11

4700 // Read data from inode. 4750 // Write data to inode.

4701 int 4751 int
4702 readi(struct inode *ip, char *dst, uint off, uint n) 4752 writei(struct inode *ip, char *src, uint off, uint n)
4703 { 4753 {
4704 uint tot, m; 4754 uint tot, m;
4705 struct buf *bp; 4755 struct buf *bp;
4706 4756
4707 if(ip>type == T_DEV){ 4757 if(ip>type == T_DEV){
4708 if(ip>major < 0 || ip>major >= NDEV || !devsw[ip>major].read) 4758 if(ip>major < 0 || ip>major >= NDEV || !devsw[ip>major].write)
4709 return 1; 4759 return 1;
4710 return devsw[ip>major].read(ip, dst, n); 4760 return devsw[ip>major].write(ip, src, n);
4711 } 4761 }
4712 4762
4713 if(off > ip>size || off + n < off) 4763 if(off > ip>size || off + n < off)
4714 return 1; 4764 return 1;
4715 if(off + n > ip>size) 4765 if(off + n > MAXFILE*BSIZE)
4716 n = ip>size off; 4766 return 1;
4717 4767
4718 for(tot=0; tot<n; tot+=m, off+=m, dst+=m){ 4768 for(tot=0; tot<n; tot+=m, off+=m, src+=m){
4719 bp = bread(ip>dev, bmap(ip, off/BSIZE)); 4769 bp = bread(ip>dev, bmap(ip, off/BSIZE));
4720 m = min(n tot, BSIZE off%BSIZE); 4770 m = min(n tot, BSIZE off%BSIZE);
4721 memmove(dst, bp>data + off%BSIZE, m); 4771 memmove(bp>data + off%BSIZE, src, m);
4722 brelse(bp); 4772 log_write(bp);
4723 } 4773 brelse(bp);
4724 return n; 4774 }
4725 } 4775
4726 4776 if(n > 0 && off > ip>size){
4727 4777 ip>size = off;
4728 4778 iupdate(ip);
4729 4779 }
4730 4780 return n;
4731 4781 }
4732 4782
4733 4783
4734 4784
4735 4785
4736 4786
4737 4787
4738 4788
4739 4789
4740 4790
4741 4791
4742 4792
4743 4793
4744 4794
4745 4795
4746 4796
4747 4797
4748 4798
4749 4799

Sheet 47 Sheet 47
Sep 5 23:39 2011 xv6/fs.c Page 12 Sep 5 23:39 2011 xv6/fs.c Page 13

4800 // Directories 4850 // Write a new directory entry (name, inum) into the directory dp.
4801 4851 int
4802 int 4852 dirlink(struct inode *dp, char *name, uint inum)
4803 namecmp(const char *s, const char *t) 4853 {
4804 { 4854 int off;
4805 return strncmp(s, t, DIRSIZ); 4855 struct dirent de;
4806 } 4856 struct inode *ip;
4807 4857
4808 // Look for a directory entry in a directory. 4858 // Check that name is not present.
4809 // If found, set *poff to byte offset of entry. 4859 if((ip = dirlookup(dp, name, 0)) != 0){
4810 // Caller must have already locked dp. 4860 iput(ip);
4811 struct inode* 4861 return 1;
4812 dirlookup(struct inode *dp, char *name, uint *poff) 4862 }
4813 { 4863
4814 uint off, inum; 4864 // Look for an empty dirent.
4815 struct dirent de; 4865 for(off = 0; off < dp>size; off += sizeof(de)){
4816 4866 if(readi(dp, (char*)&de, off, sizeof(de)) != sizeof(de))
4817 if(dp>type != T_DIR) 4867 panic("dirlink read");
4818 panic("dirlookup not DIR"); 4868 if(de.inum == 0)
4819 4869 break;
4820 for(off = 0; off < dp>size; off += sizeof(de)){ 4870 }
4821 if(readi(dp, (char*)&de, off, sizeof(de)) != sizeof(de)) 4871
4822 panic("dirlink read"); 4872 strncpy(de.name, name, DIRSIZ);
4823 if(de.inum == 0) 4873 de.inum = inum;
4824 continue; 4874 if(writei(dp, (char*)&de, off, sizeof(de)) != sizeof(de))
4825 if(namecmp(name, de.name) == 0){ 4875 panic("dirlink");
4826 // entry matches path element 4876
4827 if(poff) 4877 return 0;
4828 *poff = off; 4878 }
4829 inum = de.inum; 4879
4830 return iget(dp>dev, inum); 4880
4831 } 4881
4832 } 4882
4833 4883
4834 return 0; 4884
4835 } 4885
4836 4886
4837 4887
4838 4888
4839 4889
4840 4890
4841 4891
4842 4892
4843 4893
4844 4894
4845 4895
4846 4896
4847 4897
4848 4898
4849 4899

Sheet 48 Sheet 48
Sep 5 23:39 2011 xv6/fs.c Page 14 Sep 5 23:39 2011 xv6/fs.c Page 15

4900 // Paths 4950 // Look up and return the inode for a path name.
4901 4951 // If parent != 0, return the inode for the parent and copy the final
4902 // Copy the next path element from path into name. 4952 // path element into name, which must have room for DIRSIZ bytes.
4903 // Return a pointer to the element following the copied one. 4953 static struct inode*
4904 // The returned path has no leading slashes, 4954 namex(char *path, int nameiparent, char *name)
4905 // so the caller can check *path==\0 to see if the name is the last one. 4955 {
4906 // If no name to remove, return 0. 4956 struct inode *ip, *next;
4907 // 4957
4908 // Examples: 4958 if(*path == /)
4909 // skipelem("a/bb/c", name) = "bb/c", setting name = "a" 4959 ip = iget(ROOTDEV, ROOTINO);
4910 // skipelem("///a//bb", name) = "bb", setting name = "a" 4960 else
4911 // skipelem("a", name) = "", setting name = "a" 4961 ip = idup(proc>cwd);
4912 // skipelem("", name) = skipelem("////", name) = 0 4962
4913 // 4963 while((path = skipelem(path, name)) != 0){
4914 static char* 4964 ilock(ip);
4915 skipelem(char *path, char *name) 4965 if(ip>type != T_DIR){
4916 { 4966 iunlockput(ip);
4917 char *s; 4967 return 0;
4918 int len; 4968 }
4919 4969 if(nameiparent && *path == \0){
4920 while(*path == /) 4970 // Stop one level early.
4921 path++; 4971 iunlock(ip);
4922 if(*path == 0) 4972 return ip;
4923 return 0; 4973 }
4924 s = path; 4974 if((next = dirlookup(ip, name, 0)) == 0){
4925 while(*path != / && *path != 0) 4975 iunlockput(ip);
4926 path++; 4976 return 0;
4927 len = path s; 4977 }
4928 if(len >= DIRSIZ) 4978 iunlockput(ip);
4929 memmove(name, s, DIRSIZ); 4979 ip = next;
4930 else { 4980 }
4931 memmove(name, s, len); 4981 if(nameiparent){
4932 name[len] = 0; 4982 iput(ip);
4933 } 4983 return 0;
4934 while(*path == /) 4984 }
4935 path++; 4985 return ip;
4936 return path; 4986 }
4937 } 4987
4938 4988 struct inode*
4939 4989 namei(char *path)
4940 4990 {
4941 4991 char name[DIRSIZ];
4942 4992 return namex(path, 0, name);
4943 4993 }
4944 4994
4945 4995 struct inode*
4946 4996 nameiparent(char *path, char *name)
4947 4997 {
4948 4998 return namex(path, 1, name);
4949 4999 }

Sheet 49 Sheet 49
Sep 5 23:39 2011 xv6/file.c Page 1 Sep 5 23:39 2011 xv6/file.c Page 2

5000 #include "types.h" 5050 // Close file f. (Decrement ref count, close when reaches 0.)
5001 #include "defs.h" 5051 void
5002 #include "param.h" 5052 fileclose(struct file *f)
5003 #include "fs.h" 5053 {
5004 #include "file.h" 5054 struct file ff;
5005 #include "spinlock.h" 5055
5006 5056 acquire(&ftable.lock);
5007 struct devsw devsw[NDEV]; 5057 if(f>ref < 1)
5008 struct { 5058 panic("fileclose");
5009 struct spinlock lock; 5059 if(f>ref > 0){
5010 struct file file[NFILE]; 5060 release(&ftable.lock);
5011 } ftable; 5061 return;
5012 5062 }
5013 void 5063 ff = *f;
5014 fileinit(void) 5064 f>ref = 0;
5015 { 5065 f>type = FD_NONE;
5016 initlock(&ftable.lock, "ftable"); 5066 release(&ftable.lock);
5017 } 5067
5018 5068 if(ff.type == FD_PIPE)
5019 // Allocate a file structure. 5069 pipeclose(ff.pipe, ff.writable);
5020 struct file* 5070 else if(ff.type == FD_INODE){
5021 filealloc(void) 5071 begin_trans();
5022 { 5072 iput(ff.ip);
5023 struct file *f; 5073 commit_trans();
5024 5074 }
5025 acquire(&ftable.lock); 5075 }
5026 for(f = ftable.file; f < ftable.file + NFILE; f++){ 5076
5027 if(f>ref == 0){ 5077 // Get metadata about file f.
5028 f>ref = 1; 5078 int
5029 release(&ftable.lock); 5079 filestat(struct file *f, struct stat *st)
5030 return f; 5080 {
5031 } 5081 if(f>type == FD_INODE){
5032 } 5082 ilock(f>ip);
5033 release(&ftable.lock); 5083 stati(f>ip, st);
5034 return 0; 5084 iunlock(f>ip);
5035 } 5085 return 0;
5036 5086 }
5037 // Increment ref count for file f. 5087 return 1;
5038 struct file* 5088 }
5039 filedup(struct file *f) 5089
5040 { 5090
5041 acquire(&ftable.lock); 5091
5042 if(f>ref < 1) 5092
5043 panic("filedup"); 5093
5044 f>ref++; 5094
5045 release(&ftable.lock); 5095
5046 return f; 5096
5047 } 5097
5048 5098
5049 5099

Sheet 50 Sheet 50
Sep 5 23:39 2011 xv6/file.c Page 3 Sep 5 23:39 2011 xv6/file.c Page 4

5100 // Read from file f. Addr is kernel address. 5150 // Write to file f. Addr is kernel address.
5101 int 5151 int
5102 fileread(struct file *f, char *addr, int n) 5152 filewrite(struct file *f, char *addr, int n)
5103 { 5153 {
5104 int r; 5154 int r;
5105 5155
5106 if(f>readable == 0) 5156 if(f>writable == 0)
5107 return 1; 5157 return 1;
5108 if(f>type == FD_PIPE) 5158 if(f>type == FD_PIPE)
5109 return piperead(f>pipe, addr, n); 5159 return pipewrite(f>pipe, addr, n);
5110 if(f>type == FD_INODE){ 5160 if(f>type == FD_INODE){
5111 ilock(f>ip); 5161 // write a few blocks at a time to avoid exceeding
5112 if((r = readi(f>ip, addr, f>off, n)) > 0) 5162 // the maximum log transaction size, including
5113 f>off += r; 5163 // inode, indirect block, allocation blocks,
5114 iunlock(f>ip); 5164 // and 2 blocks of slop for nonaligned writes.
5115 return r; 5165 // this really belongs lower down, since writei()
5116 } 5166 // might be writing a device like the console.
5117 panic("fileread"); 5167 int max = ((LOGSIZE112) / 2) * 512;
5118 } 5168 int i = 0;
5119 5169 while(i < n){
5120 5170 int n1 = n i;
5121 5171 if(n1 > max)
5122 5172 n1 = max;
5123 5173
5124 5174 begin_trans();
5125 5175 ilock(f>ip);
5126 5176 if ((r = writei(f>ip, addr + i, f>off, n1)) > 0)
5127 5177 f>off += r;
5128 5178 iunlock(f>ip);
5129 5179 commit_trans();
5130 5180
5131 5181 if(r < 0)
5132 5182 break;
5133 5183 if(r != n1)
5134 5184 panic("short filewrite");
5135 5185 i += r;
5136 5186 }
5137 5187 return i == n ? n : 1;
5138 5188 }
5139 5189 panic("filewrite");
5140 5190 }
5141 5191
5142 5192
5143 5193
5144 5194
5145 5195
5146 5196
5147 5197
5148 5198
5149 5199

Sheet 51 Sheet 51
Sep 5 23:39 2011 xv6/sysfile.c Page 1 Sep 5 23:39 2011 xv6/sysfile.c Page 2

