root / qemu-doc.texi @ 6e59c1db
History | View | Annotate | Download (32.6 kB)
1 |
\input texinfo @c -*- texinfo -*- |
---|---|
2 |
|
3 |
@iftex |
4 |
@settitle QEMU CPU Emulator Reference Documentation |
5 |
@titlepage |
6 |
@sp 7 |
7 |
@center @titlefont{QEMU CPU Emulator Reference Documentation} |
8 |
@sp 3 |
9 |
@end titlepage |
10 |
@end iftex |
11 |
|
12 |
@chapter Introduction |
13 |
|
14 |
@section Features |
15 |
|
16 |
QEMU is a FAST! processor emulator. By using dynamic translation it |
17 |
achieves a reasonnable speed while being easy to port on new host |
18 |
CPUs. |
19 |
|
20 |
QEMU has two operating modes: |
21 |
|
22 |
@itemize @minus |
23 |
|
24 |
@item |
25 |
User mode emulation. In this mode, QEMU can launch Linux processes |
26 |
compiled for one CPU on another CPU. Linux system calls are converted |
27 |
because of endianness and 32/64 bit mismatches. The Wine Windows API |
28 |
emulator (@url{http://www.winehq.org}) and the DOSEMU DOS emulator |
29 |
(@url{www.dosemu.org}) are the main targets for QEMU. |
30 |
|
31 |
@item |
32 |
Full system emulation. In this mode, QEMU emulates a full |
33 |
system, including a processor and various peripherials. Currently, it |
34 |
is only used to launch an x86 Linux kernel on an x86 Linux system. It |
35 |
enables easier testing and debugging of system code. It can also be |
36 |
used to provide virtual hosting of several virtual PCs on a single |
37 |
server. |
38 |
|
39 |
@end itemize |
40 |
|
41 |
As QEMU requires no host kernel patches to run, it is very safe and |
42 |
easy to use. |
43 |
|
44 |
QEMU generic features: |
45 |
|
46 |
@itemize |
47 |
|
48 |
@item User space only or full system emulation. |
49 |
|
50 |
@item Using dynamic translation to native code for reasonnable speed. |
51 |
|
52 |
@item Working on x86 and PowerPC hosts. Being tested on ARM, Sparc32, Alpha and S390. |
53 |
|
54 |
@item Self-modifying code support. |
55 |
|
56 |
@item Precise exceptions support. |
57 |
|
58 |
@item The virtual CPU is a library (@code{libqemu}) which can be used |
59 |
in other projects. |
60 |
|
61 |
@end itemize |
62 |
|
63 |
QEMU user mode emulation features: |
64 |
@itemize |
65 |
@item Generic Linux system call converter, including most ioctls. |
66 |
|
67 |
@item clone() emulation using native CPU clone() to use Linux scheduler for threads. |
68 |
|
69 |
@item Accurate signal handling by remapping host signals to target signals. |
70 |
@end itemize |
71 |
@end itemize |
72 |
|
73 |
QEMU full system emulation features: |
74 |
@itemize |
75 |
@item Using mmap() system calls to simulate the MMU |
76 |
@end itemize |
77 |
|
78 |
@section x86 emulation |
79 |
|
80 |
QEMU x86 target features: |
81 |
|
82 |
@itemize |
83 |
|
84 |
@item The virtual x86 CPU supports 16 bit and 32 bit addressing with segmentation. |
85 |
LDT/GDT and IDT are emulated. VM86 mode is also supported to run DOSEMU. |
86 |
|
87 |
@item Support of host page sizes bigger than 4KB in user mode emulation. |
88 |
|
89 |
@item QEMU can emulate itself on x86. |
90 |
|
91 |
@item An extensive Linux x86 CPU test program is included @file{tests/test-i386}. |
92 |
It can be used to test other x86 virtual CPUs. |
93 |
|
94 |
@end itemize |
95 |
|
96 |
Current QEMU limitations: |
97 |
|
98 |
@itemize |
99 |
|
100 |
@item No SSE/MMX support (yet). |
101 |
|
102 |
@item No x86-64 support. |
103 |
|
104 |
@item IPC syscalls are missing. |
105 |
|
106 |
@item The x86 segment limits and access rights are not tested at every |
107 |
memory access. |
108 |
|
109 |
@item On non x86 host CPUs, @code{double}s are used instead of the non standard |
110 |
10 byte @code{long double}s of x86 for floating point emulation to get |
111 |
maximum performances. |
112 |
|
113 |
@item Full system emulation only works if no data are mapped above the virtual address |
114 |
0xc0000000 (yet). |
115 |
|
116 |
@item Some priviledged instructions or behaviors are missing. Only the ones |
117 |
needed for proper Linux kernel operation are emulated. |
118 |
|
119 |
@item No memory separation between the kernel and the user processes is done. |
120 |
It will be implemented very soon. |
121 |
|
122 |
@end itemize |
123 |
|
124 |
@section ARM emulation |
125 |
|
126 |
@itemize |
127 |
|
128 |
@item ARM emulation can currently launch small programs while using the |
129 |
generic dynamic code generation architecture of QEMU. |
130 |
|
131 |
@item No FPU support (yet). |
132 |
|
133 |
@item No automatic regression testing (yet). |
134 |
|
135 |
@end itemize |
136 |
|
137 |
@section SPARC emulation |
138 |
|
139 |
The SPARC emulation is currently in development. |
140 |
|
141 |
@chapter QEMU User space emulator invocation |
142 |
|
143 |
@section Quick Start |
144 |
|
145 |
If you need to compile QEMU, please read the @file{README} which gives |
146 |
the related information. |
147 |
|
148 |
In order to launch a Linux process, QEMU needs the process executable |
149 |
itself and all the target (x86) dynamic libraries used by it. |
150 |
|
151 |
@itemize |
152 |
|
153 |
@item On x86, you can just try to launch any process by using the native |
154 |
libraries: |
155 |
|
156 |
@example |
157 |
