Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | User Mode Linux HOWTO |
| 2 | User Mode Linux Core Team |
| 3 | Mon Nov 18 14:16:16 EST 2002 |
| 4 | |
| 5 | This document describes the use and abuse of Jeff Dike's User Mode |
| 6 | Linux: a port of the Linux kernel as a normal Intel Linux process. |
| 7 | ______________________________________________________________________ |
| 8 | |
| 9 | Table of Contents |
| 10 | |
| 11 | |
| 12 | |
| 13 | |
| 14 | |
| 15 | |
| 16 | |
| 17 | |
| 18 | |
| 19 | |
| 20 | |
| 21 | |
| 22 | |
| 23 | |
| 24 | |
| 25 | |
| 26 | |
| 27 | |
| 28 | |
| 29 | |
| 30 | |
| 31 | |
| 32 | |
| 33 | |
| 34 | |
| 35 | |
| 36 | |
| 37 | |
| 38 | |
| 39 | |
| 40 | |
| 41 | |
| 42 | |
| 43 | |
| 44 | |
| 45 | |
| 46 | |
| 47 | |
| 48 | |
| 49 | |
| 50 | |
| 51 | |
| 52 | |
| 53 | |
| 54 | |
| 55 | |
| 56 | |
| 57 | |
| 58 | |
| 59 | |
| 60 | |
| 61 | |
| 62 | |
| 63 | |
| 64 | |
| 65 | |
| 66 | |
| 67 | 1. Introduction |
| 68 | |
| 69 | 1.1 How is User Mode Linux Different? |
| 70 | 1.2 Why Would I Want User Mode Linux? |
| 71 | |
| 72 | 2. Compiling the kernel and modules |
| 73 | |
| 74 | 2.1 Compiling the kernel |
| 75 | 2.2 Compiling and installing kernel modules |
| 76 | 2.3 Compiling and installing uml_utilities |
| 77 | |
| 78 | 3. Running UML and logging in |
| 79 | |
| 80 | 3.1 Running UML |
| 81 | 3.2 Logging in |
| 82 | 3.3 Examples |
| 83 | |
| 84 | 4. UML on 2G/2G hosts |
| 85 | |
| 86 | 4.1 Introduction |
| 87 | 4.2 The problem |
| 88 | 4.3 The solution |
| 89 | |
| 90 | 5. Setting up serial lines and consoles |
| 91 | |
| 92 | 5.1 Specifying the device |
| 93 | 5.2 Specifying the channel |
| 94 | 5.3 Examples |
| 95 | |
| 96 | 6. Setting up the network |
| 97 | |
| 98 | 6.1 General setup |
| 99 | 6.2 Userspace daemons |
| 100 | 6.3 Specifying ethernet addresses |
| 101 | 6.4 UML interface setup |
| 102 | 6.5 Multicast |
| 103 | 6.6 TUN/TAP with the uml_net helper |
| 104 | 6.7 TUN/TAP with a preconfigured tap device |
| 105 | 6.8 Ethertap |
| 106 | 6.9 The switch daemon |
| 107 | 6.10 Slip |
| 108 | 6.11 Slirp |
| 109 | 6.12 pcap |
| 110 | 6.13 Setting up the host yourself |
| 111 | |
| 112 | 7. Sharing Filesystems between Virtual Machines |
| 113 | |
| 114 | 7.1 A warning |
| 115 | 7.2 Using layered block devices |
| 116 | 7.3 Note! |
| 117 | 7.4 Another warning |
| 118 | 7.5 uml_moo : Merging a COW file with its backing file |
| 119 | |
| 120 | 8. Creating filesystems |
| 121 | |
| 122 | 8.1 Create the filesystem file |
| 123 | 8.2 Assign the file to a UML device |
| 124 | 8.3 Creating and mounting the filesystem |
| 125 | |
| 126 | 9. Host file access |
| 127 | |
| 128 | 9.1 Using hostfs |
| 129 | 9.2 hostfs as the root filesystem |
| 130 | 9.3 Building hostfs |
| 131 | |
| 132 | 10. The Management Console |
| 133 | 10.1 version |
| 134 | 10.2 halt and reboot |
| 135 | 10.3 config |
| 136 | 10.4 remove |
| 137 | 10.5 sysrq |
| 138 | 10.6 help |
| 139 | 10.7 cad |
| 140 | 10.8 stop |
| 141 | 10.9 go |
| 142 | |
| 143 | 11. Kernel debugging |
| 144 | |
| 145 | 11.1 Starting the kernel under gdb |
| 146 | 11.2 Examining sleeping processes |
| 147 | 11.3 Running ddd on UML |
| 148 | 11.4 Debugging modules |
| 149 | 11.5 Attaching gdb to the kernel |
| 150 | 11.6 Using alternate debuggers |
| 151 | |
| 152 | 12. Kernel debugging examples |
| 153 | |
| 154 | 12.1 The case of the hung fsck |
| 155 | 12.2 Episode 2: The case of the hung fsck |
| 156 | |
| 157 | 13. What to do when UML doesn't work |
| 158 | |
| 159 | 13.1 Strange compilation errors when you build from source |
Adrian Bunk | bf6ee0a | 2006-10-03 22:17:48 +0200 | [diff] [blame^] | 160 | 13.2 (obsolete) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 161 | 13.3 A variety of panics and hangs with /tmp on a reiserfs filesystem |
| 162 | 13.4 The compile fails with errors about conflicting types for 'open', 'dup', and 'waitpid' |
| 163 | 13.5 UML doesn't work when /tmp is an NFS filesystem |
| 164 | 13.6 UML hangs on boot when compiled with gprof support |
| 165 | 13.7 syslogd dies with a SIGTERM on startup |
| 166 | 13.8 TUN/TAP networking doesn't work on a 2.4 host |
| 167 | 13.9 You can network to the host but not to other machines on the net |
| 168 | 13.10 I have no root and I want to scream |
| 169 | 13.11 UML build conflict between ptrace.h and ucontext.h |
| 170 | 13.12 The UML BogoMips is exactly half the host's BogoMips |
| 171 | 13.13 When you run UML, it immediately segfaults |
| 172 | 13.14 xterms appear, then immediately disappear |
| 173 | 13.15 Any other panic, hang, or strange behavior |
| 174 | |
| 175 | 14. Diagnosing Problems |
| 176 | |
| 177 | 14.1 Case 1 : Normal kernel panics |
| 178 | 14.2 Case 2 : Tracing thread panics |
| 179 | 14.3 Case 3 : Tracing thread panics caused by other threads |
| 180 | 14.4 Case 4 : Hangs |
| 181 | |
| 182 | 15. Thanks |
| 183 | |
| 184 | 15.1 Code and Documentation |
| 185 | 15.2 Flushing out bugs |
| 186 | 15.3 Buglets and clean-ups |
| 187 | 15.4 Case Studies |
| 188 | 15.5 Other contributions |
| 189 | |
| 190 | |
| 191 | ______________________________________________________________________ |
| 192 | |
| 193 | 11.. IInnttrroodduuccttiioonn |
| 194 | |
| 195 | Welcome to User Mode Linux. It's going to be fun. |
| 196 | |
| 197 | |
| 198 | |
| 199 | 11..11.. HHooww iiss UUsseerr MMooddee LLiinnuuxx DDiiffffeerreenntt?? |
| 200 | |
| 201 | Normally, the Linux Kernel talks straight to your hardware (video |
| 202 | card, keyboard, hard drives, etc), and any programs which run ask the |
| 203 | kernel to operate the hardware, like so: |
| 204 | |
| 205 | |
| 206 | |
| 207 | +-----------+-----------+----+ |
| 208 | | Process 1 | Process 2 | ...| |
| 209 | +-----------+-----------+----+ |
| 210 | | Linux Kernel | |
| 211 | +----------------------------+ |
| 212 | | Hardware | |
| 213 | +----------------------------+ |
| 214 | |
| 215 | |
| 216 | |
| 217 | |
| 218 | The User Mode Linux Kernel is different; instead of talking to the |
| 219 | hardware, it talks to a `real' Linux kernel (called the `host kernel' |
| 220 | from now on), like any other program. Programs can then run inside |
| 221 | User-Mode Linux as if they were running under a normal kernel, like |
| 222 | so: |
| 223 | |
| 224 | |
| 225 | |
| 226 | +----------------+ |
| 227 | | Process 2 | ...| |
| 228 | +-----------+----------------+ |
| 229 | | Process 1 | User-Mode Linux| |
| 230 | +----------------------------+ |
| 231 | | Linux Kernel | |
| 232 | +----------------------------+ |
| 233 | | Hardware | |
| 234 | +----------------------------+ |
| 235 | |
| 236 | |
| 237 | |
| 238 | |
| 239 | |
| 240 | 11..22.. WWhhyy WWoouulldd II WWaanntt UUsseerr MMooddee LLiinnuuxx?? |
| 241 | |
| 242 | |
| 243 | 1. If User Mode Linux crashes, your host kernel is still fine. |
| 244 | |
| 245 | 2. You can run a usermode kernel as a non-root user. |
| 246 | |
| 247 | 3. You can debug the User Mode Linux like any normal process. |
| 248 | |
| 249 | 4. You can run gprof (profiling) and gcov (coverage testing). |
| 250 | |
| 251 | 5. You can play with your kernel without breaking things. |
| 252 | |
| 253 | 6. You can use it as a sandbox for testing new apps. |
| 254 | |
| 255 | 7. You can try new development kernels safely. |
| 256 | |
| 257 | 8. You can run different distributions simultaneously. |
| 258 | |
| 259 | 9. It's extremely fun. |
| 260 | |
| 261 | |
| 262 | |
| 263 | |
| 264 | |
| 265 | 22.. CCoommppiilliinngg tthhee kkeerrnneell aanndd mmoodduulleess |
| 266 | |
| 267 | |
| 268 | |
| 269 | |
| 270 | 22..11.. CCoommppiilliinngg tthhee kkeerrnneell |
| 271 | |
| 272 | |
| 273 | Compiling the user mode kernel is just like compiling any other |
| 274 | kernel. Let's go through the steps, using 2.4.0-prerelease (current |
| 275 | as of this writing) as an example: |
| 276 | |
| 277 | |
| 278 | 1. Download the latest UML patch from |
| 279 | |
| 280 | the download page <http://user-mode-linux.sourceforge.net/dl- |
| 281 | sf.html> |
| 282 | |
| 283 | In this example, the file is uml-patch-2.4.0-prerelease.bz2. |
| 284 | |
| 285 | |
| 286 | 2. Download the matching kernel from your favourite kernel mirror, |
| 287 | such as: |
| 288 | |
| 289 | ftp://ftp.ca.kernel.org/pub/kernel/v2.4/linux-2.4.0-prerelease.tar.bz2 |
| 290 | <ftp://ftp.ca.kernel.org/pub/kernel/v2.4/linux-2.4.0-prerelease.tar.bz2> |
| 291 | . |
| 292 | |
| 293 | |
| 294 | 3. Make a directory and unpack the kernel into it. |
| 295 | |
| 296 | |
| 297 | |
| 298 | host% |
| 299 | mkdir ~/uml |
| 300 | |
| 301 | |
| 302 | |
| 303 | |
| 304 | |
| 305 | |
| 306 | host% |
| 307 | cd ~/uml |
| 308 | |
| 309 | |
| 310 | |
| 311 | |
| 312 | |
| 313 | |
| 314 | host% |
| 315 | tar -xzvf linux-2.4.0-prerelease.tar.bz2 |
| 316 | |
| 317 | |
| 318 | |
| 319 | |
| 320 | |
| 321 | |
| 322 | 4. Apply the patch using |
| 323 | |
| 324 | |
| 325 | |
| 326 | host% |
| 327 | cd ~/uml/linux |
| 328 | |
| 329 | |
| 330 | |
| 331 | host% |
| 332 | bzcat uml-patch-2.4.0-prerelease.bz2 | patch -p1 |
| 333 | |
| 334 | |
| 335 | |
| 336 | |
| 337 | |
| 338 | |
| 339 | 5. Run your favorite config; `make xconfig ARCH=um' is the most |
| 340 | convenient. `make config ARCH=um' and 'make menuconfig ARCH=um' |
| 341 | will work as well. The defaults will give you a useful kernel. If |
| 342 | you want to change something, go ahead, it probably won't hurt |
| 343 | anything. |
| 344 | |
| 345 | |
| 346 | Note: If the host is configured with a 2G/2G address space split |
| 347 | rather than the usual 3G/1G split, then the packaged UML binaries |
| 348 | will not run. They will immediately segfault. See ``UML on 2G/2G |
| 349 | hosts'' for the scoop on running UML on your system. |
| 350 | |
| 351 | |
| 352 | |
| 353 | 6. Finish with `make linux ARCH=um': the result is a file called |
| 354 | `linux' in the top directory of your source tree. |
| 355 | |
| 356 | Make sure that you don't build this kernel in /usr/src/linux. On some |
| 357 | distributions, /usr/include/asm is a link into this pool. The user- |
| 358 | mode build changes the other end of that link, and things that include |
| 359 | <asm/anything.h> stop compiling. |
| 360 | |
| 361 | The sources are also available from cvs at the project's cvs page, |
| 362 | which has directions on getting the sources. You can also browse the |
| 363 | CVS pool from there. |
| 364 | |
| 365 | If you get the CVS sources, you will have to check them out into an |
| 366 | empty directory. You will then have to copy each file into the |
| 367 | corresponding directory in the appropriate kernel pool. |
| 368 | |
| 369 | If you don't have the latest kernel pool, you can get the |
| 370 | corresponding user-mode sources with |
| 371 | |
| 372 | |
| 373 | host% cvs co -r v_2_3_x linux |
| 374 | |
| 375 | |
| 376 | |
| 377 | |
| 378 | where 'x' is the version in your pool. Note that you will not get the |
| 379 | bug fixes and enhancements that have gone into subsequent releases. |
| 380 | |
| 381 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 382 | 22..22.. CCoommppiilliinngg aanndd iinnssttaalllliinngg kkeerrnneell mmoodduulleess |
| 383 | |
| 384 | UML modules are built in the same way as the native kernel (with the |
| 385 | exception of the 'ARCH=um' that you always need for UML): |
| 386 | |
| 387 | |
| 388 | host% make modules ARCH=um |
| 389 | |
| 390 | |
| 391 | |
| 392 | |
| 393 | Any modules that you want to load into this kernel need to be built in |
| 394 | the user-mode pool. Modules from the native kernel won't work. |
| 395 | |
| 396 | You can install them by using ftp or something to copy them into the |
| 397 | virtual machine and dropping them into /lib/modules/`uname -r`. |
| 398 | |
| 399 | You can also get the kernel build process to install them as follows: |
| 400 | |
| 401 | 1. with the kernel not booted, mount the root filesystem in the top |
| 402 | level of the kernel pool: |
| 403 | |
| 404 | |
| 405 | host% mount root_fs mnt -o loop |
| 406 | |
| 407 | |
| 408 | |
| 409 | |
| 410 | |
| 411 | |
| 412 | 2. run |
| 413 | |
| 414 | |
| 415 | host% |
| 416 | make modules_install INSTALL_MOD_PATH=`pwd`/mnt ARCH=um |
| 417 | |
| 418 | |
| 419 | |
| 420 | |
| 421 | |
| 422 | |
| 423 | 3. unmount the filesystem |
| 424 | |
| 425 | |
| 426 | host% umount mnt |
| 427 | |
| 428 | |
| 429 | |
| 430 | |
| 431 | |
| 432 | |
| 433 | 4. boot the kernel on it |
| 434 | |
| 435 | |
| 436 | When the system is booted, you can use insmod as usual to get the |
| 437 | modules into the kernel. A number of things have been loaded into UML |
| 438 | as modules, especially filesystems and network protocols and filters, |
| 439 | so most symbols which need to be exported probably already are. |
| 440 | However, if you do find symbols that need exporting, let us |
| 441 | <http://user-mode-linux.sourceforge.net/contacts.html> know, and |
| 442 | they'll be "taken care of". |
| 443 | |
| 444 | |
| 445 | |
| 446 | 22..33.. CCoommppiilliinngg aanndd iinnssttaalllliinngg uummll__uuttiilliittiieess |
| 447 | |
| 448 | Many features of the UML kernel require a user-space helper program, |
| 449 | so a uml_utilities package is distributed separately from the kernel |
| 450 | patch which provides these helpers. Included within this is: |
| 451 | |
| 452 | +o port-helper - Used by consoles which connect to xterms or ports |
| 453 | |
| 454 | +o tunctl - Configuration tool to create and delete tap devices |
| 455 | |
| 456 | +o uml_net - Setuid binary for automatic tap device configuration |
| 457 | |
| 458 | +o uml_switch - User-space virtual switch required for daemon |
| 459 | transport |
| 460 | |
| 461 | The uml_utilities tree is compiled with: |
| 462 | |
| 463 | |
| 464 | host# |
| 465 | make && make install |
| 466 | |
| 467 | |
| 468 | |
| 469 | |
| 470 | Note that UML kernel patches may require a specific version of the |
| 471 | uml_utilities distribution. If you don't keep up with the mailing |
| 472 | lists, ensure that you have the latest release of uml_utilities if you |
| 473 | are experiencing problems with your UML kernel, particularly when |
| 474 | dealing with consoles or command-line switches to the helper programs |
| 475 | |
| 476 | |
| 477 | |
| 478 | |
| 479 | |
| 480 | |
| 481 | |
| 482 | |
| 483 | 33.. RRuunnnniinngg UUMMLL aanndd llooggggiinngg iinn |
| 484 | |
| 485 | |
| 486 | |
| 487 | 33..11.. RRuunnnniinngg UUMMLL |
| 488 | |
| 489 | It runs on 2.2.15 or later, and all 2.4 kernels. |
| 490 | |
| 491 | |
| 492 | Booting UML is straightforward. Simply run 'linux': it will try to |
| 493 | mount the file `root_fs' in the current directory. You do not need to |
| 494 | run it as root. If your root filesystem is not named `root_fs', then |
| 495 | you need to put a `ubd0=root_fs_whatever' switch on the linux command |
| 496 | line. |
| 497 | |
| 498 | |
| 499 | You will need a filesystem to boot UML from. There are a number |
| 500 | available for download from here <http://user-mode- |
| 501 | linux.sourceforge.net/dl-sf.html> . There are also several tools |
| 502 | <http://user-mode-linux.sourceforge.net/fs_making.html> which can be |
| 503 | used to generate UML-compatible filesystem images from media. |
| 504 | The kernel will boot up and present you with a login prompt. |
| 505 | |
| 506 | |
| 507 | Note: If the host is configured with a 2G/2G address space split |
| 508 | rather than the usual 3G/1G split, then the packaged UML binaries will |
| 509 | not run. They will immediately segfault. See ``UML on 2G/2G hosts'' |
| 510 | for the scoop on running UML on your system. |
| 511 | |
| 512 | |
| 513 | |
| 514 | 33..22.. LLooggggiinngg iinn |
| 515 | |
| 516 | |
| 517 | |
| 518 | The prepackaged filesystems have a root account with password 'root' |
| 519 | and a user account with password 'user'. The login banner will |
| 520 | generally tell you how to log in. So, you log in and you will find |
| 521 | yourself inside a little virtual machine. Our filesystems have a |
| 522 | variety of commands and utilities installed (and it is fairly easy to |
| 523 | add more), so you will have a lot of tools with which to poke around |
| 524 | the system. |
| 525 | |
| 526 | There are a couple of other ways to log in: |
| 527 | |
| 528 | +o On a virtual console |
| 529 | |
| 530 | |
| 531 | |
| 532 | Each virtual console that is configured (i.e. the device exists in |
| 533 | /dev and /etc/inittab runs a getty on it) will come up in its own |
| 534 | xterm. If you get tired of the xterms, read ``Setting up serial |
| 535 | lines and consoles'' to see how to attach the consoles to |
| 536 | something else, like host ptys. |
| 537 | |
| 538 | |
| 539 | |
| 540 | +o Over the serial line |
| 541 | |
| 542 | |
| 543 | In the boot output, find a line that looks like: |
| 544 | |
| 545 | |
| 546 | |
| 547 | serial line 0 assigned pty /dev/ptyp1 |
| 548 | |
| 549 | |
| 550 | |
| 551 | |
| 552 | Attach your favorite terminal program to the corresponding tty. I.e. |
| 553 | for minicom, the command would be |
| 554 | |
| 555 | |
| 556 | host% minicom -o -p /dev/ttyp1 |
| 557 | |
| 558 | |
| 559 | |
| 560 | |
| 561 | |
| 562 | |
| 563 | +o Over the net |
| 564 | |
| 565 | |
| 566 | If the network is running, then you can telnet to the virtual |
| 567 | machine and log in to it. See ``Setting up the network'' to learn |
| 568 | about setting up a virtual network. |
| 569 | |
| 570 | When you're done using it, run halt, and the kernel will bring itself |
| 571 | down and the process will exit. |
| 572 | |
| 573 | |
| 574 | 33..33.. EExxaammpplleess |
| 575 | |
| 576 | Here are some examples of UML in action: |
| 577 | |
| 578 | +o A login session <http://user-mode-linux.sourceforge.net/login.html> |
| 579 | |
| 580 | +o A virtual network <http://user-mode-linux.sourceforge.net/net.html> |
| 581 | |
| 582 | |
| 583 | |
| 584 | |
| 585 | |
| 586 | |
| 587 | |
| 588 | 44.. UUMMLL oonn 22GG//22GG hhoossttss |
| 589 | |
| 590 | |
| 591 | |
| 592 | |
| 593 | 44..11.. IInnttrroodduuccttiioonn |
| 594 | |
| 595 | |
| 596 | Most Linux machines are configured so that the kernel occupies the |
| 597 | upper 1G (0xc0000000 - 0xffffffff) of the 4G address space and |
| 598 | processes use the lower 3G (0x00000000 - 0xbfffffff). However, some |
| 599 | machine are configured with a 2G/2G split, with the kernel occupying |
| 600 | the upper 2G (0x80000000 - 0xffffffff) and processes using the lower |
| 601 | 2G (0x00000000 - 0x7fffffff). |
| 602 | |
| 603 | |
| 604 | |
| 605 | |
| 606 | 44..22.. TThhee pprroobblleemm |
| 607 | |
| 608 | |
| 609 | The prebuilt UML binaries on this site will not run on 2G/2G hosts |
| 610 | because UML occupies the upper .5G of the 3G process address space |
| 611 | (0xa0000000 - 0xbfffffff). Obviously, on 2G/2G hosts, this is right |
| 612 | in the middle of the kernel address space, so UML won't even load - it |
| 613 | will immediately segfault. |
| 614 | |
| 615 | |
| 616 | |
| 617 | |
| 618 | 44..33.. TThhee ssoolluuttiioonn |
| 619 | |
| 620 | |
| 621 | The fix for this is to rebuild UML from source after enabling |
| 622 | CONFIG_HOST_2G_2G (under 'General Setup'). This will cause UML to |
| 623 | load itself in the top .5G of that smaller process address space, |
| 624 | where it will run fine. See ``Compiling the kernel and modules'' if |
| 625 | you need help building UML from source. |
| 626 | |
| 627 | |
| 628 | |
| 629 | |
| 630 | |
| 631 | |
| 632 | |
| 633 | |
| 634 | |
| 635 | |
| 636 | 55.. SSeettttiinngg uupp sseerriiaall lliinneess aanndd ccoonnssoolleess |
| 637 | |
| 638 | |
| 639 | It is possible to attach UML serial lines and consoles to many types |
| 640 | of host I/O channels by specifying them on the command line. |
| 641 | |
| 642 | |
| 643 | You can attach them to host ptys, ttys, file descriptors, and ports. |
| 644 | This allows you to do things like |
| 645 | |
| 646 | +o have a UML console appear on an unused host console, |
| 647 | |
| 648 | +o hook two virtual machines together by having one attach to a pty |
| 649 | and having the other attach to the corresponding tty |
| 650 | |
