Statistics
| Branch: | Revision:

root / qemu-tech.texi @ a785e42e

History | View | Annotate | Download (20.1 kB)

1
\input texinfo @c -*- texinfo -*-
2
@c %**start of header
3
@setfilename qemu-tech.info
4
@settitle QEMU Internals
5
@exampleindent 0
6
@paragraphindent 0
7
@c %**end of header
8

    
9
@iftex
10
@titlepage
11
@sp 7
12
@center @titlefont{QEMU Internals}
13
@sp 3
14
@end titlepage
15
@end iftex
16

    
17
@ifnottex
18
@node Top
19
@top
20

    
21
@menu
22
* Introduction::
23
* QEMU Internals::
24
* Regression Tests::
25
* Index::
26
@end menu
27
@end ifnottex
28

    
29
@contents
30

    
31
@node Introduction
32
@chapter Introduction
33

    
34
@menu
35
* intro_features::        Features
36
* intro_x86_emulation::   x86 emulation
37
* intro_arm_emulation::   ARM emulation
38
* intro_mips_emulation::  MIPS emulation
39
* intro_ppc_emulation::   PowerPC emulation
40
* intro_sparc_emulation:: SPARC emulation
41
@end menu
42

    
43
@node intro_features
44
@section Features
45

    
46
QEMU is a FAST! processor emulator using a portable dynamic
47
translator.
48

    
49
QEMU has two operating modes:
50

    
51
@itemize @minus
52

    
53
@item
54
Full system emulation. In this mode, QEMU emulates a full system
55
(usually a PC), including a processor and various peripherals. It can
56
be used to launch an different Operating System without rebooting the
57
PC or to debug system code.
58

    
59
@item
60
User mode emulation (Linux host only). In this mode, QEMU can launch
61
Linux processes compiled for one CPU on another CPU. It can be used to
62
launch the Wine Windows API emulator (@url{http://www.winehq.org}) or
63
to ease cross-compilation and cross-debugging.
64

    
65
@end itemize
66

    
67
As QEMU requires no host kernel driver to run, it is very safe and
68
easy to use.
69

    
70
QEMU generic features:
71

    
72
@itemize
73

    
74
@item User space only or full system emulation.
75

    
76
@item Using dynamic translation to native code for reasonable speed.
77

    
78
@item Working on x86 and PowerPC hosts. Being tested on ARM, Sparc32, Alpha and S390.
79

    
80
@item Self-modifying code support.
81

    
82
@item Precise exceptions support.
83

    
84
@item The virtual CPU is a library (@code{libqemu}) which can be used
85
in other projects (look at @file{qemu/tests/qruncom.c} to have an
86
example of user mode @code{libqemu} usage).
87

    
88
@end itemize
89

    
90
QEMU user mode emulation features:
91
@itemize
92
@item Generic Linux system call converter, including most ioctls.
93

    
94
@item clone() emulation using native CPU clone() to use Linux scheduler for threads.
95

    
96
@item Accurate signal handling by remapping host signals to target signals.
97
@end itemize
98

    
99
QEMU full system emulation features:
100
@itemize
101
@item QEMU can either use a full software MMU for maximum portability or use the host system call mmap() to simulate the target MMU.
102
@end itemize
103

    
104
@node intro_x86_emulation
105
@section x86 emulation
106

    
107
QEMU x86 target features:
108

    
109
@itemize
110

    
111
@item The virtual x86 CPU supports 16 bit and 32 bit addressing with segmentation.
112
LDT/GDT and IDT are emulated. VM86 mode is also supported to run DOSEMU.
113

    
114
@item Support of host page sizes bigger than 4KB in user mode emulation.
115

    
116
@item QEMU can emulate itself on x86.
117

    
118
@item An extensive Linux x86 CPU test program is included @file{tests/test-i386}.
119
It can be used to test other x86 virtual CPUs.
120

    
121
@end itemize
122

    
123
Current QEMU limitations:
124

    
125
@itemize
126

    
127
@item No SSE/MMX support (yet).
128

    
129
@item No x86-64 support.
130

    
131
@item IPC syscalls are missing.
132

    
133
@item The x86 segment limits and access rights are not tested at every
134
memory access (yet). Hopefully, very few OSes seem to rely on that for
135
normal use.
136

    
137
@item On non x86 host CPUs, @code{double}s are used instead of the non standard
138
10 byte @code{long double}s of x86 for floating point emulation to get
139
maximum performances.
140