5200 #include "types.h" 5250 int

5201 #include "defs.h" 5251 sys_dup(void)
5202 #include "param.h" 5252 {
5203 #include "stat.h" 5253 struct file *f;
5204 #include "mmu.h" 5254 int fd;
5205 #include "proc.h" 5255
5206 #include "fs.h" 5256 if(argfd(0, 0, &f) < 0)
5207 #include "file.h" 5257 return 1;
5208 #include "fcntl.h" 5258 if((fd=fdalloc(f)) < 0)
5209 5259 return 1;
5210 // Fetch the nth wordsized system call argument as a file descriptor 5260 filedup(f);
5211 // and return both the descriptor and the corresponding struct file. 5261 return fd;
5212 static int 5262 }
5213 argfd(int n, int *pfd, struct file **pf) 5263
5214 { 5264 int
5215 int fd; 5265 sys_read(void)
5216 struct file *f; 5266 {
5217 5267 struct file *f;
5218 if(argint(n, &fd) < 0) 5268 int n;
5219 return 1; 5269 char *p;
5220 if(fd < 0 || fd >= NOFILE || (f=proc>ofile[fd]) == 0) 5270
5221 return 1; 5271 if(argfd(0, 0, &f) < 0 || argint(2, &n) < 0 || argptr(1, &p, n) < 0)
5222 if(pfd) 5272 return 1;
5223 *pfd = fd; 5273 return fileread(f, p, n);
5224 if(pf) 5274 }
5225 *pf = f; 5275
5226 return 0; 5276 int
5227 } 5277 sys_write(void)
5228 5278 {
5229 // Allocate a file descriptor for the given file. 5279 struct file *f;
5230 // Takes over file reference from caller on success. 5280 int n;
5231 static int 5281 char *p;
5232 fdalloc(struct file *f) 5282
5233 { 5283 if(argfd(0, 0, &f) < 0 || argint(2, &n) < 0 || argptr(1, &p, n) < 0)
5234 int fd; 5284 return 1;
5235 5285 return filewrite(f, p, n);
5236 for(fd = 0; fd < NOFILE; fd++){ 5286 }
5237 if(proc>ofile[fd] == 0){ 5287
5238 proc>ofile[fd] = f; 5288 int
5239 return fd; 5289 sys_close(void)
5240 } 5290 {
5241 } 5291 int fd;
5242 return 1; 5292 struct file *f;
5243 } 5293
5244 5294 if(argfd(0, &fd, &f) < 0)
5245 5295 return 1;
5246 5296 proc>ofile[fd] = 0;
5247 5297 fileclose(f);
5248 5298 return 0;
5249 5299 }

Sheet 52 Sheet 52
Sep 5 23:39 2011 xv6/sysfile.c Page 3 Sep 5 23:39 2011 xv6/sysfile.c Page 4

5300 int 5350 bad:

5301 sys_fstat(void) 5351 ilock(ip);
5302 { 5352 ip>nlink;
5303 struct file *f; 5353 iupdate(ip);
5304 struct stat *st; 5354 iunlockput(ip);
5305 5355 commit_trans();
5306 if(argfd(0, 0, &f) < 0 || argptr(1, (void*)&st, sizeof(*st)) < 0) 5356 return 1;
5307 return 1; 5357 }
5308 return filestat(f, st); 5358
5309 } 5359 // Is the directory dp empty except for "." and ".." ?
5310 5360 static int
5311 // Create the path new as a link to the same inode as old. 5361 isdirempty(struct inode *dp)
5312 int 5362 {
5313 sys_link(void) 5363 int off;
5314 { 5364 struct dirent de;
5315 char name[DIRSIZ], *new, *old; 5365
5316 struct inode *dp, *ip; 5366 for(off=2*sizeof(de); off<dp>size; off+=sizeof(de)){
5317 5367 if(readi(dp, (char*)&de, off, sizeof(de)) != sizeof(de))
5318 if(argstr(0, &old) < 0 || argstr(1, &new) < 0) 5368 panic("isdirempty: readi");
5319 return 1; 5369 if(de.inum != 0)
5320 if((ip = namei(old)) == 0) 5370 return 0;
5321 return 1; 5371 }
5322 5372 return 1;
5323 begin_trans(); 5373 }
5324 5374
5325 ilock(ip); 5375
5326 if(ip>type == T_DIR){ 5376
5327 iunlockput(ip); 5377
5328 commit_trans(); 5378
5329 return 1; 5379
5330 } 5380
5331 5381
5332 ip>nlink++; 5382
5333 iupdate(ip); 5383
5334 iunlock(ip); 5384
5335 5385
5336 if((dp = nameiparent(new, name)) == 0) 5386
5337 goto bad; 5387
5338 ilock(dp); 5388
5339 if(dp>dev != ip>dev || dirlink(dp, name, ip>inum) < 0){ 5389
5340 iunlockput(dp); 5390
5341 goto bad; 5391
5342 } 5392
5343 iunlockput(dp); 5393
5344 iput(ip); 5394
5345 5395
5346 commit_trans(); 5396
5347 5397
5348 return 0; 5398
5349 5399

Sheet 53 Sheet 53
Sep 5 23:39 2011 xv6/sysfile.c Page 5 Sep 5 23:39 2011 xv6/sysfile.c Page 6

5400 int 5450 bad:

5401 sys_unlink(void) 5451 iunlockput(dp);
5402 { 5452 commit_trans();
5403 struct inode *ip, *dp; 5453 return 1;
5404 struct dirent de; 5454 }
5405 char name[DIRSIZ], *path; 5455
5406 uint off; 5456 static struct inode*
5407 5457 create(char *path, short type, short major, short minor)
5408 if(argstr(0, &path) < 0) 5458 {
5409 return 1; 5459 uint off;
5410 if((dp = nameiparent(path, name)) == 0) 5460 struct inode *ip, *dp;
5411 return 1; 5461 char name[DIRSIZ];
5412 5462
5413 begin_trans(); 5463 if((dp = nameiparent(path, name)) == 0)
5414 5464 return 0;
5415 ilock(dp); 5465 ilock(dp);
5416 5466
5417 // Cannot unlink "." or "..". 5467 if((ip = dirlookup(dp, name, &off)) != 0){
5418 if(namecmp(name, ".") == 0 || namecmp(name, "..") == 0) 5468 iunlockput(dp);
5419 goto bad; 5469 ilock(ip);
5420 5470 if(type == T_FILE && ip>type == T_FILE)
5421 if((ip = dirlookup(dp, name, &off)) == 0) 5471 return ip;
5422 goto bad; 5472 iunlockput(ip);
5423 ilock(ip); 5473 return 0;
5424 5474 }
5425 if(ip>nlink < 1) 5475
5426 panic("unlink: nlink < 1"); 5476 if((ip = ialloc(dp>dev, type)) == 0)
5427 if(ip>type == T_DIR && !isdirempty(ip)){ 5477 panic("create: ialloc");
5428 iunlockput(ip); 5478
5429 goto bad; 5479 ilock(ip);
5430 } 5480 ip>major = major;
5431 5481 ip>minor = minor;
5432 memset(&de, 0, sizeof(de)); 5482 ip>nlink = 1;
5433 if(writei(dp, (char*)&de, off, sizeof(de)) != sizeof(de)) 5483 iupdate(ip);
5434 panic("unlink: writei"); 5484
5435 if(ip>type == T_DIR){ 5485 if(type == T_DIR){ // Create . and .. entries.
5436 dp>nlink; 5486 dp>nlink++; // for ".."
5437 iupdate(dp); 5487 iupdate(dp);
5438 } 5488 // No ip>nlink++ for ".": avoid cyclic ref count.
5439 iunlockput(dp); 5489 if(dirlink(ip, ".", ip>inum) < 0 || dirlink(ip, "..", dp>inum) < 0)
5440 5490 panic("create dots");
5441 ip>nlink; 5491 }
5442 iupdate(ip); 5492
5443 iunlockput(ip); 5493 if(dirlink(dp, name, ip>inum) < 0)
5444 5494 panic("create: dirlink");
5445 commit_trans(); 5495
5446 5496 iunlockput(dp);
5447 return 0; 5497
5448 5498 return ip;
5449 5499 }

Sheet 54 Sheet 54
Sep 5 23:39 2011 xv6/sysfile.c Page 7 Sep 5 23:39 2011 xv6/sysfile.c Page 8

5500 int 5550 int

5501 sys_open(void) 5551 sys_mkdir(void)
5502 { 5552 {
5503 char *path; 5553 char *path;
5504 int fd, omode; 5554 struct inode *ip;
5505 struct file *f; 5555
5506 struct inode *ip; 5556 begin_trans();
5507 5557 if(argstr(0, &path) < 0 || (ip = create(path, T_DIR, 0, 0)) == 0){
5508 if(argstr(0, &path) < 0 || argint(1, &omode) < 0) 5558 commit_trans();
5509 return 1; 5559 return 1;
5510 if(omode & O_CREATE){ 5560 }
5511 begin_trans(); 5561 iunlockput(ip);
5512 ip = create(path, T_FILE, 0, 0); 5562 commit_trans();
5513 commit_trans(); 5563 return 0;
5514 if(ip == 0) 5564 }
5515 return 1; 5565
5516 } else { 5566 int
5517 if((ip = namei(path)) == 0) 5567 sys_mknod(void)
5518 return 1; 5568 {
5519 ilock(ip); 5569 struct inode *ip;
5520 if(ip>type == T_DIR && omode != O_RDONLY){ 5570 char *path;
5521 iunlockput(ip); 5571 int len;
5522 return 1; 5572 int major, minor;
5523 } 5573
5524 } 5574 begin_trans();
5525 5575 if((len=argstr(0, &path)) < 0 ||
5526 if((f = filealloc()) == 0 || (fd = fdalloc(f)) < 0){ 5576 argint(1, &major) < 0 ||
5527 if(f) 5577 argint(2, &minor) < 0 ||
5528 fileclose(f); 5578 (ip = create(path, T_DEV, major, minor)) == 0){
5529 iunlockput(ip); 5579 commit_trans();
5530 return 1; 5580 return 1;
5531 } 5581 }
5532 iunlock(ip); 5582 iunlockput(ip);
5533 5583 commit_trans();
5534 f>type = FD_INODE; 5584 return 0;
5535 f>ip = ip; 5585 }
5536 f>off = 0; 5586
5537 f>readable = !(omode & O_WRONLY); 5587
5538 f>writable = (omode & O_WRONLY) || (omode & O_RDWR); 5588
5539 return fd; 5589
5540 } 5590
5541 5591
5542 5592
5543 5593
5544 5594
5545 5595
5546 5596
5547 5597
5548 5598
5549 5599

Sheet 55 Sheet 55
Sep 5 23:39 2011 xv6/sysfile.c Page 9 Sep 5 23:39 2011 xv6/sysfile.c Page 10

5600 int 5650 int

5601 sys_chdir(void) 5651 sys_pipe(void)
5602 { 5652 {
5603 char *path; 5653 int *fd;
5604 struct inode *ip; 5654 struct file *rf, *wf;
5605 5655 int fd0, fd1;
5606 if(argstr(0, &path) < 0 || (ip = namei(path)) == 0) 5656
5607 return 1; 5657 if(argptr(0, (void*)&fd, 2*sizeof(fd[0])) < 0)
5608 ilock(ip); 5658 return 1;
5609 if(ip>type != T_DIR){ 5659 if(pipealloc(&rf, &wf) < 0)
5610 iunlockput(ip); 5660 return 1;
5611 return 1; 5661 fd0 = 1;
5612 } 5662 if((fd0 = fdalloc(rf)) < 0 || (fd1 = fdalloc(wf)) < 0){
5613 iunlock(ip); 5663 if(fd0 >= 0)
5614 iput(proc>cwd); 5664 proc>ofile[fd0] = 0;
5615 proc>cwd = ip; 5665 fileclose(rf);
5616 return 0; 5666 fileclose(wf);
5617 } 5667 return 1;
5618 5668 }
5619 int 5669 fd[0] = fd0;
5620 sys_exec(void) 5670 fd[1] = fd1;
5621 { 5671 return 0;
5622 char *path, *argv[MAXARG]; 5672 }
5623 int i; 5673
5624 uint uargv, uarg; 5674
5625 5675
5626 if(argstr(0, &path) < 0 || argint(1, (int*)&uargv) < 0){ 5676
5627 return 1; 5677
5628 } 5678
5629 memset(argv, 0, sizeof(argv)); 5679
5630 for(i=0;; i++){ 5680
5631 if(i >= NELEM(argv)) 5681
5632 return 1; 5682
5633 if(fetchint(proc, uargv+4*i, (int*)&uarg) < 0) 5683
5634 return 1; 5684
5635 if(uarg == 0){ 5685
5636 argv[i] = 0; 5686
5637 break; 5687
5638 } 5688
5639 if(fetchstr(proc, uarg, &argv[i]) < 0) 5689
5640 return 1; 5690
5641 } 5691
5642 return exec(path, argv); 5692
5643 } 5693
5644 5694
5645 5695
5646 5696
5647 5697
5648 5698
5649 5699

Sheet 56 Sheet 56
Sep 5 23:39 2011 xv6/exec.c Page 1 Sep 5 23:39 2011 xv6/exec.c Page 2