| 651 | +o make a virtual machine accessible from the net by attaching a |
| 652 | console to a port on the host. |
| 653 | |
| 654 | |
| 655 | The general format of the command line option is device=channel. |
| 656 | |
| 657 | |
| 658 | |
| 659 | 55..11.. SSppeecciiffyyiinngg tthhee ddeevviiccee |
| 660 | |
| 661 | Devices are specified with "con" or "ssl" (console or serial line, |
| 662 | respectively), optionally with a device number if you are talking |
| 663 | about a specific device. |
| 664 | |
| 665 | |
| 666 | Using just "con" or "ssl" describes all of the consoles or serial |
| 667 | lines. If you want to talk about console #3 or serial line #10, they |
| 668 | would be "con3" and "ssl10", respectively. |
| 669 | |
| 670 | |
| 671 | A specific device name will override a less general "con=" or "ssl=". |
| 672 | So, for example, you can assign a pty to each of the serial lines |
| 673 | except for the first two like this: |
| 674 | |
| 675 | |
| 676 | ssl=pty ssl0=tty:/dev/tty0 ssl1=tty:/dev/tty1 |
| 677 | |
| 678 | |
| 679 | |
| 680 | |
| 681 | The specificity of the device name is all that matters; order on the |
| 682 | command line is irrelevant. |
| 683 | |
| 684 | |
| 685 | |
| 686 | 55..22.. SSppeecciiffyyiinngg tthhee cchhaannnneell |
| 687 | |
| 688 | There are a number of different types of channels to attach a UML |
| 689 | device to, each with a different way of specifying exactly what to |
| 690 | attach to. |
| 691 | |
| 692 | +o pseudo-terminals - device=pty pts terminals - device=pts |
| 693 | |
| 694 | |
| 695 | This will cause UML to allocate a free host pseudo-terminal for the |
| 696 | device. The terminal that it got will be announced in the boot |
| 697 | log. You access it by attaching a terminal program to the |
| 698 | corresponding tty: |
| 699 | |
| 700 | +o screen /dev/pts/n |
| 701 | |
| 702 | +o screen /dev/ttyxx |
| 703 | |
| 704 | +o minicom -o -p /dev/ttyxx - minicom seems not able to handle pts |
| 705 | devices |
| 706 | |
| 707 | +o kermit - start it up, 'open' the device, then 'connect' |
| 708 | |
| 709 | |
| 710 | |
| 711 | |
| 712 | |
| 713 | +o terminals - device=tty:tty device file |
| 714 | |
| 715 | |
| 716 | This will make UML attach the device to the specified tty (i.e |
| 717 | |
| 718 | |
| 719 | con1=tty:/dev/tty3 |
| 720 | |
| 721 | |
| 722 | |
| 723 | |
| 724 | will attach UML's console 1 to the host's /dev/tty3). If the tty that |
| 725 | you specify is the slave end of a tty/pty pair, something else must |
| 726 | have already opened the corresponding pty in order for this to work. |
| 727 | |
| 728 | |
| 729 | |
| 730 | |
| 731 | |
| 732 | +o xterms - device=xterm |
| 733 | |
| 734 | |
| 735 | UML will run an xterm and the device will be attached to it. |
| 736 | |
| 737 | |
| 738 | |
| 739 | |
| 740 | |
| 741 | +o Port - device=port:port number |
| 742 | |
| 743 | |
| 744 | This will attach the UML devices to the specified host port. |
| 745 | Attaching console 1 to the host's port 9000 would be done like |
| 746 | this: |
| 747 | |
| 748 | |
| 749 | con1=port:9000 |
| 750 | |
| 751 | |
| 752 | |
| 753 | |
| 754 | Attaching all the serial lines to that port would be done similarly: |
| 755 | |
| 756 | |
| 757 | ssl=port:9000 |
| 758 | |
| 759 | |
| 760 | |
| 761 | |
| 762 | You access these devices by telnetting to that port. Each active tel- |
| 763 | net session gets a different device. If there are more telnets to a |
| 764 | port than UML devices attached to it, then the extra telnet sessions |
| 765 | will block until an existing telnet detaches, or until another device |
| 766 | becomes active (i.e. by being activated in /etc/inittab). |
| 767 | |
| 768 | This channel has the advantage that you can both attach multiple UML |
| 769 | devices to it and know how to access them without reading the UML boot |
| 770 | log. It is also unique in allowing access to a UML from remote |
| 771 | machines without requiring that the UML be networked. This could be |
| 772 | useful in allowing public access to UMLs because they would be |
| 773 | accessible from the net, but wouldn't need any kind of network |
| 774 | filtering or access control because they would have no network access. |
| 775 | |
| 776 | |
| 777 | If you attach the main console to a portal, then the UML boot will |
| 778 | appear to hang. In reality, it's waiting for a telnet to connect, at |
| 779 | which point the boot will proceed. |
| 780 | |
| 781 | |
| 782 | |
| 783 | |
| 784 | |
| 785 | +o already-existing file descriptors - device=file descriptor |
| 786 | |
| 787 | |
| 788 | If you set up a file descriptor on the UML command line, you can |
| 789 | attach a UML device to it. This is most commonly used to put the |
| 790 | main console back on stdin and stdout after assigning all the other |
| 791 | consoles to something else: |
| 792 | |
| 793 | |
| 794 | con0=fd:0,fd:1 con=pts |
| 795 | |
| 796 | |
| 797 | |
| 798 | |
| 799 | |
| 800 | |
| 801 | |
| 802 | |
| 803 | +o Nothing - device=null |
| 804 | |
| 805 | |
| 806 | This allows the device to be opened, in contrast to 'none', but |
| 807 | reads will block, and writes will succeed and the data will be |
| 808 | thrown out. |
| 809 | |
| 810 | |
| 811 | |
| 812 | |
| 813 | |
| 814 | +o None - device=none |
| 815 | |
| 816 | |
Adrian Bunk | bf6ee0a | 2006-10-03 22:17:48 +0200 | [diff] [blame^] | 817 | This causes the device to disappear. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 818 | |
| 819 | |
| 820 | |
| 821 | You can also specify different input and output channels for a device |
| 822 | by putting a comma between them: |
| 823 | |
| 824 | |
| 825 | ssl3=tty:/dev/tty2,xterm |
| 826 | |
| 827 | |
| 828 | |
| 829 | |
| 830 | will cause serial line 3 to accept input on the host's /dev/tty3 and |
| 831 | display output on an xterm. That's a silly example - the most common |
| 832 | use of this syntax is to reattach the main console to stdin and stdout |
| 833 | as shown above. |
| 834 | |
| 835 | |
| 836 | If you decide to move the main console away from stdin/stdout, the |
| 837 | initial boot output will appear in the terminal that you're running |
| 838 | UML in. However, once the console driver has been officially |
| 839 | initialized, then the boot output will start appearing wherever you |
| 840 | specified that console 0 should be. That device will receive all |
| 841 | subsequent output. |
| 842 | |
| 843 | |
| 844 | |
| 845 | 55..33.. EExxaammpplleess |
| 846 | |
| 847 | There are a number of interesting things you can do with this |
| 848 | capability. |
| 849 | |
| 850 | |
| 851 | First, this is how you get rid of those bleeding console xterms by |
| 852 | attaching them to host ptys: |
| 853 | |
| 854 | |
| 855 | con=pty con0=fd:0,fd:1 |
| 856 | |
| 857 | |
| 858 | |
| 859 | |
| 860 | This will make a UML console take over an unused host virtual console, |
| 861 | so that when you switch to it, you will see the UML login prompt |
| 862 | rather than the host login prompt: |
| 863 | |
| 864 | |
| 865 | con1=tty:/dev/tty6 |
| 866 | |
| 867 | |
| 868 | |
| 869 | |
| 870 | You can attach two virtual machines together with what amounts to a |
| 871 | serial line as follows: |
| 872 | |
| 873 | Run one UML with a serial line attached to a pty - |
| 874 | |
| 875 | |
| 876 | ssl1=pty |
| 877 | |
| 878 | |
| 879 | |
| 880 | |
| 881 | Look at the boot log to see what pty it got (this example will assume |
| 882 | that it got /dev/ptyp1). |
| 883 | |
| 884 | Boot the other UML with a serial line attached to the corresponding |
| 885 | tty - |
| 886 | |
| 887 | |
| 888 | ssl1=tty:/dev/ttyp1 |
| 889 | |
| 890 | |
| 891 | |
| 892 | |
| 893 | Log in, make sure that it has no getty on that serial line, attach a |
| 894 | terminal program like minicom to it, and you should see the login |
| 895 | prompt of the other virtual machine. |
| 896 | |
| 897 | |
| 898 | 66.. SSeettttiinngg uupp tthhee nneettwwoorrkk |
| 899 | |
| 900 | |
| 901 | |
| 902 | This page describes how to set up the various transports and to |
| 903 | provide a UML instance with network access to the host, other machines |
| 904 | on the local net, and the rest of the net. |
| 905 | |
| 906 | |
| 907 | As of 2.4.5, UML networking has been completely redone to make it much |
| 908 | easier to set up, fix bugs, and add new features. |
| 909 | |
| 910 | |
| 911 | There is a new helper, uml_net, which does the host setup that |
| 912 | requires root privileges. |
| 913 | |
| 914 | |
| 915 | There are currently five transport types available for a UML virtual |
| 916 | machine to exchange packets with other hosts: |
| 917 | |
| 918 | +o ethertap |
| 919 | |
| 920 | +o TUN/TAP |
| 921 | |
| 922 | +o Multicast |
| 923 | |
| 924 | +o a switch daemon |
| 925 | |
| 926 | +o slip |
| 927 | |
| 928 | +o slirp |
| 929 | |
| 930 | +o pcap |
| 931 | |
| 932 | The TUN/TAP, ethertap, slip, and slirp transports allow a UML |
| 933 | instance to exchange packets with the host. They may be directed |
| 934 | to the host or the host may just act as a router to provide access |
| 935 | to other physical or virtual machines. |
| 936 | |
| 937 | |
| 938 | The pcap transport is a synthetic read-only interface, using the |
| 939 | libpcap binary to collect packets from interfaces on the host and |
| 940 | filter them. This is useful for building preconfigured traffic |
| 941 | monitors or sniffers. |
| 942 | |
| 943 | |
| 944 | The daemon and multicast transports provide a completely virtual |
| 945 | network to other virtual machines. This network is completely |
| 946 | disconnected from the physical network unless one of the virtual |
| 947 | machines on it is acting as a gateway. |
| 948 | |
| 949 | |
| 950 | With so many host transports, which one should you use? Here's when |
| 951 | you should use each one: |
| 952 | |
| 953 | +o ethertap - if you want access to the host networking and it is |
| 954 | running 2.2 |
| 955 | |
| 956 | +o TUN/TAP - if you want access to the host networking and it is |
| 957 | running 2.4. Also, the TUN/TAP transport is able to use a |
| 958 | preconfigured device, allowing it to avoid using the setuid uml_net |
| 959 | helper, which is a security advantage. |
| 960 | |
| 961 | +o Multicast - if you want a purely virtual network and you don't want |
| 962 | to set up anything but the UML |
| 963 | |
| 964 | +o a switch daemon - if you want a purely virtual network and you |
| 965 | don't mind running the daemon in order to get somewhat better |
| 966 | performance |
| 967 | |
| 968 | +o slip - there is no particular reason to run the slip backend unless |
| 969 | ethertap and TUN/TAP are just not available for some reason |
| 970 | |
| 971 | +o slirp - if you don't have root access on the host to setup |
| 972 | networking, or if you don't want to allocate an IP to your UML |
| 973 | |
| 974 | +o pcap - not much use for actual network connectivity, but great for |
| 975 | monitoring traffic on the host |
| 976 | |
| 977 | Ethertap is available on 2.4 and works fine. TUN/TAP is preferred |
| 978 | to it because it has better performance and ethertap is officially |
| 979 | considered obsolete in 2.4. Also, the root helper only needs to |
| 980 | run occasionally for TUN/TAP, rather than handling every packet, as |
| 981 | it does with ethertap. This is a slight security advantage since |
| 982 | it provides fewer opportunities for a nasty UML user to somehow |
| 983 | exploit the helper's root privileges. |
| 984 | |
| 985 | |
| 986 | 66..11.. GGeenneerraall sseettuupp |
| 987 | |
| 988 | First, you must have the virtual network enabled in your UML. If are |
| 989 | running a prebuilt kernel from this site, everything is already |
| 990 | enabled. If you build the kernel yourself, under the "Network device |
| 991 | support" menu, enable "Network device support", and then the three |
| 992 | transports. |
| 993 | |
| 994 | |
| 995 | The next step is to provide a network device to the virtual machine. |
| 996 | This is done by describing it on the kernel command line. |
| 997 | |
| 998 | The general format is |
| 999 | |
| 1000 | |
| 1001 | eth <n> = <transport> , <transport args> |
| 1002 | |
| 1003 | |
| 1004 | |
| 1005 | |
| 1006 | For example, a virtual ethernet device may be attached to a host |
| 1007 | ethertap device as follows: |
| 1008 | |
| 1009 | |
| 1010 | eth0=ethertap,tap0,fe:fd:0:0:0:1,192.168.0.254 |
| 1011 | |
| 1012 | |
| 1013 | |
| 1014 | |
| 1015 | This sets up eth0 inside the virtual machine to attach itself to the |
| 1016 | host /dev/tap0, assigns it an ethernet address, and assigns the host |
| 1017 | tap0 interface an IP address. |
| 1018 | |
| 1019 | |
| 1020 | |
| 1021 | Note that the IP address you assign to the host end of the tap device |
| 1022 | must be different than the IP you assign to the eth device inside UML. |
| 1023 | If you are short on IPs and don't want to comsume two per UML, then |
| 1024 | you can reuse the host's eth IP address for the host ends of the tap |
| 1025 | devices. Internally, the UMLs must still get unique IPs for their eth |
| 1026 | devices. You can also give the UMLs non-routable IPs (192.168.x.x or |
| 1027 | 10.x.x.x) and have the host masquerade them. This will let outgoing |
| 1028 | connections work, but incoming connections won't without more work, |
| 1029 | such as port forwarding from the host. |
| 1030 | Also note that when you configure the host side of an interface, it is |
| 1031 | only acting as a gateway. It will respond to pings sent to it |
| 1032 | locally, but is not useful to do that since it's a host interface. |
| 1033 | You are not talking to the UML when you ping that interface and get a |
| 1034 | response. |
| 1035 | |
| 1036 | |
| 1037 | You can also add devices to a UML and remove them at runtime. See the |
| 1038 | ``The Management Console'' page for details. |
| 1039 | |
| 1040 | |
| 1041 | The sections below describe this in more detail. |
| 1042 | |
| 1043 | |
| 1044 | Once you've decided how you're going to set up the devices, you boot |
| 1045 | UML, log in, configure the UML side of the devices, and set up routes |
| 1046 | to the outside world. At that point, you will be able to talk to any |
| 1047 | other machines, physical or virtual, on the net. |
| 1048 | |
| 1049 | |
| 1050 | If ifconfig inside UML fails and the network refuses to come up, run |
| 1051 | tell you what went wrong. |
| 1052 | |
| 1053 | |
| 1054 | |
| 1055 | 66..22.. UUsseerrssppaaccee ddaaeemmoonnss |
| 1056 | |
| 1057 | You will likely need the setuid helper, or the switch daemon, or both. |
| 1058 | They are both installed with the RPM and deb, so if you've installed |
| 1059 | either, you can skip the rest of this section. |
| 1060 | |
| 1061 | |
| 1062 | If not, then you need to check them out of CVS, build them, and |
| 1063 | install them. The helper is uml_net, in CVS /tools/uml_net, and the |
| 1064 | daemon is uml_switch, in CVS /tools/uml_router. They are both built |
| 1065 | with a plain 'make'. Both need to be installed in a directory that's |
| 1066 | in your path - /usr/bin is recommend. On top of that, uml_net needs |
| 1067 | to be setuid root. |
| 1068 | |
| 1069 | |
| 1070 | |
| 1071 | 66..33.. SSppeecciiffyyiinngg eetthheerrnneett aaddddrreesssseess |
| 1072 | |
| 1073 | Below, you will see that the TUN/TAP, ethertap, and daemon interfaces |
| 1074 | allow you to specify hardware addresses for the virtual ethernet |
| 1075 | devices. This is generally not necessary. If you don't have a |
| 1076 | specific reason to do it, you probably shouldn't. If one is not |
| 1077 | specified on the command line, the driver will assign one based on the |
| 1078 | device IP address. It will provide the address fe:fd:nn:nn:nn:nn |
| 1079 | where nn.nn.nn.nn is the device IP address. This is nearly always |
| 1080 | sufficient to guarantee a unique hardware address for the device. A |
| 1081 | couple of exceptions are: |
| 1082 | |
| 1083 | +o Another set of virtual ethernet devices are on the same network and |
| 1084 | they are assigned hardware addresses using a different scheme which |
| 1085 | may conflict with the UML IP address-based scheme |
| 1086 | |
| 1087 | +o You aren't going to use the device for IP networking, so you don't |
| 1088 | assign the device an IP address |
| 1089 | |
| 1090 | If you let the driver provide the hardware address, you should make |
| 1091 | sure that the device IP address is known before the interface is |
| 1092 | brought up. So, inside UML, this will guarantee that: |
| 1093 | |
| 1094 | |
| 1095 | |
| 1096 | UML# |
| 1097 | ifconfig eth0 192.168.0.250 up |
| 1098 | |
| 1099 | |
| 1100 | |
| 1101 | |
| 1102 | If you decide to assign the hardware address yourself, make sure that |
| 1103 | the first byte of the address is even. Addresses with an odd first |
| 1104 | byte are broadcast addresses, which you don't want assigned to a |
| 1105 | device. |
| 1106 | |
| 1107 | |
| 1108 | |
| 1109 | 66..44.. UUMMLL iinntteerrffaaccee sseettuupp |
| 1110 | |
| 1111 | Once the network devices have been described on the command line, you |
| 1112 | should boot UML and log in. |
| 1113 | |
| 1114 | |
| 1115 | The first thing to do is bring the interface up: |
| 1116 | |
| 1117 | |
| 1118 | UML# ifconfig ethn ip-address up |
| 1119 | |
| 1120 | |
| 1121 | |
| 1122 | |
| 1123 | You should be able to ping the host at this point. |
| 1124 | |
| 1125 | |
| 1126 | To reach the rest of the world, you should set a default route to the |
| 1127 | host: |
| 1128 | |
| 1129 | |
| 1130 | UML# route add default gw host ip |
| 1131 | |
| 1132 | |
| 1133 | |
| 1134 | |
| 1135 | Again, with host ip of 192.168.0.4: |
| 1136 | |
| 1137 | |
| 1138 | UML# route add default gw 192.168.0.4 |
| 1139 | |
| 1140 | |
| 1141 | |
| 1142 | |
| 1143 | This page used to recommend setting a network route to your local net. |
| 1144 | This is wrong, because it will cause UML to try to figure out hardware |
| 1145 | addresses of the local machines by arping on the interface to the |
| 1146 | host. Since that interface is basically a single strand of ethernet |
| 1147 | with two nodes on it (UML and the host) and arp requests don't cross |
| 1148 | networks, they will fail to elicit any responses. So, what you want |
| 1149 | is for UML to just blindly throw all packets at the host and let it |
| 1150 | figure out what to do with them, which is what leaving out the network |
| 1151 | route and adding the default route does. |
| 1152 | |
| 1153 | |
| 1154 | Note: If you can't communicate with other hosts on your physical |
| 1155 | ethernet, it's probably because of a network route that's |
| 1156 | automatically set up. If you run 'route -n' and see a route that |
| 1157 | looks like this: |
| 1158 | |
| 1159 | |
| 1160 | |
| 1161 | |
| 1162 | Destination Gateway Genmask Flags Metric Ref Use Iface |
| 1163 | 192.168.0.0 0.0.0.0 255.255.255.0 U 0 0 0 eth0 |
| 1164 | |
| 1165 | |
| 1166 | |
| 1167 | |
| 1168 | with a mask that's not 255.255.255.255, then replace it with a route |
| 1169 | to your host: |
| 1170 | |
| 1171 | |
| 1172 | UML# |
| 1173 | route del -net 192.168.0.0 dev eth0 netmask 255.255.255.0 |
| 1174 | |
| 1175 | |
| 1176 | |
| 1177 | |
| 1178 | |
| 1179 | |
| 1180 | UML# |
| 1181 | route add -host 192.168.0.4 dev eth0 |
| 1182 | |
| 1183 | |
| 1184 | |
| 1185 | |
| 1186 | This, plus the default route to the host, will allow UML to exchange |
| 1187 | packets with any machine on your ethernet. |
| 1188 | |
| 1189 | |
| 1190 | |
| 1191 | 66..55.. MMuullttiiccaasstt |
| 1192 | |
| 1193 | The simplest way to set up a virtual network between multiple UMLs is |
| 1194 | to use the mcast transport. This was written by Harald Welte and is |
| 1195 | present in UML version 2.4.5-5um and later. Your system must have |
| 1196 | multicast enabled in the kernel and there must be a multicast-capable |
| 1197 | network device on the host. Normally, this is eth0, but if there is |
| 1198 | no ethernet card on the host, then you will likely get strange error |
| 1199 | messages when you bring the device up inside UML. |