qemu-i386 -L / /bin/ls |
158 |
@end example |
159 |
|
160 |
@code{-L /} tells that the x86 dynamic linker must be searched with a |
161 |
@file{/} prefix. |
162 |
|
163 |
@item Since QEMU is also a linux process, you can launch qemu with qemu: |
164 |
|
165 |
@example |
166 |
qemu-i386 -L / qemu-i386 -L / /bin/ls |
167 |
@end example |
168 |
|
169 |
@item On non x86 CPUs, you need first to download at least an x86 glibc |
170 |
(@file{qemu-XXX-i386-glibc21.tar.gz} on the QEMU web page). Ensure that |
171 |
@code{LD_LIBRARY_PATH} is not set: |
172 |
|
173 |
@example |
174 |
unset LD_LIBRARY_PATH |
175 |
@end example |
176 |
|
177 |
Then you can launch the precompiled @file{ls} x86 executable: |
178 |
|
179 |
@example |
180 |
qemu-i386 /usr/local/qemu-i386/bin/ls-i386 |
181 |
@end example |
182 |
You can look at @file{/usr/local/qemu-i386/bin/qemu-conf.sh} so that |
183 |
QEMU is automatically launched by the Linux kernel when you try to |
184 |
launch x86 executables. It requires the @code{binfmt_misc} module in the |
185 |
Linux kernel. |
186 |
|
187 |
@item The x86 version of QEMU is also included. You can try weird things such as: |
188 |
@example |
189 |
qemu-i386 /usr/local/qemu-i386/bin/qemu-i386 /usr/local/qemu-i386/bin/ls-i386 |
190 |
@end example |
191 |
|
192 |
@end itemize |
193 |
|
194 |
@section Wine launch |
195 |
|
196 |
@itemize |
197 |
|
198 |
@item Ensure that you have a working QEMU with the x86 glibc |
199 |
distribution (see previous section). In order to verify it, you must be |
200 |
able to do: |
201 |
|
202 |
@example |
203 |
qemu-i386 /usr/local/qemu-i386/bin/ls-i386 |
204 |
@end example |
205 |
|
206 |
@item Download the binary x86 Wine install |
207 |
(@file{qemu-XXX-i386-wine.tar.gz} on the QEMU web page). |
208 |
|
209 |
@item Configure Wine on your account. Look at the provided script |
210 |
@file{/usr/local/qemu-i386/bin/wine-conf.sh}. Your previous |
211 |
@code{$@{HOME@}/.wine} directory is saved to @code{$@{HOME@}/.wine.org}. |
212 |
|
213 |
@item Then you can try the example @file{putty.exe}: |
214 |
|
215 |
@example |
216 |
qemu-i386 /usr/local/qemu-i386/wine/bin/wine /usr/local/qemu-i386/wine/c/Program\ Files/putty.exe |
217 |
@end example |
218 |
|
219 |
@end itemize |
220 |
|
221 |
@section Command line options |
222 |
|
223 |
@example |
224 |
usage: qemu-i386 [-h] [-d] [-L path] [-s size] program [arguments...] |
225 |
@end example |
226 |
|
227 |
@table @option |
228 |
@item -h |
229 |
Print the help |
230 |
@item -L path |
231 |
Set the x86 elf interpreter prefix (default=/usr/local/qemu-i386) |
232 |
@item -s size |
233 |
Set the x86 stack size in bytes (default=524288) |
234 |
@end table |
235 |
|
236 |
Debug options: |
237 |
|
238 |
@table @option |
239 |
@item -d |
240 |
Activate log (logfile=/tmp/qemu.log) |
241 |
@item -p pagesize |
242 |
Act as if the host page size was 'pagesize' bytes |
243 |
@end table |
244 |
|
245 |
@chapter QEMU System emulator invocation |
246 |
|
247 |
@section Introduction |
248 |
|
249 |
@c man begin DESCRIPTION |
250 |
|
251 |
The QEMU System emulator simulates a complete PC. It can either boot |
252 |
directly a Linux kernel (without any BIOS or boot loader) or boot like a |
253 |
real PC with the included BIOS. |
254 |
|
255 |
In order to meet specific user needs, two versions of QEMU are |
256 |
available: |
257 |
|
258 |
@enumerate |
259 |
|
260 |
@item |
261 |
@code{qemu} uses the host Memory Management Unit (MMU) to simulate |
262 |
the x86 MMU. It is @emph{fast} but has limitations because the whole 4 GB |
263 |
address space cannot be used and some memory mapped peripherials |
264 |
cannot be emulated accurately yet. Therefore, a specific Linux kernel |
265 |
must be used (@xref{linux_compile}). |
266 |
|
267 |
@item |
268 |
@code{qemu-softmmu} uses a software MMU. It is about @emph{two times |
269 |
slower} but gives a more accurate emulation. (XXX: Linux cannot be ran |
270 |
unpatched yet). |
271 |
|
272 |
@end enumerate |
273 |
|
274 |
QEMU emulates the following PC peripherials: |
275 |
|
276 |
@itemize @minus |
277 |
@item |
278 |
VGA (hardware level, including all non standard modes) |
279 |
@item |
280 |
PS/2 mouse and keyboard |
281 |
@item |
282 |
IDE disk interface (port=0x1f0, irq=14) |
283 |
@item |
284 |
NE2000 network adapter (port=0x300, irq=9) |
285 |
@item |
286 |
Serial port (port=0x3f8, irq=4) |
287 |
@item |
288 |
PIC (interrupt controler) |
289 |
@item |
290 |
PIT (timers) |
291 |
@item |
292 |
CMOS memory |
293 |
@end itemize |
294 |
|
295 |
@c man end |
296 |
|
297 |
@section Quick Start |
298 |
|
299 |
Download the linux image (@file{linux.img}) and type: |
300 |
|
301 |
@example |
302 |
qemu-softmmu linux.img |
303 |
@end example |
304 |
|
305 |
Linux should boot and give you a prompt. |
306 |
|
307 |
@section Direct Linux Boot and Network emulation |
308 |
|
309 |
This section explains how to launch a Linux kernel inside QEMU without |
310 |
having to make a full bootable image. It is very useful for fast Linux |
311 |
kernel testing. The QEMU network configuration is also explained. |
312 |
|
313 |
@enumerate |
314 |
@item |
315 |
Download the archive @file{linux-test-xxx.tar.gz} containing a Linux |