5700 #include "types.h" 5750 // Allocate two pages at the next page boundary.
5701 #include "param.h" 5751 // Make the first inaccessible. Use the second as the user stack.
5702 #include "memlayout.h" 5752 sz = PGROUNDUP(sz);
5703 #include "mmu.h" 5753 if((sz = allocuvm(pgdir, sz, sz + 2*PGSIZE)) == 0)
5704 #include "proc.h" 5754 goto bad;
5705 #include "defs.h" 5755 clearpteu(pgdir, (char*)(sz 2*PGSIZE));
5706 #include "x86.h" 5756 sp = sz;
5707 #include "elf.h" 5757
5708 5758 // Push argument strings, prepare rest of stack in ustack.
5709 int 5759 for(argc = 0; argv[argc]; argc++) {
5710 exec(char *path, char **argv) 5760 if(argc >= MAXARG)
5711 { 5761 goto bad;
5712 char *s, *last; 5762 sp = (sp (strlen(argv[argc]) + 1)) & ~3;
5713 int i, off; 5763 if(copyout(pgdir, sp, argv[argc], strlen(argv[argc]) + 1) < 0)
5714 uint argc, sz, sp, ustack[3+MAXARG+1]; 5764 goto bad;
5715 struct elfhdr elf; 5765 ustack[3+argc] = sp;
5716 struct inode *ip; 5766 }
5717 struct proghdr ph; 5767 ustack[3+argc] = 0;
5718 pde_t *pgdir, *oldpgdir; 5768
5719 5769 ustack[0] = 0xffffffff; // fake return PC
5720 if((ip = namei(path)) == 0) 5770 ustack[1] = argc;
5721 return 1; 5771 ustack[2] = sp (argc+1)*4; // argv pointer
5722 ilock(ip); 5772
5723 pgdir = 0; 5773 sp = (3+argc+1) * 4;
5724 5774 if(copyout(pgdir, sp, ustack, (3+argc+1)*4) < 0)
5725 // Check ELF header 5775 goto bad;
5726 if(readi(ip, (char*)&elf, 0, sizeof(elf)) < sizeof(elf)) 5776
5727 goto bad; 5777 // Save program name for debugging.
5728 if(elf.magic != ELF_MAGIC) 5778 for(last=s=path; *s; s++)
5729 goto bad; 5779 if(*s == /)
5730 5780 last = s+1;
5731 if((pgdir = setupkvm(kalloc)) == 0) 5781 safestrcpy(proc>name, last, sizeof(proc>name));
5732 goto bad; 5782
5733 5783 // Commit to the user image.
5734 // Load program into memory. 5784 oldpgdir = proc>pgdir;
5735 sz = 0; 5785 proc>pgdir = pgdir;
5736 for(i=0, off=elf.phoff; i<elf.phnum; i++, off+=sizeof(ph)){ 5786 proc>sz = sz;
5737 if(readi(ip, (char*)&ph, off, sizeof(ph)) != sizeof(ph)) 5787 proc>tf>eip = elf.entry; // main
5738 goto bad; 5788 proc>tf>esp = sp;
5739 if(ph.type != ELF_PROG_LOAD) 5789 switchuvm(proc);
5740 continue; 5790 freevm(oldpgdir);
5741 if(ph.memsz < ph.filesz) 5791 return 0;
5742 goto bad; 5792
5743 if((sz = allocuvm(pgdir, sz, ph.vaddr + ph.memsz)) == 0) 5793 bad:
5744 goto bad; 5794 if(pgdir)
5745 if(loaduvm(pgdir, (char*)ph.vaddr, ip, ph.off, ph.filesz) < 0) 5795 freevm(pgdir);
5746 goto bad; 5796 if(ip)
5747 } 5797 iunlockput(ip);
5748 iunlockput(ip); 5798 return 1;
5749 ip = 0; 5799 }

Sheet 57 Sheet 57
Sep 5 23:39 2011 xv6/pipe.c Page 1 Sep 5 23:39 2011 xv6/pipe.c Page 2

5800 #include "types.h" 5850 bad:

5801 #include "defs.h" 5851 if(p)
5802 #include "param.h" 5852 kfree((char*)p);
5803 #include "mmu.h" 5853 if(*f0)
5804 #include "proc.h" 5854 fileclose(*f0);
5805 #include "fs.h" 5855 if(*f1)
5806 #include "file.h" 5856 fileclose(*f1);
5807 #include "spinlock.h" 5857 return 1;
5808 5858 }
5809 #define PIPESIZE 512 5859
5810 5860 void
5811 struct pipe { 5861 pipeclose(struct pipe *p, int writable)
5812 struct spinlock lock; 5862 {
5813 char data[PIPESIZE]; 5863 acquire(&p>lock);
5814 uint nread; // number of bytes read 5864 if(writable){
5815 uint nwrite; // number of bytes written 5865 p>writeopen = 0;
5816 int readopen; // read fd is still open 5866 wakeup(&p>nread);
5817 int writeopen; // write fd is still open 5867 } else {
5818 }; 5868 p>readopen = 0;
5819 5869 wakeup(&p>nwrite);
5820 int 5870 }
5821 pipealloc(struct file **f0, struct file **f1) 5871 if(p>readopen == 0 && p>writeopen == 0){
5822 { 5872 release(&p>lock);
5823 struct pipe *p; 5873 kfree((char*)p);
5824 5874 } else
5825 p = 0; 5875 release(&p>lock);
5826 *f0 = *f1 = 0; 5876 }
5827 if((*f0 = filealloc()) == 0 || (*f1 = filealloc()) == 0) 5877
5828 goto bad; 5878
5829 if((p = (struct pipe*)kalloc()) == 0) 5879 int
5830 goto bad; 5880 pipewrite(struct pipe *p, char *addr, int n)
5831 p>readopen = 1; 5881 {
5832 p>writeopen = 1; 5882 int i;
5833 p>nwrite = 0; 5883
5834 p>nread = 0; 5884 acquire(&p>lock);
5835 initlock(&p>lock, "pipe"); 5885 for(i = 0; i < n; i++){
5836 (*f0)>type = FD_PIPE; 5886 while(p>nwrite == p>nread + PIPESIZE){
5837 (*f0)>readable = 1; 5887 if(p>readopen == 0 || proc>killed){
5838 (*f0)>writable = 0; 5888 release(&p>lock);
5839 (*f0)>pipe = p; 5889 return 1;
5840 (*f1)>type = FD_PIPE; 5890 }
5841 (*f1)>readable = 0; 5891 wakeup(&p>nread);
5842 (*f1)>writable = 1; 5892 sleep(&p>nwrite, &p>lock);
5843 (*f1)>pipe = p; 5893 }
5844 return 0; 5894 p>data[p>nwrite++ % PIPESIZE] = addr[i];
5845 5895 }
5846 5896 wakeup(&p>nread);
5847 5897 release(&p>lock);
5848 5898 return n;
5849 5899 }

Sheet 58 Sheet 58
Sep 5 23:39 2011 xv6/pipe.c Page 3 Sep 5 23:39 2011 xv6/string.c Page 1

5900 int 5950 #include "types.h"

5901 piperead(struct pipe *p, char *addr, int n) 5951 #include "x86.h"
5902 { 5952
5903 int i; 5953 void*
5904 5954 memset(void *dst, int c, uint n)
5905 acquire(&p>lock); 5955 {
5906 while(p>nread == p>nwrite && p>writeopen){ 5956 if ((int)dst%4 == 0 && n%4 == 0){
5907 if(proc>killed){ 5957 c &= 0xFF;
5908 release(&p>lock); 5958 stosl(dst, (c<<24)|(c<<16)|(c<<8)|c, n/4);
5909 return 1; 5959 } else
5910 } 5960 stosb(dst, c, n);
5911 sleep(&p>nread, &p>lock); 5961 return dst;
5912 } 5962 }
5913 for(i = 0; i < n; i++){ 5963
5914 if(p>nread == p>nwrite) 5964 int
5915 break; 5965 memcmp(const void *v1, const void *v2, uint n)
5916 addr[i] = p>data[p>nread++ % PIPESIZE]; 5966 {
5917 } 5967 const uchar *s1, *s2;
5918 wakeup(&p>nwrite); 5968
5919 release(&p>lock); 5969 s1 = v1;
5920 return i; 5970 s2 = v2;
5921 } 5971 while(n > 0){
5922 5972 if(*s1 != *s2)
5923 5973 return *s1 *s2;
5924 5974 s1++, s2++;
5925 5975 }
5926 5976
5927 5977 return 0;
5928 5978 }
5929 5979
5930 5980 void*
5931 5981 memmove(void *dst, const void *src, uint n)
5932 5982 {
5933 5983 const char *s;
5934 5984 char *d;
5935 5985
5936 5986 s = src;
5937 5987 d = dst;
5938 5988 if(s < d && s + n > d){
5939 5989 s += n;
5940 5990 d += n;
5941 5991 while(n > 0)
5942 5992 *d = *s;
5943 5993 } else
5944 5994 while(n > 0)
5945 5995 *d++ = *s++;
5946 5996
5947 5997 return dst;
5948 5998 }
5949 5999

Sheet 59 Sheet 59
Sep 5 23:39 2011 xv6/string.c Page 2 Sep 5 23:39 2011 xv6/string.c Page 3

6000 // memcpy exists to placate GCC. Use memmove. 6050 int

6001 void* 6051 strlen(const char *s)
6002 memcpy(void *dst, const void *src, uint n) 6052 {
6003 { 6053 int n;
6004 return memmove(dst, src, n); 6054
6005 } 6055 for(n = 0; s[n]; n++)
6006 6056 ;
6007 int 6057 return n;
6008 strncmp(const char *p, const char *q, uint n) 6058 }
6009 { 6059
6010 while(n > 0 && *p && *p == *q) 6060
6011 n, p++, q++; 6061
6012 if(n == 0) 6062
6013 return 0; 6063
6014 return (uchar)*p (uchar)*q; 6064
6015 } 6065
6016 6066
6017 char* 6067
6018 strncpy(char *s, const char *t, int n) 6068
6019 { 6069
6020 char *os; 6070
6021 6071
6022 os = s; 6072
6023 while(n > 0 && (*s++ = *t++) != 0) 6073
6024 ; 6074
6025 while(n > 0) 6075
6026 *s++ = 0; 6076
6027 return os; 6077
6028 } 6078
6029 6079
6030 // Like strncpy but guaranteed to NULterminate. 6080
6031 char* 6081
6032 safestrcpy(char *s, const char *t, int n) 6082
6033 { 6083
6034 char *os; 6084
6035 6085
6036 os = s; 6086
6037 if(n <= 0) 6087
6038 return os; 6088
6039 while(n > 0 && (*s++ = *t++) != 0) 6089
6040 ; 6090
6041 *s = 0; 6091
6042 return os; 6092
6043 } 6093
6044 6094
6045 6095
6046 6096
6047 6097
6048 6098
6049 6099

Sheet 60 Sheet 60
Sep 5 23:39 2011 xv6/mp.h Page 1 Sep 5 23:39 2011 xv6/mp.h Page 2

6100 // See MultiProcessor Specification Version 1.[14] 6150 // Table entry types
6101 6151 #define MPPROC 0x00 // One per processor
6102 struct mp { // floating pointer 6152 #define MPBUS 0x01 // One per bus
6103 uchar signature[4]; // "_MP_" 6153 #define MPIOAPIC 0x02 // One per I/O APIC
6104 void *physaddr; // phys addr of MP config table 6154 #define MPIOINTR 0x03 // One per bus interrupt source
6105 uchar length; // 1 6155 #define MPLINTR 0x04 // One per system interrupt source
6106 uchar specrev; // [14] 6156
6107 uchar checksum; // all bytes must add up to 0 6157
6108 uchar type; // MP system config type 6158
6109 uchar imcrp; 6159
6110 uchar reserved[3]; 6160
6111 }; 6161
6112 6162
6113 struct mpconf { // configuration table header 6163
6114 uchar signature[4]; // "PCMP" 6164
6115 ushort length; // total table length 6165
6116 uchar version; // [14] 6166
6117 uchar checksum; // all bytes must add up to 0 6167
6118 uchar product[20]; // product id 6168
6119 uint *oemtable; // OEM table pointer 6169
6120 ushort oemlength; // OEM table length 6170
6121 ushort entry; // entry count 6171
6122 uint *lapicaddr; // address of local APIC 6172
6123 ushort xlength; // extended table length 6173
6124 uchar xchecksum; // extended table checksum 6174
6125 uchar reserved; 6175
6126 }; 6176
6127 6177
6128 struct mpproc { // processor table entry 6178
6129 uchar type; // entry type (0) 6179
6130 uchar apicid; // local APIC id 6180
6131 uchar version; // local APIC verison 6181
6132 uchar flags; // CPU flags 6182
6133 #define MPBOOT 0x02 // This proc is the bootstrap processor. 6183
6134 uchar signature[4]; // CPU signature 6184
6135 uint feature; // feature flags from CPUID instruction 6185
6136 uchar reserved[8]; 6186
6137 }; 6187
6138 6188
6139 struct mpioapic { // I/O APIC table entry 6189
6140 uchar type; // entry type (2) 6190
6141 uchar apicno; // I/O APIC id 6191
6142 uchar version; // I/O APIC version 6192
6143 uchar flags; // I/O APIC flags 6193
6144 uint *addr; // I/O APIC address 6194
6145 }; 6195
6146 6196
6147 6197
6148 6198
6149 6199

Sheet 61 Sheet 61
Sep 5 23:39 2011 xv6/mp.c Page 1 Sep 5 23:39 2011 xv6/mp.c Page 2