    
141
@end itemize
142

    
143
@node intro_arm_emulation
144
@section ARM emulation
145

    
146
@itemize
147

    
148
@item Full ARM 7 user emulation.
149

    
150
@item NWFPE FPU support included in user Linux emulation.
151

    
152
@item Can run most ARM Linux binaries.
153

    
154
@end itemize
155

    
156
@node intro_mips_emulation
157
@section MIPS emulation
158

    
159
@itemize
160

    
161
@item The system emulation allows full MIPS32/MIPS64 Release 2 emulation,
162
including privileged instructions, FPU and MMU, in both little and big
163
endian modes.
164

    
165
@item The Linux userland emulation can run many 32 bit MIPS Linux binaries.
166

    
167
@end itemize
168

    
169
Current QEMU limitations:
170

    
171
@itemize
172

    
173
@item Self-modifying code is not always handled correctly.
174

    
175
@item 64 bit userland emulation is not implemented.
176

    
177
@item The system emulation is not complete enough to run real firmware.
178

    
179
@item The watchpoint debug facility is not implemented.
180

    
181
@end itemize
182

    
183
@node intro_ppc_emulation
184
@section PowerPC emulation
185

    
186
@itemize
187

    
188
@item Full PowerPC 32 bit emulation, including privileged instructions,
189
FPU and MMU.
190

    
191
@item Can run most PowerPC Linux binaries.
192

    
193
@end itemize
194

    
195
@node intro_sparc_emulation
196
@section SPARC emulation
197

    
198
@itemize
199

    
200
@item Full SPARC V8 emulation, including privileged
201
instructions, FPU and MMU. SPARC V9 emulation includes most privileged
202
and VIS instructions, FPU and I/D MMU. Alignment is fully enforced.
203

    
204
@item Can run most 32-bit SPARC Linux binaries, SPARC32PLUS Linux binaries and
205
some 64-bit SPARC Linux binaries.
206

    
207
@end itemize
208

    
209
Current QEMU limitations:
210

    
211
@itemize
212

    
213
@item IPC syscalls are missing.
214

    
215
@item 128-bit floating point operations are not supported, though none of the
216
real CPUs implement them either.  Floating point exception support is untested.
217

    
218
@item Atomic instructions are not correctly implemented.
219

    
220
@item Sparc64 emulators are not usable for anything yet.
221

    
222
@end itemize
223

    
224
@node QEMU Internals
225
@chapter QEMU Internals
226

    
227
@menu
228
* QEMU compared to other emulators::
229
* Portable dynamic translation::
230
* Register allocation::
231
* Condition code optimisations::
232
* CPU state optimisations::
233
* Translation cache::
234
* Direct block chaining::
235
* Self-modifying code and translated code invalidation::
236
* Exception support::
237
* MMU emulation::
238
* Hardware interrupts::
239
* User emulation specific details::
240
* Bibliography::
241
@end menu
242

    
243
@node QEMU compared to other emulators
244
@section QEMU compared to other emulators
245

    
246
Like bochs [3], QEMU emulates an x86 CPU. But QEMU is much faster than
247
bochs as it uses dynamic compilation. Bochs is closely tied to x86 PC
248
emulation while QEMU can emulate several processors.
249

    
250
Like Valgrind [2], QEMU does user space emulation and dynamic
251
translation. Valgrind is mainly a memory debugger while QEMU has no
252
support for it (QEMU could be used to detect out of bound memory
253
accesses as Valgrind, but it has no support to track uninitialised data
254
as Valgrind does). The Valgrind dynamic translator generates better code
255
than QEMU (in particular it does register allocation) but it is closely
256
tied to an x86 host and target and has no support for precise exceptions
257
and system emulation.
258

    
259
EM86 [4] is the closest project to user space QEMU (and QEMU still uses
260
some of its code, in particular the ELF file loader). EM86 was limited
261
to an alpha host and used a proprietary and slow interpreter (the
262
interpreter part of the FX!32 Digital Win32 code translator [5]).
263

    
264
TWIN [6] is a Windows API emulator like Wine. It is less accurate than
265
Wine but includes a protected mode x86 interpreter to launch x86 Windows
266
executables. Such an approach has greater potential because most of the
267
Windows API is executed natively but it is far more difficult to develop
268
because all the data structures and function parameters exchanged
269
between the API and the x86 code must be converted.
270

    
271
User mode Linux [7] was the only solution before QEMU to launch a
272
Linux kernel as a process while not needing any host kernel
273
patches. However, user mode Linux requires heavy kernel patches while
274
QEMU accepts unpatched Linux kernels. The price to pay is that QEMU is
275
slower.
276