316 |
kernel and a disk image. |
317 |
|
318 |
@item Optional: If you want network support (for example to launch X11 examples), you |
319 |
must copy the script @file{qemu-ifup} in @file{/etc} and configure |
320 |
properly @code{sudo} so that the command @code{ifconfig} contained in |
321 |
@file{qemu-ifup} can be executed as root. You must verify that your host |
322 |
kernel supports the TUN/TAP network interfaces: the device |
323 |
@file{/dev/net/tun} must be present. |
324 |
|
325 |
When network is enabled, there is a virtual network connection between |
326 |
the host kernel and the emulated kernel. The emulated kernel is seen |
327 |
from the host kernel at IP address 172.20.0.2 and the host kernel is |
328 |
seen from the emulated kernel at IP address 172.20.0.1. |
329 |
|
330 |
@item Launch @code{qemu.sh}. You should have the following output: |
331 |
|
332 |
@example |
333 |
> ./qemu.sh |
334 |
connected to host network interface: tun0 |
335 |
Uncompressing Linux... Ok, booting the kernel. |
336 |
Linux version 2.4.20 (fabrice@localhost.localdomain) (gcc version 2.96 20000731 (Red Hat Linux 7.3 2.96-110)) #22 lun jui 7 13:37:41 CEST 2003 |
337 |
BIOS-provided physical RAM map: |
338 |
BIOS-e801: 0000000000000000 - 000000000009f000 (usable) |
339 |
BIOS-e801: 0000000000100000 - 0000000002000000 (usable) |
340 |
32MB LOWMEM available. |
341 |
On node 0 totalpages: 8192 |
342 |
zone(0): 4096 pages. |
343 |
zone(1): 4096 pages. |
344 |
zone(2): 0 pages. |
345 |
Kernel command line: root=/dev/hda ide1=noprobe ide2=noprobe ide3=noprobe ide4=noprobe ide5=noprobe |
346 |
ide_setup: ide1=noprobe |
347 |
ide_setup: ide2=noprobe |
348 |
ide_setup: ide3=noprobe |
349 |
ide_setup: ide4=noprobe |
350 |
ide_setup: ide5=noprobe |
351 |
Initializing CPU#0 |
352 |
Detected 501.285 MHz processor. |
353 |
Calibrating delay loop... 989.59 BogoMIPS |
354 |
Memory: 29268k/32768k available (907k kernel code, 3112k reserved, 212k data, 52k init, 0k highmem) |
355 |
Dentry cache hash table entries: 4096 (order: 3, 32768 bytes) |
356 |
Inode cache hash table entries: 2048 (order: 2, 16384 bytes) |
357 |
Mount-cache hash table entries: 512 (order: 0, 4096 bytes) |
358 |
Buffer-cache hash table entries: 1024 (order: 0, 4096 bytes) |
359 |
Page-cache hash table entries: 8192 (order: 3, 32768 bytes) |
360 |
CPU: Intel Pentium Pro stepping 03 |
361 |
Checking 'hlt' instruction... OK. |
362 |
POSIX conformance testing by UNIFIX |
363 |
Linux NET4.0 for Linux 2.4 |
364 |
Based upon Swansea University Computer Society NET3.039 |
365 |
Initializing RT netlink socket |
366 |
apm: BIOS not found. |
367 |
Starting kswapd |
368 |
Journalled Block Device driver loaded |
369 |
pty: 256 Unix98 ptys configured |
370 |
Serial driver version 5.05c (2001-07-08) with no serial options enabled |
371 |
ttyS00 at 0x03f8 (irq = 4) is a 16450 |
372 |
Uniform Multi-Platform E-IDE driver Revision: 6.31 |
373 |
ide: Assuming 50MHz system bus speed for PIO modes; override with idebus=xx |
374 |
hda: QEMU HARDDISK, ATA DISK drive |
375 |
ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 |
376 |
hda: 12288 sectors (6 MB) w/256KiB Cache, CHS=12/16/63 |
377 |
Partition check: |
378 |
hda: unknown partition table |
379 |
ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com) |
380 |
Last modified Nov 1, 2000 by Paul Gortmaker |
381 |
NE*000 ethercard probe at 0x300: 52 54 00 12 34 56 |
382 |
eth0: NE2000 found at 0x300, using IRQ 9. |
383 |
RAMDISK driver initialized: 16 RAM disks of 4096K size 1024 blocksize |
384 |
NET4: Linux TCP/IP 1.0 for NET4.0 |
385 |
IP Protocols: ICMP, UDP, TCP, IGMP |
386 |
IP: routing cache hash table of 512 buckets, 4Kbytes |
387 |
TCP: Hash tables configured (established 2048 bind 4096) |
388 |
NET4: Unix domain sockets 1.0/SMP for Linux NET4.0. |
389 |
EXT2-fs warning: mounting unchecked fs, running e2fsck is recommended |
390 |
VFS: Mounted root (ext2 filesystem). |
391 |
Freeing unused kernel memory: 52k freed |
392 |
sh: can't access tty; job control turned off |
393 |
# |
394 |
@end example |
395 |
|
396 |
@item |
397 |
Then you can play with the kernel inside the virtual serial console. You |
398 |
can launch @code{ls} for example. Type @key{Ctrl-a h} to have an help |
399 |
about the keys you can type inside the virtual serial console. In |
400 |
particular, use @key{Ctrl-a x} to exit QEMU and use @key{Ctrl-a b} as |
401 |
the Magic SysRq key. |
402 |
|
403 |
@item |
404 |
If the network is enabled, launch the script @file{/etc/linuxrc} in the |
405 |
emulator (don't forget the leading dot): |
406 |
@example |
407 |
. /etc/linuxrc |
408 |
@end example |
409 |
|
410 |
Then enable X11 connections on your PC from the emulated Linux: |
411 |
@example |
412 |
xhost +172.20.0.2 |
413 |
@end example |
414 |
|
415 |
You can now launch @file{xterm} or @file{xlogo} and verify that you have |
416 |
a real Virtual Linux system ! |
417 |
|
418 |
@end enumerate |
419 |
|
420 |
NOTES: |
421 |
@enumerate |
422 |
@item |
423 |
A 2.5.74 kernel is also included in the archive. Just |
424 |
replace the bzImage in qemu.sh to try it. |
425 |
|
426 |
@item |
427 |
vl creates a temporary file in @var{$QEMU_TMPDIR} (@file{/tmp} is the |