6200 // Multiprocessor support 6250 // Search for the MP Floating Pointer Structure, which according to the
6201 // Search memory for MP description structures. 6251 // spec is in one of the following three locations:
6202 // http://developer.intel.com/design/pentium/datashts/24201606.pdf 6252 // 1) in the first KB of the EBDA;
6203 6253 // 2) in the last KB of system base memory;
6204 #include "types.h" 6254 // 3) in the BIOS ROM between 0xE0000 and 0xFFFFF.
6205 #include "defs.h" 6255 static struct mp*
6206 #include "param.h" 6256 mpsearch(void)
6207 #include "memlayout.h" 6257 {
6208 #include "mp.h" 6258 uchar *bda;
6209 #include "x86.h" 6259 uint p;
6210 #include "mmu.h" 6260 struct mp *mp;
6211 #include "proc.h" 6261
6212 6262 bda = (uchar *) P2V(0x400);
6213 struct cpu cpus[NCPU]; 6263 if((p = ((bda[0x0F]<<8)| bda[0x0E]) << 4)){
6214 static struct cpu *bcpu; 6264 if((mp = mpsearch1(p, 1024)))
6215 int ismp; 6265 return mp;
6216 int ncpu; 6266 } else {
6217 uchar ioapicid; 6267 p = ((bda[0x14]<<8)|bda[0x13])*1024;
6218 6268 if((mp = mpsearch1(p1024, 1024)))
6219 int 6269 return mp;
6220 mpbcpu(void) 6270 }
6221 { 6271 return mpsearch1(0xF0000, 0x10000);
6222 return bcpucpus; 6272 }
6223 } 6273
6224 6274 // Search for an MP configuration table. For now,
6225 static uchar 6275 // dont accept the default configurations (physaddr == 0).
6226 sum(uchar *addr, int len) 6276 // Check for correct signature, calculate the checksum and,
6227 { 6277 // if correct, check the version.
6228 int i, sum; 6278 // To do: check extended table checksum.
6229 6279 static struct mpconf*
6230 sum = 0; 6280 mpconfig(struct mp **pmp)
6231 for(i=0; i<len; i++) 6281 {
6232 sum += addr[i]; 6282 struct mpconf *conf;
6233 return sum; 6283 struct mp *mp;
6234 } 6284
6235 6285 if((mp = mpsearch()) == 0 || mp>physaddr == 0)
6236 // Look for an MP structure in the len bytes at addr. 6286 return 0;
6237 static struct mp* 6287 conf = (struct mpconf*) p2v((uint) mp>physaddr);
6238 mpsearch1(uint a, int len) 6288 if(memcmp(conf, "PCMP", 4) != 0)
6239 { 6289 return 0;
6240 uchar *e, *p, *addr; 6290 if(conf>version != 1 && conf>version != 4)
6241 6291 return 0;
6242 addr = p2v(a); 6292 if(sum((uchar*)conf, conf>length) != 0)
6243 e = addr+len; 6293 return 0;
6244 for(p = addr; p < e; p += sizeof(struct mp)) 6294 *pmp = mp;
6245 if(memcmp(p, "_MP_", 4) == 0 && sum(p, sizeof(struct mp)) == 0) 6295 return conf;
6246 return (struct mp*)p; 6296 }
6247 return 0; 6297
6248 } 6298
6249 6299

Sheet 62 Sheet 62
Sep 5 23:39 2011 xv6/mp.c Page 3 Sep 5 23:39 2011 xv6/mp.c Page 4

6300 void 6350 if(mp>imcrp){

6301 mpinit(void) 6351 // Bochs doesnt support IMCR, so this doesnt run on Bochs.
6302 { 6352 // But it would on real hardware.
6303 uchar *p, *e; 6353 outb(0x22, 0x70); // Select IMCR
6304 struct mp *mp; 6354 outb(0x23, inb(0x23) | 1); // Mask external interrupts.
6305 struct mpconf *conf; 6355 }
6306 struct mpproc *proc; 6356 }
6307 struct mpioapic *ioapic; 6357
6308 6358
6309 bcpu = &cpus[0]; 6359
6310 if((conf = mpconfig(&mp)) == 0) 6360
6311 return; 6361
6312 ismp = 1; 6362
6313 lapic = (uint*)conf>lapicaddr; 6363
6314 for(p=(uchar*)(conf+1), e=(uchar*)conf+conf>length; p<e; ){ 6364
6315 switch(*p){ 6365
6316 case MPPROC: 6366
6317 proc = (struct mpproc*)p; 6367
6318 if(ncpu != proc>apicid){ 6368
6319 cprintf("mpinit: ncpu=%d apicid=%d\n", ncpu, proc>apicid); 6369
6320 ismp = 0; 6370
6321 } 6371
6322 if(proc>flags & MPBOOT) 6372
6323 bcpu = &cpus[ncpu]; 6373
6324 cpus[ncpu].id = ncpu; 6374
6325 ncpu++; 6375
6326 p += sizeof(struct mpproc); 6376
6327 continue; 6377
6328 case MPIOAPIC: 6378
6329 ioapic = (struct mpioapic*)p; 6379
6330 ioapicid = ioapic>apicno; 6380
6331 p += sizeof(struct mpioapic); 6381
6332 continue; 6382
6333 case MPBUS: 6383
6334 case MPIOINTR: 6384
6335 case MPLINTR: 6385
6336 p += 8; 6386
6337 continue; 6387
6338 default: 6388
6339 cprintf("mpinit: unknown config type %x\n", *p); 6389
6340 ismp = 0; 6390
6341 } 6391
6342 } 6392
6343 if(!ismp){ 6393
6344 // Didnt like what we found; fall back to no MP. 6394
6345 ncpu = 1; 6395
6346 lapic = 0; 6396
6347 ioapicid = 0; 6397
6348 return; 6398
6349 } 6399

Sheet 63 Sheet 63
Sep 5 23:39 2011 xv6/lapic.c Page 1 Sep 5 23:39 2011 xv6/lapic.c Page 2

6400 // The local APIC manages internal (nonI/O) interrupts. 6450 void
6401 // See Chapter 8 & Appendix C of Intel processor manual volume 3. 6451 lapicinit(int c)
6402 6452 {
6403 #include "types.h" 6453 if(!lapic)
6404 #include "defs.h" 6454 return;
6405 #include "memlayout.h" 6455
6406 #include "traps.h" 6456 // Enable local APIC; set spurious interrupt vector.
6407 #include "mmu.h" 6457 lapicw(SVR, ENABLE | (T_IRQ0 + IRQ_SPURIOUS));
6408 #include "x86.h" 6458
6409 6459 // The timer repeatedly counts down at bus frequency
6410 // Local APIC registers, divided by 4 for use as uint[] indices. 6460 // from lapic[TICR] and then issues an interrupt.
6411 #define ID (0x0020/4) // ID 6461 // If xv6 cared more about precise timekeeping,
6412 #define VER (0x0030/4) // Version 6462 // TICR would be calibrated using an external time source.
6413 #define TPR (0x0080/4) // Task Priority 6463 lapicw(TDCR, X1);
6414 #define EOI (0x00B0/4) // EOI 6464 lapicw(TIMER, PERIODIC | (T_IRQ0 + IRQ_TIMER));
6415 #define SVR (0x00F0/4) // Spurious Interrupt Vector 6465 lapicw(TICR, 10000000);
6416 #define ENABLE 0x00000100 // Unit Enable 6466
6417 #define ESR (0x0280/4) // Error Status 6467 // Disable logical interrupt lines.
6418 #define ICRLO (0x0300/4) // Interrupt Command 6468 lapicw(LINT0, MASKED);
6419 #define INIT 0x00000500 // INIT/RESET 6469 lapicw(LINT1, MASKED);
6420 #define STARTUP 0x00000600 // Startup IPI 6470
6421 #define DELIVS 0x00001000 // Delivery status 6471 // Disable performance counter overflow interrupts
6422 #define ASSERT 0x00004000 // Assert interrupt (vs deassert) 6472 // on machines that provide that interrupt entry.
6423 #define DEASSERT 0x00000000 6473 if(((lapic[VER]>>16) & 0xFF) >= 4)
6424 #define LEVEL 0x00008000 // Level triggered 6474 lapicw(PCINT, MASKED);
6425 #define BCAST 0x00080000 // Send to all APICs, including self. 6475
6426 #define BUSY 0x00001000 6476 // Map error interrupt to IRQ_ERROR.
6427 #define FIXED 0x00000000 6477 lapicw(ERROR, T_IRQ0 + IRQ_ERROR);
6428 #define ICRHI (0x0310/4) // Interrupt Command [63:32] 6478
6429 #define TIMER (0x0320/4) // Local Vector Table 0 (TIMER) 6479 // Clear error status register (requires backtoback writes).
6430 #define X1 0x0000000B // divide counts by 1 6480 lapicw(ESR, 0);
6431 #define PERIODIC 0x00020000 // Periodic 6481 lapicw(ESR, 0);
6432 #define PCINT (0x0340/4) // Performance Counter LVT 6482
6433 #define LINT0 (0x0350/4) // Local Vector Table 1 (LINT0) 6483 // Ack any outstanding interrupts.
6434 #define LINT1 (0x0360/4) // Local Vector Table 2 (LINT1) 6484 lapicw(EOI, 0);
6435 #define ERROR (0x0370/4) // Local Vector Table 3 (ERROR) 6485
6436 #define MASKED 0x00010000 // Interrupt masked 6486 // Send an Init Level DeAssert to synchronise arbitration IDs.
6437 #define TICR (0x0380/4) // Timer Initial Count 6487 lapicw(ICRHI, 0);
6438 #define TCCR (0x0390/4) // Timer Current Count 6488 lapicw(ICRLO, BCAST | INIT | LEVEL);
6439 #define TDCR (0x03E0/4) // Timer Divide Configuration 6489 while(lapic[ICRLO] & DELIVS)
6440 6490 ;
6441 volatile uint *lapic; // Initialized in mp.c 6491
6442 6492 // Enable interrupts on the APIC (but not on the processor).
6443 static void 6493 lapicw(TPR, 0);
6444 lapicw(int index, int value) 6494 }
6445 { 6495
6446 lapic[index] = value; 6496
6447 lapic[ID]; // wait for write to finish, by reading 6497
6448 } 6498
6449 6499

Sheet 64 Sheet 64
Sep 5 23:39 2011 xv6/lapic.c Page 3 Sep 5 23:39 2011 xv6/lapic.c Page 4

6500 int 6550 wrv = (ushort*)P2V((0x40<<4 | 0x67)); // Warm reset vector

6501 cpunum(void) 6551 wrv[0] = 0;
6502 { 6552 wrv[1] = addr >> 4;
6503 // Cannot call cpu when interrupts are enabled: 6553
6504 // result not guaranteed to last long enough to be used! 6554 // "Universal startup algorithm."
6505 // Would prefer to panic but even printing is chancy here: 6555 // Send INIT (leveltriggered) interrupt to reset other CPU.
6506 // almost everything, including cprintf and panic, calls cpu, 6556 lapicw(ICRHI, apicid<<24);
6507 // often indirectly through acquire and release. 6557 lapicw(ICRLO, INIT | LEVEL | ASSERT);
6508 if(readeflags()&FL_IF){ 6558 microdelay(200);
6509 static int n; 6559 lapicw(ICRLO, INIT | LEVEL);
6510 if(n++ == 0) 6560 microdelay(100); // should be 10ms, but too slow in Bochs!
6511 cprintf("cpu called from %x with interrupts enabled\n", 6561
6512 __builtin_return_address(0)); 6562 // Send startup IPI (twice!) to enter code.
6513 } 6563 // Regular hardware is supposed to only accept a STARTUP
6514 6564 // when it is in the halted state due to an INIT. So the second
6515 if(lapic) 6565 // should be ignored, but it is part of the official Intel algorithm.
6516 return lapic[ID]>>24; 6566 // Bochs complains about the second one. Too bad for Bochs.
6517 return 0; 6567 for(i = 0; i < 2; i++){
6518 } 6568 lapicw(ICRHI, apicid<<24);
6519 6569 lapicw(ICRLO, STARTUP | (addr>>12));
6520 // Acknowledge interrupt. 6570 microdelay(200);
6521 void 6571 }
6522 lapiceoi(void) 6572 }
6523 { 6573
6524 if(lapic) 6574
6525 lapicw(EOI, 0); 6575
6526 } 6576
6527 6577
6528 // Spin for a given number of microseconds. 6578
6529 // On real hardware would want to tune this dynamically. 6579
6530 void 6580
6531 microdelay(int us) 6581
6532 { 6582
6533 } 6583
6534 6584
6535 #define IO_RTC 0x70 6585
6536 6586
6537 // Start additional processor running entry code at addr. 6587
6538 // See Appendix B of MultiProcessor Specification. 6588
6539 void 6589
6540 lapicstartap(uchar apicid, uint addr) 6590
6541 { 6591
6542 int i; 6592
6543 ushort *wrv; 6593
6544 6594
6545 // "The BSP must initialize CMOS shutdown code to 0AH 6595
6546 // and the warm reset vector (DWORD based at 40:67) to point at 6596
6547 // the AP startup code prior to the [universal startup algorithm]." 6597
6548 outb(IO_RTC, 0xF); // offset 0xF is shutdown code 6598
6549 outb(IO_RTC+1, 0x0A); 6599

Sheet 65 Sheet 65
Sep 5 23:39 2011 xv6/ioapic.c Page 1 Sep 5 23:39 2011 xv6/ioapic.c Page 2