| 1200 | |
| 1201 | |
| 1202 | To use it, run two UMLs with |
| 1203 | |
| 1204 | |
| 1205 | eth0=mcast |
| 1206 | |
| 1207 | |
| 1208 | |
| 1209 | |
| 1210 | on their command lines. Log in, configure the ethernet device in each |
| 1211 | machine with different IP addresses: |
| 1212 | |
| 1213 | |
| 1214 | UML1# ifconfig eth0 192.168.0.254 |
| 1215 | |
| 1216 | |
| 1217 | |
| 1218 | |
| 1219 | |
| 1220 | |
| 1221 | UML2# ifconfig eth0 192.168.0.253 |
| 1222 | |
| 1223 | |
| 1224 | |
| 1225 | |
| 1226 | and they should be able to talk to each other. |
| 1227 | |
| 1228 | The full set of command line options for this transport are |
| 1229 | |
| 1230 | |
| 1231 | |
| 1232 | ethn=mcast,ethernet address,multicast |
| 1233 | address,multicast port,ttl |
| 1234 | |
| 1235 | |
| 1236 | |
| 1237 | |
| 1238 | Harald's original README is here <http://user-mode-linux.source- |
| 1239 | forge.net/text/mcast.txt> and explains these in detail, as well as |
| 1240 | some other issues. |
| 1241 | |
| 1242 | |
| 1243 | |
| 1244 | 66..66.. TTUUNN//TTAAPP wwiitthh tthhee uummll__nneett hheellppeerr |
| 1245 | |
| 1246 | TUN/TAP is the preferred mechanism on 2.4 to exchange packets with the |
| 1247 | host. The TUN/TAP backend has been in UML since 2.4.9-3um. |
| 1248 | |
| 1249 | |
| 1250 | The easiest way to get up and running is to let the setuid uml_net |
| 1251 | helper do the host setup for you. This involves insmod-ing the tun.o |
| 1252 | module if necessary, configuring the device, and setting up IP |
| 1253 | forwarding, routing, and proxy arp. If you are new to UML networking, |
| 1254 | do this first. If you're concerned about the security implications of |
| 1255 | the setuid helper, use it to get up and running, then read the next |
| 1256 | section to see how to have UML use a preconfigured tap device, which |
| 1257 | avoids the use of uml_net. |
| 1258 | |
| 1259 | |
| 1260 | If you specify an IP address for the host side of the device, the |
| 1261 | uml_net helper will do all necessary setup on the host - the only |
| 1262 | requirement is that TUN/TAP be available, either built in to the host |
| 1263 | kernel or as the tun.o module. |
| 1264 | |
| 1265 | The format of the command line switch to attach a device to a TUN/TAP |
| 1266 | device is |
| 1267 | |
| 1268 | |
| 1269 | eth <n> =tuntap,,, <IP address> |
| 1270 | |
| 1271 | |
| 1272 | |
| 1273 | |
| 1274 | For example, this argument will attach the UML's eth0 to the next |
| 1275 | available tap device and assign an ethernet address to it based on its |
| 1276 | IP address |
| 1277 | |
| 1278 | |
| 1279 | eth0=tuntap,,,192.168.0.254 |
| 1280 | |
| 1281 | |
| 1282 | |
| 1283 | |
| 1284 | |
| 1285 | |
| 1286 | Note that the IP address that must be used for the eth device inside |
| 1287 | UML is fixed by the routing and proxy arp that is set up on the |
| 1288 | TUN/TAP device on the host. You can use a different one, but it won't |
| 1289 | work because reply packets won't reach the UML. This is a feature. |
| 1290 | It prevents a nasty UML user from doing things like setting the UML IP |
| 1291 | to the same as the network's nameserver or mail server. |
| 1292 | |
| 1293 | |
| 1294 | There are a couple potential problems with running the TUN/TAP |
| 1295 | transport on a 2.4 host kernel |
| 1296 | |
| 1297 | +o TUN/TAP seems not to work on 2.4.3 and earlier. Upgrade the host |
| 1298 | kernel or use the ethertap transport. |
| 1299 | |
| 1300 | +o With an upgraded kernel, TUN/TAP may fail with |
| 1301 | |
| 1302 | |
| 1303 | File descriptor in bad state |
| 1304 | |
| 1305 | |
| 1306 | |
| 1307 | |
| 1308 | This is due to a header mismatch between the upgraded kernel and the |
| 1309 | kernel that was originally installed on the machine. The fix is to |
| 1310 | make sure that /usr/src/linux points to the headers for the running |
| 1311 | kernel. |
| 1312 | |
| 1313 | These were pointed out by Tim Robinson <timro at trkr dot net> in |
| 1314 | <http://www.geocrawler.com/lists/3/SourceForge/597/0/> name="this uml- |
| 1315 | user post"> . |
| 1316 | |
| 1317 | |
| 1318 | |
| 1319 | 66..77.. TTUUNN//TTAAPP wwiitthh aa pprreeccoonnffiigguurreedd ttaapp ddeevviiccee |
| 1320 | |
| 1321 | If you prefer not to have UML use uml_net (which is somewhat |
| 1322 | insecure), with UML 2.4.17-11, you can set up a TUN/TAP device |
| 1323 | beforehand. The setup needs to be done as root, but once that's done, |
| 1324 | there is no need for root assistance. Setting up the device is done |
| 1325 | as follows: |
| 1326 | |
| 1327 | +o Create the device with tunctl (available from the UML utilities |
| 1328 | tarball) |
| 1329 | |
| 1330 | |
| 1331 | |
| 1332 | |
| 1333 | host# tunctl -u uid |
| 1334 | |
| 1335 | |
| 1336 | |
| 1337 | |
| 1338 | where uid is the user id or username that UML will be run as. This |
| 1339 | will tell you what device was created. |
| 1340 | |
| 1341 | +o Configure the device IP (change IP addresses and device name to |
| 1342 | suit) |
| 1343 | |
| 1344 | |
| 1345 | |
| 1346 | |
| 1347 | host# ifconfig tap0 192.168.0.254 up |
| 1348 | |
| 1349 | |
| 1350 | |
| 1351 | |
| 1352 | |
| 1353 | +o Set up routing and arping if desired - this is my recipe, there are |
| 1354 | other ways of doing the same thing |
| 1355 | |
| 1356 | |
| 1357 | host# |
| 1358 | bash -c 'echo 1 > /proc/sys/net/ipv4/ip_forward' |
| 1359 | |
| 1360 | host# |
| 1361 | route add -host 192.168.0.253 dev tap0 |
| 1362 | |
| 1363 | |
| 1364 | |
| 1365 | |
| 1366 | |
| 1367 | |
| 1368 | host# |
| 1369 | bash -c 'echo 1 > /proc/sys/net/ipv4/conf/tap0/proxy_arp' |
| 1370 | |
| 1371 | |
| 1372 | |
| 1373 | |
| 1374 | |
| 1375 | |
| 1376 | host# |
| 1377 | arp -Ds 192.168.0.253 eth0 pub |
| 1378 | |
| 1379 | |
| 1380 | |
| 1381 | |
| 1382 | Note that this must be done every time the host boots - this configu- |
| 1383 | ration is not stored across host reboots. So, it's probably a good |
| 1384 | idea to stick it in an rc file. An even better idea would be a little |
| 1385 | utility which reads the information from a config file and sets up |
| 1386 | devices at boot time. |
| 1387 | |
| 1388 | +o Rather than using up two IPs and ARPing for one of them, you can |
| 1389 | also provide direct access to your LAN by the UML by using a |
| 1390 | bridge. |
| 1391 | |
| 1392 | |
| 1393 | host# |
| 1394 | brctl addbr br0 |
| 1395 | |
| 1396 | |
| 1397 | |
| 1398 | |
| 1399 | |
| 1400 | |
| 1401 | host# |
| 1402 | ifconfig eth0 0.0.0.0 promisc up |
| 1403 | |
| 1404 | |
| 1405 | |
| 1406 | |
| 1407 | |
| 1408 | |
| 1409 | host# |
| 1410 | ifconfig tap0 0.0.0.0 promisc up |
| 1411 | |
| 1412 | |
| 1413 | |
| 1414 | |
| 1415 | |
| 1416 | |
| 1417 | host# |
| 1418 | ifconfig br0 192.168.0.1 netmask 255.255.255.0 up |
| 1419 | |
| 1420 | |
| 1421 | |
| 1422 | |
| 1423 | |
| 1424 | |
| 1425 | |
| 1426 | host# |
| 1427 | brctl stp br0 off |
| 1428 | |
| 1429 | |
| 1430 | |
| 1431 | |
| 1432 | |
| 1433 | |
| 1434 | host# |
| 1435 | brctl setfd br0 1 |
| 1436 | |
| 1437 | |
| 1438 | |
| 1439 | |
| 1440 | |
| 1441 | |
| 1442 | host# |
| 1443 | brctl sethello br0 1 |
| 1444 | |
| 1445 | |
| 1446 | |
| 1447 | |
| 1448 | |
| 1449 | |
| 1450 | host# |
| 1451 | brctl addif br0 eth0 |
| 1452 | |
| 1453 | |
| 1454 | |
| 1455 | |
| 1456 | |
| 1457 | |
| 1458 | host# |
| 1459 | brctl addif br0 tap0 |
| 1460 | |
| 1461 | |
| 1462 | |
| 1463 | |
| 1464 | Note that 'br0' should be setup using ifconfig with the existing IP |
| 1465 | address of eth0, as eth0 no longer has its own IP. |
| 1466 | |
| 1467 | +o |
| 1468 | |
| 1469 | |
| 1470 | Also, the /dev/net/tun device must be writable by the user running |
| 1471 | UML in order for the UML to use the device that's been configured |
| 1472 | for it. The simplest thing to do is |
| 1473 | |
| 1474 | |
| 1475 | host# chmod 666 /dev/net/tun |
| 1476 | |
| 1477 | |
| 1478 | |
| 1479 | |
| 1480 | Making it world-writeable looks bad, but it seems not to be |
| 1481 | exploitable as a security hole. However, it does allow anyone to cre- |
| 1482 | ate useless tap devices (useless because they can't configure them), |
| 1483 | which is a DOS attack. A somewhat more secure alternative would to be |
| 1484 | to create a group containing all the users who have preconfigured tap |
| 1485 | devices and chgrp /dev/net/tun to that group with mode 664 or 660. |
| 1486 | |
| 1487 | |
| 1488 | +o Once the device is set up, run UML with 'eth0=tuntap,device name' |
| 1489 | (i.e. 'eth0=tuntap,tap0') on the command line (or do it with the |
| 1490 | mconsole config command). |
| 1491 | |
| 1492 | +o Bring the eth device up in UML and you're in business. |
| 1493 | |
| 1494 | If you don't want that tap device any more, you can make it non- |
| 1495 | persistent with |
| 1496 | |
| 1497 | |
| 1498 | host# tunctl -d tap device |
| 1499 | |
| 1500 | |
| 1501 | |
| 1502 | |
| 1503 | Finally, tunctl has a -b (for brief mode) switch which causes it to |
| 1504 | output only the name of the tap device it created. This makes it |
| 1505 | suitable for capture by a script: |
| 1506 | |
| 1507 | |
| 1508 | host# TAP=`tunctl -u 1000 -b` |
| 1509 | |
| 1510 | |
| 1511 | |
| 1512 | |
| 1513 | |
| 1514 | |
| 1515 | 66..88.. EEtthheerrttaapp |
| 1516 | |
| 1517 | Ethertap is the general mechanism on 2.2 for userspace processes to |
| 1518 | exchange packets with the kernel. |
| 1519 | |
| 1520 | |
| 1521 | |
| 1522 | To use this transport, you need to describe the virtual network device |
| 1523 | on the UML command line. The general format for this is |
| 1524 | |
| 1525 | |
| 1526 | eth <n> =ethertap, <device> , <ethernet address> , <tap IP address> |
| 1527 | |
| 1528 | |
| 1529 | |
| 1530 | |
| 1531 | So, the previous example |
| 1532 | |
| 1533 | |
| 1534 | eth0=ethertap,tap0,fe:fd:0:0:0:1,192.168.0.254 |
| 1535 | |
| 1536 | |
| 1537 | |
| 1538 | |
| 1539 | attaches the UML eth0 device to the host /dev/tap0, assigns it the |
| 1540 | ethernet address fe:fd:0:0:0:1, and assigns the IP address |
| 1541 | 192.168.0.254 to the tap device. |
| 1542 | |
| 1543 | |
| 1544 | |
| 1545 | The tap device is mandatory, but the others are optional. If the |
| 1546 | ethernet address is omitted, one will be assigned to it. |
| 1547 | |
| 1548 | |
| 1549 | The presence of the tap IP address will cause the helper to run and do |
| 1550 | whatever host setup is needed to allow the virtual machine to |
| 1551 | communicate with the outside world. If you're not sure you know what |
| 1552 | you're doing, this is the way to go. |
| 1553 | |
| 1554 | |
| 1555 | If it is absent, then you must configure the tap device and whatever |
| 1556 | arping and routing you will need on the host. However, even in this |
| 1557 | case, the uml_net helper still needs to be in your path and it must be |
| 1558 | setuid root if you're not running UML as root. This is because the |
| 1559 | tap device doesn't support SIGIO, which UML needs in order to use |
| 1560 | something as a source of input. So, the helper is used as a |
| 1561 | convenient asynchronous IO thread. |
| 1562 | |
| 1563 | If you're using the uml_net helper, you can ignore the following host |
| 1564 | setup - uml_net will do it for you. You just need to make sure you |
| 1565 | have ethertap available, either built in to the host kernel or |
| 1566 | available as a module. |
| 1567 | |
| 1568 | |
| 1569 | If you want to set things up yourself, you need to make sure that the |
| 1570 | appropriate /dev entry exists. If it doesn't, become root and create |
| 1571 | it as follows: |
| 1572 | |
| 1573 | |
| 1574 | mknod /dev/tap <minor> c 36 <minor> + 16 |
| 1575 | |
| 1576 | |
| 1577 | |
| 1578 | |
| 1579 | For example, this is how to create /dev/tap0: |
| 1580 | |
| 1581 | |
| 1582 | mknod /dev/tap0 c 36 0 + 16 |
| 1583 | |
| 1584 | |
| 1585 | |
| 1586 | |
| 1587 | You also need to make sure that the host kernel has ethertap support. |
| 1588 | If ethertap is enabled as a module, you apparently need to insmod |
| 1589 | ethertap once for each ethertap device you want to enable. So, |
| 1590 | |
| 1591 | |
| 1592 | host# |
| 1593 | insmod ethertap |
| 1594 | |
| 1595 | |
| 1596 | |
| 1597 | |
| 1598 | will give you the tap0 interface. To get the tap1 interface, you need |
| 1599 | to run |
| 1600 | |
| 1601 | |
| 1602 | host# |
| 1603 | insmod ethertap unit=1 -o ethertap1 |
| 1604 | |
| 1605 | |
| 1606 | |
| 1607 | |
| 1608 | |
| 1609 | |
| 1610 | |
| 1611 | 66..99.. TThhee sswwiittcchh ddaaeemmoonn |
| 1612 | |
| 1613 | NNoottee: This is the daemon formerly known as uml_router, but which was |
| 1614 | renamed so the network weenies of the world would stop growling at me. |
| 1615 | |
| 1616 | |
| 1617 | The switch daemon, uml_switch, provides a mechanism for creating a |
| 1618 | totally virtual network. By default, it provides no connection to the |
| 1619 | host network (but see -tap, below). |
| 1620 | |
| 1621 | |
| 1622 | The first thing you need to do is run the daemon. Running it with no |
| 1623 | arguments will make it listen on a default pair of unix domain |
| 1624 | sockets. |
| 1625 | |
| 1626 | |
| 1627 | If you want it to listen on a different pair of sockets, use |
| 1628 | |
| 1629 | |
| 1630 | -unix control socket data socket |
| 1631 | |
| 1632 | |
| 1633 | |
| 1634 | |
| 1635 | |
| 1636 | If you want it to act as a hub rather than a switch, use |
| 1637 | |
| 1638 | |
| 1639 | -hub |
| 1640 | |
| 1641 | |
| 1642 | |
| 1643 | |
| 1644 | |
| 1645 | If you want the switch to be connected to host networking (allowing |
| 1646 | the umls to get access to the outside world through the host), use |
| 1647 | |
| 1648 | |
| 1649 | -tap tap0 |
| 1650 | |
| 1651 | |
| 1652 | |
| 1653 | |
| 1654 | |
| 1655 | Note that the tap device must be preconfigured (see "TUN/TAP with a |
| 1656 | preconfigured tap device", above). If you're using a different tap |
| 1657 | device than tap0, specify that instead of tap0. |
| 1658 | |
| 1659 | |
| 1660 | uml_switch can be backgrounded as follows |
| 1661 | |
| 1662 | |
| 1663 | host% |
| 1664 | uml_switch [ options ] < /dev/null > /dev/null |
| 1665 | |
| 1666 | |
| 1667 | |
| 1668 | |
| 1669 | The reason it doesn't background by default is that it listens to |
| 1670 | stdin for EOF. When it sees that, it exits. |
| 1671 | |
| 1672 | |
| 1673 | The general format of the kernel command line switch is |
| 1674 | |
| 1675 | |
| 1676 | |
| 1677 | ethn=daemon,ethernet address,socket |
| 1678 | type,control socket,data socket |
| 1679 | |
| 1680 | |
| 1681 | |
| 1682 | |
| 1683 | You can leave off everything except the 'daemon'. You only need to |
| 1684 | specify the ethernet address if the one that will be assigned to it |
| 1685 | isn't acceptable for some reason. The rest of the arguments describe |
| 1686 | how to communicate with the daemon. You should only specify them if |
| 1687 | you told the daemon to use different sockets than the default. So, if |
| 1688 | you ran the daemon with no arguments, running the UML on the same |
| 1689 | machine with |
| 1690 | eth0=daemon |
| 1691 | |
| 1692 | |
| 1693 | |
| 1694 | |
| 1695 | will cause the eth0 driver to attach itself to the daemon correctly. |
| 1696 | |
| 1697 | |
| 1698 | |
| 1699 | 66..1100.. SSlliipp |
| 1700 | |
| 1701 | Slip is another, less general, mechanism for a process to communicate |
| 1702 | with the host networking. In contrast to the ethertap interface, |
| 1703 | which exchanges ethernet frames with the host and can be used to |
| 1704 | transport any higher-level protocol, it can only be used to transport |
| 1705 | IP. |
| 1706 | |
| 1707 | |
| 1708 | The general format of the command line switch is |
| 1709 | |
| 1710 | |
| 1711 | |
| 1712 | ethn=slip,slip IP |
| 1713 | |
| 1714 | |
| 1715 | |
| 1716 | |
| 1717 | The slip IP argument is the IP address that will be assigned to the |
| 1718 | host end of the slip device. If it is specified, the helper will run |
| 1719 | and will set up the host so that the virtual machine can reach it and |
| 1720 | the rest of the network. |
| 1721 | |
| 1722 | |
| 1723 | There are some oddities with this interface that you should be aware |
| 1724 | of. You should only specify one slip device on a given virtual |
| 1725 | machine, and its name inside UML will be 'umn', not 'eth0' or whatever |
| 1726 | you specified on the command line. These problems will be fixed at |
| 1727 | some point. |
| 1728 | |
| 1729 | |
| 1730 | |
| 1731 | 66..1111.. SSlliirrpp |
| 1732 | |
| 1733 | slirp uses an external program, usually /usr/bin/slirp, to provide IP |
| 1734 | only networking connectivity through the host. This is similar to IP |
| 1735 | masquerading with a firewall, although the translation is performed in |
| 1736 | user-space, rather than by the kernel. As slirp does not set up any |
| 1737 | interfaces on the host, or changes routing, slirp does not require |
| 1738 | root access or setuid binaries on the host. |
| 1739 | |
| 1740 | |
| 1741 | The general format of the command line switch for slirp is: |
| 1742 | |
| 1743 | |
| 1744 | |
| 1745 | ethn=slirp,ethernet address,slirp path |
| 1746 | |
| 1747 | |
| 1748 | |
| 1749 | |
| 1750 | The ethernet address is optional, as UML will set up the interface |
| 1751 | with an ethernet address based upon the initial IP address of the |
| 1752 | interface. The slirp path is generally /usr/bin/slirp, although it |
| 1753 | will depend on distribution. |
| 1754 | |
| 1755 | |
| 1756 | The slirp program can have a number of options passed to the command |
| 1757 | line and we can't add them to the UML command line, as they will be |
| 1758 | parsed incorrectly. Instead, a wrapper shell script can be written or |
| 1759 | the options inserted into the /.slirprc file. More information on |
| 1760 | all of the slirp options can be found in its man pages. |
| 1761 | |
| 1762 | |
| 1763 | The eth0 interface on UML should be set up with the IP 10.2.0.15, |
| 1764 | although you can use anything as long as it is not used by a network |
| 1765 | you will be connecting to. The default route on UML should be set to |
| 1766 | use |
| 1767 | |
| 1768 | |
| 1769 | UML# |
| 1770 | route add default dev eth0 |
| 1771 | |
| 1772 | |
| 1773 | |
| 1774 | |
| 1775 | slirp provides a number of useful IP addresses which can be used by |
| 1776 | UML, such as 10.0.2.3 which is an alias for the DNS server specified |
| 1777 | in /etc/resolv.conf on the host or the IP given in the 'dns' option |
| 1778 | for slirp. |
| 1779 | |
| 1780 | |
| 1781 | Even with a baudrate setting higher than 115200, the slirp connection |
| 1782 | is limited to 115200. If you need it to go faster, the slirp binary |
| 1783 | needs to be compiled with FULL_BOLT defined in config.h. |
| 1784 | |
| 1785 | |
| 1786 | |
| 1787 | 66..1122.. ppccaapp |
| 1788 | |
| 1789 | The pcap transport is attached to a UML ethernet device on the command |
| 1790 | line or with uml_mconsole with the following syntax: |
| 1791 | |
| 1792 | |
| 1793 | |
| 1794 | ethn=pcap,host interface,filter |
| 1795 | expression,option1,option2 |
| 1796 | |
| 1797 | |
| 1798 | |
| 1799 | |
| 1800 | The expression and options are optional. |
| 1801 | |
| 1802 | |
| 1803 | The interface is whatever network device on the host you want to |
| 1804 | sniff. The expression is a pcap filter expression, which is also what |
| 1805 | tcpdump uses, so if you know how to specify tcpdump filters, you will |
| 1806 | use the same expressions here. The options are up to two of |
| 1807 | 'promisc', control whether pcap puts the host interface into |
| 1808 | promiscuous mode. 'optimize' and 'nooptimize' control whether the pcap |
| 1809 | expression optimizer is used. |
| 1810 | |
| 1811 | |
| 1812 | Example: |
| 1813 | |
| 1814 | |
| 1815 | |
| 1816 | eth0=pcap,eth0,tcp |
| 1817 | |
| 1818 | eth1=pcap,eth0,!tcp |
| 1819 | |
| 1820 | |
| 1821 | |
| 1822 | will cause the UML eth0 to emit all tcp packets on the host eth0 and |
| 1823 | the UML eth1 to emit all non-tcp packets on the host eth0. |
| 1824 | |
| 1825 | |
| 1826 | |
| 1827 | 66..1133.. SSeettttiinngg uupp tthhee hhoosstt yyoouurrsseellff |