428 |
default) containing all the simulated PC memory. If possible, try to use |
429 |
a temporary directory using the tmpfs filesystem to avoid too many |
430 |
unnecessary disk accesses. |
431 |
|
432 |
@item |
433 |
In order to exit cleanly for vl, you can do a @emph{shutdown} inside |
434 |
vl. vl will automatically exit when the Linux shutdown is done. |
435 |
|
436 |
@item |
437 |
You can boot slightly faster by disabling the probe of non present IDE |
438 |
interfaces. To do so, add the following options on the kernel command |
439 |
line: |
440 |
@example |
441 |
ide1=noprobe ide2=noprobe ide3=noprobe ide4=noprobe ide5=noprobe |
442 |
@end example |
443 |
|
444 |
@item |
445 |
The example disk image is a modified version of the one made by Kevin |
446 |
Lawton for the plex86 Project (@url{www.plex86.org}). |
447 |
|
448 |
@end enumerate |
449 |
|
450 |
@section Invocation |
451 |
|
452 |
@example |
453 |
@c man begin SYNOPSIS |
454 |
usage: qemu [options] [disk_image] |
455 |
@c man end |
456 |
@end example |
457 |
|
458 |
@c man begin OPTIONS |
459 |
@var{disk_image} is a raw hard image image for IDE hard disk 0. |
460 |
|
461 |
General options: |
462 |
@table @option |
463 |
@item -hda file |
464 |
@item -hdb file |
465 |
Use @var{file} as hard disk 0 or 1 image (@xref{disk_images}). |
466 |
|
467 |
@item -snapshot |
468 |
|
469 |
Write to temporary files instead of disk image files. In this case, |
470 |
the raw disk image you use is not written back. You can however force |
471 |
the write back by pressing @key{C-a s} (@xref{disk_images}). |
472 |
|
473 |
@item -m megs |
474 |
Set virtual RAM size to @var{megs} megabytes. |
475 |
|
476 |
@item -n script |
477 |
Set network init script [default=/etc/vl-ifup]. This script is |
478 |
launched to configure the host network interface (usually tun0) |
479 |
corresponding to the virtual NE2000 card. |
480 |
|
481 |
@item -initrd file |
482 |
Use @var{file} as initial ram disk. |
483 |
|
484 |
@item -tun-fd fd |
485 |
Assumes @var{fd} talks to tap/tun and use it. Read |
486 |
@url{http://bellard.org/qemu/tetrinet.html} to have an example of its |
487 |
use. |
488 |
|
489 |
@item -nographic |
490 |
|
491 |
Normally, QEMU uses SDL to display the VGA output. With this option, |
492 |
you can totally disable graphical output so that QEMU is a simple |
493 |
command line application. The emulated serial port is redirected on |
494 |
the console. Therefore, you can still use QEMU to debug a Linux kernel |
495 |
with a serial console. |
496 |
|
497 |
@end table |
498 |
|
499 |
Linux boot specific (does not require a full PC boot with a BIOS): |
500 |
@table @option |
501 |
|
502 |
@item -kernel bzImage |
503 |
Use @var{bzImage} as kernel image. |
504 |
|
505 |
@item -append cmdline |
506 |
Use @var{cmdline} as kernel command line |
507 |
|
508 |
@item -initrd file |
509 |
Use @var{file} as initial ram disk. |
510 |
|
511 |
@end table |
512 |
|
513 |
Debug options: |
514 |
@table @option |
515 |
@item -s |
516 |
Wait gdb connection to port 1234 (@xref{gdb_usage}). |
517 |
@item -p port |
518 |
Change gdb connection port. |
519 |
@item -d |
520 |
Output log in /tmp/vl.log |
521 |
@end table |
522 |
|
523 |
During emulation, use @key{C-a h} to get terminal commands: |
524 |
|
525 |
@table @key |
526 |
@item C-a h |
527 |
Print this help |
528 |
@item C-a x |
529 |
Exit emulatior |
530 |
@item C-a s |
531 |
Save disk data back to file (if -snapshot) |
532 |
@item C-a b |
533 |
Send break (magic sysrq) |
534 |
@item C-a C-a |
535 |
Send C-a |
536 |
@end table |
537 |
@c man end |
538 |
|
539 |
@ignore |
540 |
|
541 |
@setfilename qemu |
542 |
@settitle QEMU System Emulator |
543 |
|
544 |
@c man begin SEEALSO |
545 |
The HTML documentation of QEMU for more precise information and Linux |
546 |
user mode emulator invocation. |
547 |
@c man end |
548 |
|
549 |
@c man begin AUTHOR |
550 |
Fabrice Bellard |
551 |
@c man end |
552 |
|
553 |
@end ignore |
554 |
|
555 |
@end ignore |
556 |
@node disk_images |
557 |
@section Disk Images |
558 |
|
559 |
@subsection Raw disk images |
560 |
|
561 |
The disk images can simply be raw images of the hard disk. You can |
562 |
create them with the command: |
563 |
@example |
564 |
dd if=/dev/zero of=myimage bs=1024 count=mysize |
565 |
@end example |
566 |
where @var{myimage} is the image filename and @var{mysize} is its size |
567 |
in kilobytes. |
568 |
|
569 |
@subsection Snapshot mode |
570 |
|
571 |
If you use the option @option{-snapshot}, all disk images are |
572 |
considered as read only. When sectors in written, they are written in |
573 |
a temporary file created in @file{/tmp}. You can however force the |
574 |
write back to the raw disk images by pressing @key{C-a s}. |
575 |
|
576 |
NOTE: The snapshot mode only works with raw disk images. |
577 |
|
578 |
@subsection Copy On Write disk images |
579 |
|
580 |
QEMU also supports user mode Linux |
581 |
(@url{http://user-mode-linux.sourceforge.net/}) Copy On Write (COW) |
582 |
disk images. The COW disk images are much smaller than normal images |
583 |
as they store only modified sectors. They also permit the use of the |
584 |