6600 // The I/O APIC manages hardware interrupts for an SMP system. 6650 void
6601 // http://www.intel.com/design/chipsets/datashts/29056601.pdf 6651 ioapicinit(void)
6602 // See also picirq.c. 6652 {
6603 6653 int i, id, maxintr;
6604 #include "types.h" 6654
6605 #include "defs.h" 6655 if(!ismp)
6606 #include "traps.h" 6656 return;
6607 6657
6608 #define IOAPIC 0xFEC00000 // Default physical address of IO APIC 6658 ioapic = (volatile struct ioapic*)IOAPIC;
6609 6659 maxintr = (ioapicread(REG_VER) >> 16) & 0xFF;
6610 #define REG_ID 0x00 // Register index: ID 6660 id = ioapicread(REG_ID) >> 24;
6611 #define REG_VER 0x01 // Register index: version 6661 if(id != ioapicid)
6612 #define REG_TABLE 0x10 // Redirection table base 6662 cprintf("ioapicinit: id isnt equal to ioapicid; not a MP\n");
6613 6663
6614 // The redirection table starts at REG_TABLE and uses 6664 // Mark all interrupts edgetriggered, active high, disabled,
6615 // two registers to configure each interrupt. 6665 // and not routed to any CPUs.
6616 // The first (low) register in a pair contains configuration bits. 6666 for(i = 0; i <= maxintr; i++){
6617 // The second (high) register contains a bitmask telling which 6667 ioapicwrite(REG_TABLE+2*i, INT_DISABLED | (T_IRQ0 + i));
6618 // CPUs can serve that interrupt. 6668 ioapicwrite(REG_TABLE+2*i+1, 0);
6619 #define INT_DISABLED 0x00010000 // Interrupt disabled 6669 }
6620 #define INT_LEVEL 0x00008000 // Leveltriggered (vs edge) 6670 }
6621 #define INT_ACTIVELOW 0x00002000 // Active low (vs high) 6671
6622 #define INT_LOGICAL 0x00000800 // Destination is CPU id (vs APIC ID) 6672 void
6623 6673 ioapicenable(int irq, int cpunum)
6624 volatile struct ioapic *ioapic; 6674 {
6625 6675 if(!ismp)
6626 // IO APIC MMIO structure: write reg, then read or write data. 6676 return;
6627 struct ioapic { 6677
6628 uint reg; 6678 // Mark interrupt edgetriggered, active high,
6629 uint pad[3]; 6679 // enabled, and routed to the given cpunum,
6630 uint data; 6680 // which happens to be that cpus APIC ID.
6631 }; 6681 ioapicwrite(REG_TABLE+2*irq, T_IRQ0 + irq);
6632 6682 ioapicwrite(REG_TABLE+2*irq+1, cpunum << 24);
6633 static uint 6683 }
6634 ioapicread(int reg) 6684
6635 { 6685
6636 ioapic>reg = reg; 6686
6637 return ioapic>data; 6687
6638 } 6688
6639 6689
6640 static void 6690
6641 ioapicwrite(int reg, uint data) 6691
6642 { 6692
6643 ioapic>reg = reg; 6693
6644 ioapic>data = data; 6694
6645 } 6695
6646 6696
6647 6697
6648 6698
6649 6699

Sheet 66 Sheet 66
Sep 5 23:39 2011 xv6/picirq.c Page 1 Sep 5 23:39 2011 xv6/picirq.c Page 2

6700 // Intel 8259A programmable interrupt controllers. 6750 // ICW3: (master PIC) bit mask of IR lines connected to slaves
6701 6751 // (slave PIC) 3bit # of slaves connection to master
6702 #include "types.h" 6752 outb(IO_PIC1+1, 1<<IRQ_SLAVE);
6703 #include "x86.h" 6753
6704 #include "traps.h" 6754 // ICW4: 000nbmap
6705 6755 // n: 1 = special fully nested mode
6706 // I/O Addresses of the two programmable interrupt controllers 6756 // b: 1 = buffered mode
6707 #define IO_PIC1 0x20 // Master (IRQs 07) 6757 // m: 0 = slave PIC, 1 = master PIC
6708 #define IO_PIC2 0xA0 // Slave (IRQs 815) 6758 // (ignored when b is 0, as the master/slave role
6709 6759 // can be hardwired).
6710 #define IRQ_SLAVE 2 // IRQ at which slave connects to master 6760 // a: 1 = Automatic EOI mode
6711 6761 // p: 0 = MCS80/85 mode, 1 = intel x86 mode
6712 // Current IRQ mask. 6762 outb(IO_PIC1+1, 0x3);
6713 // Initial IRQ mask has interrupt 2 enabled (for slave 8259A). 6763
6714 static ushort irqmask = 0xFFFF & ~(1<<IRQ_SLAVE); 6764 // Set up slave (8259A2)
6715 6765 outb(IO_PIC2, 0x11); // ICW1
6716 static void 6766 outb(IO_PIC2+1, T_IRQ0 + 8); // ICW2
6717 picsetmask(ushort mask) 6767 outb(IO_PIC2+1, IRQ_SLAVE); // ICW3
6718 { 6768 // NB Automatic EOI mode doesnt tend to work on the slave.
6719 irqmask = mask; 6769 // Linux source code says its "to be investigated".
6720 outb(IO_PIC1+1, mask); 6770 outb(IO_PIC2+1, 0x3); // ICW4
6721 outb(IO_PIC2+1, mask >> 8); 6771
6722 } 6772 // OCW3: 0ef01prs
6723 6773 // ef: 0x = NOP, 10 = clear specific mask, 11 = set specific mask
6724 void 6774 // p: 0 = no polling, 1 = polling mode
6725 picenable(int irq) 6775 // rs: 0x = NOP, 10 = read IRR, 11 = read ISR
6726 { 6776 outb(IO_PIC1, 0x68); // clear specific mask
6727 picsetmask(irqmask & ~(1<<irq)); 6777 outb(IO_PIC1, 0x0a); // read IRR by default
6728 } 6778
6729 6779 outb(IO_PIC2, 0x68); // OCW3
6730 // Initialize the 8259A interrupt controllers. 6780 outb(IO_PIC2, 0x0a); // OCW3
6731 void 6781
6732 picinit(void) 6782 if(irqmask != 0xFFFF)
6733 { 6783 picsetmask(irqmask);
6734 // mask all interrupts 6784 }
6735 outb(IO_PIC1+1, 0xFF); 6785
6736 outb(IO_PIC2+1, 0xFF); 6786
6737 6787
6738 // Set up master (8259A1) 6788
6739 6789
6740 // ICW1: 0001g0hi 6790
6741 // g: 0 = edge triggering, 1 = level triggering 6791
6742 // h: 0 = cascaded PICs, 1 = master only 6792
6743 // i: 0 = no ICW4, 1 = ICW4 required 6793
6744 outb(IO_PIC1, 0x11); 6794
6745 6795
6746 // ICW2: Vector offset 6796
6747 outb(IO_PIC1+1, T_IRQ0); 6797
6748 6798
6749 6799

Sheet 67 Sheet 67
Sep 5 23:39 2011 xv6/kbd.h Page 1 Sep 5 23:39 2011 xv6/kbd.h Page 2

6800 // PC keyboard interface constants 6850 static uchar normalmap[256] =

6801 6851 {
6802 #define KBSTATP 0x64 // kbd controller status port(I) 6852 NO, 0x1B, 1, 2, 3, 4, 5, 6, // 0x00
6803 #define KBS_DIB 0x01 // kbd data in buffer 6853 7, 8, 9, 0, , =, \b, \t,
6804 #define KBDATAP 0x60 // kbd data port(I) 6854 q, w, e, r, t, y, u, i, // 0x10
6805 6855 o, p, [, ], \n, NO, a, s,
6806 #define NO 0 6856 d, f, g, h, j, k, l, ;, // 0x20
6807 6857 \, , NO, \\, z, x, c, v,
6808 #define SHIFT (1<<0) 6858 b, n, m, ,, ., /, NO, *, // 0x30
6809 #define CTL (1<<1) 6859 NO, , NO, NO, NO, NO, NO, NO,
6810 #define ALT (1<<2) 6860 NO, NO, NO, NO, NO, NO, NO, 7, // 0x40
6811 6861 8, 9, , 4, 5, 6, +, 1,
6812 #define CAPSLOCK (1<<3) 6862 2, 3, 0, ., NO, NO, NO, NO, // 0x50
6813 #define NUMLOCK (1<<4) 6863 [0x9C] \n, // KP_Enter
6814 #define SCROLLLOCK (1<<5) 6864 [0xB5] /, // KP_Div
6815 6865 [0xC8] KEY_UP, [0xD0] KEY_DN,
6816 #define E0ESC (1<<6) 6866 [0xC9] KEY_PGUP, [0xD1] KEY_PGDN,
6817 6867 [0xCB] KEY_LF, [0xCD] KEY_RT,
6818 // Special keycodes 6868 [0x97] KEY_HOME, [0xCF] KEY_END,
6819 #define KEY_HOME 0xE0 6869 [0xD2] KEY_INS, [0xD3] KEY_DEL
6820 #define KEY_END 0xE1 6870 };
6821 #define KEY_UP 0xE2 6871
6822 #define KEY_DN 0xE3 6872 static uchar shiftmap[256] =
6823 #define KEY_LF 0xE4 6873 {
6824 #define KEY_RT 0xE5 6874 NO, 033, !, @, #, $, %, ^, // 0x00
6825 #define KEY_PGUP 0xE6 6875 &, *, (, ), _, +, \b, \t,
6826 #define KEY_PGDN 0xE7 6876 Q, W, E, R, T, Y, U, I, // 0x10
6827 #define KEY_INS 0xE8 6877 O, P, {, }, \n, NO, A, S,
6828 #define KEY_DEL 0xE9 6878 D, F, G, H, J, K, L, :, // 0x20
6829 6879 ", ~, NO, |, Z, X, C, V,
6830 // C(A) == ControlA 6880 B, N, M, <, >, ?, NO, *, // 0x30
6831 #define C(x) (x @) 6881 NO, , NO, NO, NO, NO, NO, NO,
6832 6882 NO, NO, NO, NO, NO, NO, NO, 7, // 0x40
6833 static uchar shiftcode[256] = 6883 8, 9, , 4, 5, 6, +, 1,
6834 { 6884 2, 3, 0, ., NO, NO, NO, NO, // 0x50
6835 [0x1D] CTL, 6885 [0x9C] \n, // KP_Enter
6836 [0x2A] SHIFT, 6886 [0xB5] /, // KP_Div
6837 [0x36] SHIFT, 6887 [0xC8] KEY_UP, [0xD0] KEY_DN,
6838 [0x38] ALT, 6888 [0xC9] KEY_PGUP, [0xD1] KEY_PGDN,
6839 [0x9D] CTL, 6889 [0xCB] KEY_LF, [0xCD] KEY_RT,
6840 [0xB8] ALT 6890 [0x97] KEY_HOME, [0xCF] KEY_END,
6841 }; 6891 [0xD2] KEY_INS, [0xD3] KEY_DEL
6842 6892 };
6843 static uchar togglecode[256] = 6893
6844 { 6894
6845 [0x3A] CAPSLOCK, 6895
6846 [0x45] NUMLOCK, 6896
6847 [0x46] SCROLLLOCK 6897
6848 }; 6898
6849 6899

Sheet 68 Sheet 68
Sep 5 23:39 2011 xv6/kbd.h Page 3 Sep 5 23:39 2011 xv6/kbd.c Page 1

6900 static uchar ctlmap[256] = 6950 #include "types.h"

6901 { 6951 #include "x86.h"
6902 NO, NO, NO, NO, NO, NO, NO, NO, 6952 #include "defs.h"
6903 NO, NO, NO, NO, NO, NO, NO, NO, 6953 #include "kbd.h"
6904 C(Q), C(W), C(E), C(R), C(T), C(Y), C(U), C(I), 6954
6905 C(O), C(P), NO, NO, \r, NO, C(A), C(S), 6955 int
6906 C(D), C(F), C(G), C(H), C(J), C(K), C(L), NO, 6956 kbdgetc(void)
6907 NO, NO, NO, C(\\), C(Z), C(X), C(C), C(V), 6957 {
6908 C(B), C(N), C(M), NO, NO, C(/), NO, NO, 6958 static uint shift;
6909 [0x9C] \r, // KP_Enter 6959 static uchar *charcode[4] = {
6910 [0xB5] C(/), // KP_Div 6960 normalmap, shiftmap, ctlmap, ctlmap
6911 [0xC8] KEY_UP, [0xD0] KEY_DN, 6961 };
6912 [0xC9] KEY_PGUP, [0xD1] KEY_PGDN, 6962 uint st, data, c;
6913 [0xCB] KEY_LF, [0xCD] KEY_RT, 6963
6914 [0x97] KEY_HOME, [0xCF] KEY_END, 6964 st = inb(KBSTATP);
6915 [0xD2] KEY_INS, [0xD3] KEY_DEL 6965 if((st & KBS_DIB) == 0)
6916 }; 6966 return 1;
6917 6967 data = inb(KBDATAP);
6918 6968
6919 6969 if(data == 0xE0){
6920 6970 shift |= E0ESC;
6921 6971 return 0;
6922 6972 } else if(data & 0x80){
6923 6973 // Key released
6924 6974 data = (shift & E0ESC ? data : data & 0x7F);
6925 6975 shift &= ~(shiftcode[data] | E0ESC);
6926 6976 return 0;
6927 6977 } else if(shift & E0ESC){
6928 6978 // Last character was an E0 escape; or with 0x80
6929 6979 data |= 0x80;
6930 6980 shift &= ~E0ESC;
6931 6981 }
6932 6982
6933 6983 shift |= shiftcode[data];
6934 6984 shift ^= togglecode[data];
6935 6985 c = charcode[shift & (CTL | SHIFT)][data];
6936 6986 if(shift & CAPSLOCK){
6937 6987 if(a <= c && c <= z)
6938 6988 c += A a;
6939 6989 else if(A <= c && c <= Z)
6940 6990 c += a A;
6941 6991 }
6942 6992 return c;
6943 6993 }
6944 6994
6945 6995 void
6946 6996 kbdintr(void)
6947 6997 {
6948 6998 consoleintr(kbdgetc);
6949 6999 }