    
277
The new Plex86 [8] PC virtualizer is done in the same spirit as the
278
qemu-fast system emulator. It requires a patched Linux kernel to work
279
(you cannot launch the same kernel on your PC), but the patches are
280
really small. As it is a PC virtualizer (no emulation is done except
281
for some priveledged instructions), it has the potential of being
282
faster than QEMU. The downside is that a complicated (and potentially
283
unsafe) host kernel patch is needed.
284

    
285
The commercial PC Virtualizers (VMWare [9], VirtualPC [10], TwoOStwo
286
[11]) are faster than QEMU, but they all need specific, proprietary
287
and potentially unsafe host drivers. Moreover, they are unable to
288
provide cycle exact simulation as an emulator can.
289

    
290
@node Portable dynamic translation
291
@section Portable dynamic translation
292

    
293
QEMU is a dynamic translator. When it first encounters a piece of code,
294
it converts it to the host instruction set. Usually dynamic translators
295
are very complicated and highly CPU dependent. QEMU uses some tricks
296
which make it relatively easily portable and simple while achieving good
297
performances.
298

    
299
The basic idea is to split every x86 instruction into fewer simpler
300
instructions. Each simple instruction is implemented by a piece of C
301
code (see @file{target-i386/op.c}). Then a compile time tool
302
(@file{dyngen}) takes the corresponding object file (@file{op.o})
303
to generate a dynamic code generator which concatenates the simple
304
instructions to build a function (see @file{op.h:dyngen_code()}).
305

    
306
In essence, the process is similar to [1], but more work is done at
307
compile time.
308

    
309
A key idea to get optimal performances is that constant parameters can
310
be passed to the simple operations. For that purpose, dummy ELF
311
relocations are generated with gcc for each constant parameter. Then,
312
the tool (@file{dyngen}) can locate the relocations and generate the
313
appriopriate C code to resolve them when building the dynamic code.
314

    
315
That way, QEMU is no more difficult to port than a dynamic linker.
316

    
317
To go even faster, GCC static register variables are used to keep the
318
state of the virtual CPU.
319

    
320
@node Register allocation
321
@section Register allocation
322

    
323
Since QEMU uses fixed simple instructions, no efficient register
324
allocation can be done. However, because RISC CPUs have a lot of
325
register, most of the virtual CPU state can be put in registers without
326
doing complicated register allocation.
327

    
328
@node Condition code optimisations
329
@section Condition code optimisations
330

    
331
Good CPU condition codes emulation (@code{EFLAGS} register on x86) is a
332
critical point to get good performances. QEMU uses lazy condition code
333
evaluation: instead of computing the condition codes after each x86
334
instruction, it just stores one operand (called @code{CC_SRC}), the
335
result (called @code{CC_DST}) and the type of operation (called
336
@code{CC_OP}).
337

    
338
@code{CC_OP} is almost never explicitely set in the generated code
339
because it is known at translation time.
340

    
341
In order to increase performances, a backward pass is performed on the
342
generated simple instructions (see
343
@code{target-i386/translate.c:optimize_flags()}). When it can be proved that
344
the condition codes are not needed by the next instructions, no
345
condition codes are computed at all.
346

    
347
@node CPU state optimisations
348
@section CPU state optimisations
349

    
350
The x86 CPU has many internal states which change the way it evaluates
351
instructions. In order to achieve a good speed, the translation phase
352
considers that some state information of the virtual x86 CPU cannot
353
change in it. For example, if the SS, DS and ES segments have a zero
354
base, then the translator does not even generate an addition for the
355
segment base.
356

    
357
[The FPU stack pointer register is not handled that way yet].
358

    
359
@node Translation cache
360
@section Translation cache
361

    
362
A 16 MByte cache holds the most recently used translations. For
363
simplicity, it is completely flushed when it is full. A translation unit
364
contains just a single basic block (a block of x86 instructions
365
terminated by a jump or by a virtual CPU state change which the
366
translator cannot deduce statically).
367

    
368
@node Direct block chaining
369
@section Direct block chaining
370

    
371
After each translated basic block is executed, QEMU uses the simulated
372
Program Counter (PC) and other cpu state informations (such as the CS
373
segment base value) to find the next basic block.
374

    
375
In order to accelerate the most common cases where the new simulated PC
376
is known, QEMU can patch a basic block so that it jumps directly to the
377
next one.
378

    
379
The most portable code uses an indirect jump. An indirect jump makes
380
it easier to make the jump target modification atomic. On some host
381
architectures (such as x86 or PowerPC), the @code{JUMP} opcode is
382
directly patched so that the block chaining has no overhead.
383