| 1828 | |
| 1829 | If you don't specify an address for the host side of the ethertap or |
| 1830 | slip device, UML won't do any setup on the host. So this is what is |
| 1831 | needed to get things working (the examples use a host-side IP of |
| 1832 | 192.168.0.251 and a UML-side IP of 192.168.0.250 - adjust to suit your |
| 1833 | own network): |
| 1834 | |
| 1835 | +o The device needs to be configured with its IP address. Tap devices |
| 1836 | are also configured with an mtu of 1484. Slip devices are |
| 1837 | configured with a point-to-point address pointing at the UML ip |
| 1838 | address. |
| 1839 | |
| 1840 | |
| 1841 | host# ifconfig tap0 arp mtu 1484 192.168.0.251 up |
| 1842 | |
| 1843 | |
| 1844 | |
| 1845 | |
| 1846 | |
| 1847 | |
| 1848 | host# |
| 1849 | ifconfig sl0 192.168.0.251 pointopoint 192.168.0.250 up |
| 1850 | |
| 1851 | |
| 1852 | |
| 1853 | |
| 1854 | |
| 1855 | +o If a tap device is being set up, a route is set to the UML IP. |
| 1856 | |
| 1857 | |
| 1858 | UML# route add -host 192.168.0.250 gw 192.168.0.251 |
| 1859 | |
| 1860 | |
| 1861 | |
| 1862 | |
| 1863 | |
| 1864 | +o To allow other hosts on your network to see the virtual machine, |
| 1865 | proxy arp is set up for it. |
| 1866 | |
| 1867 | |
| 1868 | host# arp -Ds 192.168.0.250 eth0 pub |
| 1869 | |
| 1870 | |
| 1871 | |
| 1872 | |
| 1873 | |
| 1874 | +o Finally, the host is set up to route packets. |
| 1875 | |
| 1876 | |
| 1877 | host# echo 1 > /proc/sys/net/ipv4/ip_forward |
| 1878 | |
| 1879 | |
| 1880 | |
| 1881 | |
| 1882 | |
| 1883 | |
| 1884 | |
| 1885 | |
| 1886 | |
| 1887 | |
| 1888 | 77.. SShhaarriinngg FFiilleessyysstteemmss bbeettwweeeenn VViirrttuuaall MMaacchhiinneess |
| 1889 | |
| 1890 | |
| 1891 | |
| 1892 | |
| 1893 | 77..11.. AA wwaarrnniinngg |
| 1894 | |
| 1895 | Don't attempt to share filesystems simply by booting two UMLs from the |
| 1896 | same file. That's the same thing as booting two physical machines |
| 1897 | from a shared disk. It will result in filesystem corruption. |
| 1898 | |
| 1899 | |
| 1900 | |
| 1901 | 77..22.. UUssiinngg llaayyeerreedd bblloocckk ddeevviicceess |
| 1902 | |
| 1903 | The way to share a filesystem between two virtual machines is to use |
| 1904 | the copy-on-write (COW) layering capability of the ubd block driver. |
| 1905 | As of 2.4.6-2um, the driver supports layering a read-write private |
| 1906 | device over a read-only shared device. A machine's writes are stored |
| 1907 | in the private device, while reads come from either device - the |
| 1908 | private one if the requested block is valid in it, the shared one if |
| 1909 | not. Using this scheme, the majority of data which is unchanged is |
| 1910 | shared between an arbitrary number of virtual machines, each of which |
| 1911 | has a much smaller file containing the changes that it has made. With |
| 1912 | a large number of UMLs booting from a large root filesystem, this |
| 1913 | leads to a huge disk space saving. It will also help performance, |
| 1914 | since the host will be able to cache the shared data using a much |
| 1915 | smaller amount of memory, so UML disk requests will be served from the |
| 1916 | host's memory rather than its disks. |
| 1917 | |
| 1918 | |
| 1919 | |
| 1920 | |
| 1921 | To add a copy-on-write layer to an existing block device file, simply |
| 1922 | add the name of the COW file to the appropriate ubd switch: |
| 1923 | |
| 1924 | |
| 1925 | ubd0=root_fs_cow,root_fs_debian_22 |
| 1926 | |
| 1927 | |
| 1928 | |
| 1929 | |
| 1930 | where 'root_fs_cow' is the private COW file and 'root_fs_debian_22' is |
| 1931 | the existing shared filesystem. The COW file need not exist. If it |
| 1932 | doesn't, the driver will create and initialize it. Once the COW file |
| 1933 | has been initialized, it can be used on its own on the command line: |
| 1934 | |
| 1935 | |
| 1936 | ubd0=root_fs_cow |
| 1937 | |
| 1938 | |
| 1939 | |
| 1940 | |
| 1941 | The name of the backing file is stored in the COW file header, so it |
| 1942 | would be redundant to continue specifying it on the command line. |
| 1943 | |
| 1944 | |
| 1945 | |
| 1946 | 77..33.. NNoottee!! |
| 1947 | |
| 1948 | When checking the size of the COW file in order to see the gobs of |
| 1949 | space that you're saving, make sure you use 'ls -ls' to see the actual |
| 1950 | disk consumption rather than the length of the file. The COW file is |
| 1951 | sparse, so the length will be very different from the disk usage. |
| 1952 | Here is a 'ls -l' of a COW file and backing file from one boot and |
| 1953 | shutdown: |
| 1954 | host% ls -l cow.debian debian2.2 |
| 1955 | -rw-r--r-- 1 jdike jdike 492504064 Aug 6 21:16 cow.debian |
| 1956 | -rwxrw-rw- 1 jdike jdike 537919488 Aug 6 20:42 debian2.2 |
| 1957 | |
| 1958 | |
| 1959 | |
| 1960 | |
| 1961 | Doesn't look like much saved space, does it? Well, here's 'ls -ls': |
| 1962 | |
| 1963 | |
| 1964 | host% ls -ls cow.debian debian2.2 |
| 1965 | 880 -rw-r--r-- 1 jdike jdike 492504064 Aug 6 21:16 cow.debian |
| 1966 | 525832 -rwxrw-rw- 1 jdike jdike 537919488 Aug 6 20:42 debian2.2 |
| 1967 | |
| 1968 | |
| 1969 | |
| 1970 | |
| 1971 | Now, you can see that the COW file has less than a meg of disk, rather |
| 1972 | than 492 meg. |
| 1973 | |
| 1974 | |
| 1975 | |
| 1976 | 77..44.. AAnnootthheerr wwaarrnniinngg |
| 1977 | |
| 1978 | Once a filesystem is being used as a readonly backing file for a COW |
| 1979 | file, do not boot directly from it or modify it in any way. Doing so |
| 1980 | will invalidate any COW files that are using it. The mtime and size |
| 1981 | of the backing file are stored in the COW file header at its creation, |
| 1982 | and they must continue to match. If they don't, the driver will |
| 1983 | refuse to use the COW file. |
| 1984 | |
| 1985 | |
| 1986 | |
| 1987 | |
| 1988 | If you attempt to evade this restriction by changing either the |
| 1989 | backing file or the COW header by hand, you will get a corrupted |
| 1990 | filesystem. |
| 1991 | |
| 1992 | |
| 1993 | |
| 1994 | |
| 1995 | Among other things, this means that upgrading the distribution in a |
| 1996 | backing file and expecting that all of the COW files using it will see |
| 1997 | the upgrade will not work. |
| 1998 | |
| 1999 | |
| 2000 | |
| 2001 | |
| 2002 | 77..55.. uummll__mmoooo :: MMeerrggiinngg aa CCOOWW ffiillee wwiitthh iittss bbaacckkiinngg ffiillee |
| 2003 | |
| 2004 | Depending on how you use UML and COW devices, it may be advisable to |
| 2005 | merge the changes in the COW file into the backing file every once in |
| 2006 | a while. |
| 2007 | |
| 2008 | |
| 2009 | |
| 2010 | |
| 2011 | The utility that does this is uml_moo. Its usage is |
| 2012 | |
| 2013 | |
| 2014 | host% uml_moo COW file new backing file |
| 2015 | |
| 2016 | |
| 2017 | |
| 2018 | |
| 2019 | There's no need to specify the backing file since that information is |
| 2020 | already in the COW file header. If you're paranoid, boot the new |
| 2021 | merged file, and if you're happy with it, move it over the old backing |
| 2022 | file. |
| 2023 | |
| 2024 | |
| 2025 | |
| 2026 | |
| 2027 | uml_moo creates a new backing file by default as a safety measure. It |
| 2028 | also has a destructive merge option which will merge the COW file |
| 2029 | directly into its current backing file. This is really only usable |
| 2030 | when the backing file only has one COW file associated with it. If |
| 2031 | there are multiple COWs associated with a backing file, a -d merge of |
| 2032 | one of them will invalidate all of the others. However, it is |
| 2033 | convenient if you're short of disk space, and it should also be |
| 2034 | noticably faster than a non-destructive merge. |
| 2035 | |
| 2036 | |
| 2037 | |
| 2038 | |
| 2039 | uml_moo is installed with the UML deb and RPM. If you didn't install |
| 2040 | UML from one of those packages, you can also get it from the UML |
| 2041 | utilities <http://user-mode-linux.sourceforge.net/dl-sf.html#UML |
| 2042 | utilities> tar file in tools/moo. |
| 2043 | |
| 2044 | |
| 2045 | |
| 2046 | |
| 2047 | |
| 2048 | |
| 2049 | |
| 2050 | |
| 2051 | 88.. CCrreeaattiinngg ffiilleessyysstteemmss |
| 2052 | |
| 2053 | |
| 2054 | You may want to create and mount new UML filesystems, either because |
| 2055 | your root filesystem isn't large enough or because you want to use a |
| 2056 | filesystem other than ext2. |
| 2057 | |
| 2058 | |
| 2059 | This was written on the occasion of reiserfs being included in the |
| 2060 | 2.4.1 kernel pool, and therefore the 2.4.1 UML, so the examples will |
| 2061 | talk about reiserfs. This information is generic, and the examples |
| 2062 | should be easy to translate to the filesystem of your choice. |
| 2063 | |
| 2064 | |
| 2065 | 88..11.. CCrreeaattee tthhee ffiilleessyysstteemm ffiillee |
| 2066 | |
| 2067 | dd is your friend. All you need to do is tell dd to create an empty |
| 2068 | file of the appropriate size. I usually make it sparse to save time |
| 2069 | and to avoid allocating disk space until it's actually used. For |
| 2070 | example, the following command will create a sparse 100 meg file full |
| 2071 | of zeroes. |
| 2072 | |
| 2073 | |
| 2074 | host% |
| 2075 | dd if=/dev/zero of=new_filesystem seek=100 count=1 bs=1M |
| 2076 | |
| 2077 | |
| 2078 | |
| 2079 | |
| 2080 | |
| 2081 | |
| 2082 | 88..22.. AAssssiiggnn tthhee ffiillee ttoo aa UUMMLL ddeevviiccee |
| 2083 | |
| 2084 | Add an argument like the following to the UML command line: |
| 2085 | |
| 2086 | ubd4=new_filesystem |
| 2087 | |
| 2088 | |
| 2089 | |
| 2090 | |
| 2091 | making sure that you use an unassigned ubd device number. |
| 2092 | |
| 2093 | |
| 2094 | |
| 2095 | 88..33.. CCrreeaattiinngg aanndd mmoouunnttiinngg tthhee ffiilleessyysstteemm |
| 2096 | |
| 2097 | Make sure that the filesystem is available, either by being built into |
| 2098 | the kernel, or available as a module, then boot up UML and log in. If |
| 2099 | the root filesystem doesn't have the filesystem utilities (mkfs, fsck, |
| 2100 | etc), then get them into UML by way of the net or hostfs. |
| 2101 | |
| 2102 | |
| 2103 | Make the new filesystem on the device assigned to the new file: |
| 2104 | |
| 2105 | |
| 2106 | host# mkreiserfs /dev/ubd/4 |
| 2107 | |
| 2108 | |
| 2109 | <----------- MKREISERFSv2 -----------> |
| 2110 | |
| 2111 | ReiserFS version 3.6.25 |
| 2112 | Block size 4096 bytes |
| 2113 | Block count 25856 |
| 2114 | Used blocks 8212 |
| 2115 | Journal - 8192 blocks (18-8209), journal header is in block 8210 |
| 2116 | Bitmaps: 17 |
| 2117 | Root block 8211 |
| 2118 | Hash function "r5" |
| 2119 | ATTENTION: ALL DATA WILL BE LOST ON '/dev/ubd/4'! (y/n)y |
| 2120 | journal size 8192 (from 18) |
| 2121 | Initializing journal - 0%....20%....40%....60%....80%....100% |
| 2122 | Syncing..done. |
| 2123 | |
| 2124 | |
| 2125 | |
| 2126 | |
| 2127 | Now, mount it: |
| 2128 | |
| 2129 | |
| 2130 | UML# |
| 2131 | mount /dev/ubd/4 /mnt |
| 2132 | |
| 2133 | |
| 2134 | |
| 2135 | |
| 2136 | and you're in business. |
| 2137 | |
| 2138 | |
| 2139 | |
| 2140 | |
| 2141 | |
| 2142 | |
| 2143 | |
| 2144 | |
| 2145 | |
| 2146 | 99.. HHoosstt ffiillee aacccceessss |
| 2147 | |
| 2148 | |
| 2149 | If you want to access files on the host machine from inside UML, you |
| 2150 | can treat it as a separate machine and either nfs mount directories |
| 2151 | from the host or copy files into the virtual machine with scp or rcp. |
Tobias Klauser | d533f67 | 2005-09-10 00:26:46 -0700 | [diff] [blame] | 2152 | However, since UML is running on the host, it can access those |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2153 | files just like any other process and make them available inside the |
| 2154 | virtual machine without needing to use the network. |
| 2155 | |
| 2156 | |
| 2157 | This is now possible with the hostfs virtual filesystem. With it, you |
| 2158 | can mount a host directory into the UML filesystem and access the |
| 2159 | files contained in it just as you would on the host. |
| 2160 | |
| 2161 | |
| 2162 | 99..11.. UUssiinngg hhoossttffss |
| 2163 | |
| 2164 | To begin with, make sure that hostfs is available inside the virtual |
| 2165 | machine with |
| 2166 | |
| 2167 | |
| 2168 | UML# cat /proc/filesystems |
| 2169 | |
| 2170 | |
| 2171 | |
| 2172 | . hostfs should be listed. If it's not, either rebuild the kernel |
| 2173 | with hostfs configured into it or make sure that hostfs is built as a |
| 2174 | module and available inside the virtual machine, and insmod it. |
| 2175 | |
| 2176 | |
| 2177 | Now all you need to do is run mount: |
| 2178 | |
| 2179 | |
| 2180 | UML# mount none /mnt/host -t hostfs |
| 2181 | |
| 2182 | |
| 2183 | |
| 2184 | |
| 2185 | will mount the host's / on the virtual machine's /mnt/host. |
| 2186 | |
| 2187 | |
| 2188 | If you don't want to mount the host root directory, then you can |
| 2189 | specify a subdirectory to mount with the -o switch to mount: |
| 2190 | |
| 2191 | |
| 2192 | UML# mount none /mnt/home -t hostfs -o /home |
| 2193 | |
| 2194 | |
| 2195 | |
| 2196 | |
| 2197 | will mount the hosts's /home on the virtual machine's /mnt/home. |
| 2198 | |
| 2199 | |
| 2200 | |
| 2201 | 99..22.. hhoossttffss aass tthhee rroooott ffiilleessyysstteemm |
| 2202 | |
| 2203 | It's possible to boot from a directory hierarchy on the host using |
| 2204 | hostfs rather than using the standard filesystem in a file. |
| 2205 | |
| 2206 | To start, you need that hierarchy. The easiest way is to loop mount |
| 2207 | an existing root_fs file: |
| 2208 | |
| 2209 | |
| 2210 | host# mount root_fs uml_root_dir -o loop |
| 2211 | |
| 2212 | |
| 2213 | |
| 2214 | |
| 2215 | You need to change the filesystem type of / in etc/fstab to be |
| 2216 | 'hostfs', so that line looks like this: |
| 2217 | |
| 2218 | /dev/ubd/0 / hostfs defaults 1 1 |
| 2219 | |
| 2220 | |
| 2221 | |
| 2222 | |
| 2223 | Then you need to chown to yourself all the files in that directory |
| 2224 | that are owned by root. This worked for me: |
| 2225 | |
| 2226 | |
| 2227 | host# find . -uid 0 -exec chown jdike {} \; |
| 2228 | |
| 2229 | |
| 2230 | |
| 2231 | |
| 2232 | Next, make sure that your UML kernel has hostfs compiled in, not as a |
| 2233 | module. Then run UML with the boot device pointing at that directory: |
| 2234 | |
| 2235 | |
| 2236 | ubd0=/path/to/uml/root/directory |
| 2237 | |
| 2238 | |
| 2239 | |
| 2240 | |
| 2241 | UML should then boot as it does normally. |
| 2242 | |
| 2243 | |
| 2244 | 99..33.. BBuuiillddiinngg hhoossttffss |
| 2245 | |
| 2246 | If you need to build hostfs because it's not in your kernel, you have |
| 2247 | two choices: |
| 2248 | |
| 2249 | |
| 2250 | |
| 2251 | +o Compiling hostfs into the kernel: |
| 2252 | |
| 2253 | |
| 2254 | Reconfigure the kernel and set the 'Host filesystem' option under |
| 2255 | |
| 2256 | |
| 2257 | +o Compiling hostfs as a module: |
| 2258 | |
| 2259 | |
| 2260 | Reconfigure the kernel and set the 'Host filesystem' option under |
| 2261 | be in arch/um/fs/hostfs/hostfs.o. Install that in |
| 2262 | /lib/modules/`uname -r`/fs in the virtual machine, boot it up, and |
| 2263 | |
| 2264 | |
| 2265 | UML# insmod hostfs |
| 2266 | |
| 2267 | |
| 2268 | |
| 2269 | |
| 2270 | |
| 2271 | |
| 2272 | |
| 2273 | |
| 2274 | |
| 2275 | |
| 2276 | |
| 2277 | |
| 2278 | 1100.. TThhee MMaannaaggeemmeenntt CCoonnssoollee |
| 2279 | |
| 2280 | |
| 2281 | |
| 2282 | The UML management console is a low-level interface to the kernel, |
| 2283 | somewhat like the i386 SysRq interface. Since there is a full-blown |
| 2284 | operating system under UML, there is much greater flexibility possible |
| 2285 | than with the SysRq mechanism. |
| 2286 | |
| 2287 | |
| 2288 | There are a number of things you can do with the mconsole interface: |
| 2289 | |
| 2290 | +o get the kernel version |
| 2291 | |
| 2292 | +o add and remove devices |
| 2293 | |
| 2294 | +o halt or reboot the machine |
| 2295 | |
| 2296 | +o Send SysRq commands |
| 2297 | |
| 2298 | +o Pause and resume the UML |
| 2299 | |
| 2300 | |
| 2301 | You need the mconsole client (uml_mconsole) which is present in CVS |
| 2302 | (/tools/mconsole) in 2.4.5-9um and later, and will be in the RPM in |
| 2303 | 2.4.6. |
| 2304 | |
| 2305 | |
| 2306 | You also need CONFIG_MCONSOLE (under 'General Setup') enabled in UML. |
| 2307 | When you boot UML, you'll see a line like: |
| 2308 | |
| 2309 | |
| 2310 | mconsole initialized on /home/jdike/.uml/umlNJ32yL/mconsole |
| 2311 | |
| 2312 | |
| 2313 | |
| 2314 | |
| 2315 | If you specify a unique machine id one the UML command line, i.e. |
| 2316 | |
| 2317 | |
| 2318 | umid=debian |
| 2319 | |
| 2320 | |
| 2321 | |
| 2322 | |
| 2323 | you'll see this |
| 2324 | |
| 2325 | |
| 2326 | mconsole initialized on /home/jdike/.uml/debian/mconsole |
| 2327 | |
| 2328 | |
| 2329 | |
| 2330 | |
| 2331 | That file is the socket that uml_mconsole will use to communicate with |
| 2332 | UML. Run it with either the umid or the full path as its argument: |
| 2333 | |
| 2334 | |
| 2335 | host% uml_mconsole debian |
| 2336 | |
| 2337 | |
| 2338 | |
| 2339 | |
| 2340 | or |
| 2341 | |
| 2342 | |
| 2343 | host% uml_mconsole /home/jdike/.uml/debian/mconsole |
| 2344 | |
| 2345 | |
| 2346 | |
| 2347 | |
| 2348 | You'll get a prompt, at which you can run one of these commands: |
| 2349 | |
| 2350 | +o version |
| 2351 | |
| 2352 | +o halt |
| 2353 | |
| 2354 | +o reboot |
| 2355 | |
| 2356 | +o config |
| 2357 | |
| 2358 | +o remove |
| 2359 | |
| 2360 | +o sysrq |
| 2361 | |
| 2362 | +o help |
| 2363 | |
| 2364 | +o cad |
| 2365 | |
| 2366 | +o stop |
| 2367 | |
| 2368 | +o go |
| 2369 | |
| 2370 | |
| 2371 | 1100..11.. vveerrssiioonn |
| 2372 | |
| 2373 | This takes no arguments. It prints the UML version. |
| 2374 | |
| 2375 | |
| 2376 | (mconsole) version |
| 2377 | OK Linux usermode 2.4.5-9um #1 Wed Jun 20 22:47:08 EDT 2001 i686 |
| 2378 | |
| 2379 | |
| 2380 | |
| 2381 | |
| 2382 | There are a couple actual uses for this. It's a simple no-op which |
| 2383 | can be used to check that a UML is running. It's also a way of |
| 2384 | sending an interrupt to the UML. This is sometimes useful on SMP |
| 2385 | hosts, where there's a bug which causes signals to UML to be lost, |
| 2386 | often causing it to appear to hang. Sending such a UML the mconsole |
| 2387 | version command is a good way to 'wake it up' before networking has |
| 2388 | been enabled, as it does not do anything to the function of the UML. |
| 2389 | |
| 2390 | |
| 2391 | |
| 2392 | 1100..22.. hhaalltt aanndd rreebboooott |
| 2393 | |
| 2394 | These take no arguments. They shut the machine down immediately, with |
| 2395 | no syncing of disks and no clean shutdown of userspace. So, they are |
| 2396 | pretty close to crashing the machine. |
| 2397 | |
| 2398 | |
| 2399 | (mconsole) halt |
| 2400 | OK |
| 2401 | |
| 2402 | |
| 2403 | |
| 2404 | |
| 2405 | |
| 2406 | |
| 2407 | 1100..33.. ccoonnffiigg |
| 2408 | |
| 2409 | "config" adds a new device to the virtual machine. Currently the ubd |
| 2410 | and network drivers support this. It takes one argument, which is the |
| 2411 | device to add, with the same syntax as the kernel command line. |
| 2412 | |
| 2413 | |
| 2414 | |
| 2415 | |
| 2416 | (mconsole) |
| 2417 | config ubd3=/home/jdike/incoming/roots/root_fs_debian22 |
| 2418 | |
| 2419 | OK |
| 2420 | (mconsole) config eth1=mcast |
| 2421 | OK |
| 2422 | |
| 2423 | |
| 2424 | |
| 2425 | |
| 2426 | |
| 2427 | |
| 2428 | 1100..44.. rreemmoovvee |
| 2429 | |
| 2430 | "remove" deletes a device from the system. Its argument is just the |
| 2431 | name of the device to be removed. The device must be idle in whatever |
| 2432 | sense the driver considers necessary. In the case of the ubd driver, |
| 2433 | the removed block device must not be mounted, swapped on, or otherwise |
| 2434 | open, and in the case of the network driver, the device must be down. |
| 2435 | |
| 2436 | |
| 2437 | (mconsole) remove ubd3 |
| 2438 | OK |
| 2439 | (mconsole) remove eth1 |
| 2440 | OK |
| 2441 | |