same disk image template for many users. |
585 |
|
586 |
To create a COW disk images, use the command: |
587 |
|
588 |
@example |
589 |
qemu-mkcow -f myrawimage.bin mycowimage.cow |
590 |
@end example |
591 |
|
592 |
@file{myrawimage.bin} is a raw image you want to use as original disk |
593 |
image. It will never be written to. |
594 |
|
595 |
@file{mycowimage.cow} is the COW disk image which is created by |
596 |
@code{qemu-mkcow}. You can use it directly with the @option{-hdx} |
597 |
options. You must not modify the original raw disk image if you use |
598 |
COW images, as COW images only store the modified sectors from the raw |
599 |
disk image. QEMU stores the original raw disk image name and its |
600 |
modified time in the COW disk image so that chances of mistakes are |
601 |
reduced. |
602 |
|
603 |
If the raw disk image is not read-only, by pressing @key{C-a s} you |
604 |
can flush the COW disk image back into the raw disk image, as in |
605 |
snapshot mode. |
606 |
|
607 |
COW disk images can also be created without a corresponding raw disk |
608 |
image. It is useful to have a big initial virtual disk image without |
609 |
using much disk space. Use: |
610 |
|
611 |
@example |
612 |
qemu-mkcow mycowimage.cow 1024 |
613 |
@end example |
614 |
|
615 |
to create a 1 gigabyte empty COW disk image. |
616 |
|
617 |
NOTES: |
618 |
@enumerate |
619 |
@item |
620 |
COW disk images must be created on file systems supporting |
621 |
@emph{holes} such as ext2 or ext3. |
622 |
@item |
623 |
Since holes are used, the displayed size of the COW disk image is not |
624 |
the real one. To know it, use the @code{ls -ls} command. |
625 |
@end enumerate |
626 |
|
627 |
@node linux_compile |
628 |
@section Linux Kernel Compilation |
629 |
|
630 |
You should be able to use any kernel with QEMU provided you make the |
631 |
following changes (only 2.4.x and 2.5.x were tested): |
632 |
|
633 |
@enumerate |
634 |
@item |
635 |
The kernel must be mapped at 0x90000000 (the default is |
636 |
0xc0000000). You must modify only two lines in the kernel source: |
637 |
|
638 |
In @file{include/asm/page.h}, replace |
639 |
@example |
640 |
#define __PAGE_OFFSET (0xc0000000) |
641 |
@end example |
642 |
by |
643 |
@example |
644 |
#define __PAGE_OFFSET (0x90000000) |
645 |
@end example |
646 |
|
647 |
And in @file{arch/i386/vmlinux.lds}, replace |
648 |
@example |
649 |
. = 0xc0000000 + 0x100000; |
650 |
@end example |
651 |
by |
652 |
@example |
653 |
. = 0x90000000 + 0x100000; |
654 |
@end example |
655 |
|
656 |
@item |
657 |
If you want to enable SMP (Symmetric Multi-Processing) support, you |
658 |
must make the following change in @file{include/asm/fixmap.h}. Replace |
659 |
@example |
660 |
#define FIXADDR_TOP (0xffffX000UL) |
661 |
@end example |
662 |
by |
663 |
@example |
664 |
#define FIXADDR_TOP (0xa7ffX000UL) |
665 |
@end example |
666 |
(X is 'e' or 'f' depending on the kernel version). Although you can |
667 |
use an SMP kernel with QEMU, it only supports one CPU. |
668 |
|
669 |
@item |
670 |
If you are not using a 2.5 kernel as host kernel but if you use a target |
671 |
2.5 kernel, you must also ensure that the 'HZ' define is set to 100 |
672 |
(1000 is the default) as QEMU cannot currently emulate timers at |
673 |
frequencies greater than 100 Hz on host Linux systems < 2.5. In |
674 |
@file{include/asm/param.h}, replace: |
675 |
|
676 |
@example |
677 |
# define HZ 1000 /* Internal kernel timer frequency */ |
678 |
@end example |
679 |
by |
680 |
@example |
681 |
# define HZ 100 /* Internal kernel timer frequency */ |
682 |
@end example |
683 |
|
684 |
@end enumerate |
685 |
|
686 |
The file config-2.x.x gives the configuration of the example kernels. |
687 |
|
688 |
Just type |
689 |
@example |
690 |
make bzImage |
691 |
@end example |
692 |
|
693 |
As you would do to make a real kernel. Then you can use with QEMU |
694 |
exactly the same kernel as you would boot on your PC (in |
695 |
@file{arch/i386/boot/bzImage}). |
696 |
|
697 |
@node gdb_usage |
698 |
@section GDB usage |
699 |
|
700 |
QEMU has a primitive support to work with gdb, so that you can do |
701 |
'Ctrl-C' while the virtual machine is running and inspect its state. |
702 |
|
703 |
In order to use gdb, launch vl with the '-s' option. It will wait for a |
704 |
gdb connection: |
705 |
@example |
706 |
> vl -s arch/i386/boot/bzImage -hda root-2.4.20.img root=/dev/hda |
707 |
Connected to host network interface: tun0 |
708 |
Waiting gdb connection on port 1234 |
709 |
@end example |
710 |
|
711 |
Then launch gdb on the 'vmlinux' executable: |
712 |
@example |
713 |
> gdb vmlinux |
714 |
@end example |
715 |
|
716 |
In gdb, connect to QEMU: |
717 |
@example |
718 |
(gdb) target remote locahost:1234 |
719 |
@end example |
720 |
|
721 |
Then you can use gdb normally. For example, type 'c' to launch the kernel: |
722 |
@example |
723 |
(gdb) c |
724 |
@end example |
725 |
|
726 |
WARNING: breakpoints and single stepping are not yet supported. |
727 |
|
728 |
Here are some useful tips in order to use gdb on system code: |
729 |
|
730 |
@enumerate |
731 |
@item |
732 |
Use @code{info reg} to display all the CPU registers. |
733 |