Sheet 69 Sheet 69
Sep 5 23:39 2011 xv6/console.c Page 1 Sep 5 23:39 2011 xv6/console.c Page 2

7000 // Console input and output. 7050 // Print to the console. only understands %d, %x, %p, %s.
7001 // Input is from the keyboard or serial port. 7051 void
7002 // Output is written to the screen and serial port. 7052 cprintf(char *fmt, ...)
7003 7053 {
7004 #include "types.h" 7054 int i, c, state, locking;
7005 #include "defs.h" 7055 uint *argp;
7006 #include "param.h" 7056 char *s;
7007 #include "traps.h" 7057
7008 #include "spinlock.h" 7058 locking = cons.locking;
7009 #include "fs.h" 7059 if(locking)
7010 #include "file.h" 7060 acquire(&cons.lock);
7011 #include "memlayout.h" 7061
7012 #include "mmu.h" 7062 if (fmt == 0)
7013 #include "proc.h" 7063 panic("null fmt");
7014 #include "x86.h" 7064
7015 7065 argp = (uint*)(void*)(&fmt + 1);
7016 static void consputc(int); 7066 state = 0;
7017 7067 for(i = 0; (c = fmt[i] & 0xff) != 0; i++){
7018 static int panicked = 0; 7068 if(c != %){
7019 7069 consputc(c);
7020 static struct { 7070 continue;
7021 struct spinlock lock; 7071 }
7022 int locking; 7072 c = fmt[++i] & 0xff;
7023 } cons; 7073 if(c == 0)
7024 7074 break;
7025 static void 7075 switch(c){
7026 printint(int xx, int base, int sign) 7076 case d:
7027 { 7077 printint(*argp++, 10, 1);
7028 static char digits[] = "0123456789abcdef"; 7078 break;
7029 char buf[16]; 7079 case x:
7030 int i; 7080 case p:
7031 uint x; 7081 printint(*argp++, 16, 0);
7032 7082 break;
7033 if(sign && (sign = xx < 0)) 7083 case s:
7034 x = xx; 7084 if((s = (char*)*argp++) == 0)
7035 else 7085 s = "(null)";
7036 x = xx; 7086 for(; *s; s++)
7037 7087 consputc(*s);
7038 i = 0; 7088 break;
7039 do{ 7089 case %:
7040 buf[i++] = digits[x % base]; 7090 consputc(%);
7041 }while((x /= base) != 0); 7091 break;
7042 7092 default:
7043 if(sign) 7093 // Print unknown % sequence to draw attention.
7044 buf[i++] = ; 7094 consputc(%);
7045 7095 consputc(c);
7046 while(i >= 0) 7096 break;
7047 consputc(buf[i]); 7097 }
7048 } 7098 }
7049 7099

Sheet 70 Sheet 70
Sep 5 23:39 2011 xv6/console.c Page 3 Sep 5 23:39 2011 xv6/console.c Page 4

7100 if(locking) 7150 #define BACKSPACE 0x100

7101 release(&cons.lock); 7151 #define CRTPORT 0x3d4
7102 } 7152 static ushort *crt = (ushort*)P2V(0xb8000); // CGA memory
7103 7153
7104 void 7154 static void
7105 panic(char *s) 7155 cgaputc(int c)
7106 { 7156 {
7107 int i; 7157 int pos;
7108 uint pcs[10]; 7158
7109 7159 // Cursor position: col + 80*row.
7110 cli(); 7160 outb(CRTPORT, 14);
7111 cons.locking = 0; 7161 pos = inb(CRTPORT+1) << 8;
7112 cprintf("cpu%d: panic: ", cpu>id); 7162 outb(CRTPORT, 15);
7113 cprintf(s); 7163 pos |= inb(CRTPORT+1);
7114 cprintf("\n"); 7164
7115 getcallerpcs(&s, pcs); 7165 if(c == \n)
7116 for(i=0; i<10; i++) 7166 pos += 80 pos%80;
7117 cprintf(" %p", pcs[i]); 7167 else if(c == BACKSPACE){
7118 panicked = 1; // freeze other CPU 7168 if(pos > 0) pos;
7119 for(;;) 7169 } else
7120 ; 7170 crt[pos++] = (c&0xff) | 0x0700; // black on white
7121 } 7171
7122 7172 if((pos/80) >= 24){ // Scroll up.
7123 7173 memmove(crt, crt+80, sizeof(crt[0])*23*80);
7124 7174 pos = 80;
7125 7175 memset(crt+pos, 0, sizeof(crt[0])*(24*80 pos));
7126 7176 }
7127 7177
7128 7178 outb(CRTPORT, 14);
7129 7179 outb(CRTPORT+1, pos>>8);
7130 7180 outb(CRTPORT, 15);
7131 7181 outb(CRTPORT+1, pos);
7132 7182 crt[pos] = | 0x0700;
7133 7183 }
7134 7184
7135 7185 void
7136 7186 consputc(int c)
7137 7187 {
7138 7188 if(panicked){
7139 7189 cli();
7140 7190 for(;;)
7141 7191 ;
7142 7192 }
7143 7193
7144 7194 if(c == BACKSPACE){
7145 7195 uartputc(\b); uartputc( ); uartputc(\b);
7146 7196 } else
7147 7197 uartputc(c);
7148 7198 cgaputc(c);
7149 7199 }

Sheet 71 Sheet 71
Sep 5 23:39 2011 xv6/console.c Page 5 Sep 5 23:39 2011 xv6/console.c Page 6

7200 #define INPUT_BUF 128 7250 int

7201 struct { 7251 consoleread(struct inode *ip, char *dst, int n)
7202 struct spinlock lock; 7252 {
7203 char buf[INPUT_BUF]; 7253 uint target;
7204 uint r; // Read index 7254 int c;
7205 uint w; // Write index 7255
7206 uint e; // Edit index 7256 iunlock(ip);
7207 } input; 7257 target = n;
7208 7258 acquire(&input.lock);
7209 #define C(x) ((x)@) // Controlx 7259 while(n > 0){
7210 7260 while(input.r == input.w){
7211 void 7261 if(proc>killed){
7212 consoleintr(int (*getc)(void)) 7262 release(&input.lock);
7213 { 7263 ilock(ip);
7214 int c; 7264 return 1;
7215 7265 }
7216 acquire(&input.lock); 7266 sleep(&input.r, &input.lock);
7217 while((c = getc()) >= 0){ 7267 }
7218 switch(c){ 7268 c = input.buf[input.r++ % INPUT_BUF];
7219 case C(P): // Process listing. 7269 if(c == C(D)){ // EOF
7220 procdump(); 7270 if(n < target){
7221 break; 7271 // Save ^D for next time, to make sure
7222 case C(U): // Kill line. 7272 // caller gets a 0byte result.
7223 while(input.e != input.w && 7273 input.r;
7224 input.buf[(input.e1) % INPUT_BUF] != \n){ 7274 }
7225 input.e; 7275 break;
7226 consputc(BACKSPACE); 7276 }
7227 } 7277 *dst++ = c;
7228 break; 7278 n;
7229 case C(H): case \x7f: // Backspace 7279 if(c == \n)
7230 if(input.e != input.w){ 7280 break;
7231 input.e; 7281 }
7232 consputc(BACKSPACE); 7282 release(&input.lock);
7233 } 7283 ilock(ip);
7234 break; 7284
7235 default: 7285 return target n;
7236 if(c != 0 && input.einput.r < INPUT_BUF){ 7286 }
7237 c = (c == \r) ? \n : c; 7287
7238 input.buf[input.e++ % INPUT_BUF] = c; 7288
7239 consputc(c); 7289
7240 if(c == \n || c == C(D) || input.e == input.r+INPUT_BUF){ 7290
7241 input.w = input.e; 7291
7242 wakeup(&input.r); 7292
7243 } 7293
7244 } 7294
7245 break; 7295
7246 } 7296
7247 } 7297
7248 release(&input.lock); 7298
7249 } 7299

Sheet 72 Sheet 72
Sep 5 23:39 2011 xv6/console.c Page 7 Sep 5 23:39 2011 xv6/timer.c Page 1

7300 int 7350 // Intel 8253/8254/82C54 Programmable Interval Timer (PIT).

7301 consolewrite(struct inode *ip, char *buf, int n) 7351 // Only used on uniprocessors;
7302 { 7352 // SMP machines use the local APIC timer.
7303 int i; 7353
7304 7354 #include "types.h"
7305 iunlock(ip); 7355 #include "defs.h"
7306 acquire(&cons.lock); 7356 #include "traps.h"
7307 for(i = 0; i < n; i++) 7357 #include "x86.h"
7308 consputc(buf[i] & 0xff); 7358
7309 release(&cons.lock); 7359 #define IO_TIMER1 0x040 // 8253 Timer #1
7310 ilock(ip); 7360
7311 7361 // Frequency of all three countdown timers;
7312 return n; 7362 // (TIMER_FREQ/freq) is the appropriate count
7313 } 7363 // to generate a frequency of freq Hz.
7314 7364
7315 void 7365 #define TIMER_FREQ 1193182
7316 consoleinit(void) 7366 #define TIMER_DIV(x) ((TIMER_FREQ+(x)/2)/(x))
7317 { 7367
7318 initlock(&cons.lock, "console"); 7368 #define TIMER_MODE (IO_TIMER1 + 3) // timer mode port
7319 initlock(&input.lock, "input"); 7369 #define TIMER_SEL0 0x00 // select counter 0
7320 7370 #define TIMER_RATEGEN 0x04 // mode 2, rate generator
7321 devsw[CONSOLE].write = consolewrite; 7371 #define TIMER_16BIT 0x30 // r/w counter 16 bits, LSB first
7322 devsw[CONSOLE].read = consoleread; 7372
7323 cons.locking = 1; 7373 void
7324 7374 timerinit(void)
7325 picenable(IRQ_KBD); 7375 {
7326 ioapicenable(IRQ_KBD, 0); 7376 // Interrupt 100 times/sec.
7327 } 7377 outb(TIMER_MODE, TIMER_SEL0 | TIMER_RATEGEN | TIMER_16BIT);
7328 7378 outb(IO_TIMER1, TIMER_DIV(100) % 256);
7329 7379 outb(IO_TIMER1, TIMER_DIV(100) / 256);
7330 7380 picenable(IRQ_TIMER);
7331 7381 }
7332 7382
7333 7383
7334 7384
7335 7385
7336 7386
7337 7387
7338 7388
7339 7389
7340 7390
7341 7391
7342 7392
7343 7393
7344 7394
7345 7395
7346 7396
7347 7397
7348 7398
7349 7399

Sheet 73 Sheet 73
Sep 5 23:39 2011 xv6/uart.c Page 1 Sep 5 23:39 2011 xv6/uart.c Page 2

7400 // Intel 8250 serial port (UART). 7450 void

7401 7451 uartputc(int c)
7402 #include "types.h" 7452 {
7403 #include "defs.h" 7453 int i;
7404 #include "param.h" 7454
7405 #include "traps.h" 7455 if(!uart)
7406 #include "spinlock.h" 7456 return;
7407 #include "fs.h" 7457 for(i = 0; i < 128 && !(inb(COM1+5) & 0x20); i++)
7408 #include "file.h" 7458 microdelay(10);
7409 #include "mmu.h" 7459 outb(COM1+0, c);
7410 #include "proc.h" 7460 }
7411 #include "x86.h" 7461
7412 7462 static int
7413 #define COM1 0x3f8 7463 uartgetc(void)
7414 7464 {
7415 static int uart; // is there a uart? 7465 if(!uart)
7416 7466 return 1;
7417 void 7467 if(!(inb(COM1+5) & 0x01))
7418 uartinit(void) 7468 return 1;
7419 { 7469 return inb(COM1+0);
7420 char *p; 7470 }
7421 7471
7422 // Turn off the FIFO 7472 void
7423 outb(COM1+2, 0); 7473 uartintr(void)
7424 7474 {
7425 // 9600 baud, 8 data bits, 1 stop bit, parity off. 7475 consoleintr(uartgetc);
7426 outb(COM1+3, 0x80); // Unlock divisor 7476 }
7427 outb(COM1+0, 115200/9600); 7477
7428 outb(COM1+1, 0); 7478
7429 outb(COM1+3, 0x03); // Lock divisor, 8 data bits. 7479
7430 outb(COM1+4, 0); 7480
7431 outb(COM1+1, 0x01); // Enable receive interrupts. 7481
7432 7482
7433 // If status is 0xFF, no serial port. 7483
7434 if(inb(COM1+5) == 0xFF) 7484
7435 return; 7485
7436 uart = 1; 7486
7437 7487
7438 // Acknowledge preexisting interrupt conditions; 7488
7439 // enable interrupts. 7489
7440 inb(COM1+2); 7490
7441 inb(COM1+0); 7491
7442 picenable(IRQ_COM1); 7492
7443 ioapicenable(IRQ_COM1, 0); 7493
7444 7494
7445 // Announce that were here. 7495
7446 for(p="xv6...\n"; *p; p++) 7496
7447 uartputc(*p); 7497
7448 } 7498
7449 7499