| 2442 | |
| 2443 | |
| 2444 | |
| 2445 | |
| 2446 | |
| 2447 | 1100..55.. ssyyssrrqq |
| 2448 | |
| 2449 | This takes one argument, which is a single letter. It calls the |
| 2450 | generic kernel's SysRq driver, which does whatever is called for by |
| 2451 | that argument. See the SysRq documentation in Documentation/sysrq.txt |
| 2452 | in your favorite kernel tree to see what letters are valid and what |
| 2453 | they do. |
| 2454 | |
| 2455 | |
| 2456 | |
| 2457 | 1100..66.. hheellpp |
| 2458 | |
| 2459 | "help" returns a string listing the valid commands and what each one |
| 2460 | does. |
| 2461 | |
| 2462 | |
| 2463 | |
| 2464 | 1100..77.. ccaadd |
| 2465 | |
| 2466 | This invokes the Ctl-Alt-Del action on init. What exactly this ends |
| 2467 | up doing is up to /etc/inittab. Normally, it reboots the machine. |
| 2468 | With UML, this is usually not desired, so if a halt would be better, |
| 2469 | then find the section of inittab that looks like this |
| 2470 | |
| 2471 | |
| 2472 | # What to do when CTRL-ALT-DEL is pressed. |
| 2473 | ca:12345:ctrlaltdel:/sbin/shutdown -t1 -a -r now |
| 2474 | |
| 2475 | |
| 2476 | |
| 2477 | |
| 2478 | and change the command to halt. |
| 2479 | |
| 2480 | |
| 2481 | |
| 2482 | 1100..88.. ssttoopp |
| 2483 | |
| 2484 | This puts the UML in a loop reading mconsole requests until a 'go' |
| 2485 | mconsole command is received. This is very useful for making backups |
| 2486 | of UML filesystems, as the UML can be stopped, then synced via 'sysrq |
| 2487 | s', so that everything is written to the filesystem. You can then copy |
| 2488 | the filesystem and then send the UML 'go' via mconsole. |
| 2489 | |
| 2490 | |
| 2491 | Note that a UML running with more than one CPU will have problems |
| 2492 | after you send the 'stop' command, as only one CPU will be held in a |
| 2493 | mconsole loop and all others will continue as normal. This is a bug, |
| 2494 | and will be fixed. |
| 2495 | |
| 2496 | |
| 2497 | |
| 2498 | 1100..99.. ggoo |
| 2499 | |
| 2500 | This resumes a UML after being paused by a 'stop' command. Note that |
| 2501 | when the UML has resumed, TCP connections may have timed out and if |
| 2502 | the UML is paused for a long period of time, crond might go a little |
| 2503 | crazy, running all the jobs it didn't do earlier. |
| 2504 | |
| 2505 | |
| 2506 | |
| 2507 | |
| 2508 | |
| 2509 | |
| 2510 | |
| 2511 | |
| 2512 | 1111.. KKeerrnneell ddeebbuuggggiinngg |
| 2513 | |
| 2514 | |
| 2515 | NNoottee:: The interface that makes debugging, as described here, possible |
| 2516 | is present in 2.4.0-test6 kernels and later. |
| 2517 | |
| 2518 | |
| 2519 | Since the user-mode kernel runs as a normal Linux process, it is |
| 2520 | possible to debug it with gdb almost like any other process. It is |
| 2521 | slightly different because the kernel's threads are already being |
| 2522 | ptraced for system call interception, so gdb can't ptrace them. |
| 2523 | However, a mechanism has been added to work around that problem. |
| 2524 | |
| 2525 | |
| 2526 | In order to debug the kernel, you need build it from source. See |
| 2527 | ``Compiling the kernel and modules'' for information on doing that. |
| 2528 | Make sure that you enable CONFIG_DEBUGSYM and CONFIG_PT_PROXY during |
| 2529 | the config. These will compile the kernel with -g, and enable the |
| 2530 | ptrace proxy so that gdb works with UML, respectively. |
| 2531 | |
| 2532 | |
| 2533 | |
| 2534 | |
| 2535 | 1111..11.. SSttaarrttiinngg tthhee kkeerrnneell uunnddeerr ggddbb |
| 2536 | |
| 2537 | You can have the kernel running under the control of gdb from the |
| 2538 | beginning by putting 'debug' on the command line. You will get an |
| 2539 | xterm with gdb running inside it. The kernel will send some commands |
| 2540 | to gdb which will leave it stopped at the beginning of start_kernel. |
| 2541 | At this point, you can get things going with 'next', 'step', or |
| 2542 | 'cont'. |
| 2543 | |
| 2544 | |
| 2545 | There is a transcript of a debugging session here <debug- |
| 2546 | session.html> , with breakpoints being set in the scheduler and in an |
| 2547 | interrupt handler. |
| 2548 | 1111..22.. EExxaammiinniinngg sslleeeeppiinngg pprroocceesssseess |
| 2549 | |
| 2550 | Not every bug is evident in the currently running process. Sometimes, |
| 2551 | processes hang in the kernel when they shouldn't because they've |
| 2552 | deadlocked on a semaphore or something similar. In this case, when |
| 2553 | you ^C gdb and get a backtrace, you will see the idle thread, which |
| 2554 | isn't very relevant. |
| 2555 | |
| 2556 | |
| 2557 | What you want is the stack of whatever process is sleeping when it |
| 2558 | shouldn't be. You need to figure out which process that is, which is |
| 2559 | generally fairly easy. Then you need to get its host process id, |
| 2560 | which you can do either by looking at ps on the host or at |
| 2561 | task.thread.extern_pid in gdb. |
| 2562 | |
| 2563 | |
| 2564 | Now what you do is this: |
| 2565 | |
| 2566 | +o detach from the current thread |
| 2567 | |
| 2568 | |
| 2569 | (UML gdb) det |
| 2570 | |
| 2571 | |
| 2572 | |
| 2573 | |
| 2574 | |
| 2575 | +o attach to the thread you are interested in |
| 2576 | |
| 2577 | |
| 2578 | (UML gdb) att <host pid> |
| 2579 | |
| 2580 | |
| 2581 | |
| 2582 | |
| 2583 | |
| 2584 | +o look at its stack and anything else of interest |
| 2585 | |
| 2586 | |
| 2587 | (UML gdb) bt |
| 2588 | |
| 2589 | |
| 2590 | |
| 2591 | |
| 2592 | Note that you can't do anything at this point that requires that a |
| 2593 | process execute, e.g. calling a function |
| 2594 | |
| 2595 | +o when you're done looking at that process, reattach to the current |
| 2596 | thread and continue it |
| 2597 | |
| 2598 | |
| 2599 | (UML gdb) |
| 2600 | att 1 |
| 2601 | |
| 2602 | |
| 2603 | |
| 2604 | |
| 2605 | |
| 2606 | |
| 2607 | (UML gdb) |
| 2608 | c |
| 2609 | |
| 2610 | |
| 2611 | |
| 2612 | |
| 2613 | Here, specifying any pid which is not the process id of a UML thread |
| 2614 | will cause gdb to reattach to the current thread. I commonly use 1, |
| 2615 | but any other invalid pid would work. |
| 2616 | |
| 2617 | |
| 2618 | |
| 2619 | 1111..33.. RRuunnnniinngg dddddd oonn UUMMLL |
| 2620 | |
| 2621 | ddd works on UML, but requires a special kludge. The process goes |
| 2622 | like this: |
| 2623 | |
| 2624 | +o Start ddd |
| 2625 | |
| 2626 | |
| 2627 | host% ddd linux |
| 2628 | |
| 2629 | |
| 2630 | |
| 2631 | |
| 2632 | |
| 2633 | +o With ps, get the pid of the gdb that ddd started. You can ask the |
| 2634 | gdb to tell you, but for some reason that confuses things and |
| 2635 | causes a hang. |
| 2636 | |
| 2637 | +o run UML with 'debug=parent gdb-pid=<pid>' added to the command line |
| 2638 | - it will just sit there after you hit return |
| 2639 | |
| 2640 | +o type 'att 1' to the ddd gdb and you will see something like |
| 2641 | |
| 2642 | |
| 2643 | 0xa013dc51 in __kill () |
| 2644 | |
| 2645 | |
| 2646 | (gdb) |
| 2647 | |
| 2648 | |
| 2649 | |
| 2650 | |
| 2651 | |
| 2652 | +o At this point, type 'c', UML will boot up, and you can use ddd just |
| 2653 | as you do on any other process. |
| 2654 | |
| 2655 | |
| 2656 | |
| 2657 | 1111..44.. DDeebbuuggggiinngg mmoodduulleess |
| 2658 | |
| 2659 | gdb has support for debugging code which is dynamically loaded into |
| 2660 | the process. This support is what is needed to debug kernel modules |
| 2661 | under UML. |
| 2662 | |
| 2663 | |
| 2664 | Using that support is somewhat complicated. You have to tell gdb what |
| 2665 | object file you just loaded into UML and where in memory it is. Then, |
| 2666 | it can read the symbol table, and figure out where all the symbols are |
| 2667 | from the load address that you provided. It gets more interesting |
| 2668 | when you load the module again (i.e. after an rmmod). You have to |
| 2669 | tell gdb to forget about all its symbols, including the main UML ones |
| 2670 | for some reason, then load then all back in again. |
| 2671 | |
| 2672 | |
| 2673 | There's an easy way and a hard way to do this. The easy way is to use |
| 2674 | the umlgdb expect script written by Chandan Kudige. It basically |
| 2675 | automates the process for you. |
| 2676 | |
| 2677 | |
| 2678 | First, you must tell it where your modules are. There is a list in |
| 2679 | the script that looks like this: |
| 2680 | set MODULE_PATHS { |
| 2681 | "fat" "/usr/src/uml/linux-2.4.18/fs/fat/fat.o" |
| 2682 | "isofs" "/usr/src/uml/linux-2.4.18/fs/isofs/isofs.o" |
| 2683 | "minix" "/usr/src/uml/linux-2.4.18/fs/minix/minix.o" |
| 2684 | } |
| 2685 | |
| 2686 | |
| 2687 | |
| 2688 | |
| 2689 | You change that to list the names and paths of the modules that you |
| 2690 | are going to debug. Then you run it from the toplevel directory of |
| 2691 | your UML pool and it basically tells you what to do: |
| 2692 | |
| 2693 | |
| 2694 | |
| 2695 | |
| 2696 | ******** GDB pid is 21903 ******** |
| 2697 | Start UML as: ./linux <kernel switches> debug gdb-pid=21903 |
| 2698 | |
| 2699 | |
| 2700 | |
| 2701 | GNU gdb 5.0rh-5 Red Hat Linux 7.1 |
| 2702 | Copyright 2001 Free Software Foundation, Inc. |
| 2703 | GDB is free software, covered by the GNU General Public License, and you are |
| 2704 | welcome to change it and/or distribute copies of it under certain conditions. |
| 2705 | Type "show copying" to see the conditions. |
| 2706 | There is absolutely no warranty for GDB. Type "show warranty" for details. |
| 2707 | This GDB was configured as "i386-redhat-linux"... |
| 2708 | (gdb) b sys_init_module |
| 2709 | Breakpoint 1 at 0xa0011923: file module.c, line 349. |
| 2710 | (gdb) att 1 |
| 2711 | |
| 2712 | |
| 2713 | |
| 2714 | |
| 2715 | After you run UML and it sits there doing nothing, you hit return at |
| 2716 | the 'att 1' and continue it: |
| 2717 | |
| 2718 | |
| 2719 | Attaching to program: /home/jdike/linux/2.4/um/./linux, process 1 |
| 2720 | 0xa00f4221 in __kill () |
| 2721 | (UML gdb) c |
| 2722 | Continuing. |
| 2723 | |
| 2724 | |
| 2725 | |
| 2726 | |
| 2727 | At this point, you debug normally. When you insmod something, the |
| 2728 | expect magic will kick in and you'll see something like: |
| 2729 | |
| 2730 | |
| 2731 | |
| 2732 | |
| 2733 | |
| 2734 | |
| 2735 | |
| 2736 | |
| 2737 | |
| 2738 | |
| 2739 | |
| 2740 | |
| 2741 | |
| 2742 | |
| 2743 | |
| 2744 | |
| 2745 | |
| 2746 | *** Module hostfs loaded *** |
| 2747 | Breakpoint 1, sys_init_module (name_user=0x805abb0 "hostfs", |
| 2748 | mod_user=0x8070e00) at module.c:349 |
| 2749 | 349 char *name, *n_name, *name_tmp = NULL; |
| 2750 | (UML gdb) finish |
| 2751 | Run till exit from #0 sys_init_module (name_user=0x805abb0 "hostfs", |
| 2752 | mod_user=0x8070e00) at module.c:349 |
| 2753 | 0xa00e2e23 in execute_syscall (r=0xa8140284) at syscall_kern.c:411 |
| 2754 | 411 else res = EXECUTE_SYSCALL(syscall, regs); |
| 2755 | Value returned is $1 = 0 |
| 2756 | (UML gdb) |
| 2757 | p/x (int)module_list + module_list->size_of_struct |
| 2758 | |
| 2759 | $2 = 0xa9021054 |
| 2760 | (UML gdb) symbol-file ./linux |
| 2761 | Load new symbol table from "./linux"? (y or n) y |
| 2762 | Reading symbols from ./linux... |
| 2763 | done. |
| 2764 | (UML gdb) |
| 2765 | add-symbol-file /home/jdike/linux/2.4/um/arch/um/fs/hostfs/hostfs.o 0xa9021054 |
| 2766 | |
| 2767 | add symbol table from file "/home/jdike/linux/2.4/um/arch/um/fs/hostfs/hostfs.o" at |
| 2768 | .text_addr = 0xa9021054 |
| 2769 | (y or n) y |
| 2770 | |
| 2771 | Reading symbols from /home/jdike/linux/2.4/um/arch/um/fs/hostfs/hostfs.o... |
| 2772 | done. |
| 2773 | (UML gdb) p *module_list |
| 2774 | $1 = {size_of_struct = 84, next = 0xa0178720, name = 0xa9022de0 "hostfs", |
| 2775 | size = 9016, uc = {usecount = {counter = 0}, pad = 0}, flags = 1, |
| 2776 | nsyms = 57, ndeps = 0, syms = 0xa9023170, deps = 0x0, refs = 0x0, |
| 2777 | init = 0xa90221f0 <init_hostfs>, cleanup = 0xa902222c <exit_hostfs>, |
| 2778 | ex_table_start = 0x0, ex_table_end = 0x0, persist_start = 0x0, |
| 2779 | persist_end = 0x0, can_unload = 0, runsize = 0, kallsyms_start = 0x0, |
| 2780 | kallsyms_end = 0x0, |
| 2781 | archdata_start = 0x1b855 <Address 0x1b855 out of bounds>, |
| 2782 | archdata_end = 0xe5890000 <Address 0xe5890000 out of bounds>, |
| 2783 | kernel_data = 0xf689c35d <Address 0xf689c35d out of bounds>} |
| 2784 | >> Finished loading symbols for hostfs ... |
| 2785 | |
| 2786 | |
| 2787 | |
| 2788 | |
| 2789 | That's the easy way. It's highly recommended. The hard way is |
| 2790 | described below in case you're interested in what's going on. |
| 2791 | |
| 2792 | |
| 2793 | Boot the kernel under the debugger and load the module with insmod or |
| 2794 | modprobe. With gdb, do: |
| 2795 | |
| 2796 | |
| 2797 | (UML gdb) p module_list |
| 2798 | |
| 2799 | |
| 2800 | |
| 2801 | |
| 2802 | This is a list of modules that have been loaded into the kernel, with |
| 2803 | the most recently loaded module first. Normally, the module you want |
| 2804 | is at module_list. If it's not, walk down the next links, looking at |
| 2805 | the name fields until find the module you want to debug. Take the |
| 2806 | address of that structure, and add module.size_of_struct (which in |
| 2807 | 2.4.10 kernels is 96 (0x60)) to it. Gdb can make this hard addition |
| 2808 | for you :-): |
| 2809 | |
| 2810 | |
| 2811 | |
| 2812 | (UML gdb) |
| 2813 | printf "%#x\n", (int)module_list module_list->size_of_struct |
| 2814 | |
| 2815 | |
| 2816 | |
| 2817 | |
| 2818 | The offset from the module start occasionally changes (before 2.4.0, |
| 2819 | it was module.size_of_struct + 4), so it's a good idea to check the |
| 2820 | init and cleanup addresses once in a while, as describe below. Now |
| 2821 | do: |
| 2822 | |
| 2823 | |
| 2824 | (UML gdb) |
| 2825 | add-symbol-file /path/to/module/on/host that_address |
| 2826 | |
| 2827 | |
| 2828 | |
| 2829 | |
| 2830 | Tell gdb you really want to do it, and you're in business. |
| 2831 | |
| 2832 | |
| 2833 | If there's any doubt that you got the offset right, like breakpoints |
| 2834 | appear not to work, or they're appearing in the wrong place, you can |
| 2835 | check it by looking at the module structure. The init and cleanup |
| 2836 | fields should look like: |
| 2837 | |
| 2838 | |
| 2839 | init = 0x588066b0 <init_hostfs>, cleanup = 0x588066c0 <exit_hostfs> |
| 2840 | |
| 2841 | |
| 2842 | |
| 2843 | |
| 2844 | with no offsets on the symbol names. If the names are right, but they |
| 2845 | are offset, then the offset tells you how much you need to add to the |
| 2846 | address you gave to add-symbol-file. |
| 2847 | |
| 2848 | |
| 2849 | When you want to load in a new version of the module, you need to get |
| 2850 | gdb to forget about the old one. The only way I've found to do that |
| 2851 | is to tell gdb to forget about all symbols that it knows about: |
| 2852 | |
| 2853 | |
| 2854 | (UML gdb) symbol-file |
| 2855 | |
| 2856 | |
| 2857 | |
| 2858 | |
| 2859 | Then reload the symbols from the kernel binary: |
| 2860 | |
| 2861 | |
| 2862 | (UML gdb) symbol-file /path/to/kernel |
| 2863 | |
| 2864 | |
| 2865 | |
| 2866 | |
| 2867 | and repeat the process above. You'll also need to re-enable break- |
| 2868 | points. They were disabled when you dumped all the symbols because |
| 2869 | gdb couldn't figure out where they should go. |
| 2870 | |
| 2871 | |
| 2872 | |
| 2873 | 1111..55.. AAttttaacchhiinngg ggddbb ttoo tthhee kkeerrnneell |
| 2874 | |
| 2875 | If you don't have the kernel running under gdb, you can attach gdb to |
| 2876 | it later by sending the tracing thread a SIGUSR1. The first line of |
| 2877 | the console output identifies its pid: |
| 2878 | tracing thread pid = 20093 |
| 2879 | |
| 2880 | |
| 2881 | |
| 2882 | |
| 2883 | When you send it the signal: |
| 2884 | |
| 2885 | |
| 2886 | host% kill -USR1 20093 |
| 2887 | |
| 2888 | |
| 2889 | |
| 2890 | |
| 2891 | you will get an xterm with gdb running in it. |
| 2892 | |
| 2893 | |
| 2894 | If you have the mconsole compiled into UML, then the mconsole client |
| 2895 | can be used to start gdb: |
| 2896 | |
| 2897 | |
| 2898 | (mconsole) (mconsole) config gdb=xterm |
| 2899 | |
| 2900 | |
| 2901 | |
| 2902 | |
| 2903 | will fire up an xterm with gdb running in it. |
| 2904 | |
| 2905 | |
| 2906 | |
| 2907 | 1111..66.. UUssiinngg aalltteerrnnaattee ddeebbuuggggeerrss |
| 2908 | |
| 2909 | UML has support for attaching to an already running debugger rather |
| 2910 | than starting gdb itself. This is present in CVS as of 17 Apr 2001. |
| 2911 | I sent it to Alan for inclusion in the ac tree, and it will be in my |
| 2912 | 2.4.4 release. |
| 2913 | |
| 2914 | |
| 2915 | This is useful when gdb is a subprocess of some UI, such as emacs or |
| 2916 | ddd. It can also be used to run debuggers other than gdb on UML. |
| 2917 | Below is an example of using strace as an alternate debugger. |
| 2918 | |
| 2919 | |
| 2920 | To do this, you need to get the pid of the debugger and pass it in |
| 2921 | with the |
| 2922 | |
| 2923 | |
| 2924 | If you are using gdb under some UI, then tell it to 'att 1', and |
| 2925 | you'll find yourself attached to UML. |
| 2926 | |
| 2927 | |
| 2928 | If you are using something other than gdb as your debugger, then |
| 2929 | you'll need to get it to do the equivalent of 'att 1' if it doesn't do |
| 2930 | it automatically. |
| 2931 | |
| 2932 | |
| 2933 | An example of an alternate debugger is strace. You can strace the |
| 2934 | actual kernel as follows: |
| 2935 | |
| 2936 | +o Run the following in a shell |
| 2937 | |
| 2938 | |
| 2939 | host% |
| 2940 | sh -c 'echo pid=$$; echo -n hit return; read x; exec strace -p 1 -o strace.out' |
| 2941 | |
| 2942 | |
| 2943 | |
| 2944 | +o Run UML with 'debug' and 'gdb-pid=<pid>' with the pid printed out |
| 2945 | by the previous command |
| 2946 | |
| 2947 | +o Hit return in the shell, and UML will start running, and strace |
| 2948 | output will start accumulating in the output file. |
| 2949 | |
| 2950 | Note that this is different from running |
| 2951 | |
| 2952 | |
| 2953 | host% strace ./linux |
| 2954 | |
| 2955 | |
| 2956 | |
| 2957 | |
| 2958 | That will strace only the main UML thread, the tracing thread, which |
| 2959 | doesn't do any of the actual kernel work. It just oversees the vir- |
| 2960 | tual machine. In contrast, using strace as described above will show |
| 2961 | you the low-level activity of the virtual machine. |
| 2962 | |
| 2963 | |
| 2964 | |
| 2965 | |
| 2966 | |
| 2967 | 1122.. KKeerrnneell ddeebbuuggggiinngg eexxaammpplleess |
| 2968 | |
| 2969 | 1122..11.. TThhee ccaassee ooff tthhee hhuunngg ffsscckk |
| 2970 | |
| 2971 | When booting up the kernel, fsck failed, and dropped me into a shell |
| 2972 | to fix things up. I ran fsck -y, which hung: |
| 2973 | |
| 2974 | |
| 2975 | |
| 2976 | |
| 2977 | |
| 2978 | |
| 2979 | |
| 2980 | |
| 2981 | |
| 2982 | |
| 2983 | |
| 2984 | |
| 2985 | |
| 2986 | |
| 2987 | |
| 2988 | |
| 2989 | |
| 2990 | |
| 2991 | |
| 2992 | |
| 2993 | |
| 2994 | |
| 2995 | |
| 2996 | |
| 2997 | |
| 2998 | |
| 2999 | |
| 3000 | |
| 3001 | |
| 3002 | |
| 3003 | |
| 3004 | |
| 3005 | |
| 3006 | |
| 3007 | |
| 3008 | |
| 3009 | |
| 3010 | Setting hostname uml [ OK ] |
| 3011 | Checking root filesystem |
| 3012 | /dev/fhd0 was not cleanly unmounted, check forced. |
| 3013 | Error reading block 86894 (Attempt to read block from filesystem resulted in short read) while reading indirect blocks of inode 19780. |
| 3014 | |
| 3015 | /dev/fhd0: UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY. |
| 3016 | (i.e., without -a or -p options) |
| 3017 | [ FAILED ] |
| 3018 | |
| 3019 | *** An error occurred during the file system check. |
| 3020 | *** Dropping you to a shell; the system will reboot |
| 3021 | *** when you leave the shell. |
| 3022 | Give root password for maintenance |
| 3023 | (or type Control-D for normal startup): |
| 3024 | |
| 3025 | [root@uml /root]# fsck -y /dev/fhd0 |
| 3026 | fsck -y /dev/fhd0 |
| 3027 | Parallelizing fsck version 1.14 (9-Jan-1999) |