@item |
734 |
Use @code{x/10i $eip} to display the code at the PC position. |
735 |
@item |
736 |
Use @code{set architecture i8086} to dump 16 bit code. Then use |
737 |
@code{x/10i $cs*16+*eip} to dump the code at the PC position. |
738 |
@end enumerate |
739 |
|
740 |
@chapter QEMU Internals |
741 |
|
742 |
@section QEMU compared to other emulators |
743 |
|
744 |
Like bochs [3], QEMU emulates an x86 CPU. But QEMU is much faster than |
745 |
bochs as it uses dynamic compilation and because it uses the host MMU to |
746 |
simulate the x86 MMU. The downside is that currently the emulation is |
747 |
not as accurate as bochs (for example, you cannot currently run Windows |
748 |
inside QEMU). |
749 |
|
750 |
Like Valgrind [2], QEMU does user space emulation and dynamic |
751 |
translation. Valgrind is mainly a memory debugger while QEMU has no |
752 |
support for it (QEMU could be used to detect out of bound memory |
753 |
accesses as Valgrind, but it has no support to track uninitialised data |
754 |
as Valgrind does). The Valgrind dynamic translator generates better code |
755 |
than QEMU (in particular it does register allocation) but it is closely |
756 |
tied to an x86 host and target and has no support for precise exceptions |
757 |
and system emulation. |
758 |
|
759 |
EM86 [4] is the closest project to user space QEMU (and QEMU still uses |
760 |
some of its code, in particular the ELF file loader). EM86 was limited |
761 |
to an alpha host and used a proprietary and slow interpreter (the |
762 |
interpreter part of the FX!32 Digital Win32 code translator [5]). |
763 |
|
764 |
TWIN [6] is a Windows API emulator like Wine. It is less accurate than |
765 |
Wine but includes a protected mode x86 interpreter to launch x86 Windows |
766 |
executables. Such an approach as greater potential because most of the |
767 |
Windows API is executed natively but it is far more difficult to develop |
768 |
because all the data structures and function parameters exchanged |
769 |
between the API and the x86 code must be converted. |
770 |
|
771 |
User mode Linux [7] was the only solution before QEMU to launch a Linux |
772 |
kernel as a process while not needing any host kernel patches. However, |
773 |
user mode Linux requires heavy kernel patches while QEMU accepts |
774 |
unpatched Linux kernels. It would be interesting to compare the |
775 |
performance of the two approaches. |
776 |
|
777 |
The new Plex86 [8] PC virtualizer is done in the same spirit as the QEMU |
778 |
system emulator. It requires a patched Linux kernel to work (you cannot |
779 |
launch the same kernel on your PC), but the patches are really small. As |
780 |
it is a PC virtualizer (no emulation is done except for some priveledged |
781 |
instructions), it has the potential of being faster than QEMU. The |
782 |
downside is that a complicated (and potentially unsafe) host kernel |
783 |
patch is needed. |
784 |
|
785 |
@section Portable dynamic translation |
786 |
|
787 |
QEMU is a dynamic translator. When it first encounters a piece of code, |
788 |
it converts it to the host instruction set. Usually dynamic translators |
789 |
are very complicated and highly CPU dependent. QEMU uses some tricks |
790 |
which make it relatively easily portable and simple while achieving good |
791 |
performances. |
792 |
|
793 |
The basic idea is to split every x86 instruction into fewer simpler |
794 |
instructions. Each simple instruction is implemented by a piece of C |
795 |
code (see @file{op-i386.c}). Then a compile time tool (@file{dyngen}) |
796 |
takes the corresponding object file (@file{op-i386.o}) to generate a |
797 |
dynamic code generator which concatenates the simple instructions to |
798 |
build a function (see @file{op-i386.h:dyngen_code()}). |
799 |
|
800 |
In essence, the process is similar to [1], but more work is done at |
801 |
compile time. |
802 |
|
803 |
A key idea to get optimal performances is that constant parameters can |
804 |
be passed to the simple operations. For that purpose, dummy ELF |
805 |
relocations are generated with gcc for each constant parameter. Then, |
806 |
the tool (@file{dyngen}) can locate the relocations and generate the |
807 |
appriopriate C code to resolve them when building the dynamic code. |
808 |
|
809 |
That way, QEMU is no more difficult to port than a dynamic linker. |
810 |
|
811 |
To go even faster, GCC static register variables are used to keep the |
812 |
state of the virtual CPU. |
813 |
|
814 |
@section Register allocation |
815 |
|
816 |
Since QEMU uses fixed simple instructions, no efficient register |
817 |
allocation can be done. However, because RISC CPUs have a lot of |
818 |
register, most of the virtual CPU state can be put in registers without |
819 |
doing complicated register allocation. |
820 |
|
821 |
@section Condition code optimisations |
822 |
|
823 |
Good CPU condition codes emulation (@code{EFLAGS} register on x86) is a |
824 |
critical point to get good performances. QEMU uses lazy condition code |
825 |
evaluation: instead of computing the condition codes after each x86 |
826 |
instruction, it just stores one operand (called @code{CC_SRC}), the |
827 |