    
384
@node Self-modifying code and translated code invalidation
385
@section Self-modifying code and translated code invalidation
386

    
387
Self-modifying code is a special challenge in x86 emulation because no
388
instruction cache invalidation is signaled by the application when code
389
is modified.
390

    
391
When translated code is generated for a basic block, the corresponding
392
host page is write protected if it is not already read-only (with the
393
system call @code{mprotect()}). Then, if a write access is done to the
394
page, Linux raises a SEGV signal. QEMU then invalidates all the
395
translated code in the page and enables write accesses to the page.
396

    
397
Correct translated code invalidation is done efficiently by maintaining
398
a linked list of every translated block contained in a given page. Other
399
linked lists are also maintained to undo direct block chaining.
400

    
401
Although the overhead of doing @code{mprotect()} calls is important,
402
most MSDOS programs can be emulated at reasonnable speed with QEMU and
403
DOSEMU.
404

    
405
Note that QEMU also invalidates pages of translated code when it detects
406
that memory mappings are modified with @code{mmap()} or @code{munmap()}.
407

    
408
When using a software MMU, the code invalidation is more efficient: if
409
a given code page is invalidated too often because of write accesses,
410
then a bitmap representing all the code inside the page is
411
built. Every store into that page checks the bitmap to see if the code
412
really needs to be invalidated. It avoids invalidating the code when
413
only data is modified in the page.
414

    
415
@node Exception support
416
@section Exception support
417

    
418
longjmp() is used when an exception such as division by zero is
419
encountered.
420

    
421
The host SIGSEGV and SIGBUS signal handlers are used to get invalid
422
memory accesses. The exact CPU state can be retrieved because all the
423
x86 registers are stored in fixed host registers. The simulated program
424
counter is found by retranslating the corresponding basic block and by
425
looking where the host program counter was at the exception point.
426

    
427
The virtual CPU cannot retrieve the exact @code{EFLAGS} register because
428
in some cases it is not computed because of condition code
429
optimisations. It is not a big concern because the emulated code can
430
still be restarted in any cases.
431

    
432
@node MMU emulation
433
@section MMU emulation
434

    
435
For system emulation, QEMU uses the mmap() system call to emulate the
436
target CPU MMU. It works as long the emulated OS does not use an area
437
reserved by the host OS (such as the area above 0xc0000000 on x86
438
Linux).
439

    
440
In order to be able to launch any OS, QEMU also supports a soft
441
MMU. In that mode, the MMU virtual to physical address translation is
442
done at every memory access. QEMU uses an address translation cache to
443
speed up the translation.
444

    
445
In order to avoid flushing the translated code each time the MMU
446
mappings change, QEMU uses a physically indexed translation cache. It
447
means that each basic block is indexed with its physical address.
448

    
449
When MMU mappings change, only the chaining of the basic blocks is
450
reset (i.e. a basic block can no longer jump directly to another one).
451

    
452
@node Hardware interrupts
453
@section Hardware interrupts
454

    
455
In order to be faster, QEMU does not check at every basic block if an
456
hardware interrupt is pending. Instead, the user must asynchrously
457
call a specific function to tell that an interrupt is pending. This
458
function resets the chaining of the currently executing basic
459
block. It ensures that the execution will return soon in the main loop
460
of the CPU emulator. Then the main loop can test if the interrupt is
461
pending and handle it.
462

    
463
@node User emulation specific details
464
@section User emulation specific details
465

    
466
@subsection Linux system call translation
467

    
468
QEMU includes a generic system call translator for Linux. It means that
469
the parameters of the system calls can be converted to fix the
470
endianness and 32/64 bit issues. The IOCTLs are converted with a generic
471
type description system (see @file{ioctls.h} and @file{thunk.c}).
472

    
473
QEMU supports host CPUs which have pages bigger than 4KB. It records all
474
the mappings the process does and try to emulated the @code{mmap()}
475
system calls in cases where the host @code{mmap()} call would fail
476
because of bad page alignment.
477

    
478
@subsection Linux signals
479

    
480
Normal and real-time signals are queued along with their information
481
(@code{siginfo_t}) as it is done in the Linux kernel. Then an interrupt
482
request is done to the virtual CPU. When it is interrupted, one queued
483
signal is handled by generating a stack frame in the virtual CPU as the
484
Linux kernel does. The @code{sigreturn()} system call is emulated to return
485
from the virtual signal handler.
486

    
487
Some signals (such as SIGALRM) directly come from the host. Other
488
signals are synthetized from the virtual CPU exceptions such as SIGFPE
489
when a division by zero is done (see @code{main.c:cpu_loop()}).
490