| 3028 | e2fsck 1.14, 9-Jan-1999 for EXT2 FS 0.5b, 95/08/09 |
| 3029 | /dev/fhd0 contains a file system with errors, check forced. |
| 3030 | Pass 1: Checking inodes, blocks, and sizes |
| 3031 | Error reading block 86894 (Attempt to read block from filesystem resulted in short read) while reading indirect blocks of inode 19780. Ignore error? yes |
| 3032 | |
| 3033 | Inode 19780, i_blocks is 1548, should be 540. Fix? yes |
| 3034 | |
| 3035 | Pass 2: Checking directory structure |
| 3036 | Error reading block 49405 (Attempt to read block from filesystem resulted in short read). Ignore error? yes |
| 3037 | |
| 3038 | Directory inode 11858, block 0, offset 0: directory corrupted |
| 3039 | Salvage? yes |
| 3040 | |
| 3041 | Missing '.' in directory inode 11858. |
| 3042 | Fix? yes |
| 3043 | |
| 3044 | Missing '..' in directory inode 11858. |
| 3045 | Fix? yes |
| 3046 | |
| 3047 | |
| 3048 | |
| 3049 | |
| 3050 | |
| 3051 | The standard drill in this sort of situation is to fire up gdb on the |
| 3052 | signal thread, which, in this case, was pid 1935. In another window, |
| 3053 | I run gdb and attach pid 1935. |
| 3054 | |
| 3055 | |
| 3056 | |
| 3057 | |
| 3058 | ~/linux/2.3.26/um 1016: gdb linux |
| 3059 | GNU gdb 4.17.0.11 with Linux support |
| 3060 | Copyright 1998 Free Software Foundation, Inc. |
| 3061 | GDB is free software, covered by the GNU General Public License, and you are |
| 3062 | welcome to change it and/or distribute copies of it under certain conditions. |
| 3063 | Type "show copying" to see the conditions. |
| 3064 | There is absolutely no warranty for GDB. Type "show warranty" for details. |
| 3065 | This GDB was configured as "i386-redhat-linux"... |
| 3066 | |
| 3067 | (gdb) att 1935 |
| 3068 | Attaching to program `/home/dike/linux/2.3.26/um/linux', Pid 1935 |
| 3069 | 0x100756d9 in __wait4 () |
| 3070 | |
| 3071 | |
| 3072 | |
| 3073 | |
| 3074 | |
| 3075 | |
| 3076 | Let's see what's currently running: |
| 3077 | |
| 3078 | |
| 3079 | |
| 3080 | (gdb) p current_task.pid |
| 3081 | $1 = 0 |
| 3082 | |
| 3083 | |
| 3084 | |
| 3085 | |
| 3086 | |
| 3087 | It's the idle thread, which means that fsck went to sleep for some |
| 3088 | reason and never woke up. |
| 3089 | |
| 3090 | |
| 3091 | Let's guess that the last process in the process list is fsck: |
| 3092 | |
| 3093 | |
| 3094 | |
| 3095 | (gdb) p current_task.prev_task.comm |
| 3096 | $13 = "fsck.ext2\000\000\000\000\000\000" |
| 3097 | |
| 3098 | |
| 3099 | |
| 3100 | |
| 3101 | |
| 3102 | It is, so let's see what it thinks it's up to: |
| 3103 | |
| 3104 | |
| 3105 | |
| 3106 | (gdb) p current_task.prev_task.thread |
| 3107 | $14 = {extern_pid = 1980, tracing = 0, want_tracing = 0, forking = 0, |
| 3108 | kernel_stack_page = 0, signal_stack = 1342627840, syscall = {id = 4, args = { |
| 3109 | 3, 134973440, 1024, 0, 1024}, have_result = 0, result = 50590720}, |
| 3110 | request = {op = 2, u = {exec = {ip = 1350467584, sp = 2952789424}, fork = { |
| 3111 | regs = {1350467584, 2952789424, 0 <repeats 15 times>}, sigstack = 0, |
| 3112 | pid = 0}, switch_to = 0x507e8000, thread = {proc = 0x507e8000, |
| 3113 | arg = 0xaffffdb0, flags = 0, new_pid = 0}, input_request = { |
| 3114 | op = 1350467584, fd = -1342177872, proc = 0, pid = 0}}}} |
| 3115 | |
| 3116 | |
| 3117 | |
| 3118 | |
| 3119 | |
| 3120 | The interesting things here are the fact that its .thread.syscall.id |
| 3121 | is __NR_write (see the big switch in arch/um/kernel/syscall_kern.c or |
| 3122 | the defines in include/asm-um/arch/unistd.h), and that it never |
| 3123 | returned. Also, its .request.op is OP_SWITCH (see |
| 3124 | arch/um/include/user_util.h). These mean that it went into a write, |
| 3125 | and, for some reason, called schedule(). |
| 3126 | |
| 3127 | |
| 3128 | The fact that it never returned from write means that its stack should |
| 3129 | be fairly interesting. Its pid is 1980 (.thread.extern_pid). That |
| 3130 | process is being ptraced by the signal thread, so it must be detached |
| 3131 | before gdb can attach it: |
| 3132 | |
| 3133 | |
| 3134 | |
| 3135 | |
| 3136 | |
| 3137 | |
| 3138 | |
| 3139 | |
| 3140 | |
| 3141 | |
| 3142 | (gdb) call detach(1980) |
| 3143 | |
| 3144 | Program received signal SIGSEGV, Segmentation fault. |
| 3145 | <function called from gdb> |
| 3146 | The program being debugged stopped while in a function called from GDB. |
| 3147 | When the function (detach) is done executing, GDB will silently |
| 3148 | stop (instead of continuing to evaluate the expression containing |
| 3149 | the function call). |
| 3150 | (gdb) call detach(1980) |
| 3151 | $15 = 0 |
| 3152 | |
| 3153 | |
| 3154 | |
| 3155 | |
| 3156 | |
| 3157 | The first detach segfaults for some reason, and the second one |
| 3158 | succeeds. |
| 3159 | |
| 3160 | |
| 3161 | Now I detach from the signal thread, attach to the fsck thread, and |
| 3162 | look at its stack: |
| 3163 | |
| 3164 | |
| 3165 | (gdb) det |
| 3166 | Detaching from program: /home/dike/linux/2.3.26/um/linux Pid 1935 |
| 3167 | (gdb) att 1980 |
| 3168 | Attaching to program `/home/dike/linux/2.3.26/um/linux', Pid 1980 |
| 3169 | 0x10070451 in __kill () |
| 3170 | (gdb) bt |
| 3171 | #0 0x10070451 in __kill () |
| 3172 | #1 0x10068ccd in usr1_pid (pid=1980) at process.c:30 |
| 3173 | #2 0x1006a03f in _switch_to (prev=0x50072000, next=0x507e8000) |
| 3174 | at process_kern.c:156 |
| 3175 | #3 0x1006a052 in switch_to (prev=0x50072000, next=0x507e8000, last=0x50072000) |
| 3176 | at process_kern.c:161 |
| 3177 | #4 0x10001d12 in schedule () at sched.c:777 |
| 3178 | #5 0x1006a744 in __down (sem=0x507d241c) at semaphore.c:71 |
| 3179 | #6 0x1006aa10 in __down_failed () at semaphore.c:157 |
| 3180 | #7 0x1006c5d8 in segv_handler (sc=0x5006e940) at trap_user.c:174 |
| 3181 | #8 0x1006c5ec in kern_segv_handler (sig=11) at trap_user.c:182 |
| 3182 | #9 <signal handler called> |
| 3183 | #10 0x10155404 in errno () |
| 3184 | #11 0x1006c0aa in segv (address=1342179328, is_write=2) at trap_kern.c:50 |
| 3185 | #12 0x1006c5d8 in segv_handler (sc=0x5006eaf8) at trap_user.c:174 |
| 3186 | #13 0x1006c5ec in kern_segv_handler (sig=11) at trap_user.c:182 |
| 3187 | #14 <signal handler called> |
| 3188 | #15 0xc0fd in ?? () |
| 3189 | #16 0x10016647 in sys_write (fd=3, |
| 3190 | buf=0x80b8800 <Address 0x80b8800 out of bounds>, count=1024) |
| 3191 | at read_write.c:159 |
| 3192 | #17 0x1006d5b3 in execute_syscall (syscall=4, args=0x5006ef08) |
| 3193 | at syscall_kern.c:254 |
| 3194 | #18 0x1006af87 in really_do_syscall (sig=12) at syscall_user.c:35 |
| 3195 | #19 <signal handler called> |
| 3196 | #20 0x400dc8b0 in ?? () |
| 3197 | |
| 3198 | |
| 3199 | |
| 3200 | |
| 3201 | |
| 3202 | The interesting things here are : |
| 3203 | |
| 3204 | +o There are two segfaults on this stack (frames 9 and 14) |
| 3205 | |
| 3206 | +o The first faulting address (frame 11) is 0x50000800 |
| 3207 | |
| 3208 | (gdb) p (void *)1342179328 |
| 3209 | $16 = (void *) 0x50000800 |
| 3210 | |
| 3211 | |
| 3212 | |
| 3213 | |
| 3214 | |
| 3215 | The initial faulting address is interesting because it is on the idle |
| 3216 | thread's stack. I had been seeing the idle thread segfault for no |
| 3217 | apparent reason, and the cause looked like stack corruption. In hopes |
| 3218 | of catching the culprit in the act, I had turned off all protections |
| 3219 | to that stack while the idle thread wasn't running. This apparently |
| 3220 | tripped that trap. |
| 3221 | |
| 3222 | |
| 3223 | However, the more immediate problem is that second segfault and I'm |
| 3224 | going to concentrate on that. First, I want to see where the fault |
| 3225 | happened, so I have to go look at the sigcontent struct in frame 8: |
| 3226 | |
| 3227 | |
| 3228 | |
| 3229 | (gdb) up |
| 3230 | #1 0x10068ccd in usr1_pid (pid=1980) at process.c:30 |
| 3231 | 30 kill(pid, SIGUSR1); |
| 3232 | (gdb) |
| 3233 | #2 0x1006a03f in _switch_to (prev=0x50072000, next=0x507e8000) |
| 3234 | at process_kern.c:156 |
| 3235 | 156 usr1_pid(getpid()); |
| 3236 | (gdb) |
| 3237 | #3 0x1006a052 in switch_to (prev=0x50072000, next=0x507e8000, last=0x50072000) |
| 3238 | at process_kern.c:161 |
| 3239 | 161 _switch_to(prev, next); |
| 3240 | (gdb) |
| 3241 | #4 0x10001d12 in schedule () at sched.c:777 |
| 3242 | 777 switch_to(prev, next, prev); |
| 3243 | (gdb) |
| 3244 | #5 0x1006a744 in __down (sem=0x507d241c) at semaphore.c:71 |
| 3245 | 71 schedule(); |
| 3246 | (gdb) |
| 3247 | #6 0x1006aa10 in __down_failed () at semaphore.c:157 |
| 3248 | 157 } |
| 3249 | (gdb) |
| 3250 | #7 0x1006c5d8 in segv_handler (sc=0x5006e940) at trap_user.c:174 |
| 3251 | 174 segv(sc->cr2, sc->err & 2); |
| 3252 | (gdb) |
| 3253 | #8 0x1006c5ec in kern_segv_handler (sig=11) at trap_user.c:182 |
| 3254 | 182 segv_handler(sc); |
| 3255 | (gdb) p *sc |
| 3256 | Cannot access memory at address 0x0. |
| 3257 | |
| 3258 | |
| 3259 | |
| 3260 | |
| 3261 | That's not very useful, so I'll try a more manual method: |
| 3262 | |
| 3263 | |
| 3264 | (gdb) p *((struct sigcontext *) (&sig + 1)) |
| 3265 | $19 = {gs = 0, __gsh = 0, fs = 0, __fsh = 0, es = 43, __esh = 0, ds = 43, |
| 3266 | __dsh = 0, edi = 1342179328, esi = 1350378548, ebp = 1342630440, |
| 3267 | esp = 1342630420, ebx = 1348150624, edx = 1280, ecx = 0, eax = 0, |
| 3268 | trapno = 14, err = 4, eip = 268480945, cs = 35, __csh = 0, eflags = 66118, |
| 3269 | esp_at_signal = 1342630420, ss = 43, __ssh = 0, fpstate = 0x0, oldmask = 0, |
| 3270 | cr2 = 1280} |
| 3271 | |
| 3272 | |
| 3273 | |
| 3274 | The ip is in handle_mm_fault: |
| 3275 | |
| 3276 | |
| 3277 | (gdb) p (void *)268480945 |
| 3278 | $20 = (void *) 0x1000b1b1 |
| 3279 | (gdb) i sym $20 |
| 3280 | handle_mm_fault + 57 in section .text |
| 3281 | |
| 3282 | |
| 3283 | |
| 3284 | |
| 3285 | |
| 3286 | Specifically, it's in pte_alloc: |
| 3287 | |
| 3288 | |
| 3289 | (gdb) i line *$20 |
| 3290 | Line 124 of "/home/dike/linux/2.3.26/um/include/asm/pgalloc.h" |
| 3291 | starts at address 0x1000b1b1 <handle_mm_fault+57> |
| 3292 | and ends at 0x1000b1b7 <handle_mm_fault+63>. |
| 3293 | |
| 3294 | |
| 3295 | |
| 3296 | |
| 3297 | |
| 3298 | To find where in handle_mm_fault this is, I'll jump forward in the |
| 3299 | code until I see an address in that procedure: |
| 3300 | |
| 3301 | |
| 3302 | |
| 3303 | (gdb) i line *0x1000b1c0 |
| 3304 | Line 126 of "/home/dike/linux/2.3.26/um/include/asm/pgalloc.h" |
| 3305 | starts at address 0x1000b1b7 <handle_mm_fault+63> |
| 3306 | and ends at 0x1000b1c3 <handle_mm_fault+75>. |
| 3307 | (gdb) i line *0x1000b1d0 |
| 3308 | Line 131 of "/home/dike/linux/2.3.26/um/include/asm/pgalloc.h" |
| 3309 | starts at address 0x1000b1d0 <handle_mm_fault+88> |
| 3310 | and ends at 0x1000b1da <handle_mm_fault+98>. |
| 3311 | (gdb) i line *0x1000b1e0 |
| 3312 | Line 61 of "/home/dike/linux/2.3.26/um/include/asm/pgalloc.h" |
| 3313 | starts at address 0x1000b1da <handle_mm_fault+98> |
| 3314 | and ends at 0x1000b1e1 <handle_mm_fault+105>. |
| 3315 | (gdb) i line *0x1000b1f0 |
| 3316 | Line 134 of "/home/dike/linux/2.3.26/um/include/asm/pgalloc.h" |
| 3317 | starts at address 0x1000b1f0 <handle_mm_fault+120> |
| 3318 | and ends at 0x1000b200 <handle_mm_fault+136>. |
| 3319 | (gdb) i line *0x1000b200 |
| 3320 | Line 135 of "/home/dike/linux/2.3.26/um/include/asm/pgalloc.h" |
| 3321 | starts at address 0x1000b200 <handle_mm_fault+136> |
| 3322 | and ends at 0x1000b208 <handle_mm_fault+144>. |
| 3323 | (gdb) i line *0x1000b210 |
| 3324 | Line 139 of "/home/dike/linux/2.3.26/um/include/asm/pgalloc.h" |
| 3325 | starts at address 0x1000b210 <handle_mm_fault+152> |
| 3326 | and ends at 0x1000b219 <handle_mm_fault+161>. |
| 3327 | (gdb) i line *0x1000b220 |
| 3328 | Line 1168 of "memory.c" starts at address 0x1000b21e <handle_mm_fault+166> |
| 3329 | and ends at 0x1000b222 <handle_mm_fault+170>. |
| 3330 | |
| 3331 | |
| 3332 | |
| 3333 | |
| 3334 | |
| 3335 | Something is apparently wrong with the page tables or vma_structs, so |
| 3336 | lets go back to frame 11 and have a look at them: |
| 3337 | |
| 3338 | |
| 3339 | |
| 3340 | #11 0x1006c0aa in segv (address=1342179328, is_write=2) at trap_kern.c:50 |
| 3341 | 50 handle_mm_fault(current, vma, address, is_write); |
| 3342 | (gdb) call pgd_offset_proc(vma->vm_mm, address) |
| 3343 | $22 = (pgd_t *) 0x80a548c |
| 3344 | |
| 3345 | |
| 3346 | |
| 3347 | |
| 3348 | |
| 3349 | That's pretty bogus. Page tables aren't supposed to be in process |
| 3350 | text or data areas. Let's see what's in the vma: |
| 3351 | |
| 3352 | |
| 3353 | (gdb) p *vma |
| 3354 | $23 = {vm_mm = 0x507d2434, vm_start = 0, vm_end = 134512640, |
| 3355 | vm_next = 0x80a4f8c, vm_page_prot = {pgprot = 0}, vm_flags = 31200, |
| 3356 | vm_avl_height = 2058, vm_avl_left = 0x80a8c94, vm_avl_right = 0x80d1000, |
| 3357 | vm_next_share = 0xaffffdb0, vm_pprev_share = 0xaffffe63, |
| 3358 | vm_ops = 0xaffffe7a, vm_pgoff = 2952789626, vm_file = 0xafffffec, |
| 3359 | vm_private_data = 0x62} |
| 3360 | (gdb) p *vma.vm_mm |
| 3361 | $24 = {mmap = 0x507d2434, mmap_avl = 0x0, mmap_cache = 0x8048000, |
| 3362 | pgd = 0x80a4f8c, mm_users = {counter = 0}, mm_count = {counter = 134904288}, |
| 3363 | map_count = 134909076, mmap_sem = {count = {counter = 135073792}, |
| 3364 | sleepers = -1342177872, wait = {lock = <optimized out or zero length>, |
| 3365 | task_list = {next = 0xaffffe63, prev = 0xaffffe7a}, |
| 3366 | __magic = -1342177670, __creator = -1342177300}, __magic = 98}, |
| 3367 | page_table_lock = {}, context = 138, start_code = 0, end_code = 0, |
| 3368 | start_data = 0, end_data = 0, start_brk = 0, brk = 0, start_stack = 0, |
| 3369 | arg_start = 0, arg_end = 0, env_start = 0, env_end = 0, rss = 1350381536, |
| 3370 | total_vm = 0, locked_vm = 0, def_flags = 0, cpu_vm_mask = 0, swap_cnt = 0, |
| 3371 | swap_address = 0, segments = 0x0} |
| 3372 | |
| 3373 | |
| 3374 | |
| 3375 | |
| 3376 | |
| 3377 | This also pretty bogus. With all of the 0x80xxxxx and 0xaffffxxx |
| 3378 | addresses, this is looking like a stack was plonked down on top of |
| 3379 | these structures. Maybe it's a stack overflow from the next page: |
| 3380 | |
| 3381 | |
| 3382 | |
| 3383 | (gdb) p vma |
| 3384 | $25 = (struct vm_area_struct *) 0x507d2434 |
| 3385 | |
| 3386 | |
| 3387 | |
| 3388 | |
| 3389 | |
| 3390 | That's towards the lower quarter of the page, so that would have to |
| 3391 | have been pretty heavy stack overflow: |
| 3392 | |
| 3393 | |
| 3394 | |
| 3395 | |
| 3396 | |
| 3397 | |
| 3398 | |
| 3399 | |
| 3400 | |
| 3401 | |
| 3402 | |
| 3403 | |
| 3404 | |
| 3405 | |
| 3406 | (gdb) x/100x $25 |
| 3407 | 0x507d2434: 0x507d2434 0x00000000 0x08048000 0x080a4f8c |
| 3408 | 0x507d2444: 0x00000000 0x080a79e0 0x080a8c94 0x080d1000 |
| 3409 | 0x507d2454: 0xaffffdb0 0xaffffe63 0xaffffe7a 0xaffffe7a |
| 3410 | 0x507d2464: 0xafffffec 0x00000062 0x0000008a 0x00000000 |
| 3411 | 0x507d2474: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3412 | 0x507d2484: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3413 | 0x507d2494: 0x00000000 0x00000000 0x507d2fe0 0x00000000 |
| 3414 | 0x507d24a4: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3415 | 0x507d24b4: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3416 | 0x507d24c4: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3417 | 0x507d24d4: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3418 | 0x507d24e4: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3419 | 0x507d24f4: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3420 | 0x507d2504: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3421 | 0x507d2514: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3422 | 0x507d2524: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3423 | 0x507d2534: 0x00000000 0x00000000 0x507d25dc 0x00000000 |
| 3424 | 0x507d2544: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3425 | 0x507d2554: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3426 | 0x507d2564: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3427 | 0x507d2574: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3428 | 0x507d2584: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3429 | 0x507d2594: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3430 | 0x507d25a4: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3431 | 0x507d25b4: 0x00000000 0x00000000 0x00000000 0x00000000 |
| 3432 | |
| 3433 | |
| 3434 | |
| 3435 | |
| 3436 | |
| 3437 | It's not stack overflow. The only "stack-like" piece of this data is |
| 3438 | the vma_struct itself. |
| 3439 | |
| 3440 | |
| 3441 | At this point, I don't see any avenues to pursue, so I just have to |
| 3442 | admit that I have no idea what's going on. What I will do, though, is |
| 3443 | stick a trap on the segfault handler which will stop if it sees any |
| 3444 | writes to the idle thread's stack. That was the thing that happened |
| 3445 | first, and it may be that if I can catch it immediately, what's going |
| 3446 | on will be somewhat clearer. |
| 3447 | |
| 3448 | |
| 3449 | 1122..22.. EEppiissooddee 22:: TThhee ccaassee ooff tthhee hhuunngg ffsscckk |
| 3450 | |
| 3451 | After setting a trap in the SEGV handler for accesses to the signal |
| 3452 | thread's stack, I reran the kernel. |
| 3453 | |
| 3454 | |
| 3455 | fsck hung again, this time by hitting the trap: |
| 3456 | |
| 3457 | |
| 3458 | |
| 3459 | |
| 3460 | |
| 3461 | |
| 3462 | |
| 3463 | |
| 3464 | |
| 3465 | |
| 3466 | |
| 3467 | |
| 3468 | |
| 3469 | |
| 3470 | |
| 3471 | |
| 3472 | Setting hostname uml [ OK ] |
| 3473 | Checking root filesystem |
| 3474 | /dev/fhd0 contains a file system with errors, check forced. |
| 3475 | Error reading block 86894 (Attempt to read block from filesystem resulted in short read) while reading indirect blocks of inode 19780. |
| 3476 | |
| 3477 | /dev/fhd0: UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY. |
| 3478 | (i.e., without -a or -p options) |
| 3479 | [ FAILED ] |
| 3480 | |
| 3481 | *** An error occurred during the file system check. |
| 3482 | *** Dropping you to a shell; the system will reboot |
| 3483 | *** when you leave the shell. |
| 3484 | Give root password for maintenance |
| 3485 | (or type Control-D for normal startup): |
| 3486 | |
| 3487 | [root@uml /root]# fsck -y /dev/fhd0 |
| 3488 | fsck -y /dev/fhd0 |
| 3489 | Parallelizing fsck version 1.14 (9-Jan-1999) |
| 3490 | e2fsck 1.14, 9-Jan-1999 for EXT2 FS 0.5b, 95/08/09 |
| 3491 | /dev/fhd0 contains a file system with errors, check forced. |
| 3492 | Pass 1: Checking inodes, blocks, and sizes |
| 3493 | Error reading block 86894 (Attempt to read block from filesystem resulted in short read) while reading indirect blocks of inode 19780. Ignore error? yes |
| 3494 | |
| 3495 | Pass 2: Checking directory structure |
| 3496 | Error reading block 49405 (Attempt to read block from filesystem resulted in short read). Ignore error? yes |
| 3497 | |
| 3498 | Directory inode 11858, block 0, offset 0: directory corrupted |
| 3499 | Salvage? yes |
| 3500 | |
| 3501 | Missing '.' in directory inode 11858. |
| 3502 | Fix? yes |
| 3503 | |
| 3504 | Missing '..' in directory inode 11858. |
| 3505 | Fix? yes |
| 3506 | |
| 3507 | Untested (4127) [100fe44c]: trap_kern.c line 31 |
| 3508 | |
| 3509 | |
| 3510 | |
| 3511 | |
| 3512 | |
| 3513 | I need to get the signal thread to detach from pid 4127 so that I can |
| 3514 | attach to it with gdb. This is done by sending it a SIGUSR1, which is |
| 3515 | caught by the signal thread, which detaches the process: |
| 3516 | |
| 3517 | |
| 3518 | kill -USR1 4127 |
| 3519 | |
| 3520 | |
| 3521 | |
| 3522 | |
| 3523 | |
| 3524 | Now I can run gdb on it: |
| 3525 | |
| 3526 | |
| 3527 | |
| 3528 | |
| 3529 | |
| 3530 | |
| 3531 | |
| 3532 | |
| 3533 | |
| 3534 | |
| 3535 | |
| 3536 | |
| 3537 | |
| 3538 | ~/linux/2.3.26/um 1034: gdb linux |
| 3539 | GNU gdb 4.17.0.11 with Linux support |
| 3540 | Copyright 1998 Free Software Foundation, Inc. |
| 3541 | GDB is free software, covered by the GNU General Public License, and you are |
| 3542 | welcome to change it and/or distribute copies of it under certain conditions. |