result (called @code{CC_DST}) and the type of operation (called |
828 |
@code{CC_OP}). |
829 |
|
830 |
@code{CC_OP} is almost never explicitely set in the generated code |
831 |
because it is known at translation time. |
832 |
|
833 |
In order to increase performances, a backward pass is performed on the |
834 |
generated simple instructions (see |
835 |
@code{translate-i386.c:optimize_flags()}). When it can be proved that |
836 |
the condition codes are not needed by the next instructions, no |
837 |
condition codes are computed at all. |
838 |
|
839 |
@section CPU state optimisations |
840 |
|
841 |
The x86 CPU has many internal states which change the way it evaluates |
842 |
instructions. In order to achieve a good speed, the translation phase |
843 |
considers that some state information of the virtual x86 CPU cannot |
844 |
change in it. For example, if the SS, DS and ES segments have a zero |
845 |
base, then the translator does not even generate an addition for the |
846 |
segment base. |
847 |
|
848 |
[The FPU stack pointer register is not handled that way yet]. |
849 |
|
850 |
@section Translation cache |
851 |
|
852 |
A 2MByte cache holds the most recently used translations. For |
853 |
simplicity, it is completely flushed when it is full. A translation unit |
854 |
contains just a single basic block (a block of x86 instructions |
855 |
terminated by a jump or by a virtual CPU state change which the |
856 |
translator cannot deduce statically). |
857 |
|
858 |
@section Direct block chaining |
859 |
|
860 |
After each translated basic block is executed, QEMU uses the simulated |
861 |
Program Counter (PC) and other cpu state informations (such as the CS |
862 |
segment base value) to find the next basic block. |
863 |
|
864 |
In order to accelerate the most common cases where the new simulated PC |
865 |
is known, QEMU can patch a basic block so that it jumps directly to the |
866 |
next one. |
867 |
|
868 |
The most portable code uses an indirect jump. An indirect jump makes it |
869 |
easier to make the jump target modification atomic. On some |
870 |
architectures (such as PowerPC), the @code{JUMP} opcode is directly |
871 |
patched so that the block chaining has no overhead. |
872 |
|
873 |
@section Self-modifying code and translated code invalidation |
874 |
|
875 |
Self-modifying code is a special challenge in x86 emulation because no |
876 |
instruction cache invalidation is signaled by the application when code |
877 |
is modified. |
878 |
|
879 |
When translated code is generated for a basic block, the corresponding |
880 |
host page is write protected if it is not already read-only (with the |
881 |
system call @code{mprotect()}). Then, if a write access is done to the |
882 |
page, Linux raises a SEGV signal. QEMU then invalidates all the |
883 |
translated code in the page and enables write accesses to the page. |
884 |
|
885 |
Correct translated code invalidation is done efficiently by maintaining |
886 |
a linked list of every translated block contained in a given page. Other |
887 |
linked lists are also maintained to undo direct block chaining. |
888 |
|
889 |
Although the overhead of doing @code{mprotect()} calls is important, |
890 |
most MSDOS programs can be emulated at reasonnable speed with QEMU and |
891 |
DOSEMU. |
892 |
|
893 |
Note that QEMU also invalidates pages of translated code when it detects |
894 |
that memory mappings are modified with @code{mmap()} or @code{munmap()}. |
895 |
|
896 |
@section Exception support |
897 |
|
898 |
longjmp() is used when an exception such as division by zero is |
899 |
encountered. |
900 |
|
901 |
The host SIGSEGV and SIGBUS signal handlers are used to get invalid |
902 |
memory accesses. The exact CPU state can be retrieved because all the |
903 |
x86 registers are stored in fixed host registers. The simulated program |
904 |
counter is found by retranslating the corresponding basic block and by |
905 |
looking where the host program counter was at the exception point. |
906 |
|
907 |
The virtual CPU cannot retrieve the exact @code{EFLAGS} register because |
908 |
in some cases it is not computed because of condition code |
909 |
optimisations. It is not a big concern because the emulated code can |
910 |
still be restarted in any cases. |
911 |
|
912 |
@section Linux system call translation |
913 |
|
914 |
QEMU includes a generic system call translator for Linux. It means that |
915 |
the parameters of the system calls can be converted to fix the |
916 |
endianness and 32/64 bit issues. The IOCTLs are converted with a generic |
917 |
type description system (see @file{ioctls.h} and @file{thunk.c}). |
918 |
|
919 |
QEMU supports host CPUs which have pages bigger than 4KB. It records all |
920 |
the mappings the process does and try to emulated the @code{mmap()} |
921 |
system calls in cases where the host @code{mmap()} call would fail |
922 |
because of bad page alignment. |
923 |
|
924 |
@section Linux signals |
925 |
|
926 |
Normal and real-time signals are queued along with their information |
927 |
(@code{siginfo_t}) as it is done in the Linux kernel. Then an interrupt |
928 |
request is done to the virtual CPU. When it is interrupted, one queued |