    
491
The blocked signal mask is still handled by the host Linux kernel so
492
that most signal system calls can be redirected directly to the host
493
Linux kernel. Only the @code{sigaction()} and @code{sigreturn()} system
494
calls need to be fully emulated (see @file{signal.c}).
495

    
496
@subsection clone() system call and threads
497

    
498
The Linux clone() system call is usually used to create a thread. QEMU
499
uses the host clone() system call so that real host threads are created
500
for each emulated thread. One virtual CPU instance is created for each
501
thread.
502

    
503
The virtual x86 CPU atomic operations are emulated with a global lock so
504
that their semantic is preserved.
505

    
506
Note that currently there are still some locking issues in QEMU. In
507
particular, the translated cache flush is not protected yet against
508
reentrancy.
509

    
510
@subsection Self-virtualization
511

    
512
QEMU was conceived so that ultimately it can emulate itself. Although
513
it is not very useful, it is an important test to show the power of the
514
emulator.
515

    
516
Achieving self-virtualization is not easy because there may be address
517
space conflicts. QEMU solves this problem by being an executable ELF
518
shared object as the ld-linux.so ELF interpreter. That way, it can be
519
relocated at load time.
520

    
521
@node Bibliography
522
@section Bibliography
523

    
524
@table @asis
525

    
526
@item [1]
527
@url{http://citeseer.nj.nec.com/piumarta98optimizing.html}, Optimizing
528
direct threaded code by selective inlining (1998) by Ian Piumarta, Fabio
529
Riccardi.
530

    
531
@item [2]
532
@url{http://developer.kde.org/~sewardj/}, Valgrind, an open-source
533
memory debugger for x86-GNU/Linux, by Julian Seward.
534

    
535
@item [3]
536
@url{http://bochs.sourceforge.net/}, the Bochs IA-32 Emulator Project,
537
by Kevin Lawton et al.
538

    
539
@item [4]
540
@url{http://www.cs.rose-hulman.edu/~donaldlf/em86/index.html}, the EM86
541
x86 emulator on Alpha-Linux.
542

    
543
@item [5]
544
@url{http://www.usenix.org/publications/library/proceedings/usenix-nt97/@/full_papers/chernoff/chernoff.pdf},
545
DIGITAL FX!32: Running 32-Bit x86 Applications on Alpha NT, by Anton
546
Chernoff and Ray Hookway.
547

    
548
@item [6]
549
@url{http://www.willows.com/}, Windows API library emulation from
550
Willows Software.
551

    
552
@item [7]
553
@url{http://user-mode-linux.sourceforge.net/},
554
The User-mode Linux Kernel.
555

    
556
@item [8]
557
@url{http://www.plex86.org/},
558
The new Plex86 project.
559

    
560
@item [9]
561
@url{http://www.vmware.com/},
562
The VMWare PC virtualizer.
563

    
564
@item [10]
565
@url{http://www.microsoft.com/windowsxp/virtualpc/},
566
The VirtualPC PC virtualizer.
567

    
568
@item [11]
569
@url{http://www.twoostwo.org/},
570
The TwoOStwo PC virtualizer.
571

    
572
@end table
573

    
574
@node Regression Tests
575
@chapter Regression Tests
576

    
577
In the directory @file{tests/}, various interesting testing programs
578
are available. They are used for regression testing.
579

    
580
@menu
581
* test-i386::
582
* linux-test::
583
* qruncom.c::
584
@end menu
585

    
586
@node test-i386
587
@section @file{test-i386}
588

    
589
This program executes most of the 16 bit and 32 bit x86 instructions and
590
generates a text output. It can be compared with the output obtained with
591
a real CPU or another emulator. The target @code{make test} runs this
592
program and a @code{diff} on the generated output.
593

    
594
The Linux system call @code{modify_ldt()} is used to create x86 selectors
595
to test some 16 bit addressing and 32 bit with segmentation cases.
596

    
597
The Linux system call @code{vm86()} is used to test vm86 emulation.
598

    
599
Various exceptions are raised to test most of the x86 user space
600
exception reporting.
601

    
602
@node linux-test
603
@section @file{linux-test}
604

    
605
This program tests various Linux system calls. It is used to verify
606
that the system call parameters are correctly converted between target
607
and host CPUs.
608

    
609
@node qruncom.c
610
@section @file{qruncom.c}
611

    
612
Example of usage of @code{libqemu} to emulate a user mode i386 CPU.
613

    
614
@node Index
615
@chapter Index
616
@printindex cp
617

    
618
@bye