| 3543 | Type "show copying" to see the conditions. |
| 3544 | There is absolutely no warranty for GDB. Type "show warranty" for details. |
| 3545 | This GDB was configured as "i386-redhat-linux"... |
| 3546 | (gdb) att 4127 |
| 3547 | Attaching to program `/home/dike/linux/2.3.26/um/linux', Pid 4127 |
| 3548 | 0x10075891 in __libc_nanosleep () |
| 3549 | |
| 3550 | |
| 3551 | |
| 3552 | |
| 3553 | |
| 3554 | The backtrace shows that it was in a write and that the fault address |
| 3555 | (address in frame 3) is 0x50000800, which is right in the middle of |
| 3556 | the signal thread's stack page: |
| 3557 | |
| 3558 | |
| 3559 | (gdb) bt |
| 3560 | #0 0x10075891 in __libc_nanosleep () |
| 3561 | #1 0x1007584d in __sleep (seconds=1000000) |
| 3562 | at ../sysdeps/unix/sysv/linux/sleep.c:78 |
| 3563 | #2 0x1006ce9a in stop () at user_util.c:191 |
| 3564 | #3 0x1006bf88 in segv (address=1342179328, is_write=2) at trap_kern.c:31 |
| 3565 | #4 0x1006c628 in segv_handler (sc=0x5006eaf8) at trap_user.c:174 |
| 3566 | #5 0x1006c63c in kern_segv_handler (sig=11) at trap_user.c:182 |
| 3567 | #6 <signal handler called> |
| 3568 | #7 0xc0fd in ?? () |
| 3569 | #8 0x10016647 in sys_write (fd=3, buf=0x80b8800 "R.", count=1024) |
| 3570 | at read_write.c:159 |
| 3571 | #9 0x1006d603 in execute_syscall (syscall=4, args=0x5006ef08) |
| 3572 | at syscall_kern.c:254 |
| 3573 | #10 0x1006af87 in really_do_syscall (sig=12) at syscall_user.c:35 |
| 3574 | #11 <signal handler called> |
| 3575 | #12 0x400dc8b0 in ?? () |
| 3576 | #13 <signal handler called> |
| 3577 | #14 0x400dc8b0 in ?? () |
| 3578 | #15 0x80545fd in ?? () |
| 3579 | #16 0x804daae in ?? () |
| 3580 | #17 0x8054334 in ?? () |
| 3581 | #18 0x804d23e in ?? () |
| 3582 | #19 0x8049632 in ?? () |
| 3583 | #20 0x80491d2 in ?? () |
| 3584 | #21 0x80596b5 in ?? () |
| 3585 | (gdb) p (void *)1342179328 |
| 3586 | $3 = (void *) 0x50000800 |
| 3587 | |
| 3588 | |
| 3589 | |
| 3590 | |
| 3591 | |
| 3592 | Going up the stack to the segv_handler frame and looking at where in |
| 3593 | the code the access happened shows that it happened near line 110 of |
| 3594 | block_dev.c: |
| 3595 | |
| 3596 | |
| 3597 | |
| 3598 | |
| 3599 | |
| 3600 | |
| 3601 | |
| 3602 | |
| 3603 | |
| 3604 | (gdb) up |
| 3605 | #1 0x1007584d in __sleep (seconds=1000000) |
| 3606 | at ../sysdeps/unix/sysv/linux/sleep.c:78 |
| 3607 | ../sysdeps/unix/sysv/linux/sleep.c:78: No such file or directory. |
| 3608 | (gdb) |
| 3609 | #2 0x1006ce9a in stop () at user_util.c:191 |
| 3610 | 191 while(1) sleep(1000000); |
| 3611 | (gdb) |
| 3612 | #3 0x1006bf88 in segv (address=1342179328, is_write=2) at trap_kern.c:31 |
| 3613 | 31 KERN_UNTESTED(); |
| 3614 | (gdb) |
| 3615 | #4 0x1006c628 in segv_handler (sc=0x5006eaf8) at trap_user.c:174 |
| 3616 | 174 segv(sc->cr2, sc->err & 2); |
| 3617 | (gdb) p *sc |
| 3618 | $1 = {gs = 0, __gsh = 0, fs = 0, __fsh = 0, es = 43, __esh = 0, ds = 43, |
| 3619 | __dsh = 0, edi = 1342179328, esi = 134973440, ebp = 1342631484, |
| 3620 | esp = 1342630864, ebx = 256, edx = 0, ecx = 256, eax = 1024, trapno = 14, |
| 3621 | err = 6, eip = 268550834, cs = 35, __csh = 0, eflags = 66070, |
| 3622 | esp_at_signal = 1342630864, ss = 43, __ssh = 0, fpstate = 0x0, oldmask = 0, |
| 3623 | cr2 = 1342179328} |
| 3624 | (gdb) p (void *)268550834 |
| 3625 | $2 = (void *) 0x1001c2b2 |
| 3626 | (gdb) i sym $2 |
| 3627 | block_write + 1090 in section .text |
| 3628 | (gdb) i line *$2 |
| 3629 | Line 209 of "/home/dike/linux/2.3.26/um/include/asm/arch/string.h" |
| 3630 | starts at address 0x1001c2a1 <block_write+1073> |
| 3631 | and ends at 0x1001c2bf <block_write+1103>. |
| 3632 | (gdb) i line *0x1001c2c0 |
| 3633 | Line 110 of "block_dev.c" starts at address 0x1001c2bf <block_write+1103> |
| 3634 | and ends at 0x1001c2e3 <block_write+1139>. |
| 3635 | |
| 3636 | |
| 3637 | |
| 3638 | |
| 3639 | |
| 3640 | Looking at the source shows that the fault happened during a call to |
| 3641 | copy_to_user to copy the data into the kernel: |
| 3642 | |
| 3643 | |
| 3644 | 107 count -= chars; |
| 3645 | 108 copy_from_user(p,buf,chars); |
| 3646 | 109 p += chars; |
| 3647 | 110 buf += chars; |
| 3648 | |
| 3649 | |
| 3650 | |
| 3651 | |
| 3652 | |
| 3653 | p is the pointer which must contain 0x50000800, since buf contains |
| 3654 | 0x80b8800 (frame 8 above). It is defined as: |
| 3655 | |
| 3656 | |
| 3657 | p = offset + bh->b_data; |
| 3658 | |
| 3659 | |
| 3660 | |
| 3661 | |
| 3662 | |
| 3663 | I need to figure out what bh is, and it just so happens that bh is |
| 3664 | passed as an argument to mark_buffer_uptodate and mark_buffer_dirty a |
| 3665 | few lines later, so I do a little disassembly: |
| 3666 | |
| 3667 | |
| 3668 | |
| 3669 | |
| 3670 | (gdb) disas 0x1001c2bf 0x1001c2e0 |
| 3671 | Dump of assembler code from 0x1001c2bf to 0x1001c2d0: |
| 3672 | 0x1001c2bf <block_write+1103>: addl %eax,0xc(%ebp) |
| 3673 | 0x1001c2c2 <block_write+1106>: movl 0xfffffdd4(%ebp),%edx |
| 3674 | 0x1001c2c8 <block_write+1112>: btsl $0x0,0x18(%edx) |
| 3675 | 0x1001c2cd <block_write+1117>: btsl $0x1,0x18(%edx) |
| 3676 | 0x1001c2d2 <block_write+1122>: sbbl %ecx,%ecx |
| 3677 | 0x1001c2d4 <block_write+1124>: testl %ecx,%ecx |
| 3678 | 0x1001c2d6 <block_write+1126>: jne 0x1001c2e3 <block_write+1139> |
| 3679 | 0x1001c2d8 <block_write+1128>: pushl $0x0 |
| 3680 | 0x1001c2da <block_write+1130>: pushl %edx |
| 3681 | 0x1001c2db <block_write+1131>: call 0x1001819c <__mark_buffer_dirty> |
| 3682 | End of assembler dump. |
| 3683 | |
| 3684 | |
| 3685 | |
| 3686 | |
| 3687 | |
| 3688 | At that point, bh is in %edx (address 0x1001c2da), which is calculated |
| 3689 | at 0x1001c2c2 as %ebp + 0xfffffdd4, so I figure exactly what that is, |
| 3690 | taking %ebp from the sigcontext_struct above: |
| 3691 | |
| 3692 | |
| 3693 | (gdb) p (void *)1342631484 |
| 3694 | $5 = (void *) 0x5006ee3c |
| 3695 | (gdb) p 0x5006ee3c+0xfffffdd4 |
| 3696 | $6 = 1342630928 |
| 3697 | (gdb) p (void *)$6 |
| 3698 | $7 = (void *) 0x5006ec10 |
| 3699 | (gdb) p *((void **)$7) |
| 3700 | $8 = (void *) 0x50100200 |
| 3701 | |
| 3702 | |
| 3703 | |
| 3704 | |
| 3705 | |
| 3706 | Now, I look at the structure to see what's in it, and particularly, |
| 3707 | what its b_data field contains: |
| 3708 | |
| 3709 | |
| 3710 | (gdb) p *((struct buffer_head *)0x50100200) |
| 3711 | $13 = {b_next = 0x50289380, b_blocknr = 49405, b_size = 1024, b_list = 0, |
| 3712 | b_dev = 15872, b_count = {counter = 1}, b_rdev = 15872, b_state = 24, |
| 3713 | b_flushtime = 0, b_next_free = 0x501001a0, b_prev_free = 0x50100260, |
| 3714 | b_this_page = 0x501001a0, b_reqnext = 0x0, b_pprev = 0x507fcf58, |
| 3715 | b_data = 0x50000800 "", b_page = 0x50004000, |
| 3716 | b_end_io = 0x10017f60 <end_buffer_io_sync>, b_dev_id = 0x0, |
| 3717 | b_rsector = 98810, b_wait = {lock = <optimized out or zero length>, |
| 3718 | task_list = {next = 0x50100248, prev = 0x50100248}, __magic = 1343226448, |
| 3719 | __creator = 0}, b_kiobuf = 0x0} |
| 3720 | |
| 3721 | |
| 3722 | |
| 3723 | |
| 3724 | |
| 3725 | The b_data field is indeed 0x50000800, so the question becomes how |
| 3726 | that happened. The rest of the structure looks fine, so this probably |
| 3727 | is not a case of data corruption. It happened on purpose somehow. |
| 3728 | |
| 3729 | |
| 3730 | The b_page field is a pointer to the page_struct representing the |
| 3731 | 0x50000000 page. Looking at it shows the kernel's idea of the state |
| 3732 | of that page: |
| 3733 | |
| 3734 | |
| 3735 | |
| 3736 | (gdb) p *$13.b_page |
| 3737 | $17 = {list = {next = 0x50004a5c, prev = 0x100c5174}, mapping = 0x0, |
| 3738 | index = 0, next_hash = 0x0, count = {counter = 1}, flags = 132, lru = { |
| 3739 | next = 0x50008460, prev = 0x50019350}, wait = { |
| 3740 | lock = <optimized out or zero length>, task_list = {next = 0x50004024, |
| 3741 | prev = 0x50004024}, __magic = 1342193708, __creator = 0}, |
| 3742 | pprev_hash = 0x0, buffers = 0x501002c0, virtual = 1342177280, |
| 3743 | zone = 0x100c5160} |
| 3744 | |
| 3745 | |
| 3746 | |
| 3747 | |
| 3748 | |
| 3749 | Some sanity-checking: the virtual field shows the "virtual" address of |
| 3750 | this page, which in this kernel is the same as its "physical" address, |
| 3751 | and the page_struct itself should be mem_map[0], since it represents |
| 3752 | the first page of memory: |
| 3753 | |
| 3754 | |
| 3755 | |
| 3756 | (gdb) p (void *)1342177280 |
| 3757 | $18 = (void *) 0x50000000 |
| 3758 | (gdb) p mem_map |
| 3759 | $19 = (mem_map_t *) 0x50004000 |
| 3760 | |
| 3761 | |
| 3762 | |
| 3763 | |
| 3764 | |
| 3765 | These check out fine. |
| 3766 | |
| 3767 | |
| 3768 | Now to check out the page_struct itself. In particular, the flags |
| 3769 | field shows whether the page is considered free or not: |
| 3770 | |
| 3771 | |
| 3772 | (gdb) p (void *)132 |
| 3773 | $21 = (void *) 0x84 |
| 3774 | |
| 3775 | |
| 3776 | |
| 3777 | |
| 3778 | |
| 3779 | The "reserved" bit is the high bit, which is definitely not set, so |
| 3780 | the kernel considers the signal stack page to be free and available to |
| 3781 | be used. |
| 3782 | |
| 3783 | |
| 3784 | At this point, I jump to conclusions and start looking at my early |
| 3785 | boot code, because that's where that page is supposed to be reserved. |
| 3786 | |
| 3787 | |
| 3788 | In my setup_arch procedure, I have the following code which looks just |
| 3789 | fine: |
| 3790 | |
| 3791 | |
| 3792 | |
| 3793 | bootmap_size = init_bootmem(start_pfn, end_pfn - start_pfn); |
| 3794 | free_bootmem(__pa(low_physmem) + bootmap_size, high_physmem - low_physmem); |
| 3795 | |
| 3796 | |
| 3797 | |
| 3798 | |
| 3799 | |
| 3800 | Two stack pages have already been allocated, and low_physmem points to |
| 3801 | the third page, which is the beginning of free memory. |
| 3802 | The init_bootmem call declares the entire memory to the boot memory |
| 3803 | manager, which marks it all reserved. The free_bootmem call frees up |
| 3804 | all of it, except for the first two pages. This looks correct to me. |
| 3805 | |
| 3806 | |
| 3807 | So, I decide to see init_bootmem run and make sure that it is marking |
| 3808 | those first two pages as reserved. I never get that far. |
| 3809 | |
| 3810 | |
| 3811 | Stepping into init_bootmem, and looking at bootmem_map before looking |
| 3812 | at what it contains shows the following: |
| 3813 | |
| 3814 | |
| 3815 | |
| 3816 | (gdb) p bootmem_map |
| 3817 | $3 = (void *) 0x50000000 |
| 3818 | |
| 3819 | |
| 3820 | |
| 3821 | |
| 3822 | |
| 3823 | Aha! The light dawns. That first page is doing double duty as a |
| 3824 | stack and as the boot memory map. The last thing that the boot memory |
| 3825 | manager does is to free the pages used by its memory map, so this page |
| 3826 | is getting freed even its marked as reserved. |
| 3827 | |
| 3828 | |
| 3829 | The fix was to initialize the boot memory manager before allocating |
| 3830 | those two stack pages, and then allocate them through the boot memory |
| 3831 | manager. After doing this, and fixing a couple of subsequent buglets, |
| 3832 | the stack corruption problem disappeared. |
| 3833 | |
| 3834 | |
| 3835 | |
| 3836 | |
| 3837 | |
| 3838 | 1133.. WWhhaatt ttoo ddoo wwhheenn UUMMLL ddooeessnn''tt wwoorrkk |
| 3839 | |
| 3840 | |
| 3841 | |
| 3842 | |
| 3843 | 1133..11.. SSttrraannggee ccoommppiillaattiioonn eerrrroorrss wwhheenn yyoouu bbuuiilldd ffrroomm ssoouurrccee |
| 3844 | |
| 3845 | As of test11, it is necessary to have "ARCH=um" in the environment or |
| 3846 | on the make command line for all steps in building UML, including |
| 3847 | clean, distclean, or mrproper, config, menuconfig, or xconfig, dep, |
| 3848 | and linux. If you forget for any of them, the i386 build seems to |
| 3849 | contaminate the UML build. If this happens, start from scratch with |
| 3850 | |
| 3851 | |
| 3852 | host% |
| 3853 | make mrproper ARCH=um |
| 3854 | |
| 3855 | |
| 3856 | |
| 3857 | |
| 3858 | and repeat the build process with ARCH=um on all the steps. |
| 3859 | |
| 3860 | |
| 3861 | See ``Compiling the kernel and modules'' for more details. |
| 3862 | |
| 3863 | |
| 3864 | Another cause of strange compilation errors is building UML in |
| 3865 | /usr/src/linux. If you do this, the first thing you need to do is |
| 3866 | clean up the mess you made. The /usr/src/linux/asm link will now |
| 3867 | point to /usr/src/linux/asm-um. Make it point back to |
| 3868 | /usr/src/linux/asm-i386. Then, move your UML pool someplace else and |
| 3869 | build it there. Also see below, where a more specific set of symptoms |
| 3870 | is described. |
| 3871 | |
| 3872 | |
| 3873 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3874 | 1133..33.. AA vvaarriieettyy ooff ppaanniiccss aanndd hhaannggss wwiitthh //ttmmpp oonn aa rreeiisseerrffss ffiilleessyyss-- |
| 3875 | tteemm |
| 3876 | |
| 3877 | I saw this on reiserfs 3.5.21 and it seems to be fixed in 3.5.27. |
| 3878 | Panics preceded by |
| 3879 | |
| 3880 | |
| 3881 | Detaching pid nnnn |
| 3882 | |
| 3883 | |
| 3884 | |
| 3885 | are diagnostic of this problem. This is a reiserfs bug which causes a |
| 3886 | thread to occasionally read stale data from a mmapped page shared with |
| 3887 | another thread. The fix is to upgrade the filesystem or to have /tmp |
| 3888 | be an ext2 filesystem. |
| 3889 | |
| 3890 | |
| 3891 | |
| 3892 | 1133..44.. TThhee ccoommppiillee ffaaiillss wwiitthh eerrrroorrss aabboouutt ccoonnfflliiccttiinngg ttyyppeess ffoorr |
| 3893 | ''ooppeenn'',, ''dduupp'',, aanndd ''wwaaiittppiidd'' |
| 3894 | |
| 3895 | This happens when you build in /usr/src/linux. The UML build makes |
| 3896 | the include/asm link point to include/asm-um. /usr/include/asm points |
| 3897 | to /usr/src/linux/include/asm, so when that link gets moved, files |
| 3898 | which need to include the asm-i386 versions of headers get the |
| 3899 | incompatible asm-um versions. The fix is to move the include/asm link |
| 3900 | back to include/asm-i386 and to do UML builds someplace else. |
| 3901 | |
| 3902 | |
| 3903 | |
| 3904 | 1133..55.. UUMMLL ddooeessnn''tt wwoorrkk wwhheenn //ttmmpp iiss aann NNFFSS ffiilleessyysstteemm |
| 3905 | |
| 3906 | This seems to be a similar situation with the resierfs problem above. |
| 3907 | Some versions of NFS seems not to handle mmap correctly, which UML |
| 3908 | depends on. The workaround is have /tmp be non-NFS directory. |
| 3909 | |
| 3910 | |
| 3911 | 1133..66.. UUMMLL hhaannggss oonn bboooott wwhheenn ccoommppiilleedd wwiitthh ggpprrooff ssuuppppoorrtt |
| 3912 | |
| 3913 | If you build UML with gprof support and, early in the boot, it does |
| 3914 | this |
| 3915 | |
| 3916 | |
| 3917 | kernel BUG at page_alloc.c:100! |
| 3918 | |
| 3919 | |
| 3920 | |
| 3921 | |
| 3922 | you have a buggy gcc. You can work around the problem by removing |
| 3923 | UM_FASTCALL from CFLAGS in arch/um/Makefile-i386. This will open up |
| 3924 | another bug, but that one is fairly hard to reproduce. |
| 3925 | |
| 3926 | |
| 3927 | |
| 3928 | 1133..77.. ssyyssllooggdd ddiieess wwiitthh aa SSIIGGTTEERRMM oonn ssttaarrttuupp |
| 3929 | |
| 3930 | The exact boot error depends on the distribution that you're booting, |
| 3931 | but Debian produces this: |
| 3932 | |
| 3933 | |
| 3934 | /etc/rc2.d/S10sysklogd: line 49: 93 Terminated |
| 3935 | start-stop-daemon --start --quiet --exec /sbin/syslogd -- $SYSLOGD |
| 3936 | |
| 3937 | |
| 3938 | |
| 3939 | |
| 3940 | This is a syslogd bug. There's a race between a parent process |
| 3941 | installing a signal handler and its child sending the signal. See |
| 3942 | this uml-devel post <http://www.geocrawler.com/lists/3/Source- |
| 3943 | Forge/709/0/6612801> for the details. |
| 3944 | |
| 3945 | |
| 3946 | |
| 3947 | 1133..88.. TTUUNN//TTAAPP nneettwwoorrkkiinngg ddooeessnn''tt wwoorrkk oonn aa 22..44 hhoosstt |
| 3948 | |
| 3949 | There are a couple of problems which were |
| 3950 | <http://www.geocrawler.com/lists/3/SourceForge/597/0/> name="pointed |
| 3951 | out"> by Tim Robinson <timro at trkr dot net> |
| 3952 | |
| 3953 | +o It doesn't work on hosts running 2.4.7 (or thereabouts) or earlier. |
| 3954 | The fix is to upgrade to something more recent and then read the |
| 3955 | next item. |
| 3956 | |
| 3957 | +o If you see |
| 3958 | |
| 3959 | |
| 3960 | File descriptor in bad state |
| 3961 | |
| 3962 | |
| 3963 | |
| 3964 | when you bring up the device inside UML, you have a header mismatch |
| 3965 | between the original kernel and the upgraded one. Make /usr/src/linux |
| 3966 | point at the new headers. This will only be a problem if you build |
| 3967 | uml_net yourself. |
| 3968 | |
| 3969 | |
| 3970 | |
| 3971 | 1133..99.. YYoouu ccaann nneettwwoorrkk ttoo tthhee hhoosstt bbuutt nnoott ttoo ootthheerr mmaacchhiinneess oonn tthhee |
| 3972 | nneett |
| 3973 | |
| 3974 | If you can connect to the host, and the host can connect to UML, but |
| 3975 | you can not connect to any other machines, then you may need to enable |
| 3976 | IP Masquerading on the host. Usually this is only experienced when |
| 3977 | using private IP addresses (192.168.x.x or 10.x.x.x) for host/UML |
| 3978 | networking, rather than the public address space that your host is |
| 3979 | connected to. UML does not enable IP Masquerading, so you will need |
| 3980 | to create a static rule to enable it: |
| 3981 | |
| 3982 | |
| 3983 | host% |
| 3984 | iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE |
| 3985 | |
| 3986 | |
| 3987 | |
| 3988 | |
| 3989 | Replace eth0 with the interface that you use to talk to the rest of |
| 3990 | the world. |
| 3991 | |
| 3992 | |
| 3993 | Documentation on IP Masquerading, and SNAT, can be found at |
| 3994 | www.netfilter.org <http://www.netfilter.org> . |
| 3995 | |
| 3996 | |
| 3997 | If you can reach the local net, but not the outside Internet, then |
| 3998 | that is usually a routing problem. The UML needs a default route: |
| 3999 | |
| 4000 | |
| 4001 | UML# |
| 4002 | route add default gw gateway IP |
| 4003 | |
| 4004 | |
| 4005 | |
| 4006 | |
| 4007 | The gateway IP can be any machine on the local net that knows how to |
| 4008 | reach the outside world. Usually, this is the host or the local net- |
| 4009 | work's gateway. |
| 4010 | |
| 4011 | |
| 4012 | Occasionally, we hear from someone who can reach some machines, but |
| 4013 | not others on the same net, or who can reach some ports on other |
| 4014 | machines, but not others. These are usually caused by strange |
| 4015 | firewalling somewhere between the UML and the other box. You track |
| 4016 | this down by running tcpdump on every interface the packets travel |
| 4017 | over and see where they disappear. When you find a machine that takes |
| 4018 | the packets in, but does not send them onward, that's the culprit. |
| 4019 | |
| 4020 | |
| 4021 | |
| 4022 | 1133..1100.. II hhaavvee nnoo rroooott aanndd II wwaanntt ttoo ssccrreeaamm |
| 4023 | |
| 4024 | Thanks to Birgit Wahlich for telling me about this strange one. It |
| 4025 | turns out that there's a limit of six environment variables on the |
| 4026 | kernel command line. When that limit is reached or exceeded, argument |
| 4027 | processing stops, which means that the 'root=' argument that UML |
| 4028 | usually adds is not seen. So, the filesystem has no idea what the |
| 4029 | root device is, so it panics. |
| 4030 | |
| 4031 | |
| 4032 | The fix is to put less stuff on the command line. Glomming all your |
| 4033 | setup variables into one is probably the best way to go. |
| 4034 | |
| 4035 | |
| 4036 | |
| 4037 | 1133..1111.. UUMMLL bbuuiilldd ccoonnfflliicctt bbeettwweeeenn ppttrraaccee..hh aanndd uuccoonntteexxtt..hh |
| 4038 | |
| 4039 | On some older systems, /usr/include/asm/ptrace.h and |
| 4040 | /usr/include/sys/ucontext.h define the same names. So, when they're |
| 4041 | included together, the defines from one completely mess up the parsing |
| 4042 | of the other, producing errors like: |
| 4043 | /usr/include/sys/ucontext.h:47: parse error before |
| 4044 | `10' |
| 4045 | |
| 4046 | |
| 4047 | |