929 |
signal is handled by generating a stack frame in the virtual CPU as the |
930 |
Linux kernel does. The @code{sigreturn()} system call is emulated to return |
931 |
from the virtual signal handler. |
932 |
|
933 |
Some signals (such as SIGALRM) directly come from the host. Other |
934 |
signals are synthetized from the virtual CPU exceptions such as SIGFPE |
935 |
when a division by zero is done (see @code{main.c:cpu_loop()}). |
936 |
|
937 |
The blocked signal mask is still handled by the host Linux kernel so |
938 |
that most signal system calls can be redirected directly to the host |
939 |
Linux kernel. Only the @code{sigaction()} and @code{sigreturn()} system |
940 |
calls need to be fully emulated (see @file{signal.c}). |
941 |
|
942 |
@section clone() system call and threads |
943 |
|
944 |
The Linux clone() system call is usually used to create a thread. QEMU |
945 |
uses the host clone() system call so that real host threads are created |
946 |
for each emulated thread. One virtual CPU instance is created for each |
947 |
thread. |
948 |
|
949 |
The virtual x86 CPU atomic operations are emulated with a global lock so |
950 |
that their semantic is preserved. |
951 |
|
952 |
Note that currently there are still some locking issues in QEMU. In |
953 |
particular, the translated cache flush is not protected yet against |
954 |
reentrancy. |
955 |
|
956 |
@section Self-virtualization |
957 |
|
958 |
QEMU was conceived so that ultimately it can emulate itself. Although |
959 |
it is not very useful, it is an important test to show the power of the |
960 |
emulator. |
961 |
|
962 |
Achieving self-virtualization is not easy because there may be address |
963 |
space conflicts. QEMU solves this problem by being an executable ELF |
964 |
shared object as the ld-linux.so ELF interpreter. That way, it can be |
965 |
relocated at load time. |
966 |
|
967 |
@section MMU emulation |
968 |
|
969 |
For system emulation, QEMU uses the mmap() system call to emulate the |
970 |
target CPU MMU. It works as long the emulated OS does not use an area |
971 |
reserved by the host OS (such as the area above 0xc0000000 on x86 |
972 |
Linux). |
973 |
|
974 |
It is planned to add a slower but more precise MMU emulation |
975 |
with a software MMU. |
976 |
|
977 |
@section Bibliography |
978 |
|
979 |
@table @asis |
980 |
|
981 |
@item [1] |
982 |
@url{http://citeseer.nj.nec.com/piumarta98optimizing.html}, Optimizing |
983 |
direct threaded code by selective inlining (1998) by Ian Piumarta, Fabio |
984 |
Riccardi. |
985 |
|
986 |
@item [2] |
987 |
@url{http://developer.kde.org/~sewardj/}, Valgrind, an open-source |
988 |
memory debugger for x86-GNU/Linux, by Julian Seward. |
989 |
|
990 |
@item [3] |
991 |
@url{http://bochs.sourceforge.net/}, the Bochs IA-32 Emulator Project, |
992 |
by Kevin Lawton et al. |
993 |
|
994 |
@item [4] |
995 |
@url{http://www.cs.rose-hulman.edu/~donaldlf/em86/index.html}, the EM86 |
996 |
x86 emulator on Alpha-Linux. |
997 |
|
998 |
@item [5] |
999 |
@url{http://www.usenix.org/publications/library/proceedings/usenix-nt97/full_papers/chernoff/chernoff.pdf}, |
1000 |
DIGITAL FX!32: Running 32-Bit x86 Applications on Alpha NT, by Anton |
1001 |
Chernoff and Ray Hookway. |
1002 |
|
1003 |
@item [6] |
1004 |
@url{http://www.willows.com/}, Windows API library emulation from |
1005 |
Willows Software. |
1006 |
|
1007 |
@item [7] |
1008 |
@url{http://user-mode-linux.sourceforge.net/}, |
1009 |
The User-mode Linux Kernel. |
1010 |
|
1011 |
@item [8] |
1012 |
@url{http://www.plex86.org/}, |
1013 |
The new Plex86 project. |
1014 |
|
1015 |
@end table |
1016 |
|
1017 |
@chapter Regression Tests |
1018 |
|
1019 |
In the directory @file{tests/}, various interesting testing programs |
1020 |
are available. There are used for regression testing. |
1021 |
|
1022 |
@section @file{hello-i386} |
1023 |
|
1024 |
Very simple statically linked x86 program, just to test QEMU during a |
1025 |
port to a new host CPU. |
1026 |
|
1027 |
@section @file{hello-arm} |
1028 |
|
1029 |
Very simple statically linked ARM program, just to test QEMU during a |
1030 |
port to a new host CPU. |
1031 |
|
1032 |
@section @file{test-i386} |
1033 |
|
1034 |
This program executes most of the 16 bit and 32 bit x86 instructions and |
1035 |
generates a text output. It can be compared with the output obtained with |
1036 |
a real CPU or another emulator. The target @code{make test} runs this |
1037 |
program and a @code{diff} on the generated output. |
1038 |
|
1039 |
The Linux system call @code{modify_ldt()} is used to create x86 selectors |
1040 |
to test some 16 bit addressing and 32 bit with segmentation cases. |
1041 |
|
1042 |
The Linux system call @code{vm86()} is used to test vm86 emulation. |
1043 |
|
1044 |
Various exceptions are raised to test most of the x86 user space |
1045 |
exception reporting. |
1046 |
|
1047 |
@section @file{sha1} |
1048 |
|
1049 |
It is a simple benchmark. Care must be taken to interpret the results |
1050 |
because it mostly tests the ability of the virtual CPU to optimize the |
1051 |
@code{rol} x86 instruction and the condition code computations. |
1052 |
|