Sheet 74 Sheet 74
Sep 5 23:39 2011 xv6/initcode.S Page 1 Sep 5 23:39 2011 xv6/usys.S Page 1

7500 # Initial process execs /init. 7550 #include "syscall.h"

7501 7551 #include "traps.h"
7502 #include "syscall.h" 7552
7503 #include "traps.h" 7553 #define SYSCALL(name) \
7504 7554 .globl name; \
7505 7555 name: \
7506 # exec(init, argv) 7556 movl $SYS_ ## name, %eax; \
7507 .globl start 7557 int $T_SYSCALL; \
7508 start: 7558 ret
7509 pushl $argv 7559
7510 pushl $init 7560 SYSCALL(fork)
7511 pushl $0 // where caller pc would be 7561 SYSCALL(exit)
7512 movl $SYS_exec, %eax 7562 SYSCALL(wait)
7513 int $T_SYSCALL 7563 SYSCALL(pipe)
7514 7564 SYSCALL(read)
7515 # for(;;) exit(); 7565 SYSCALL(write)
7516 exit: 7566 SYSCALL(close)
7517 movl $SYS_exit, %eax 7567 SYSCALL(kill)
7518 int $T_SYSCALL 7568 SYSCALL(exec)
7519 jmp exit 7569 SYSCALL(open)
7520 7570 SYSCALL(mknod)
7521 # char init[] = "/init\0"; 7571 SYSCALL(unlink)
7522 init: 7572 SYSCALL(fstat)
7523 .string "/init\0" 7573 SYSCALL(link)
7524 7574 SYSCALL(mkdir)
7525 # char *argv[] = { init, 0 }; 7575 SYSCALL(chdir)
7526 .p2align 2 7576 SYSCALL(dup)
7527 argv: 7577 SYSCALL(getpid)
7528 .long init 7578 SYSCALL(sbrk)
7529 .long 0 7579 SYSCALL(sleep)
7530 7580 SYSCALL(uptime)
7531 7581
7532 7582
7533 7583
7534 7584
7535 7585
7536 7586
7537 7587
7538 7588
7539 7589
7540 7590
7541 7591
7542 7592
7543 7593
7544 7594
7545 7595
7546 7596
7547 7597
7548 7598
7549 7599

Sheet 75 Sheet 75
Sep 5 23:39 2011 xv6/init.c Page 1 Sep 5 23:39 2011 xv6/sh.c Page 1

7600 // init: The initial userlevel program 7650 // Shell.

7601 7651
7602 #include "types.h" 7652 #include "types.h"
7603 #include "stat.h" 7653 #include "user.h"
7604 #include "user.h" 7654 #include "fcntl.h"
7605 #include "fcntl.h" 7655
7606 7656 // Parsed command representation
7607 char *argv[] = { "sh", 0 }; 7657 #define EXEC 1
7608 7658 #define REDIR 2
7609 int 7659 #define PIPE 3
7610 main(void) 7660 #define LIST 4
7611 { 7661 #define BACK 5
7612 int pid, wpid; 7662
7613 7663 #define MAXARGS 10
7614 if(open("console", O_RDWR) < 0){ 7664
7615 mknod("console", 1, 1); 7665 struct cmd {
7616 open("console", O_RDWR); 7666 int type;
7617 } 7667 };
7618 dup(0); // stdout 7668
7619 dup(0); // stderr 7669 struct execcmd {
7620 7670 int type;
7621 for(;;){ 7671 char *argv[MAXARGS];
7622 printf(1, "init: starting sh\n"); 7672 char *eargv[MAXARGS];
7623 pid = fork(); 7673 };
7624 if(pid < 0){ 7674
7625 printf(1, "init: fork failed\n"); 7675 struct redircmd {
7626 exit(); 7676 int type;
7627 } 7677 struct cmd *cmd;
7628 if(pid == 0){ 7678 char *file;
7629 exec("sh", argv); 7679 char *efile;
7630 printf(1, "init: exec sh failed\n"); 7680 int mode;
7631 exit(); 7681 int fd;
7632 } 7682 };
7633 while((wpid=wait()) >= 0 && wpid != pid) 7683
7634 printf(1, "zombie!\n"); 7684 struct pipecmd {
7635 } 7685 int type;
7636 } 7686 struct cmd *left;
7637 7687 struct cmd *right;
7638 7688 };
7639 7689
7640 7690 struct listcmd {
7641 7691 int type;
7642 7692 struct cmd *left;
7643 7693 struct cmd *right;
7644 7694 };
7645 7695
7646 7696 struct backcmd {
7647 7697 int type;
7648 7698 struct cmd *cmd;
7649 7699 };

Sheet 76 Sheet 76
Sep 5 23:39 2011 xv6/sh.c Page 2 Sep 5 23:39 2011 xv6/sh.c Page 3

7700 int fork1(void); // Fork but panics on failure. 7750 case PIPE:
7701 void panic(char*); 7751 pcmd = (struct pipecmd*)cmd;
7702 struct cmd *parsecmd(char*); 7752 if(pipe(p) < 0)
7703 7753 panic("pipe");
7704 // Execute cmd. Never returns. 7754 if(fork1() == 0){
7705 void 7755 close(1);
7706 runcmd(struct cmd *cmd) 7756 dup(p[1]);
7707 { 7757 close(p[0]);
7708 int p[2]; 7758 close(p[1]);
7709 struct backcmd *bcmd; 7759 runcmd(pcmd>left);
7710 struct execcmd *ecmd; 7760 }
7711 struct listcmd *lcmd; 7761 if(fork1() == 0){
7712 struct pipecmd *pcmd; 7762 close(0);
7713 struct redircmd *rcmd; 7763 dup(p[0]);
7714 7764 close(p[0]);
7715 if(cmd == 0) 7765 close(p[1]);
7716 exit(); 7766 runcmd(pcmd>right);
7717 7767 }
7718 switch(cmd>type){ 7768 close(p[0]);
7719 default: 7769 close(p[1]);
7720 panic("runcmd"); 7770 wait();
7721 7771 wait();
7722 case EXEC: 7772 break;
7723 ecmd = (struct execcmd*)cmd; 7773
7724 if(ecmd>argv[0] == 0) 7774 case BACK:
7725 exit(); 7775 bcmd = (struct backcmd*)cmd;
7726 exec(ecmd>argv[0], ecmd>argv); 7776 if(fork1() == 0)
7727 printf(2, "exec %s failed\n", ecmd>argv[0]); 7777 runcmd(bcmd>cmd);
7728 break; 7778 break;
7729 7779 }
7730 case REDIR: 7780 exit();
7731 rcmd = (struct redircmd*)cmd; 7781 }
7732 close(rcmd>fd); 7782
7733 if(open(rcmd>file, rcmd>mode) < 0){ 7783 int
7734 printf(2, "open %s failed\n", rcmd>file); 7784 getcmd(char *buf, int nbuf)
7735 exit(); 7785 {
7736 } 7786 printf(2, "$ ");
7737 runcmd(rcmd>cmd); 7787 memset(buf, 0, nbuf);
7738 break; 7788 gets(buf, nbuf);
7739 7789 if(buf[0] == 0) // EOF
7740 case LIST: 7790 return 1;
7741 lcmd = (struct listcmd*)cmd; 7791 return 0;
7742 if(fork1() == 0) 7792 }
7743 runcmd(lcmd>left); 7793
7744 wait(); 7794
7745 runcmd(lcmd>right); 7795
7746 break; 7796
7747 7797
7748 7798
7749 7799

Sheet 77 Sheet 77
Sep 5 23:39 2011 xv6/sh.c Page 4 Sep 5 23:39 2011 xv6/sh.c Page 5

7800 int 7850 // Constructors

7801 main(void) 7851
7802 { 7852 struct cmd*
7803 static char buf[100]; 7853 execcmd(void)
7804 int fd; 7854 {
7805 7855 struct execcmd *cmd;
7806 // Assumes three file descriptors open. 7856
7807 while((fd = open("console", O_RDWR)) >= 0){ 7857 cmd = malloc(sizeof(*cmd));
7808 if(fd >= 3){ 7858 memset(cmd, 0, sizeof(*cmd));
7809 close(fd); 7859 cmd>type = EXEC;
7810 break; 7860 return (struct cmd*)cmd;
7811 } 7861 }
7812 } 7862
7813 7863 struct cmd*
7814 // Read and run input commands. 7864 redircmd(struct cmd *subcmd, char *file, char *efile, int mode, int fd)
7815 while(getcmd(buf, sizeof(buf)) >= 0){ 7865 {
7816 if(buf[0] == c && buf[1] == d && buf[2] == ){ 7866 struct redircmd *cmd;
7817 // Clumsy but will have to do for now. 7867
7818 // Chdir has no effect on the parent if run in the child. 7868 cmd = malloc(sizeof(*cmd));
7819 buf[strlen(buf)1] = 0; // chop \n 7869 memset(cmd, 0, sizeof(*cmd));
7820 if(chdir(buf+3) < 0) 7870 cmd>type = REDIR;
7821 printf(2, "cannot cd %s\n", buf+3); 7871 cmd>cmd = subcmd;
7822 continue; 7872 cmd>file = file;
7823 } 7873 cmd>efile = efile;
7824 if(fork1() == 0) 7874 cmd>mode = mode;
7825 runcmd(parsecmd(buf)); 7875 cmd>fd = fd;
7826 wait(); 7876 return (struct cmd*)cmd;
7827 } 7877 }
7828 exit(); 7878
7829 } 7879 struct cmd*
7830 7880 pipecmd(struct cmd *left, struct cmd *right)
7831 void 7881 {
7832 panic(char *s) 7882 struct pipecmd *cmd;
7833 { 7883
7834 printf(2, "%s\n", s); 7884 cmd = malloc(sizeof(*cmd));
7835 exit(); 7885 memset(cmd, 0, sizeof(*cmd));
7836 } 7886 cmd>type = PIPE;
7837 7887 cmd>left = left;
7838 int 7888 cmd>right = right;
7839 fork1(void) 7889 return (struct cmd*)cmd;
7840 { 7890 }
7841 int pid; 7891
7842 7892
7843 pid = fork(); 7893
7844 if(pid == 1) 7894
7845 panic("fork"); 7895
7846 return pid; 7896
7847 } 7897
7848 7898
7849 7899

Sheet 78 Sheet 78
Sep 5 23:39 2011 xv6/sh.c Page 6 Sep 5 23:39 2011 xv6/sh.c Page 7

7900 struct cmd* 7950 // Parsing

7901 listcmd(struct cmd *left, struct cmd *right) 7951
7902 { 7952 char whitespace[] = " \t\r\n\v";
7903 struct listcmd *cmd; 7953 char symbols[] = "<|>&;()";
7904 7954
7905 cmd = malloc(sizeof(*cmd)); 7955 int
7906 memset(cmd, 0, sizeof(*cmd)); 7956 gettoken(char **ps, char *es, char **q, char **eq)
7907 cmd>type = LIST; 7957 {
7908 cmd>left = left; 7958 char *s;
7909 cmd>right = right; 7959 int ret;
7910 return (struct cmd*)cmd; 7960
7911 } 7961 s = *ps;
7912 7962 while(s < es && strchr(whitespace, *s))
7913 struct cmd* 7963 s++;
7914 backcmd(struct cmd *subcmd) 7964 if(q)
7915 { 7965 *q = s;
7916 struct backcmd *cmd; 7966 ret = *s;
7917 7967 switch(*s){
7918 cmd = malloc(sizeof(*cmd)); 7968 case 0:
7919 memset(cmd, 0, sizeof(*cmd)); 7969 break;
7920 cmd>type = BACK; 7970 case |:
7921 cmd>cmd = subcmd; 7971 case (:
7922 return (struct cmd*)cmd; 7972 case ):
7923 } 7973 case ;:
7924 7974 case &:
7925 7975 case <:
7926 7976 s++;
7927 7977 break;
7928 7978 case >:
7929 7979 s++;
7930 7980 if(*s == >){
7931 7981 ret = +;
7932 7982 s++;
7933 7983 }
7934 7984 break;
7935 7985 default:
7936 7986 ret = a;
7937 7987 while(s < es && !strchr(whitespace, *s) && !strchr(symbols, *s))
7938 7988 s++;
7939 7989 break;
7940 7990 }
7941 7991 if(eq)
7942 7992 *eq = s;
7943 7993
7944 7994 while(s < es && strchr(whitespace, *s))
7945 7995 s++;
7946 7996 *ps = s;
7947 7997 return ret;
7948 7998 }
7949 7999

Sheet 79 Sheet 79
Sep 5 23:39 2011 xv6/sh.c Page 8 Sep 5 23:39 2011 xv6/sh.c Page 9