| 4048 | |
| 4049 | plus a pile of warnings. |
| 4050 | |
| 4051 | |
| 4052 | This is a libc botch, which has since been fixed, and I don't see any |
| 4053 | way around it besides upgrading. |
| 4054 | |
| 4055 | |
| 4056 | |
| 4057 | 1133..1122.. TThhee UUMMLL BBooggooMMiippss iiss eexxaaccttllyy hhaallff tthhee hhoosstt''ss BBooggooMMiippss |
| 4058 | |
| 4059 | On i386 kernels, there are two ways of running the loop that is used |
| 4060 | to calculate the BogoMips rating, using the TSC if it's there or using |
| 4061 | a one-instruction loop. The TSC produces twice the BogoMips as the |
| 4062 | loop. UML uses the loop, since it has nothing resembling a TSC, and |
| 4063 | will get almost exactly the same BogoMips as a host using the loop. |
| 4064 | However, on a host with a TSC, its BogoMips will be double the loop |
| 4065 | BogoMips, and therefore double the UML BogoMips. |
| 4066 | |
| 4067 | |
| 4068 | |
| 4069 | 1133..1133.. WWhheenn yyoouu rruunn UUMMLL,, iitt iimmmmeeddiiaatteellyy sseeggffaauullttss |
| 4070 | |
| 4071 | If the host is configured with the 2G/2G address space split, that's |
| 4072 | why. See ``UML on 2G/2G hosts'' for the details on getting UML to |
| 4073 | run on your host. |
| 4074 | |
| 4075 | |
| 4076 | |
| 4077 | 1133..1144.. xxtteerrmmss aappppeeaarr,, tthheenn iimmmmeeddiiaatteellyy ddiissaappppeeaarr |
| 4078 | |
| 4079 | If you're running an up to date kernel with an old release of |
| 4080 | uml_utilities, the port-helper program will not work properly, so |
| 4081 | xterms will exit straight after they appear. The solution is to |
| 4082 | upgrade to the latest release of uml_utilities. Usually this problem |
| 4083 | occurs when you have installed a packaged release of UML then compiled |
| 4084 | your own development kernel without upgrading the uml_utilities from |
| 4085 | the source distribution. |
| 4086 | |
| 4087 | |
| 4088 | |
| 4089 | 1133..1155.. AAnnyy ootthheerr ppaanniicc,, hhaanngg,, oorr ssttrraannggee bbeehhaavviioorr |
| 4090 | |
| 4091 | If you're seeing truly strange behavior, such as hangs or panics that |
| 4092 | happen in random places, or you try running the debugger to see what's |
| 4093 | happening and it acts strangely, then it could be a problem in the |
| 4094 | host kernel. If you're not running a stock Linus or -ac kernel, then |
| 4095 | try that. An early version of the preemption patch and a 2.4.10 SuSE |
| 4096 | kernel have caused very strange problems in UML. |
| 4097 | |
| 4098 | |
| 4099 | Otherwise, let me know about it. Send a message to one of the UML |
| 4100 | mailing lists - either the developer list - user-mode-linux-devel at |
| 4101 | lists dot sourceforge dot net (subscription info) or the user list - |
| 4102 | user-mode-linux-user at lists dot sourceforge do net (subscription |
| 4103 | info), whichever you prefer. Don't assume that everyone knows about |
| 4104 | it and that a fix is imminent. |
| 4105 | |
| 4106 | |
| 4107 | If you want to be super-helpful, read ``Diagnosing Problems'' and |
| 4108 | follow the instructions contained therein. |
| 4109 | 1144.. DDiiaaggnnoossiinngg PPrroobblleemmss |
| 4110 | |
| 4111 | |
| 4112 | If you get UML to crash, hang, or otherwise misbehave, you should |
| 4113 | report this on one of the project mailing lists, either the developer |
| 4114 | list - user-mode-linux-devel at lists dot sourceforge dot net |
| 4115 | (subscription info) or the user list - user-mode-linux-user at lists |
| 4116 | dot sourceforge dot net (subscription info). When you do, it is |
| 4117 | likely that I will want more information. So, it would be helpful to |
| 4118 | read the stuff below, do whatever is applicable in your case, and |
| 4119 | report the results to the list. |
| 4120 | |
| 4121 | |
| 4122 | For any diagnosis, you're going to need to build a debugging kernel. |
| 4123 | The binaries from this site aren't debuggable. If you haven't done |
| 4124 | this before, read about ``Compiling the kernel and modules'' and |
| 4125 | ``Kernel debugging'' UML first. |
| 4126 | |
| 4127 | |
| 4128 | 1144..11.. CCaassee 11 :: NNoorrmmaall kkeerrnneell ppaanniiccss |
| 4129 | |
| 4130 | The most common case is for a normal thread to panic. To debug this, |
| 4131 | you will need to run it under the debugger (add 'debug' to the command |
| 4132 | line). An xterm will start up with gdb running inside it. Continue |
| 4133 | it when it stops in start_kernel and make it crash. Now ^C gdb and |
| 4134 | |
| 4135 | |
| 4136 | If the panic was a "Kernel mode fault", then there will be a segv |
| 4137 | frame on the stack and I'm going to want some more information. The |
| 4138 | stack might look something like this: |
| 4139 | |
| 4140 | |
| 4141 | (UML gdb) backtrace |
| 4142 | #0 0x1009bf76 in __sigprocmask (how=1, set=0x5f347940, oset=0x0) |
| 4143 | at ../sysdeps/unix/sysv/linux/sigprocmask.c:49 |
| 4144 | #1 0x10091411 in change_sig (signal=10, on=1) at process.c:218 |
| 4145 | #2 0x10094785 in timer_handler (sig=26) at time_kern.c:32 |
| 4146 | #3 0x1009bf38 in __restore () |
| 4147 | at ../sysdeps/unix/sysv/linux/i386/sigaction.c:125 |
| 4148 | #4 0x1009534c in segv (address=8, ip=268849158, is_write=2, is_user=0) |
| 4149 | at trap_kern.c:66 |
| 4150 | #5 0x10095c04 in segv_handler (sig=11) at trap_user.c:285 |
| 4151 | #6 0x1009bf38 in __restore () |
| 4152 | |
| 4153 | |
| 4154 | |
| 4155 | |
| 4156 | I'm going to want to see the symbol and line information for the value |
| 4157 | of ip in the segv frame. In this case, you would do the following: |
| 4158 | |
| 4159 | |
| 4160 | (UML gdb) i sym 268849158 |
| 4161 | |
| 4162 | |
| 4163 | |
| 4164 | |
| 4165 | and |
| 4166 | |
| 4167 | |
| 4168 | (UML gdb) i line *268849158 |
| 4169 | |
| 4170 | |
| 4171 | |
| 4172 | |
| 4173 | The reason for this is the __restore frame right above the segv_han- |
| 4174 | dler frame is hiding the frame that actually segfaulted. So, I have |
| 4175 | to get that information from the faulting ip. |
| 4176 | |
| 4177 | |
| 4178 | 1144..22.. CCaassee 22 :: TTrraacciinngg tthhrreeaadd ppaanniiccss |
| 4179 | |
| 4180 | The less common and more painful case is when the tracing thread |
| 4181 | panics. In this case, the kernel debugger will be useless because it |
| 4182 | needs a healthy tracing thread in order to work. The first thing to |
| 4183 | do is get a backtrace from the tracing thread. This is done by |
| 4184 | figuring out what its pid is, firing up gdb, and attaching it to that |
| 4185 | pid. You can figure out the tracing thread pid by looking at the |
| 4186 | first line of the console output, which will look like this: |
| 4187 | |
| 4188 | |
| 4189 | tracing thread pid = 15851 |
| 4190 | |
| 4191 | |
| 4192 | |
| 4193 | |
| 4194 | or by running ps on the host and finding the line that looks like |
| 4195 | this: |
| 4196 | |
| 4197 | |
| 4198 | jdike 15851 4.5 0.4 132568 1104 pts/0 S 21:34 0:05 ./linux [(tracing thread)] |
| 4199 | |
| 4200 | |
| 4201 | |
| 4202 | |
| 4203 | If the panic was 'segfault in signals', then follow the instructions |
| 4204 | above for collecting information about the location of the seg fault. |
| 4205 | |
| 4206 | |
| 4207 | If the tracing thread flaked out all by itself, then send that |
| 4208 | backtrace in and wait for our crack debugging team to fix the problem. |
| 4209 | |
| 4210 | |
| 4211 | 1144..33.. CCaassee 33 :: TTrraacciinngg tthhrreeaadd ppaanniiccss ccaauusseedd bbyy ootthheerr tthhrreeaaddss |
| 4212 | |
| 4213 | However, there are cases where the misbehavior of another thread |
| 4214 | caused the problem. The most common panic of this type is: |
| 4215 | |
| 4216 | |
| 4217 | wait_for_stop failed to wait for <pid> to stop with <signal number> |
| 4218 | |
| 4219 | |
| 4220 | |
| 4221 | |
| 4222 | In this case, you'll need to get a backtrace from the process men- |
| 4223 | tioned in the panic, which is complicated by the fact that the kernel |
| 4224 | debugger is defunct and without some fancy footwork, another gdb can't |
| 4225 | attach to it. So, this is how the fancy footwork goes: |
| 4226 | |
| 4227 | In a shell: |
| 4228 | |
| 4229 | |
| 4230 | host% kill -STOP pid |
| 4231 | |
| 4232 | |
| 4233 | |
| 4234 | |
| 4235 | Run gdb on the tracing thread as described in case 2 and do: |
| 4236 | |
| 4237 | |
| 4238 | (host gdb) call detach(pid) |
| 4239 | |
| 4240 | |
| 4241 | If you get a segfault, do it again. It always works the second time. |
| 4242 | |
| 4243 | Detach from the tracing thread and attach to that other thread: |
| 4244 | |
| 4245 | |
| 4246 | (host gdb) detach |
| 4247 | |
| 4248 | |
| 4249 | |
| 4250 | |
| 4251 | |
| 4252 | |
| 4253 | (host gdb) attach pid |
| 4254 | |
| 4255 | |
| 4256 | |
| 4257 | |
| 4258 | If gdb hangs when attaching to that process, go back to a shell and |
| 4259 | do: |
| 4260 | |
| 4261 | |
| 4262 | host% |
| 4263 | kill -CONT pid |
| 4264 | |
| 4265 | |
| 4266 | |
| 4267 | |
| 4268 | And then get the backtrace: |
| 4269 | |
| 4270 | |
| 4271 | (host gdb) backtrace |
| 4272 | |
| 4273 | |
| 4274 | |
| 4275 | |
| 4276 | |
| 4277 | 1144..44.. CCaassee 44 :: HHaannggss |
| 4278 | |
| 4279 | Hangs seem to be fairly rare, but they sometimes happen. When a hang |
| 4280 | happens, we need a backtrace from the offending process. Run the |
| 4281 | kernel debugger as described in case 1 and get a backtrace. If the |
| 4282 | current process is not the idle thread, then send in the backtrace. |
| 4283 | You can tell that it's the idle thread if the stack looks like this: |
| 4284 | |
| 4285 | |
| 4286 | #0 0x100b1401 in __libc_nanosleep () |
| 4287 | #1 0x100a2885 in idle_sleep (secs=10) at time.c:122 |
| 4288 | #2 0x100a546f in do_idle () at process_kern.c:445 |
| 4289 | #3 0x100a5508 in cpu_idle () at process_kern.c:471 |
| 4290 | #4 0x100ec18f in start_kernel () at init/main.c:592 |
| 4291 | #5 0x100a3e10 in start_kernel_proc (unused=0x0) at um_arch.c:71 |
| 4292 | #6 0x100a383f in signal_tramp (arg=0x100a3dd8) at trap_user.c:50 |
| 4293 | |
| 4294 | |
| 4295 | |
| 4296 | |
| 4297 | If this is the case, then some other process is at fault, and went to |
| 4298 | sleep when it shouldn't have. Run ps on the host and figure out which |
| 4299 | process should not have gone to sleep and stayed asleep. Then attach |
| 4300 | to it with gdb and get a backtrace as described in case 3. |
| 4301 | |
| 4302 | |
| 4303 | |
| 4304 | |
| 4305 | |
| 4306 | |
| 4307 | 1155.. TThhaannkkss |
| 4308 | |
| 4309 | |
| 4310 | A number of people have helped this project in various ways, and this |
| 4311 | page gives recognition where recognition is due. |
| 4312 | |
| 4313 | |
| 4314 | If you're listed here and you would prefer a real link on your name, |
| 4315 | or no link at all, instead of the despammed email address pseudo-link, |
| 4316 | let me know. |
| 4317 | |
| 4318 | |
| 4319 | If you're not listed here and you think maybe you should be, please |
| 4320 | let me know that as well. I try to get everyone, but sometimes my |
| 4321 | bookkeeping lapses and I forget about contributions. |
| 4322 | |
| 4323 | |
| 4324 | 1155..11.. CCooddee aanndd DDooccuummeennttaattiioonn |
| 4325 | |
| 4326 | Rusty Russell <rusty at linuxcare.com.au> - |
| 4327 | |
| 4328 | +o wrote the HOWTO <http://user-mode- |
| 4329 | linux.sourceforge.net/UserModeLinux-HOWTO.html> |
| 4330 | |
| 4331 | +o prodded me into making this project official and putting it on |
| 4332 | SourceForge |
| 4333 | |
| 4334 | +o came up with the way cool UML logo <http://user-mode- |
| 4335 | linux.sourceforge.net/uml-small.png> |
| 4336 | |
| 4337 | +o redid the config process |
| 4338 | |
| 4339 | |
| 4340 | Peter Moulder <reiter at netspace.net.au> - Fixed my config and build |
| 4341 | processes, and added some useful code to the block driver |
| 4342 | |
| 4343 | |
| 4344 | Bill Stearns <wstearns at pobox.com> - |
| 4345 | |
| 4346 | +o HOWTO updates |
| 4347 | |
| 4348 | +o lots of bug reports |
| 4349 | |
| 4350 | +o lots of testing |
| 4351 | |
| 4352 | +o dedicated a box (uml.ists.dartmouth.edu) to support UML development |
| 4353 | |
| 4354 | +o wrote the mkrootfs script, which allows bootable filesystems of |
| 4355 | RPM-based distributions to be cranked out |
| 4356 | |
| 4357 | +o cranked out a large number of filesystems with said script |
| 4358 | |
| 4359 | |
| 4360 | Jim Leu <jleu at mindspring.com> - Wrote the virtual ethernet driver |
| 4361 | and associated usermode tools |
| 4362 | |
| 4363 | Lars Brinkhoff <http://lars.nocrew.org/> - Contributed the ptrace |
| 4364 | proxy from his own project <http://a386.nocrew.org/> to allow easier |
| 4365 | kernel debugging |
| 4366 | |
| 4367 | |
| 4368 | Andrea Arcangeli <andrea at suse.de> - Redid some of the early boot |
| 4369 | code so that it would work on machines with Large File Support |
| 4370 | |
| 4371 | |
| 4372 | Chris Emerson <http://www.chiark.greenend.org.uk/~cemerson/> - Did |
| 4373 | the first UML port to Linux/ppc |
| 4374 | |
| 4375 | |
| 4376 | Harald Welte <laforge at gnumonks.org> - Wrote the multicast |
| 4377 | transport for the network driver |
| 4378 | |
| 4379 | |
| 4380 | Jorgen Cederlof - Added special file support to hostfs |
| 4381 | |
| 4382 | |
| 4383 | Greg Lonnon <glonnon at ridgerun dot com> - Changed the ubd driver |
| 4384 | to allow it to layer a COW file on a shared read-only filesystem and |
| 4385 | wrote the iomem emulation support |
| 4386 | |
| 4387 | |
| 4388 | Henrik Nordstrom <http://hem.passagen.se/hno/> - Provided a variety |
| 4389 | of patches, fixes, and clues |
| 4390 | |
| 4391 | |
| 4392 | Lennert Buytenhek - Contributed various patches, a rewrite of the |
| 4393 | network driver, the first implementation of the mconsole driver, and |
| 4394 | did the bulk of the work needed to get SMP working again. |
| 4395 | |
| 4396 | |
| 4397 | Yon Uriarte - Fixed the TUN/TAP network backend while I slept. |
| 4398 | |
| 4399 | |
| 4400 | Adam Heath - Made a bunch of nice cleanups to the initialization code, |
| 4401 | plus various other small patches. |
| 4402 | |
| 4403 | |
| 4404 | Matt Zimmerman - Matt volunteered to be the UML Debian maintainer and |
| 4405 | is doing a real nice job of it. He also noticed and fixed a number of |
| 4406 | actually and potentially exploitable security holes in uml_net. Plus |
| 4407 | the occasional patch. I like patches. |
| 4408 | |
| 4409 | |
| 4410 | James McMechan - James seems to have taken over maintenance of the ubd |
| 4411 | driver and is doing a nice job of it. |
| 4412 | |
| 4413 | |
| 4414 | Chandan Kudige - wrote the umlgdb script which automates the reloading |
| 4415 | of module symbols. |
| 4416 | |
| 4417 | |
| 4418 | Steve Schmidtke - wrote the UML slirp transport and hostaudio drivers, |
| 4419 | enabling UML processes to access audio devices on the host. He also |
| 4420 | submitted patches for the slip transport and lots of other things. |
| 4421 | |
| 4422 | |
| 4423 | David Coulson <http://davidcoulson.net> - |
| 4424 | |
| 4425 | +o Set up the usermodelinux.org <http://usermodelinux.org> site, |
| 4426 | which is a great way of keeping the UML user community on top of |
| 4427 | UML goings-on. |
| 4428 | |
| 4429 | +o Site documentation and updates |
| 4430 | |
| 4431 | +o Nifty little UML management daemon UMLd |
| 4432 | <http://uml.openconsultancy.com/umld/> |
| 4433 | |
| 4434 | +o Lots of testing and bug reports |
| 4435 | |
| 4436 | |
| 4437 | |
| 4438 | |
| 4439 | 1155..22.. FFlluusshhiinngg oouutt bbuuggss |
| 4440 | |
| 4441 | |
| 4442 | |
| 4443 | +o Yuri Pudgorodsky |
| 4444 | |
| 4445 | +o Gerald Britton |
| 4446 | |
| 4447 | +o Ian Wehrman |
| 4448 | |
| 4449 | +o Gord Lamb |
| 4450 | |
| 4451 | +o Eugene Koontz |
| 4452 | |
| 4453 | +o John H. Hartman |
| 4454 | |
| 4455 | +o Anders Karlsson |
| 4456 | |
| 4457 | +o Daniel Phillips |
| 4458 | |
| 4459 | +o John Fremlin |
| 4460 | |
| 4461 | +o Rainer Burgstaller |
| 4462 | |
| 4463 | +o James Stevenson |
| 4464 | |
| 4465 | +o Matt Clay |
| 4466 | |
| 4467 | +o Cliff Jefferies |
| 4468 | |
| 4469 | +o Geoff Hoff |
| 4470 | |
| 4471 | +o Lennert Buytenhek |
| 4472 | |
| 4473 | +o Al Viro |
| 4474 | |
| 4475 | +o Frank Klingenhoefer |
| 4476 | |
| 4477 | +o Livio Baldini Soares |
| 4478 | |
| 4479 | +o Jon Burgess |
| 4480 | |
| 4481 | +o Petru Paler |
| 4482 | |
| 4483 | +o Paul |
| 4484 | |
| 4485 | +o Chris Reahard |
| 4486 | |
| 4487 | +o Sverker Nilsson |
| 4488 | |
| 4489 | +o Gong Su |
| 4490 | |
| 4491 | +o johan verrept |
| 4492 | |
| 4493 | +o Bjorn Eriksson |
| 4494 | |
| 4495 | +o Lorenzo Allegrucci |
| 4496 | |
| 4497 | +o Muli Ben-Yehuda |
| 4498 | |
| 4499 | +o David Mansfield |
| 4500 | |
| 4501 | +o Howard Goff |
| 4502 | |
| 4503 | +o Mike Anderson |
| 4504 | |
| 4505 | +o John Byrne |
| 4506 | |
| 4507 | +o Sapan J. Batia |
| 4508 | |
| 4509 | +o Iris Huang |
| 4510 | |
| 4511 | +o Jan Hudec |
| 4512 | |
| 4513 | +o Voluspa |
| 4514 | |
| 4515 | |
| 4516 | |
| 4517 | |
| 4518 | 1155..33.. BBuugglleettss aanndd cclleeaann--uuppss |
| 4519 | |
| 4520 | |
| 4521 | |
| 4522 | +o Dave Zarzycki |
| 4523 | |
| 4524 | +o Adam Lazur |
| 4525 | |
| 4526 | +o Boria Feigin |
| 4527 | |
| 4528 | +o Brian J. Murrell |
| 4529 | |
| 4530 | +o JS |
| 4531 | |
| 4532 | +o Roman Zippel |
| 4533 | |
| 4534 | +o Wil Cooley |
| 4535 | |
| 4536 | +o Ayelet Shemesh |
| 4537 | |
| 4538 | +o Will Dyson |
| 4539 | |
| 4540 | +o Sverker Nilsson |
| 4541 | |
| 4542 | +o dvorak |
| 4543 | |
| 4544 | +o v.naga srinivas |
| 4545 | |
| 4546 | +o Shlomi Fish |
| 4547 | |
| 4548 | +o Roger Binns |
| 4549 | |
| 4550 | +o johan verrept |
| 4551 | |
| 4552 | +o MrChuoi |
| 4553 | |
| 4554 | +o Peter Cleve |
| 4555 | |
| 4556 | +o Vincent Guffens |
| 4557 | |
| 4558 | +o Nathan Scott |
| 4559 | |
| 4560 | +o Patrick Caulfield |
| 4561 | |
| 4562 | +o jbearce |
| 4563 | |
| 4564 | +o Catalin Marinas |
| 4565 | |
| 4566 | +o Shane Spencer |
| 4567 | |
| 4568 | +o Zou Min |
| 4569 | |
| 4570 | |
| 4571 | +o Ryan Boder |
| 4572 | |
| 4573 | +o Lorenzo Colitti |
| 4574 | |
| 4575 | +o Gwendal Grignou |
| 4576 | |
| 4577 | +o Andre' Breiler |
| 4578 | |
| 4579 | +o Tsutomu Yasuda |
| 4580 | |
| 4581 | |
| 4582 | |
| 4583 | 1155..44.. CCaassee SSttuuddiieess |
| 4584 | |
| 4585 | |
| 4586 | +o Jon Wright |
| 4587 | |
| 4588 | +o William McEwan |
| 4589 | |
| 4590 | +o Michael Richardson |
| 4591 | |
| 4592 | |
| 4593 | |
| 4594 | 1155..55.. OOtthheerr ccoonnttrriibbuuttiioonnss |
| 4595 | |
| 4596 | |
| 4597 | Bill Carr <Bill.Carr at compaq.com> made the Red Hat mkrootfs script |
| 4598 | work with RH 6.2. |
| 4599 | |
| 4600 | Michael Jennings <mikejen at hevanet.com> sent in some material which |
| 4601 | is now gracing the top of the index page <http://user-mode- |
| 4602 | linux.sourceforge.net/index.html> of this site. |
| 4603 | |
| 4604 | SGI <http://www.sgi.com> (and more specifically Ralf Baechle <ralf at |
| 4605 | uni-koblenz.de> ) gave me an account on oss.sgi.com |
| 4606 | <http://www.oss.sgi.com> . The bandwidth there made it possible to |
| 4607 | produce most of the filesystems available on the project download |
| 4608 | page. |
| 4609 | |
| 4610 | Laurent Bonnaud <Laurent.Bonnaud at inpg.fr> took the old grotty |
| 4611 | Debian filesystem that I've been distributing and updated it to 2.2. |
| 4612 | It is now available by itself here. |
| 4613 | |
| 4614 | Rik van Riel gave me some ftp space on ftp.nl.linux.org so I can make |
| 4615 | releases even when Sourceforge is broken. |
| 4616 | |
| 4617 | Rodrigo de Castro looked at my broken pte code and told me what was |
| 4618 | wrong with it, letting me fix a long-standing (several weeks) and |
| 4619 | serious set of bugs. |
| 4620 | |
| 4621 | Chris Reahard built a specialized root filesystem for running a DNS |
| 4622 | server jailed inside UML. It's available from the download |
| 4623 | <http://user-mode-linux.sourceforge.net/dl-sf.html> page in the Jail |
| 4624 | Filesysems section. |
| 4625 | |
| 4626 | |
| 4627 | |
| 4628 | |
| 4629 | |
| 4630 | |
| 4631 | |
| 4632 | |
| 4633 | |
| 4634 | |
| 4635 | |
| 4636 | |