8000 int 8050 struct cmd*

8001 peek(char **ps, char *es, char *toks) 8051 parsepipe(char **ps, char *es)
8002 { 8052 {
8003 char *s; 8053 struct cmd *cmd;
8004 8054
8005 s = *ps; 8055 cmd = parseexec(ps, es);
8006 while(s < es && strchr(whitespace, *s)) 8056 if(peek(ps, es, "|")){
8007 s++; 8057 gettoken(ps, es, 0, 0);
8008 *ps = s; 8058 cmd = pipecmd(cmd, parsepipe(ps, es));
8009 return *s && strchr(toks, *s); 8059 }
8010 } 8060 return cmd;
8011 8061 }
8012 struct cmd *parseline(char**, char*); 8062
8013 struct cmd *parsepipe(char**, char*); 8063 struct cmd*
8014 struct cmd *parseexec(char**, char*); 8064 parseredirs(struct cmd *cmd, char **ps, char *es)
8015 struct cmd *nulterminate(struct cmd*); 8065 {
8016 8066 int tok;
8017 struct cmd* 8067 char *q, *eq;
8018 parsecmd(char *s) 8068
8019 { 8069 while(peek(ps, es, "<>")){
8020 char *es; 8070 tok = gettoken(ps, es, 0, 0);
8021 struct cmd *cmd; 8071 if(gettoken(ps, es, &q, &eq) != a)
8022 8072 panic("missing file for redirection");
8023 es = s + strlen(s); 8073 switch(tok){
8024 cmd = parseline(&s, es); 8074 case <:
8025 peek(&s, es, ""); 8075 cmd = redircmd(cmd, q, eq, O_RDONLY, 0);
8026 if(s != es){ 8076 break;
8027 printf(2, "leftovers: %s\n", s); 8077 case >:
8028 panic("syntax"); 8078 cmd = redircmd(cmd, q, eq, O_WRONLY|O_CREATE, 1);
8029 } 8079 break;
8030 nulterminate(cmd); 8080 case +: // >>
8031 return cmd; 8081 cmd = redircmd(cmd, q, eq, O_WRONLY|O_CREATE, 1);
8032 } 8082 break;
8033 8083 }
8034 struct cmd* 8084 }
8035 parseline(char **ps, char *es) 8085 return cmd;
8036 { 8086 }
8037 struct cmd *cmd; 8087
8038 8088
8039 cmd = parsepipe(ps, es); 8089
8040 while(peek(ps, es, "&")){ 8090
8041 gettoken(ps, es, 0, 0); 8091
8042 cmd = backcmd(cmd); 8092
8043 } 8093
8044 if(peek(ps, es, ";")){ 8094
8045 gettoken(ps, es, 0, 0); 8095
8046 cmd = listcmd(cmd, parseline(ps, es)); 8096
8047 } 8097
8048 return cmd; 8098
8049 } 8099

Sheet 80 Sheet 80
Sep 5 23:39 2011 xv6/sh.c Page 10 Sep 5 23:39 2011 xv6/sh.c Page 11

8100 struct cmd* 8150 // NULterminate all the counted strings.

8101 parseblock(char **ps, char *es) 8151 struct cmd*
8102 { 8152 nulterminate(struct cmd *cmd)
8103 struct cmd *cmd; 8153 {
8104 8154 int i;
8105 if(!peek(ps, es, "(")) 8155 struct backcmd *bcmd;
8106 panic("parseblock"); 8156 struct execcmd *ecmd;
8107 gettoken(ps, es, 0, 0); 8157 struct listcmd *lcmd;
8108 cmd = parseline(ps, es); 8158 struct pipecmd *pcmd;
8109 if(!peek(ps, es, ")")) 8159 struct redircmd *rcmd;
8110 panic("syntax missing )"); 8160
8111 gettoken(ps, es, 0, 0); 8161 if(cmd == 0)
8112 cmd = parseredirs(cmd, ps, es); 8162 return 0;
8113 return cmd; 8163
8114 } 8164 switch(cmd>type){
8115 8165 case EXEC:
8116 struct cmd* 8166 ecmd = (struct execcmd*)cmd;
8117 parseexec(char **ps, char *es) 8167 for(i=0; ecmd>argv[i]; i++)
8118 { 8168 *ecmd>eargv[i] = 0;
8119 char *q, *eq; 8169 break;
8120 int tok, argc; 8170
8121 struct execcmd *cmd; 8171 case REDIR:
8122 struct cmd *ret; 8172 rcmd = (struct redircmd*)cmd;
8123 8173 nulterminate(rcmd>cmd);
8124 if(peek(ps, es, "(")) 8174 *rcmd>efile = 0;
8125 return parseblock(ps, es); 8175 break;
8126 8176
8127 ret = execcmd(); 8177 case PIPE:
8128 cmd = (struct execcmd*)ret; 8178 pcmd = (struct pipecmd*)cmd;
8129 8179 nulterminate(pcmd>left);
8130 argc = 0; 8180 nulterminate(pcmd>right);
8131 ret = parseredirs(ret, ps, es); 8181 break;
8132 while(!peek(ps, es, "|)&;")){ 8182
8133 if((tok=gettoken(ps, es, &q, &eq)) == 0) 8183 case LIST:
8134 break; 8184 lcmd = (struct listcmd*)cmd;
8135 if(tok != a) 8185 nulterminate(lcmd>left);
8136 panic("syntax"); 8186 nulterminate(lcmd>right);
8137 cmd>argv[argc] = q; 8187 break;
8138 cmd>eargv[argc] = eq; 8188
8139 argc++; 8189 case BACK:
8140 if(argc >= MAXARGS) 8190 bcmd = (struct backcmd*)cmd;
8141 panic("too many args"); 8191 nulterminate(bcmd>cmd);
8142 ret = parseredirs(ret, ps, es); 8192 break;
8143 } 8193 }
8144 cmd>argv[argc] = 0; 8194 return cmd;
8145 cmd>eargv[argc] = 0; 8195 }
8146 return ret; 8196
8147 } 8197
8148 8198
8149 8199

Sheet 81 Sheet 81
Sep 5 23:39 2011 xv6/bootasm.S Page 1 Sep 5 23:39 2011 xv6/bootasm.S Page 2

8200 #include "asm.h" 8250 # Complete transition to 32bit protected mode by using long jmp
8201 #include "memlayout.h" 8251 # to reload %cs and %eip. The segment descriptors are set up with no
8202 #include "mmu.h" 8252 # translation, so that the mapping is still the identity mapping.
8203 8253 ljmp $(SEG_KCODE<<3), $start32
8204 # Start the first CPU: switch to 32bit protected mode, jump into C. 8254
8205 # The BIOS loads this code from the first sector of the hard disk into 8255 .code32 # Tell assembler to generate 32bit code now.
8206 # memory at physical address 0x7c00 and starts executing in real mode 8256 start32:
8207 # with %cs=0 %ip=7c00. 8257 # Set up the protectedmode data segment registers
8208 8258 movw $(SEG_KDATA<<3), %ax # Our data segment selector
8209 .code16 # Assemble for 16bit mode 8259 movw %ax, %ds # > DS: Data Segment
8210 .globl start 8260 movw %ax, %es # > ES: Extra Segment
8211 start: 8261 movw %ax, %ss # > SS: Stack Segment
8212 cli # BIOS enabled interrupts; disable 8262 movw $0, %ax # Zero segments not ready for use
8213 8263 movw %ax, %fs # > FS
8214 # Set up the important data segment registers (DS, ES, SS). 8264 movw %ax, %gs # > GS
8215 xorw %ax,%ax # Segment number zero 8265
8216 movw %ax,%ds # > Data Segment 8266 # Set up the stack pointer and call into C.
8217 movw %ax,%es # > Extra Segment 8267 movl $start, %esp
8218 movw %ax,%ss # > Stack Segment 8268 call bootmain
8219 8269
8220 # Physical address line A20 is tied to zero so that the first PCs 8270 # If bootmain returns (it shouldnt), trigger a Bochs
8221 # with 2 MB would run software that assumed 1 MB. Undo that. 8271 # breakpoint if running under Bochs, then loop.
8222 seta20.1: 8272 movw $0x8a00, %ax # 0x8a00 > port 0x8a00
8223 inb $0x64,%al # Wait for not busy 8273 movw %ax, %dx
8224 testb $0x2,%al 8274 outw %ax, %dx
8225 jnz seta20.1 8275 movw $0x8ae0, %ax # 0x8ae0 > port 0x8a00
8226 8276 outw %ax, %dx
8227 movb $0xd1,%al # 0xd1 > port 0x64 8277 spin:
8228 outb %al,$0x64 8278 jmp spin
8229 8279
8230 seta20.2: 8280 # Bootstrap GDT
8231 inb $0x64,%al # Wait for not busy 8281 .p2align 2 # force 4 byte alignment
8232 testb $0x2,%al 8282 gdt:
8233 jnz seta20.2 8283 SEG_NULLASM # null seg
8234 8284 SEG_ASM(STA_X|STA_R, 0x0, 0xffffffff) # code seg
8235 movb $0xdf,%al # 0xdf > port 0x60 8285 SEG_ASM(STA_W, 0x0, 0xffffffff) # data seg
8236 outb %al,$0x60 8286
8237 8287 gdtdesc:
8238 # Switch from real to protected mode. Use a bootstrap GDT that makes 8288 .word (gdtdesc gdt 1) # sizeof(gdt) 1
8239 # virtual addresses map dierctly to physical addresses so that the 8289 .long gdt # address gdt
8240 # effective memory map doesnt change during the transition. 8290
8241 lgdt gdtdesc 8291
8242 movl %cr0, %eax 8292
8243 orl $CR0_PE, %eax 8293
8244 movl %eax, %cr0 8294
8245 8295
8246 8296
8247 8297
8248 8298
8249 8299

Sheet 82 Sheet 82
Sep 5 23:39 2011 xv6/bootmain.c Page 1 Sep 5 23:39 2011 xv6/bootmain.c Page 2

8300 // Boot loader. 8350 void

8301 // 8351 waitdisk(void)
8302 // Part of the boot sector, along with bootasm.S, which calls bootmain(). 8352 {
8303 // bootasm.S has put the processor into protected 32bit mode. 8353 // Wait for disk ready.
8304 // bootmain() loads an ELF kernel image from the disk starting at 8354 while((inb(0x1F7) & 0xC0) != 0x40)
8305 // sector 1 and then jumps to the kernel entry routine. 8355 ;
8306 8356 }
8307 #include "types.h" 8357
8308 #include "elf.h" 8358 // Read a single sector at offset into dst.
8309 #include "x86.h" 8359 void
8310 #include "memlayout.h" 8360 readsect(void *dst, uint offset)
8311 8361 {
8312 #define SECTSIZE 512 8362 // Issue command.
8313 8363 waitdisk();
8314 void readseg(uchar*, uint, uint); 8364 outb(0x1F2, 1); // count = 1
8315 8365 outb(0x1F3, offset);
8316 void 8366 outb(0x1F4, offset >> 8);
8317 bootmain(void) 8367 outb(0x1F5, offset >> 16);
8318 { 8368 outb(0x1F6, (offset >> 24) | 0xE0);
8319 struct elfhdr *elf; 8369 outb(0x1F7, 0x20); // cmd 0x20 read sectors
8320 struct proghdr *ph, *eph; 8370
8321 void (*entry)(void); 8371 // Read data.
8322 uchar* pa; 8372 waitdisk();
8323 8373 insl(0x1F0, dst, SECTSIZE/4);
8324 elf = (struct elfhdr*)0x10000; // scratch space 8374 }
8325 8375
8326 // Read 1st page off disk 8376 // Read count bytes at offset from kernel into physical address pa.
8327 readseg((uchar*)elf, 4096, 0); 8377 // Might copy more than asked.
8328 8378 void
8329 // Is this an ELF executable? 8379 readseg(uchar* pa, uint count, uint offset)
8330 if(elf>magic != ELF_MAGIC) 8380 {
8331 return; // let bootasm.S handle error 8381 uchar* epa;
8332 8382
8333 // Load each program segment (ignores ph flags). 8383 epa = pa + count;
8334 ph = (struct proghdr*)((uchar*)elf + elf>phoff); 8384
8335 eph = ph + elf>phnum; 8385 // Round down to sector boundary.
8336 for(; ph < eph; ph++){ 8386 pa = offset % SECTSIZE;
8337 pa = (uchar*)ph>paddr; 8387
8338 readseg(pa, ph>filesz, ph>off); 8388 // Translate from bytes to sectors; kernel starts at sector 1.
8339 if(ph>memsz > ph>filesz) 8389 offset = (offset / SECTSIZE) + 1;
8340 stosb(pa + ph>filesz, 0, ph>memsz ph>filesz); 8390
8341 } 8391 // If this is too slow, we could read lots of sectors at a time.
8342 8392 // Wed write more to memory than asked, but it doesnt matter
8343 // Call the entry point from the ELF header. 8393 // we load in increasing order.
8344 // Does not return! 8394 for(; pa < epa; pa += SECTSIZE, offset++)
8345 entry = (void(*)(void))(elf>entry); 8395 readsect(pa, offset);
8346 entry(); 8396 }
8347 } 8397
8348 8398
8349 8399

Sheet 83 Sheet 83